码迷,mamicode.com
首页 > 其他好文 > 详细

DecimalFormat详解

时间:2018-10-31 14:03:25      阅读:146      评论:0      收藏:0      [点我收藏+]

标签:iterator   tde   elf   parse   next   ctf   第一个   har   item   

  DecimalFormat继承自NumberFormat,可以使用它将十进制的数以不同形式格式化为字符串形式,可以控制前导和尾随0、前缀、后缀、分组(千)、小数分隔符等,如果要更改格式符号(例如小数点分隔符)。

它有各种各样的设计使其能够解析和格式化数字的特征区域设置,包括对西方、阿拉伯和印度数字的支持。它也支持不同类型的数字,包括整数(123),定点数字(123.4),科学符号(1.23 e4),百分比(12%),以及货币金额(123美元)。所有这些都可以本地化。

可以将其DecimalFormatSymbolsDecimalFormat一起使用这些类在数字格式化方面提供了极大的灵活性,但它们可以代码更加复杂。

 

包:import java.text.DecimalFormat;

 

构建模式

  使用DecimalFormat模式指定格式设置String的模式属性pattern:

DecimalFormat df = new DecimalFormat(pattern);

关于pattern,有:

数字格式模式语法

 

可以按照以下BNF(巴科斯范式)图表指定的规则为数字设计自己的格式模式:

pattern: subpattern {; subpattern}
subpattern(子模式): {prefix} integer {.fraction(分数)} {suffix}

prefix(前缀):‘\\ u0000‘..‘\\ uFFFD‘ - specialCharacters suffix(后缀):‘\\ u0000‘..‘\\ uFFFD‘ - specialCharacters integer:‘#‘*‘0‘*‘0‘ 分数:  ‘0‘*‘#‘*

上图中使用的符号如下表所示:

符号描述
X* 0个或更多个X实例
(X | Y) X或Y
X..Y 从X到Y的任何字符,包括在内
S - T S中的字符,除了T中的字符
{X} X是可选的
模式中的特殊符号
符号  描述
0 一个数字
# 一个数字,0显示为缺省(即空字符) 
. 小数点
分组分隔符
E 指数表示的底数指数分隔符
分隔格式
- 负号前缀
% 乘以100并显示为百分比
乘以1000并显示为千分比
¤ 货币符号; 用货币符号代替; 如果加倍,用国际货币符号代替; 如果存在于模式中,则使用货币小数分隔符而不是小数分隔符
X 任何其他字符都可以在前缀或者后缀中使用
用于引号前缀或后缀中的特殊符号

 

示例:

DecimalFormatDemo程序输出
valuepatternoutput说明
123456.789 ###,###。### 123,456.789 井号(#)表示一个数字,逗号是分组分隔符的占位符,句点是小数分隔符的占位符。
123456.789 ###。## 123456.79 value有三个数字的小数点右边,但pattern只有两个。format方法通过四舍五入来处理这个问题。
123.78 000000.000 000123.780 pattern指定前导和尾随零,因为0字符被用来代替井号(#)。
12345.67 $ ###,###。### $ 12,345.67 pattern美元符号($)中的第一个字符请注意,它紧接在格式化的最左边的数字之前output
12345.67 \ u00A5 ###,###。### ¥12,345.67 pattern与Unicode值00A5指定为日元(¥)货币符号。
 1 DecimalFormat df = new DecimalFormat("000.###");
 2         String s = df.format( 1.11111);
 3         System.out.println(s);
 4         //output:   001.111
 5 
 6         df.applyPattern("000.000");  //重新应用新的模式
 7         System.out.println(df.format(1.11111));
 8         //output: 001.111
 9 
10         df.applyPattern("000E000");
11         System.out.println(df.format(1111111));
12         //output: 111E004  (1111111 约等于111 X 10^4)
13         df.applyPattern("0E0");
14         System.out.println(df.format(1000000000));
15         //output: 1E9
16 
17         df.applyPattern("###,###,###.###");
18         System.out.println(df.format(11111111.11));
19         //output:11,111,111.11
20 
21         df.applyPattern("0.0%");
22         System.out.println(df.format(0.1));
23         //output: 10.0%
24 
25          df.applyPattern("sd.00");  //可添加其他常规字符
26         System.out.println(df.format(0.1));
27         //output: sd.10

 

 

区域敏感格式

前面的示例DecimalFormat为默认值创建了一个对象Locale如果您想要一个DecimalFormat非默认对象Locale,则实例化a NumberFormat然后将其强制转换为DecimalFormat这是一个例子:

NumberFormat nf = NumberFormat.getNumberInstance(loc);
DecimalFormat df =(DecimalFormat)nf;
df.applyPattern(图案);
String output = df.format(value);
System.out.println(pattern +“”+ output +“”+ loc.toString());

运行上一个代码示例将导致后面的输出。格式化的数字位于第二列,具体取决于Locale

###,###。### 123,456.789 zh_CN
###,###。### 123.456,789 de_DE
###,###。### 123 456,789 fr_FR

到目前为止,这里讨论的格式模式遵循美国英语的惯例。例如,在模式###,###。##中,逗号是千位分隔符,句点表示小数点。如果您的最终用户没有接触到它,那么这个约定很好。但是,某些应用程序(如电子表格和报表生成器)允许最终用户定义自己的格式设置模式。对于这些应用程序,最终用户指定的格式模式应使用本地化表示法。在这些情况下,您将要调用applyLocalizedPatternDecimalFormat对象上的方法

改变格式符号

您可以使用 DecimalFormatSymbols类更改format方法生成的格式化数字中出现的符号这些符号包括小数分隔符,分组分隔符,减号和百分号等。

下一个示例DecimalFormatSymbols通过将奇怪的格式应用于数字来演示该类。这种不寻常的格式是调用的结果setDecimalSeparatorsetGroupingSeparatorsetGroupingSize方法。

DecimalFormatSymbols unusualSymbols = new DecimalFormatSymbols(currentLocale);
unusualSymbols.setDecimalSeparator( ‘|‘);
unusualSymbols.setGroupingSeparator( ‘^‘);

String strange =“#,## 0。###”;
DecimalFormat weirdFormatter = new DecimalFormat(strange,unusualSymbols);
weirdFormatter.setGroupingSize(4);

String bizarre = weirdFormatter.format(12345.678);
的System.out.println(奇异);

运行时,此示例以奇怪的格式打印数字:

1 ^ 2345 | 678


date:
2018-10-31  12:57:51

参考:
https://docs.oracle.com/javase/tutorial/i18n/format/decimalFormat.html#numberpattern

附DecimalFormat类源码:
   1 /*
   2  * Copyright (c) 1996, 2017, Oracle and/or its affiliates. All rights reserved.
   3  * ORACLE PROPRIETARY/CONFIDENTIAL. Use is subject to license terms.
   4  */
   5 
   6 /*
   7  * (C) Copyright Taligent, Inc. 1996, 1997 - All Rights Reserved
   8  * (C) Copyright IBM Corp. 1996 - 1998 - All Rights Reserved
   9  *
  10  *   The original version of this source code and documentation is copyrighted
  11  * and owned by Taligent, Inc., a wholly-owned subsidiary of IBM. These
  12  * materials are provided under terms of a License Agreement between Taligent
  13  * and Sun. This technology is protected by multiple US and International
  14  * patents. This notice and attribution to Taligent may not be removed.
  15  *   Taligent is a registered trademark of Taligent, Inc.
  16  *
  17  */
  18 
  19 package java.text;
  20 
  21 import java.io.IOException;
  22 import java.io.InvalidObjectException;
  23 import java.io.ObjectInputStream;
  24 import java.math.BigDecimal;
  25 import java.math.BigInteger;
  26 import java.math.RoundingMode;
  27 import java.text.spi.NumberFormatProvider;
  28 import java.util.ArrayList;
  29 import java.util.Currency;
  30 import java.util.Locale;
  31 import java.util.ResourceBundle;
  32 import java.util.concurrent.ConcurrentHashMap;
  33 import java.util.concurrent.ConcurrentMap;
  34 import java.util.concurrent.atomic.AtomicInteger;
  35 import java.util.concurrent.atomic.AtomicLong;
  36 import sun.util.locale.provider.LocaleProviderAdapter;
  37 import sun.util.locale.provider.ResourceBundleBasedAdapter;
  38 
  39 /**
  40  * <code>DecimalFormat</code> is a concrete subclass of
  41  * <code>NumberFormat</code> that formats decimal numbers. It has a variety of
  42  * features designed to make it possible to parse and format numbers in any
  43  * locale, including support for Western, Arabic, and Indic digits.  It also
  44  * supports different kinds of numbers, including integers (123), fixed-point
  45  * numbers (123.4), scientific notation (1.23E4), percentages (12%), and
  46  * currency amounts ($123).  All of these can be localized.
  47  *
  48  * <p>To obtain a <code>NumberFormat</code> for a specific locale, including the
  49  * default locale, call one of <code>NumberFormat</code>‘s factory methods, such
  50  * as <code>getInstance()</code>.  In general, do not call the
  51  * <code>DecimalFormat</code> constructors directly, since the
  52  * <code>NumberFormat</code> factory methods may return subclasses other than
  53  * <code>DecimalFormat</code>. If you need to customize the format object, do
  54  * something like this:
  55  *
  56  * <blockquote><pre>
  57  * NumberFormat f = NumberFormat.getInstance(loc);
  58  * if (f instanceof DecimalFormat) {
  59  *     ((DecimalFormat) f).setDecimalSeparatorAlwaysShown(true);
  60  * }
  61  * </pre></blockquote>
  62  *
  63  * <p>A <code>DecimalFormat</code> comprises a <em>pattern</em> and a set of
  64  * <em>symbols</em>.  The pattern may be set directly using
  65  * <code>applyPattern()</code>, or indirectly using the API methods.  The
  66  * symbols are stored in a <code>DecimalFormatSymbols</code> object.  When using
  67  * the <code>NumberFormat</code> factory methods, the pattern and symbols are
  68  * read from localized <code>ResourceBundle</code>s.
  69  *
  70  * <h3>Patterns</h3>
  71  *
  72  * <code>DecimalFormat</code> patterns have the following syntax:
  73  * <blockquote><pre>
  74  * <i>Pattern:</i>
  75  *         <i>PositivePattern</i>
  76  *         <i>PositivePattern</i> ; <i>NegativePattern</i>
  77  * <i>PositivePattern:</i>
  78  *         <i>Prefix<sub>opt</sub></i> <i>Number</i> <i>Suffix<sub>opt</sub></i>
  79  * <i>NegativePattern:</i>
  80  *         <i>Prefix<sub>opt</sub></i> <i>Number</i> <i>Suffix<sub>opt</sub></i>
  81  * <i>Prefix:</i>
  82  *         any Unicode characters except &#92;uFFFE, &#92;uFFFF, and special characters
  83  * <i>Suffix:</i>
  84  *         any Unicode characters except &#92;uFFFE, &#92;uFFFF, and special characters
  85  * <i>Number:</i>
  86  *         <i>Integer</i> <i>Exponent<sub>opt</sub></i>
  87  *         <i>Integer</i> . <i>Fraction</i> <i>Exponent<sub>opt</sub></i>
  88  * <i>Integer:</i>
  89  *         <i>MinimumInteger</i>
  90  *         #
  91  *         # <i>Integer</i>
  92  *         # , <i>Integer</i>
  93  * <i>MinimumInteger:</i>
  94  *         0
  95  *         0 <i>MinimumInteger</i>
  96  *         0 , <i>MinimumInteger</i>
  97  * <i>Fraction:</i>
  98  *         <i>MinimumFraction<sub>opt</sub></i> <i>OptionalFraction<sub>opt</sub></i>
  99  * <i>MinimumFraction:</i>
 100  *         0 <i>MinimumFraction<sub>opt</sub></i>
 101  * <i>OptionalFraction:</i>
 102  *         # <i>OptionalFraction<sub>opt</sub></i>
 103  * <i>Exponent:</i>
 104  *         E <i>MinimumExponent</i>
 105  * <i>MinimumExponent:</i>
 106  *         0 <i>MinimumExponent<sub>opt</sub></i>
 107  * </pre></blockquote>
 108  *
 109  * <p>A <code>DecimalFormat</code> pattern contains a positive and negative
 110  * subpattern, for example, <code>"#,##0.00;(#,##0.00)"</code>.  Each
 111  * subpattern has a prefix, numeric part, and suffix. The negative subpattern
 112  * is optional; if absent, then the positive subpattern prefixed with the
 113  * localized minus sign (<code>‘-‘</code> in most locales) is used as the
 114  * negative subpattern. That is, <code>"0.00"</code> alone is equivalent to
 115  * <code>"0.00;-0.00"</code>.  If there is an explicit negative subpattern, it
 116  * serves only to specify the negative prefix and suffix; the number of digits,
 117  * minimal digits, and other characteristics are all the same as the positive
 118  * pattern. That means that <code>"#,##0.0#;(#)"</code> produces precisely
 119  * the same behavior as <code>"#,##0.0#;(#,##0.0#)"</code>.
 120  *
 121  * <p>The prefixes, suffixes, and various symbols used for infinity, digits,
 122  * thousands separators, decimal separators, etc. may be set to arbitrary
 123  * values, and they will appear properly during formatting.  However, care must
 124  * be taken that the symbols and strings do not conflict, or parsing will be
 125  * unreliable.  For example, either the positive and negative prefixes or the
 126  * suffixes must be distinct for <code>DecimalFormat.parse()</code> to be able
 127  * to distinguish positive from negative values.  (If they are identical, then
 128  * <code>DecimalFormat</code> will behave as if no negative subpattern was
 129  * specified.)  Another example is that the decimal separator and thousands
 130  * separator should be distinct characters, or parsing will be impossible.
 131  *
 132  * <p>The grouping separator is commonly used for thousands, but in some
 133  * countries it separates ten-thousands. The grouping size is a constant number
 134  * of digits between the grouping characters, such as 3 for 100,000,000 or 4 for
 135  * 1,0000,0000.  If you supply a pattern with multiple grouping characters, the
 136  * interval between the last one and the end of the integer is the one that is
 137  * used. So <code>"#,##,###,####"</code> == <code>"######,####"</code> ==
 138  * <code>"##,####,####"</code>.
 139  *
 140  * <h4>Special Pattern Characters</h4>
 141  *
 142  * <p>Many characters in a pattern are taken literally; they are matched during
 143  * parsing and output unchanged during formatting.  Special characters, on the
 144  * other hand, stand for other characters, strings, or classes of characters.
 145  * They must be quoted, unless noted otherwise, if they are to appear in the
 146  * prefix or suffix as literals.
 147  *
 148  * <p>The characters listed here are used in non-localized patterns.  Localized
 149  * patterns use the corresponding characters taken from this formatter‘s
 150  * <code>DecimalFormatSymbols</code> object instead, and these characters lose
 151  * their special status.  Two exceptions are the currency sign and quote, which
 152  * are not localized.
 153  *
 154  * <blockquote>
 155  * <table class="striped">
 156  * <caption style="display:none">Chart showing symbol, location, localized, and meaning.</caption>
 157  * <thead>
 158  *     <tr>
 159  *          <th style="text-align:left">Symbol
 160  *          <th style="text-align:left">Location
 161  *          <th style="text-align:left">Localized?
 162  *          <th style="text-align:left">Meaning
 163  * </thead>
 164  * <tbody>
 165  *     <tr style="vertical-align:top">
 166  *          <td><code>0</code>
 167  *          <td>Number
 168  *          <td>Yes
 169  *          <td>Digit
 170  *     <tr style="vertical-align: top">
 171  *          <td><code>#</code>
 172  *          <td>Number
 173  *          <td>Yes
 174  *          <td>Digit, zero shows as absent
 175  *     <tr style="vertical-align:top">
 176  *          <td><code>.</code>
 177  *          <td>Number
 178  *          <td>Yes
 179  *          <td>Decimal separator or monetary decimal separator
 180  *     <tr style="vertical-align: top">
 181  *          <td><code>-</code>
 182  *          <td>Number
 183  *          <td>Yes
 184  *          <td>Minus sign
 185  *     <tr style="vertical-align:top">
 186  *          <td><code>,</code>
 187  *          <td>Number
 188  *          <td>Yes
 189  *          <td>Grouping separator
 190  *     <tr style="vertical-align: top">
 191  *          <td><code>E</code>
 192  *          <td>Number
 193  *          <td>Yes
 194  *          <td>Separates mantissa and exponent in scientific notation.
 195  *              <em>Need not be quoted in prefix or suffix.</em>
 196  *     <tr style="vertical-align:top">
 197  *          <td><code>;</code>
 198  *          <td>Subpattern boundary
 199  *          <td>Yes
 200  *          <td>Separates positive and negative subpatterns
 201  *     <tr style="vertical-align: top">
 202  *          <td><code>%</code>
 203  *          <td>Prefix or suffix
 204  *          <td>Yes
 205  *          <td>Multiply by 100 and show as percentage
 206  *     <tr style="vertical-align:top">
 207  *          <td><code>&#92;u2030</code>
 208  *          <td>Prefix or suffix
 209  *          <td>Yes
 210  *          <td>Multiply by 1000 and show as per mille value
 211  *     <tr style="vertical-align: top">
 212  *          <td><code>&#164;</code> (<code>&#92;u00A4</code>)
 213  *          <td>Prefix or suffix
 214  *          <td>No
 215  *          <td>Currency sign, replaced by currency symbol.  If
 216  *              doubled, replaced by international currency symbol.
 217  *              If present in a pattern, the monetary decimal separator
 218  *              is used instead of the decimal separator.
 219  *     <tr style="vertical-align:top">
 220  *          <td><code>‘</code>
 221  *          <td>Prefix or suffix
 222  *          <td>No
 223  *          <td>Used to quote special characters in a prefix or suffix,
 224  *              for example, <code>"‘#‘#"</code> formats 123 to
 225  *              <code>"#123"</code>.  To create a single quote
 226  *              itself, use two in a row: <code>"# o‘‘clock"</code>.
 227  * </tbody>
 228  * </table>
 229  * </blockquote>
 230  *
 231  * <h4>Scientific Notation</h4>
 232  *
 233  * <p>Numbers in scientific notation are expressed as the product of a mantissa
 234  * and a power of ten, for example, 1234 can be expressed as 1.234 x 10^3.  The
 235  * mantissa is often in the range 1.0 &le; x {@literal <} 10.0, but it need not
 236  * be.
 237  * <code>DecimalFormat</code> can be instructed to format and parse scientific
 238  * notation <em>only via a pattern</em>; there is currently no factory method
 239  * that creates a scientific notation format.  In a pattern, the exponent
 240  * character immediately followed by one or more digit characters indicates
 241  * scientific notation.  Example: <code>"0.###E0"</code> formats the number
 242  * 1234 as <code>"1.234E3"</code>.
 243  *
 244  * <ul>
 245  * <li>The number of digit characters after the exponent character gives the
 246  * minimum exponent digit count.  There is no maximum.  Negative exponents are
 247  * formatted using the localized minus sign, <em>not</em> the prefix and suffix
 248  * from the pattern.  This allows patterns such as <code>"0.###E0 m/s"</code>.
 249  *
 250  * <li>The minimum and maximum number of integer digits are interpreted
 251  * together:
 252  *
 253  * <ul>
 254  * <li>If the maximum number of integer digits is greater than their minimum number
 255  * and greater than 1, it forces the exponent to be a multiple of the maximum
 256  * number of integer digits, and the minimum number of integer digits to be
 257  * interpreted as 1.  The most common use of this is to generate
 258  * <em>engineering notation</em>, in which the exponent is a multiple of three,
 259  * e.g., <code>"##0.#####E0"</code>. Using this pattern, the number 12345
 260  * formats to <code>"12.345E3"</code>, and 123456 formats to
 261  * <code>"123.456E3"</code>.
 262  *
 263  * <li>Otherwise, the minimum number of integer digits is achieved by adjusting the
 264  * exponent.  Example: 0.00123 formatted with <code>"00.###E0"</code> yields
 265  * <code>"12.3E-4"</code>.
 266  * </ul>
 267  *
 268  * <li>The number of significant digits in the mantissa is the sum of the
 269  * <em>minimum integer</em> and <em>maximum fraction</em> digits, and is
 270  * unaffected by the maximum integer digits.  For example, 12345 formatted with
 271  * <code>"##0.##E0"</code> is <code>"12.3E3"</code>. To show all digits, set
 272  * the significant digits count to zero.  The number of significant digits
 273  * does not affect parsing.
 274  *
 275  * <li>Exponential patterns may not contain grouping separators.
 276  * </ul>
 277  *
 278  * <h4>Rounding</h4>
 279  *
 280  * <code>DecimalFormat</code> provides rounding modes defined in
 281  * {@link java.math.RoundingMode} for formatting.  By default, it uses
 282  * {@link java.math.RoundingMode#HALF_EVEN RoundingMode.HALF_EVEN}.
 283  *
 284  * <h4>Digits</h4>
 285  *
 286  * For formatting, <code>DecimalFormat</code> uses the ten consecutive
 287  * characters starting with the localized zero digit defined in the
 288  * <code>DecimalFormatSymbols</code> object as digits. For parsing, these
 289  * digits as well as all Unicode decimal digits, as defined by
 290  * {@link Character#digit Character.digit}, are recognized.
 291  *
 292  * <h4>Special Values</h4>
 293  *
 294  * <p><code>NaN</code> is formatted as a string, which typically has a single character
 295  * <code>&#92;uFFFD</code>.  This string is determined by the
 296  * <code>DecimalFormatSymbols</code> object.  This is the only value for which
 297  * the prefixes and suffixes are not used.
 298  *
 299  * <p>Infinity is formatted as a string, which typically has a single character
 300  * <code>&#92;u221E</code>, with the positive or negative prefixes and suffixes
 301  * applied.  The infinity string is determined by the
 302  * <code>DecimalFormatSymbols</code> object.
 303  *
 304  * <p>Negative zero (<code>"-0"</code>) parses to
 305  * <ul>
 306  * <li><code>BigDecimal(0)</code> if <code>isParseBigDecimal()</code> is
 307  * true,
 308  * <li><code>Long(0)</code> if <code>isParseBigDecimal()</code> is false
 309  *     and <code>isParseIntegerOnly()</code> is true,
 310  * <li><code>Double(-0.0)</code> if both <code>isParseBigDecimal()</code>
 311  * and <code>isParseIntegerOnly()</code> are false.
 312  * </ul>
 313  *
 314  * <h4><a id="synchronization">Synchronization</a></h4>
 315  *
 316  * <p>
 317  * Decimal formats are generally not synchronized.
 318  * It is recommended to create separate format instances for each thread.
 319  * If multiple threads access a format concurrently, it must be synchronized
 320  * externally.
 321  *
 322  * <h4>Example</h4>
 323  *
 324  * <blockquote><pre>{@code
 325  * <strong>// Print out a number using the localized number, integer, currency,
 326  * // and percent format for each locale</strong>
 327  * Locale[] locales = NumberFormat.getAvailableLocales();
 328  * double myNumber = -1234.56;
 329  * NumberFormat form;
 330  * for (int j = 0; j < 4; ++j) {
 331  *     System.out.println("FORMAT");
 332  *     for (int i = 0; i < locales.length; ++i) {
 333  *         if (locales[i].getCountry().length() == 0) {
 334  *            continue; // Skip language-only locales
 335  *         }
 336  *         System.out.print(locales[i].getDisplayName());
 337  *         switch (j) {
 338  *         case 0:
 339  *             form = NumberFormat.getInstance(locales[i]); break;
 340  *         case 1:
 341  *             form = NumberFormat.getIntegerInstance(locales[i]); break;
 342  *         case 2:
 343  *             form = NumberFormat.getCurrencyInstance(locales[i]); break;
 344  *         default:
 345  *             form = NumberFormat.getPercentInstance(locales[i]); break;
 346  *         }
 347  *         if (form instanceof DecimalFormat) {
 348  *             System.out.print(": " + ((DecimalFormat) form).toPattern());
 349  *         }
 350  *         System.out.print(" -> " + form.format(myNumber));
 351  *         try {
 352  *             System.out.println(" -> " + form.parse(form.format(myNumber)));
 353  *         } catch (ParseException e) {}
 354  *     }
 355  * }
 356  * }</pre></blockquote>
 357  *
 358  * @see          <a href="http://docs.oracle.com/javase/tutorial/i18n/format/decimalFormat.html">Java Tutorial</a>
 359  * @see          NumberFormat
 360  * @see          DecimalFormatSymbols
 361  * @see          ParsePosition
 362  * @author       Mark Davis
 363  * @author       Alan Liu
 364  * @since 1.1
 365  */
 366 public class DecimalFormat extends NumberFormat {
 367 
 368     /**
 369      * Creates a DecimalFormat using the default pattern and symbols
 370      * for the default {@link java.util.Locale.Category#FORMAT FORMAT} locale.
 371      * This is a convenient way to obtain a
 372      * DecimalFormat when internationalization is not the main concern.
 373      * <p>
 374      * To obtain standard formats for a given locale, use the factory methods
 375      * on NumberFormat such as getNumberInstance. These factories will
 376      * return the most appropriate sub-class of NumberFormat for a given
 377      * locale.
 378      *
 379      * @see java.text.NumberFormat#getInstance
 380      * @see java.text.NumberFormat#getNumberInstance
 381      * @see java.text.NumberFormat#getCurrencyInstance
 382      * @see java.text.NumberFormat#getPercentInstance
 383      */
 384     public DecimalFormat() {
 385         // Get the pattern for the default locale.
 386         Locale def = Locale.getDefault(Locale.Category.FORMAT);
 387         LocaleProviderAdapter adapter = LocaleProviderAdapter.getAdapter(NumberFormatProvider.class, def);
 388         if (!(adapter instanceof ResourceBundleBasedAdapter)) {
 389             adapter = LocaleProviderAdapter.getResourceBundleBased();
 390         }
 391         String[] all = adapter.getLocaleResources(def).getNumberPatterns();
 392 
 393         // Always applyPattern after the symbols are set
 394         this.symbols = DecimalFormatSymbols.getInstance(def);
 395         applyPattern(all[0], false);
 396     }
 397 
 398 
 399     /**
 400      * Creates a DecimalFormat using the given pattern and the symbols
 401      * for the default {@link java.util.Locale.Category#FORMAT FORMAT} locale.
 402      * This is a convenient way to obtain a
 403      * DecimalFormat when internationalization is not the main concern.
 404      * <p>
 405      * To obtain standard formats for a given locale, use the factory methods
 406      * on NumberFormat such as getNumberInstance. These factories will
 407      * return the most appropriate sub-class of NumberFormat for a given
 408      * locale.
 409      *
 410      * @param pattern a non-localized pattern string.
 411      * @exception NullPointerException if <code>pattern</code> is null
 412      * @exception IllegalArgumentException if the given pattern is invalid.
 413      * @see java.text.NumberFormat#getInstance
 414      * @see java.text.NumberFormat#getNumberInstance
 415      * @see java.text.NumberFormat#getCurrencyInstance
 416      * @see java.text.NumberFormat#getPercentInstance
 417      */
 418     public DecimalFormat(String pattern) {
 419         // Always applyPattern after the symbols are set
 420         this.symbols = DecimalFormatSymbols.getInstance(Locale.getDefault(Locale.Category.FORMAT));
 421         applyPattern(pattern, false);
 422     }
 423 
 424 
 425     /**
 426      * Creates a DecimalFormat using the given pattern and symbols.
 427      * Use this constructor when you need to completely customize the
 428      * behavior of the format.
 429      * <p>
 430      * To obtain standard formats for a given
 431      * locale, use the factory methods on NumberFormat such as
 432      * getInstance or getCurrencyInstance. If you need only minor adjustments
 433      * to a standard format, you can modify the format returned by
 434      * a NumberFormat factory method.
 435      *
 436      * @param pattern a non-localized pattern string
 437      * @param symbols the set of symbols to be used
 438      * @exception NullPointerException if any of the given arguments is null
 439      * @exception IllegalArgumentException if the given pattern is invalid
 440      * @see java.text.NumberFormat#getInstance
 441      * @see java.text.NumberFormat#getNumberInstance
 442      * @see java.text.NumberFormat#getCurrencyInstance
 443      * @see java.text.NumberFormat#getPercentInstance
 444      * @see java.text.DecimalFormatSymbols
 445      */
 446     public DecimalFormat (String pattern, DecimalFormatSymbols symbols) {
 447         // Always applyPattern after the symbols are set
 448         this.symbols = (DecimalFormatSymbols)symbols.clone();
 449         applyPattern(pattern, false);
 450     }
 451 
 452 
 453     // Overrides
 454     /**
 455      * Formats a number and appends the resulting text to the given string
 456      * buffer.
 457      * The number can be of any subclass of {@link java.lang.Number}.
 458      * <p>
 459      * This implementation uses the maximum precision permitted.
 460      * @param number     the number to format
 461      * @param toAppendTo the <code>StringBuffer</code> to which the formatted
 462      *                   text is to be appended
 463      * @param pos        On input: an alignment field, if desired.
 464      *                   On output: the offsets of the alignment field.
 465      * @return           the value passed in as <code>toAppendTo</code>
 466      * @exception        IllegalArgumentException if <code>number</code> is
 467      *                   null or not an instance of <code>Number</code>.
 468      * @exception        NullPointerException if <code>toAppendTo</code> or
 469      *                   <code>pos</code> is null
 470      * @exception        ArithmeticException if rounding is needed with rounding
 471      *                   mode being set to RoundingMode.UNNECESSARY
 472      * @see              java.text.FieldPosition
 473      */
 474     @Override
 475     public final StringBuffer format(Object number,
 476                                      StringBuffer toAppendTo,
 477                                      FieldPosition pos) {
 478         if (number instanceof Long || number instanceof Integer ||
 479                    number instanceof Short || number instanceof Byte ||
 480                    number instanceof AtomicInteger ||
 481                    number instanceof AtomicLong ||
 482                    (number instanceof BigInteger &&
 483                     ((BigInteger)number).bitLength () < 64)) {
 484             return format(((Number)number).longValue(), toAppendTo, pos);
 485         } else if (number instanceof BigDecimal) {
 486             return format((BigDecimal)number, toAppendTo, pos);
 487         } else if (number instanceof BigInteger) {
 488             return format((BigInteger)number, toAppendTo, pos);
 489         } else if (number instanceof Number) {
 490             return format(((Number)number).doubleValue(), toAppendTo, pos);
 491         } else {
 492             throw new IllegalArgumentException("Cannot format given Object as a Number");
 493         }
 494     }
 495 
 496     /**
 497      * Formats a double to produce a string.
 498      * @param number    The double to format
 499      * @param result    where the text is to be appended
 500      * @param fieldPosition    On input: an alignment field, if desired.
 501      * On output: the offsets of the alignment field.
 502      * @exception NullPointerException if {@code result} or
 503      *            {@code fieldPosition} is {@code null}
 504      * @exception ArithmeticException if rounding is needed with rounding
 505      *            mode being set to RoundingMode.UNNECESSARY
 506      * @return The formatted number string
 507      * @see java.text.FieldPosition
 508      */
 509     @Override
 510     public StringBuffer format(double number, StringBuffer result,
 511                                FieldPosition fieldPosition) {
 512         // If fieldPosition is a DontCareFieldPosition instance we can
 513         // try to go to fast-path code.
 514         boolean tryFastPath = false;
 515         if (fieldPosition == DontCareFieldPosition.INSTANCE)
 516             tryFastPath = true;
 517         else {
 518             fieldPosition.setBeginIndex(0);
 519             fieldPosition.setEndIndex(0);
 520         }
 521 
 522         if (tryFastPath) {
 523             String tempResult = fastFormat(number);
 524             if (tempResult != null) {
 525                 result.append(tempResult);
 526                 return result;
 527             }
 528         }
 529 
 530         // if fast-path could not work, we fallback to standard code.
 531         return format(number, result, fieldPosition.getFieldDelegate());
 532     }
 533 
 534     /**
 535      * Formats a double to produce a string.
 536      * @param number    The double to format
 537      * @param result    where the text is to be appended
 538      * @param delegate notified of locations of sub fields
 539      * @exception       ArithmeticException if rounding is needed with rounding
 540      *                  mode being set to RoundingMode.UNNECESSARY
 541      * @return The formatted number string
 542      */
 543     private StringBuffer format(double number, StringBuffer result,
 544                                 FieldDelegate delegate) {
 545         if (Double.isNaN(number) ||
 546            (Double.isInfinite(number) && multiplier == 0)) {
 547             int iFieldStart = result.length();
 548             result.append(symbols.getNaN());
 549             delegate.formatted(INTEGER_FIELD, Field.INTEGER, Field.INTEGER,
 550                                iFieldStart, result.length(), result);
 551             return result;
 552         }
 553 
 554         /* Detecting whether a double is negative is easy with the exception of
 555          * the value -0.0.  This is a double which has a zero mantissa (and
 556          * exponent), but a negative sign bit.  It is semantically distinct from
 557          * a zero with a positive sign bit, and this distinction is important
 558          * to certain kinds of computations.  However, it‘s a little tricky to
 559          * detect, since (-0.0 == 0.0) and !(-0.0 < 0.0).  How then, you may
 560          * ask, does it behave distinctly from +0.0?  Well, 1/(-0.0) ==
 561          * -Infinity.  Proper detection of -0.0 is needed to deal with the
 562          * issues raised by bugs 4106658, 4106667, and 4147706.  Liu 7/6/98.
 563          */
 564         boolean isNegative = ((number < 0.0) || (number == 0.0 && 1/number < 0.0)) ^ (multiplier < 0);
 565 
 566         if (multiplier != 1) {
 567             number *= multiplier;
 568         }
 569 
 570         if (Double.isInfinite(number)) {
 571             if (isNegative) {
 572                 append(result, negativePrefix, delegate,
 573                        getNegativePrefixFieldPositions(), Field.SIGN);
 574             } else {
 575                 append(result, positivePrefix, delegate,
 576                        getPositivePrefixFieldPositions(), Field.SIGN);
 577             }
 578 
 579             int iFieldStart = result.length();
 580             result.append(symbols.getInfinity());
 581             delegate.formatted(INTEGER_FIELD, Field.INTEGER, Field.INTEGER,
 582                                iFieldStart, result.length(), result);
 583 
 584             if (isNegative) {
 585                 append(result, negativeSuffix, delegate,
 586                        getNegativeSuffixFieldPositions(), Field.SIGN);
 587             } else {
 588                 append(result, positiveSuffix, delegate,
 589                        getPositiveSuffixFieldPositions(), Field.SIGN);
 590             }
 591 
 592             return result;
 593         }
 594 
 595         if (isNegative) {
 596             number = -number;
 597         }
 598 
 599         // at this point we are guaranteed a nonnegative finite number.
 600         assert(number >= 0 && !Double.isInfinite(number));
 601 
 602         synchronized(digitList) {
 603             int maxIntDigits = super.getMaximumIntegerDigits();
 604             int minIntDigits = super.getMinimumIntegerDigits();
 605             int maxFraDigits = super.getMaximumFractionDigits();
 606             int minFraDigits = super.getMinimumFractionDigits();
 607 
 608             digitList.set(isNegative, number, useExponentialNotation ?
 609                           maxIntDigits + maxFraDigits : maxFraDigits,
 610                           !useExponentialNotation);
 611             return subformat(result, delegate, isNegative, false,
 612                        maxIntDigits, minIntDigits, maxFraDigits, minFraDigits);
 613         }
 614     }
 615 
 616     /**
 617      * Format a long to produce a string.
 618      * @param number    The long to format
 619      * @param result    where the text is to be appended
 620      * @param fieldPosition    On input: an alignment field, if desired.
 621      * On output: the offsets of the alignment field.
 622      * @exception       NullPointerException if {@code result} or
 623      *                  {@code fieldPosition} is {@code null}
 624      * @exception       ArithmeticException if rounding is needed with rounding
 625      *                  mode being set to RoundingMode.UNNECESSARY
 626      * @return The formatted number string
 627      * @see java.text.FieldPosition
 628      */
 629     @Override
 630     public StringBuffer format(long number, StringBuffer result,
 631                                FieldPosition fieldPosition) {
 632         fieldPosition.setBeginIndex(0);
 633         fieldPosition.setEndIndex(0);
 634 
 635         return format(number, result, fieldPosition.getFieldDelegate());
 636     }
 637 
 638     /**
 639      * Format a long to produce a string.
 640      * @param number    The long to format
 641      * @param result    where the text is to be appended
 642      * @param delegate notified of locations of sub fields
 643      * @return The formatted number string
 644      * @exception        ArithmeticException if rounding is needed with rounding
 645      *                   mode being set to RoundingMode.UNNECESSARY
 646      * @see java.text.FieldPosition
 647      */
 648     private StringBuffer format(long number, StringBuffer result,
 649                                FieldDelegate delegate) {
 650         boolean isNegative = (number < 0);
 651         if (isNegative) {
 652             number = -number;
 653         }
 654 
 655         // In general, long values always represent real finite numbers, so
 656         // we don‘t have to check for +/- Infinity or NaN.  However, there
 657         // is one case we have to be careful of:  The multiplier can push
 658         // a number near MIN_VALUE or MAX_VALUE outside the legal range.  We
 659         // check for this before multiplying, and if it happens we use
 660         // BigInteger instead.
 661         boolean useBigInteger = false;
 662         if (number < 0) { // This can only happen if number == Long.MIN_VALUE.
 663             if (multiplier != 0) {
 664                 useBigInteger = true;
 665             }
 666         } else if (multiplier != 1 && multiplier != 0) {
 667             long cutoff = Long.MAX_VALUE / multiplier;
 668             if (cutoff < 0) {
 669                 cutoff = -cutoff;
 670             }
 671             useBigInteger = (number > cutoff);
 672         }
 673 
 674         if (useBigInteger) {
 675             if (isNegative) {
 676                 number = -number;
 677             }
 678             BigInteger bigIntegerValue = BigInteger.valueOf(number);
 679             return format(bigIntegerValue, result, delegate, true);
 680         }
 681 
 682         number *= multiplier;
 683         if (number == 0) {
 684             isNegative = false;
 685         } else {
 686             if (multiplier < 0) {
 687                 number = -number;
 688                 isNegative = !isNegative;
 689             }
 690         }
 691 
 692         synchronized(digitList) {
 693             int maxIntDigits = super.getMaximumIntegerDigits();
 694             int minIntDigits = super.getMinimumIntegerDigits();
 695             int maxFraDigits = super.getMaximumFractionDigits();
 696             int minFraDigits = super.getMinimumFractionDigits();
 697 
 698             digitList.set(isNegative, number,
 699                      useExponentialNotation ? maxIntDigits + maxFraDigits : 0);
 700 
 701             return subformat(result, delegate, isNegative, true,
 702                        maxIntDigits, minIntDigits, maxFraDigits, minFraDigits);
 703         }
 704     }
 705 
 706     /**
 707      * Formats a BigDecimal to produce a string.
 708      * @param number    The BigDecimal to format
 709      * @param result    where the text is to be appended
 710      * @param fieldPosition    On input: an alignment field, if desired.
 711      * On output: the offsets of the alignment field.
 712      * @return The formatted number string
 713      * @exception        ArithmeticException if rounding is needed with rounding
 714      *                   mode being set to RoundingMode.UNNECESSARY
 715      * @see java.text.FieldPosition
 716      */
 717     private StringBuffer format(BigDecimal number, StringBuffer result,
 718                                 FieldPosition fieldPosition) {
 719         fieldPosition.setBeginIndex(0);
 720         fieldPosition.setEndIndex(0);
 721         return format(number, result, fieldPosition.getFieldDelegate());
 722     }
 723 
 724     /**
 725      * Formats a BigDecimal to produce a string.
 726      * @param number    The BigDecimal to format
 727      * @param result    where the text is to be appended
 728      * @param delegate notified of locations of sub fields
 729      * @exception        ArithmeticException if rounding is needed with rounding
 730      *                   mode being set to RoundingMode.UNNECESSARY
 731      * @return The formatted number string
 732      */
 733     private StringBuffer format(BigDecimal number, StringBuffer result,
 734                                 FieldDelegate delegate) {
 735         if (multiplier != 1) {
 736             number = number.multiply(getBigDecimalMultiplier());
 737         }
 738         boolean isNegative = number.signum() == -1;
 739         if (isNegative) {
 740             number = number.negate();
 741         }
 742 
 743         synchronized(digitList) {
 744             int maxIntDigits = getMaximumIntegerDigits();
 745             int minIntDigits = getMinimumIntegerDigits();
 746             int maxFraDigits = getMaximumFractionDigits();
 747             int minFraDigits = getMinimumFractionDigits();
 748             int maximumDigits = maxIntDigits + maxFraDigits;
 749 
 750             digitList.set(isNegative, number, useExponentialNotation ?
 751                 ((maximumDigits < 0) ? Integer.MAX_VALUE : maximumDigits) :
 752                 maxFraDigits, !useExponentialNotation);
 753 
 754             return subformat(result, delegate, isNegative, false,
 755                 maxIntDigits, minIntDigits, maxFraDigits, minFraDigits);
 756         }
 757     }
 758 
 759     /**
 760      * Format a BigInteger to produce a string.
 761      * @param number    The BigInteger to format
 762      * @param result    where the text is to be appended
 763      * @param fieldPosition    On input: an alignment field, if desired.
 764      * On output: the offsets of the alignment field.
 765      * @return The formatted number string
 766      * @exception        ArithmeticException if rounding is needed with rounding
 767      *                   mode being set to RoundingMode.UNNECESSARY
 768      * @see java.text.FieldPosition
 769      */
 770     private StringBuffer format(BigInteger number, StringBuffer result,
 771                                FieldPosition fieldPosition) {
 772         fieldPosition.setBeginIndex(0);
 773         fieldPosition.setEndIndex(0);
 774 
 775         return format(number, result, fieldPosition.getFieldDelegate(), false);
 776     }
 777 
 778     /**
 779      * Format a BigInteger to produce a string.
 780      * @param number    The BigInteger to format
 781      * @param result    where the text is to be appended
 782      * @param delegate notified of locations of sub fields
 783      * @return The formatted number string
 784      * @exception        ArithmeticException if rounding is needed with rounding
 785      *                   mode being set to RoundingMode.UNNECESSARY
 786      * @see java.text.FieldPosition
 787      */
 788     private StringBuffer format(BigInteger number, StringBuffer result,
 789                                FieldDelegate delegate, boolean formatLong) {
 790         if (multiplier != 1) {
 791             number = number.multiply(getBigIntegerMultiplier());
 792         }
 793         boolean isNegative = number.signum() == -1;
 794         if (isNegative) {
 795             number = number.negate();
 796         }
 797 
 798         synchronized(digitList) {
 799             int maxIntDigits, minIntDigits, maxFraDigits, minFraDigits, maximumDigits;
 800             if (formatLong) {
 801                 maxIntDigits = super.getMaximumIntegerDigits();
 802                 minIntDigits = super.getMinimumIntegerDigits();
 803                 maxFraDigits = super.getMaximumFractionDigits();
 804                 minFraDigits = super.getMinimumFractionDigits();
 805                 maximumDigits = maxIntDigits + maxFraDigits;
 806             } else {
 807                 maxIntDigits = getMaximumIntegerDigits();
 808                 minIntDigits = getMinimumIntegerDigits();
 809                 maxFraDigits = getMaximumFractionDigits();
 810                 minFraDigits = getMinimumFractionDigits();
 811                 maximumDigits = maxIntDigits + maxFraDigits;
 812                 if (maximumDigits < 0) {
 813                     maximumDigits = Integer.MAX_VALUE;
 814                 }
 815             }
 816 
 817             digitList.set(isNegative, number,
 818                           useExponentialNotation ? maximumDigits : 0);
 819 
 820             return subformat(result, delegate, isNegative, true,
 821                 maxIntDigits, minIntDigits, maxFraDigits, minFraDigits);
 822         }
 823     }
 824 
 825     /**
 826      * Formats an Object producing an <code>AttributedCharacterIterator</code>.
 827      * You can use the returned <code>AttributedCharacterIterator</code>
 828      * to build the resulting String, as well as to determine information
 829      * about the resulting String.
 830      * <p>
 831      * Each attribute key of the AttributedCharacterIterator will be of type
 832      * <code>NumberFormat.Field</code>, with the attribute value being the
 833      * same as the attribute key.
 834      *
 835      * @exception NullPointerException if obj is null.
 836      * @exception IllegalArgumentException when the Format cannot format the
 837      *            given object.
 838      * @exception        ArithmeticException if rounding is needed with rounding
 839      *                   mode being set to RoundingMode.UNNECESSARY
 840      * @param obj The object to format
 841      * @return AttributedCharacterIterator describing the formatted value.
 842      * @since 1.4
 843      */
 844     @Override
 845     public AttributedCharacterIterator formatToCharacterIterator(Object obj) {
 846         CharacterIteratorFieldDelegate delegate =
 847                          new CharacterIteratorFieldDelegate();
 848         StringBuffer sb = new StringBuffer();
 849 
 850         if (obj instanceof Double || obj instanceof Float) {
 851             format(((Number)obj).doubleValue(), sb, delegate);
 852         } else if (obj instanceof Long || obj instanceof Integer ||
 853                    obj instanceof Short || obj instanceof Byte ||
 854                    obj instanceof AtomicInteger || obj instanceof AtomicLong) {
 855             format(((Number)obj).longValue(), sb, delegate);
 856         } else if (obj instanceof BigDecimal) {
 857             format((BigDecimal)obj, sb, delegate);
 858         } else if (obj instanceof BigInteger) {
 859             format((BigInteger)obj, sb, delegate, false);
 860         } else if (obj == null) {
 861             throw new NullPointerException(
 862                 "formatToCharacterIterator must be passed non-null object");
 863         } else {
 864             throw new IllegalArgumentException(
 865                 "Cannot format given Object as a Number");
 866         }
 867         return delegate.getIterator(sb.toString());
 868     }
 869 
 870     // ==== Begin fast-path formating logic for double =========================
 871 
 872     /* Fast-path formatting will be used for format(double ...) methods iff a
 873      * number of conditions are met (see checkAndSetFastPathStatus()):
 874      * - Only if instance properties meet the right predefined conditions.
 875      * - The abs value of the double to format is <= Integer.MAX_VALUE.
 876      *
 877      * The basic approach is to split the binary to decimal conversion of a
 878      * double value into two phases:
 879      * * The conversion of the integer portion of the double.
 880      * * The conversion of the fractional portion of the double
 881      *   (limited to two or three digits).
 882      *
 883      * The isolation and conversion of the integer portion of the double is
 884      * straightforward. The conversion of the fraction is more subtle and relies
 885      * on some rounding properties of double to the decimal precisions in
 886      * question.  Using the terminology of BigDecimal, this fast-path algorithm
 887      * is applied when a double value has a magnitude less than Integer.MAX_VALUE
 888      * and rounding is to nearest even and the destination format has two or
 889      * three digits of *scale* (digits after the decimal point).
 890      *
 891      * Under a rounding to nearest even policy, the returned result is a digit
 892      * string of a number in the (in this case decimal) destination format
 893      * closest to the exact numerical value of the (in this case binary) input
 894      * value.  If two destination format numbers are equally distant, the one
 895      * with the last digit even is returned.  To compute such a correctly rounded
 896      * value, some information about digits beyond the smallest returned digit
 897      * position needs to be consulted.
 898      *
 899      * In general, a guard digit, a round digit, and a sticky *bit* are needed
 900      * beyond the returned digit position.  If the discarded portion of the input
 901      * is sufficiently large, the returned digit string is incremented.  In round
 902      * to nearest even, this threshold to increment occurs near the half-way
 903      * point between digits.  The sticky bit records if there are any remaining
 904      * trailing digits of the exact input value in the new format; the sticky bit
 905      * is consulted only in close to half-way rounding cases.
 906      *
 907      * Given the computation of the digit and bit values, rounding is then
 908      * reduced to a table lookup problem.  For decimal, the even/odd cases look
 909      * like this:
 910      *
 911      * Last   Round   Sticky
 912      * 6      5       0      => 6   // exactly halfway, return even digit.
 913      * 6      5       1      => 7   // a little bit more than halfway, round up.
 914      * 7      5       0      => 8   // exactly halfway, round up to even.
 915      * 7      5       1      => 8   // a little bit more than halfway, round up.
 916      * With analogous entries for other even and odd last-returned digits.
 917      *
 918      * However, decimal negative powers of 5 smaller than 0.5 are *not* exactly
 919      * representable as binary fraction.  In particular, 0.005 (the round limit
 920      * for a two-digit scale) and 0.0005 (the round limit for a three-digit
 921      * scale) are not representable. Therefore, for input values near these cases
 922      * the sticky bit is known to be set which reduces the rounding logic to:
 923      *
 924      * Last   Round   Sticky
 925      * 6      5       1      => 7   // a little bit more than halfway, round up.
 926      * 7      5       1      => 8   // a little bit more than halfway, round up.
 927      *
 928      * In other words, if the round digit is 5, the sticky bit is known to be
 929      * set.  If the round digit is something other than 5, the sticky bit is not
 930      * relevant.  Therefore, some of the logic about whether or not to increment
 931      * the destination *decimal* value can occur based on tests of *binary*
 932      * computations of the binary input number.
 933      */
 934 
 935     /**
 936      * Check validity of using fast-path for this instance. If fast-path is valid
 937      * for this instance, sets fast-path state as true and initializes fast-path
 938      * utility fields as needed.
 939      *
 940      * This method is supposed to be called rarely, otherwise that will break the
 941      * fast-path performance. That means avoiding frequent changes of the
 942      * properties of the instance, since for most properties, each time a change
 943      * happens, a call to this method is needed at the next format call.
 944      *
 945      * FAST-PATH RULES:
 946      *  Similar to the default DecimalFormat instantiation case.
 947      *  More precisely:
 948      *  - HALF_EVEN rounding mode,
 949      *  - isGroupingUsed() is true,
 950      *  - groupingSize of 3,
 951      *  - multiplier is 1,
 952      *  - Decimal separator not mandatory,
 953      *  - No use of exponential notation,
 954      *  - minimumIntegerDigits is exactly 1 and maximumIntegerDigits at least 10
 955      *  - For number of fractional digits, the exact values found in the default case:
 956      *     Currency : min = max = 2.
 957      *     Decimal  : min = 0. max = 3.
 958      *
 959      */
 960     private boolean checkAndSetFastPathStatus() {
 961 
 962         boolean fastPathWasOn = isFastPath;
 963 
 964         if ((roundingMode == RoundingMode.HALF_EVEN) &&
 965             (isGroupingUsed()) &&
 966             (groupingSize == 3) &&
 967             (multiplier == 1) &&
 968             (!decimalSeparatorAlwaysShown) &&
 969             (!useExponentialNotation)) {
 970 
 971             // The fast-path algorithm is semi-hardcoded against
 972             //  minimumIntegerDigits and maximumIntegerDigits.
 973             isFastPath = ((minimumIntegerDigits == 1) &&
 974                           (maximumIntegerDigits >= 10));
 975 
 976             // The fast-path algorithm is hardcoded against
 977             //  minimumFractionDigits and maximumFractionDigits.
 978             if (isFastPath) {
 979                 if (isCurrencyFormat) {
 980                     if ((minimumFractionDigits != 2) ||
 981                         (maximumFractionDigits != 2))
 982                         isFastPath = false;
 983                 } else if ((minimumFractionDigits != 0) ||
 984                            (maximumFractionDigits != 3))
 985                     isFastPath = false;
 986             }
 987         } else
 988             isFastPath = false;
 989 
 990         resetFastPathData(fastPathWasOn);
 991         fastPathCheckNeeded = false;
 992 
 993         /*
 994          * Returns true after successfully checking the fast path condition and
 995          * setting the fast path data. The return value is used by the
 996          * fastFormat() method to decide whether to call the resetFastPathData
 997          * method to reinitialize fast path data or is it already initialized
 998          * in this method.
 999          */
1000         return true;
1001     }
1002 
1003     private void resetFastPathData(boolean fastPathWasOn) {
1004         // Since some instance properties may have changed while still falling
1005         // in the fast-path case, we need to reinitialize fastPathData anyway.
1006         if (isFastPath) {
1007             // We need to instantiate fastPathData if not already done.
1008             if (fastPathData == null) {
1009                 fastPathData = new FastPathData();
1010             }
1011 
1012             // Sets up the locale specific constants used when formatting.
1013             // ‘0‘ is our default representation of zero.
1014             fastPathData.zeroDelta = symbols.getZeroDigit() - ‘0‘;
1015             fastPathData.groupingChar = symbols.getGroupingSeparator();
1016 
1017             // Sets up fractional constants related to currency/decimal pattern.
1018             fastPathData.fractionalMaxIntBound = (isCurrencyFormat)
1019                     ? 99 : 999;
1020             fastPathData.fractionalScaleFactor = (isCurrencyFormat)
1021                     ? 100.0d : 1000.0d;
1022 
1023             // Records the need for adding prefix or suffix
1024             fastPathData.positiveAffixesRequired
1025                     = (positivePrefix.length() != 0)
1026                         || (positiveSuffix.length() != 0);
1027             fastPathData.negativeAffixesRequired
1028                     = (negativePrefix.length() != 0)
1029                         || (negativeSuffix.length() != 0);
1030 
1031             // Creates a cached char container for result, with max possible size.
1032             int maxNbIntegralDigits = 10;
1033             int maxNbGroups = 3;
1034             int containerSize
1035                     = Math.max(positivePrefix.length(), negativePrefix.length())
1036                     + maxNbIntegralDigits + maxNbGroups + 1
1037                     + maximumFractionDigits
1038                     + Math.max(positiveSuffix.length(), negativeSuffix.length());
1039 
1040             fastPathData.fastPathContainer = new char[containerSize];
1041 
1042             // Sets up prefix and suffix char arrays constants.
1043             fastPathData.charsPositiveSuffix = positiveSuffix.toCharArray();
1044             fastPathData.charsNegativeSuffix = negativeSuffix.toCharArray();
1045             fastPathData.charsPositivePrefix = positivePrefix.toCharArray();
1046             fastPathData.charsNegativePrefix = negativePrefix.toCharArray();
1047 
1048             // Sets up fixed index positions for integral and fractional digits.
1049             // Sets up decimal point in cached result container.
1050             int longestPrefixLength
1051                     = Math.max(positivePrefix.length(),
1052                             negativePrefix.length());
1053             int decimalPointIndex
1054                     = maxNbIntegralDigits + maxNbGroups + longestPrefixLength;
1055 
1056             fastPathData.integralLastIndex = decimalPointIndex - 1;
1057             fastPathData.fractionalFirstIndex = decimalPointIndex + 1;
1058             fastPathData.fastPathContainer[decimalPointIndex]
1059                     = isCurrencyFormat
1060                             ? symbols.getMonetaryDecimalSeparator()
1061                             : symbols.getDecimalSeparator();
1062 
1063         } else if (fastPathWasOn) {
1064             // Previous state was fast-path and is no more.
1065             // Resets cached array constants.
1066             fastPathData.fastPathContainer = null;
1067             fastPathData.charsPositiveSuffix = null;
1068             fastPathData.charsNegativeSuffix = null;
1069             fastPathData.charsPositivePrefix = null;
1070             fastPathData.charsNegativePrefix = null;
1071         }
1072     }
1073 
1074     /**
1075      * Returns true if rounding-up must be done on {@code scaledFractionalPartAsInt},
1076      * false otherwise.
1077      *
1078      * This is a utility method that takes correct half-even rounding decision on
1079      * passed fractional value at the scaled decimal point (2 digits for currency
1080      * case and 3 for decimal case), when the approximated fractional part after
1081      * scaled decimal point is exactly 0.5d.  This is done by means of exact
1082      * calculations on the {@code fractionalPart} floating-point value.
1083      *
1084      * This method is supposed to be called by private {@code fastDoubleFormat}
1085      * method only.
1086      *
1087      * The algorithms used for the exact calculations are :
1088      *
1089      * The <b><i>FastTwoSum</i></b> algorithm, from T.J.Dekker, described in the
1090      * papers  "<i>A  Floating-Point   Technique  for  Extending  the  Available
1091      * Precision</i>"  by Dekker, and  in "<i>Adaptive  Precision Floating-Point
1092      * Arithmetic and Fast Robust Geometric Predicates</i>" from J.Shewchuk.
1093      *
1094      * A modified version of <b><i>Sum2S</i></b> cascaded summation described in
1095      * "<i>Accurate Sum and Dot Product</i>" from Takeshi Ogita and All.  As
1096      * Ogita says in this paper this is an equivalent of the Kahan-Babuska‘s
1097      * summation algorithm because we order the terms by magnitude before summing
1098      * them. For this reason we can use the <i>FastTwoSum</i> algorithm rather
1099      * than the more expensive Knuth‘s <i>TwoSum</i>.
1100      *
1101      * We do this to avoid a more expensive exact "<i>TwoProduct</i>" algorithm,
1102      * like those described in Shewchuk‘s paper above. See comments in the code
1103      * below.
1104      *
1105      * @param  fractionalPart The  fractional value  on which  we  take rounding
1106      * decision.
1107      * @param scaledFractionalPartAsInt The integral part of the scaled
1108      * fractional value.
1109      *
1110      * @return the decision that must be taken regarding half-even rounding.
1111      */
1112     private boolean exactRoundUp(double fractionalPart,
1113                                  int scaledFractionalPartAsInt) {
1114 
1115         /* exactRoundUp() method is called by fastDoubleFormat() only.
1116          * The precondition expected to be verified by the passed parameters is :
1117          * scaledFractionalPartAsInt ==
1118          *     (int) (fractionalPart * fastPathData.fractionalScaleFactor).
1119          * This is ensured by fastDoubleFormat() code.
1120          */
1121 
1122         /* We first calculate roundoff error made by fastDoubleFormat() on
1123          * the scaled fractional part. We do this with exact calculation on the
1124          * passed fractionalPart. Rounding decision will then be taken from roundoff.
1125          */
1126 
1127         /* ---- TwoProduct(fractionalPart, scale factor (i.e. 1000.0d or 100.0d)).
1128          *
1129          * The below is an optimized exact "TwoProduct" calculation of passed
1130          * fractional part with scale factor, using Ogita‘s Sum2S cascaded
1131          * summation adapted as Kahan-Babuska equivalent by using FastTwoSum
1132          * (much faster) rather than Knuth‘s TwoSum.
1133          *
1134          * We can do this because we order the summation from smallest to
1135          * greatest, so that FastTwoSum can be used without any additional error.
1136          *
1137          * The "TwoProduct" exact calculation needs 17 flops. We replace this by
1138          * a cascaded summation of FastTwoSum calculations, each involving an
1139          * exact multiply by a power of 2.
1140          *
1141          * Doing so saves overall 4 multiplications and 1 addition compared to
1142          * using traditional "TwoProduct".
1143          *
1144          * The scale factor is either 100 (currency case) or 1000 (decimal case).
1145          * - when 1000, we replace it by (1024 - 16 - 8) = 1000.
1146          * - when 100,  we replace it by (128  - 32 + 4) =  100.
1147          * Every multiplication by a power of 2 (1024, 128, 32, 16, 8, 4) is exact.
1148          *
1149          */
1150         double approxMax;    // Will always be positive.
1151         double approxMedium; // Will always be negative.
1152         double approxMin;
1153 
1154         double fastTwoSumApproximation = 0.0d;
1155         double fastTwoSumRoundOff = 0.0d;
1156         double bVirtual = 0.0d;
1157 
1158         if (isCurrencyFormat) {
1159             // Scale is 100 = 128 - 32 + 4.
1160             // Multiply by 2**n is a shift. No roundoff. No error.
1161             approxMax    = fractionalPart * 128.00d;
1162             approxMedium = - (fractionalPart * 32.00d);
1163             approxMin    = fractionalPart * 4.00d;
1164         } else {
1165             // Scale is 1000 = 1024 - 16 - 8.
1166             // Multiply by 2**n is a shift. No roundoff. No error.
1167             approxMax    = fractionalPart * 1024.00d;
1168             approxMedium = - (fractionalPart * 16.00d);
1169             approxMin    = - (fractionalPart * 8.00d);
1170         }
1171 
1172         // Shewchuk/Dekker‘s FastTwoSum(approxMedium, approxMin).
1173         assert(-approxMedium >= Math.abs(approxMin));
1174         fastTwoSumApproximation = approxMedium + approxMin;
1175         bVirtual = fastTwoSumApproximation - approxMedium;
1176         fastTwoSumRoundOff = approxMin - bVirtual;
1177         double approxS1 = fastTwoSumApproximation;
1178         double roundoffS1 = fastTwoSumRoundOff;
1179 
1180         // Shewchuk/Dekker‘s FastTwoSum(approxMax, approxS1);
1181         assert(approxMax >= Math.abs(approxS1));
1182         fastTwoSumApproximation = approxMax + approxS1;
1183         bVirtual = fastTwoSumApproximation - approxMax;
1184         fastTwoSumRoundOff = approxS1 - bVirtual;
1185         double roundoff1000 = fastTwoSumRoundOff;
1186         double approx1000 = fastTwoSumApproximation;
1187         double roundoffTotal = roundoffS1 + roundoff1000;
1188 
1189         // Shewchuk/Dekker‘s FastTwoSum(approx1000, roundoffTotal);
1190         assert(approx1000 >= Math.abs(roundoffTotal));
1191         fastTwoSumApproximation = approx1000 + roundoffTotal;
1192         bVirtual = fastTwoSumApproximation - approx1000;
1193 
1194         // Now we have got the roundoff for the scaled fractional
1195         double scaledFractionalRoundoff = roundoffTotal - bVirtual;
1196 
1197         // ---- TwoProduct(fractionalPart, scale (i.e. 1000.0d or 100.0d)) end.
1198 
1199         /* ---- Taking the rounding decision
1200          *
1201          * We take rounding decision based on roundoff and half-even rounding
1202          * rule.
1203          *
1204          * The above TwoProduct gives us the exact roundoff on the approximated
1205          * scaled fractional, and we know that this approximation is exactly
1206          * 0.5d, since that has already been tested by the caller
1207          * (fastDoubleFormat).
1208          *
1209          * Decision comes first from the sign of the calculated exact roundoff.
1210          * - Since being exact roundoff, it cannot be positive with a scaled
1211          *   fractional less than 0.5d, as well as negative with a scaled
1212          *   fractional greater than 0.5d. That leaves us with following 3 cases.
1213          * - positive, thus scaled fractional == 0.500....0fff ==> round-up.
1214          * - negative, thus scaled fractional == 0.499....9fff ==> don‘t round-up.
1215          * - is zero,  thus scaled fractioanl == 0.5 ==> half-even rounding applies :
1216          *    we round-up only if the integral part of the scaled fractional is odd.
1217          *
1218          */
1219         if (scaledFractionalRoundoff > 0.0) {
1220             return true;
1221         } else if (scaledFractionalRoundoff < 0.0) {
1222             return false;
1223         } else if ((scaledFractionalPartAsInt & 1) != 0) {
1224             return true;
1225         }
1226 
1227         return false;
1228 
1229         // ---- Taking the rounding decision end
1230     }
1231 
1232     /**
1233      * Collects integral digits from passed {@code number}, while setting
1234      * grouping chars as needed. Updates {@code firstUsedIndex} accordingly.
1235      *
1236      * Loops downward starting from {@code backwardIndex} position (inclusive).
1237      *
1238      * @param number  The int value from which we collect digits.
1239      * @param digitsBuffer The char array container where digits and grouping chars
1240      *  are stored.
1241      * @param backwardIndex the position from which we start storing digits in
1242      *  digitsBuffer.
1243      *
1244      */
1245     private void collectIntegralDigits(int number,
1246                                        char[] digitsBuffer,
1247                                        int backwardIndex) {
1248         int index = backwardIndex;
1249         int q;
1250         int r;
1251         while (number > 999) {
1252             // Generates 3 digits per iteration.
1253             q = number / 1000;
1254             r = number - (q << 10) + (q << 4) + (q << 3); // -1024 +16 +8 = 1000.
1255             number = q;
1256 
1257             digitsBuffer[index--] = DigitArrays.DigitOnes1000[r];
1258             digitsBuffer[index--] = DigitArrays.DigitTens1000[r];
1259             digitsBuffer[index--] = DigitArrays.DigitHundreds1000[r];
1260             digitsBuffer[index--] = fastPathData.groupingChar;
1261         }
1262 
1263         // Collects last 3 or less digits.
1264         digitsBuffer[index] = DigitArrays.DigitOnes1000[number];
1265         if (number > 9) {
1266             digitsBuffer[--index]  = DigitArrays.DigitTens1000[number];
1267             if (number > 99)
1268                 digitsBuffer[--index]   = DigitArrays.DigitHundreds1000[number];
1269         }
1270 
1271         fastPathData.firstUsedIndex = index;
1272     }
1273 
1274     /**
1275      * Collects the 2 (currency) or 3 (decimal) fractional digits from passed
1276      * {@code number}, starting at {@code startIndex} position
1277      * inclusive.  There is no punctuation to set here (no grouping chars).
1278      * Updates {@code fastPathData.lastFreeIndex} accordingly.
1279      *
1280      *
1281      * @param number  The int value from which we collect digits.
1282      * @param digitsBuffer The char array container where digits are stored.
1283      * @param startIndex the position from which we start storing digits in
1284      *  digitsBuffer.
1285      *
1286      */
1287     private void collectFractionalDigits(int number,
1288                                          char[] digitsBuffer,
1289                                          int startIndex) {
1290         int index = startIndex;
1291 
1292         char digitOnes = DigitArrays.DigitOnes1000[number];
1293         char digitTens = DigitArrays.DigitTens1000[number];
1294 
1295         if (isCurrencyFormat) {
1296             // Currency case. Always collects fractional digits.
1297             digitsBuffer[index++] = digitTens;
1298             digitsBuffer[index++] = digitOnes;
1299         } else if (number != 0) {
1300             // Decimal case. Hundreds will always be collected
1301             digitsBuffer[index++] = DigitArrays.DigitHundreds1000[number];
1302 
1303             // Ending zeros won‘t be collected.
1304             if (digitOnes != ‘0‘) {
1305                 digitsBuffer[index++] = digitTens;
1306                 digitsBuffer[index++] = digitOnes;
1307             } else if (digitTens != ‘0‘)
1308                 digitsBuffer[index++] = digitTens;
1309 
1310         } else
1311             // This is decimal pattern and fractional part is zero.
1312             // We must remove decimal point from result.
1313             index--;
1314 
1315         fastPathData.lastFreeIndex = index;
1316     }
1317 
1318     /**
1319      * Internal utility.
1320      * Adds the passed {@code prefix} and {@code suffix} to {@code container}.
1321      *
1322      * @param container  Char array container which to prepend/append the
1323      *  prefix/suffix.
1324      * @param prefix     Char sequence to prepend as a prefix.
1325      * @param suffix     Char sequence to append as a suffix.
1326      *
1327      */
1328     //    private void addAffixes(boolean isNegative, char[] container) {
1329     private void addAffixes(char[] container, char[] prefix, char[] suffix) {
1330 
1331         // We add affixes only if needed (affix length > 0).
1332         int pl = prefix.length;
1333         int sl = suffix.length;
1334         if (pl != 0) prependPrefix(prefix, pl, container);
1335         if (sl != 0) appendSuffix(suffix, sl, container);
1336 
1337     }
1338 
1339     /**
1340      * Prepends the passed {@code prefix} chars to given result
1341      * {@code container}.  Updates {@code fastPathData.firstUsedIndex}
1342      * accordingly.
1343      *
1344      * @param prefix The prefix characters to prepend to result.
1345      * @param len The number of chars to prepend.
1346      * @param container Char array container which to prepend the prefix
1347      */
1348     private void prependPrefix(char[] prefix,
1349                                int len,
1350                                char[] container) {
1351 
1352         fastPathData.firstUsedIndex -= len;
1353         int startIndex = fastPathData.firstUsedIndex;
1354 
1355         // If prefix to prepend is only 1 char long, just assigns this char.
1356         // If prefix is less or equal 4, we use a dedicated algorithm that
1357         //  has shown to run faster than System.arraycopy.
1358         // If more than 4, we use System.arraycopy.
1359         if (len == 1)
1360             container[startIndex] = prefix[0];
1361         else if (len <= 4) {
1362             int dstLower = startIndex;
1363             int dstUpper = dstLower + len - 1;
1364             int srcUpper = len - 1;
1365             container[dstLower] = prefix[0];
1366             container[dstUpper] = prefix[srcUpper];
1367 
1368             if (len > 2)
1369                 container[++dstLower] = prefix[1];
1370             if (len == 4)
1371                 container[--dstUpper] = prefix[2];
1372         } else
1373             System.arraycopy(prefix, 0, container, startIndex, len);
1374     }
1375 
1376     /**
1377      * Appends the passed {@code suffix} chars to given result
1378      * {@code container}.  Updates {@code fastPathData.lastFreeIndex}
1379      * accordingly.
1380      *
1381      * @param suffix The suffix characters to append to result.
1382      * @param len The number of chars to append.
1383      * @param container Char array container which to append the suffix
1384      */
1385     private void appendSuffix(char[] suffix,
1386                               int len,
1387                               char[] container) {
1388 
1389         int startIndex = fastPathData.lastFreeIndex;
1390 
1391         // If suffix to append is only 1 char long, just assigns this char.
1392         // If suffix is less or equal 4, we use a dedicated algorithm that
1393         //  has shown to run faster than System.arraycopy.
1394         // If more than 4, we use System.arraycopy.
1395         if (len == 1)
1396             container[startIndex] = suffix[0];
1397         else if (len <= 4) {
1398             int dstLower = startIndex;
1399             int dstUpper = dstLower + len - 1;
1400             int srcUpper = len - 1;
1401             container[dstLower] = suffix[0];
1402             container[dstUpper] = suffix[srcUpper];
1403 
1404             if (len > 2)
1405                 container[++dstLower] = suffix[1];
1406             if (len == 4)
1407                 container[--dstUpper] = suffix[2];
1408         } else
1409             System.arraycopy(suffix, 0, container, startIndex, len);
1410 
1411         fastPathData.lastFreeIndex += len;
1412     }
1413 
1414     /**
1415      * Converts digit chars from {@code digitsBuffer} to current locale.
1416      *
1417      * Must be called before adding affixes since we refer to
1418      * {@code fastPathData.firstUsedIndex} and {@code fastPathData.lastFreeIndex},
1419      * and do not support affixes (for speed reason).
1420      *
1421      * We loop backward starting from last used index in {@code fastPathData}.
1422      *
1423      * @param digitsBuffer The char array container where the digits are stored.
1424      */
1425     private void localizeDigits(char[] digitsBuffer) {
1426 
1427         // We will localize only the digits, using the groupingSize,
1428         // and taking into account fractional part.
1429 
1430         // First take into account fractional part.
1431         int digitsCounter =
1432             fastPathData.lastFreeIndex - fastPathData.fractionalFirstIndex;
1433 
1434         // The case when there is no fractional digits.
1435         if (digitsCounter < 0)
1436             digitsCounter = groupingSize;
1437 
1438         // Only the digits remains to localize.
1439         for (int cursor = fastPathData.lastFreeIndex - 1;
1440              cursor >= fastPathData.firstUsedIndex;
1441              cursor--) {
1442             if (digitsCounter != 0) {
1443                 // This is a digit char, we must localize it.
1444                 digitsBuffer[cursor] += fastPathData.zeroDelta;
1445                 digitsCounter--;
1446             } else {
1447                 // Decimal separator or grouping char. Reinit counter only.
1448                 digitsCounter = groupingSize;
1449             }
1450         }
1451     }
1452 
1453     /**
1454      * This is the main entry point for the fast-path format algorithm.
1455      *
1456      * At this point we are sure to be in the expected conditions to run it.
1457      * This algorithm builds the formatted result and puts it in the dedicated
1458      * {@code fastPathData.fastPathContainer}.
1459      *
1460      * @param d the double value to be formatted.
1461      * @param negative Flag precising if {@code d} is negative.
1462      */
1463     private void fastDoubleFormat(double d,
1464                                   boolean negative) {
1465 
1466         char[] container = fastPathData.fastPathContainer;
1467 
1468         /*
1469          * The principle of the algorithm is to :
1470          * - Break the passed double into its integral and fractional parts
1471          *    converted into integers.
1472          * - Then decide if rounding up must be applied or not by following
1473          *    the half-even rounding rule, first using approximated scaled
1474          *    fractional part.
1475          * - For the difficult cases (approximated scaled fractional part
1476          *    being exactly 0.5d), we refine the rounding decision by calling
1477          *    exactRoundUp utility method that both calculates the exact roundoff
1478          *    on the approximation and takes correct rounding decision.
1479          * - We round-up the fractional part if needed, possibly propagating the
1480          *    rounding to integral part if we meet a "all-nine" case for the
1481          *    scaled fractional part.
1482          * - We then collect digits from the resulting integral and fractional
1483          *   parts, also setting the required grouping chars on the fly.
1484          * - Then we localize the collected digits if needed, and
1485          * - Finally prepend/append prefix/suffix if any is needed.
1486          */
1487 
1488         // Exact integral part of d.
1489         int integralPartAsInt = (int) d;
1490 
1491         // Exact fractional part of d (since we subtract it‘s integral part).
1492         double exactFractionalPart = d - (double) integralPartAsInt;
1493 
1494         // Approximated scaled fractional part of d (due to multiplication).
1495         double scaledFractional =
1496             exactFractionalPart * fastPathData.fractionalScaleFactor;
1497 
1498         // Exact integral part of scaled fractional above.
1499         int fractionalPartAsInt = (int) scaledFractional;
1500 
1501         // Exact fractional part of scaled fractional above.
1502         scaledFractional = scaledFractional - (double) fractionalPartAsInt;
1503 
1504         // Only when scaledFractional is exactly 0.5d do we have to do exact
1505         // calculations and take fine-grained rounding decision, since
1506         // approximated results above may lead to incorrect decision.
1507         // Otherwise comparing against 0.5d (strictly greater or less) is ok.
1508         boolean roundItUp = false;
1509         if (scaledFractional >= 0.5d) {
1510             if (scaledFractional == 0.5d)
1511                 // Rounding need fine-grained decision.
1512                 roundItUp = exactRoundUp(exactFractionalPart, fractionalPartAsInt);
1513             else
1514                 roundItUp = true;
1515 
1516             if (roundItUp) {
1517                 // Rounds up both fractional part (and also integral if needed).
1518                 if (fractionalPartAsInt < fastPathData.fractionalMaxIntBound) {
1519                     fractionalPartAsInt++;
1520                 } else {
1521                     // Propagates rounding to integral part since "all nines" case.
1522                     fractionalPartAsInt = 0;
1523                     integralPartAsInt++;
1524                 }
1525             }
1526         }
1527 
1528         // Collecting digits.
1529         collectFractionalDigits(fractionalPartAsInt, container,
1530                                 fastPathData.fractionalFirstIndex);
1531         collectIntegralDigits(integralPartAsInt, container,
1532                               fastPathData.integralLastIndex);
1533 
1534         // Localizing digits.
1535         if (fastPathData.zeroDelta != 0)
1536             localizeDigits(container);
1537 
1538         // Adding prefix and suffix.
1539         if (negative) {
1540             if (fastPathData.negativeAffixesRequired)
1541                 addAffixes(container,
1542                            fastPathData.charsNegativePrefix,
1543                            fastPathData.charsNegativeSuffix);
1544         } else if (fastPathData.positiveAffixesRequired)
1545             addAffixes(container,
1546                        fastPathData.charsPositivePrefix,
1547                        fastPathData.charsPositiveSuffix);
1548     }
1549 
1550     /**
1551      * A fast-path shortcut of format(double) to be called by NumberFormat, or by
1552      * format(double, ...) public methods.
1553      *
1554      * If instance can be applied fast-path and passed double is not NaN or
1555      * Infinity, is in the integer range, we call {@code fastDoubleFormat}
1556      * after changing {@code d} to its positive value if necessary.
1557      *
1558      * Otherwise returns null by convention since fast-path can‘t be exercized.
1559      *
1560      * @param d The double value to be formatted
1561      *
1562      * @return the formatted result for {@code d} as a string.
1563      */
1564     String fastFormat(double d) {
1565         boolean isDataSet = false;
1566         // (Re-)Evaluates fast-path status if needed.
1567         if (fastPathCheckNeeded) {
1568             isDataSet = checkAndSetFastPathStatus();
1569         }
1570 
1571         if (!isFastPath )
1572             // DecimalFormat instance is not in a fast-path state.
1573             return null;
1574 
1575         if (!Double.isFinite(d))
1576             // Should not use fast-path for Infinity and NaN.
1577             return null;
1578 
1579         // Extracts and records sign of double value, possibly changing it
1580         // to a positive one, before calling fastDoubleFormat().
1581         boolean negative = false;
1582         if (d < 0.0d) {
1583             negative = true;
1584             d = -d;
1585         } else if (d == 0.0d) {
1586             negative = (Math.copySign(1.0d, d) == -1.0d);
1587             d = +0.0d;
1588         }
1589 
1590         if (d > MAX_INT_AS_DOUBLE)
1591             // Filters out values that are outside expected fast-path range
1592             return null;
1593         else {
1594             if (!isDataSet) {
1595                 /*
1596                  * If the fast path data is not set through
1597                  * checkAndSetFastPathStatus() and fulfil the
1598                  * fast path conditions then reset the data
1599                  * directly through resetFastPathData()
1600                  */
1601                 resetFastPathData(isFastPath);
1602             }
1603             fastDoubleFormat(d, negative);
1604 
1605         }
1606 
1607 
1608         // Returns a new string from updated fastPathContainer.
1609         return new String(fastPathData.fastPathContainer,
1610                           fastPathData.firstUsedIndex,
1611                           fastPathData.lastFreeIndex - fastPathData.firstUsedIndex);
1612 
1613     }
1614 
1615     // ======== End fast-path formating logic for double =========================
1616 
1617     /**
1618      * Complete the formatting of a finite number.  On entry, the digitList must
1619      * be filled in with the correct digits.
1620      */
1621     private StringBuffer subformat(StringBuffer result, FieldDelegate delegate,
1622                                    boolean isNegative, boolean isInteger,
1623                                    int maxIntDigits, int minIntDigits,
1624                                    int maxFraDigits, int minFraDigits) {
1625         // NOTE: This isn‘t required anymore because DigitList takes care of this.
1626         //
1627         //  // The negative of the exponent represents the number of leading
1628         //  // zeros between the decimal and the first non-zero digit, for
1629         //  // a value < 0.1 (e.g., for 0.00123, -fExponent == 2).  If this
1630         //  // is more than the maximum fraction digits, then we have an underflow
1631         //  // for the printed representation.  We recognize this here and set
1632         //  // the DigitList representation to zero in this situation.
1633         //
1634         //  if (-digitList.decimalAt >= getMaximumFractionDigits())
1635         //  {
1636         //      digitList.count = 0;
1637         //  }
1638 
1639         char zero = symbols.getZeroDigit();
1640         int zeroDelta = zero - ‘0‘; // ‘0‘ is the DigitList representation of zero
1641         char grouping = symbols.getGroupingSeparator();
1642         char decimal = isCurrencyFormat ?
1643             symbols.getMonetaryDecimalSeparator() :
1644             symbols.getDecimalSeparator();
1645 
1646         /* Per bug 4147706, DecimalFormat must respect the sign of numbers which
1647          * format as zero.  This allows sensible computations and preserves
1648          * relations such as signum(1/x) = signum(x), where x is +Infinity or
1649          * -Infinity.  Prior to this fix, we always formatted zero values as if
1650          * they were positive.  Liu 7/6/98.
1651          */
1652         if (digitList.isZero()) {
1653             digitList.decimalAt = 0; // Normalize
1654         }
1655 
1656         if (isNegative) {
1657             append(result, negativePrefix, delegate,
1658                    getNegativePrefixFieldPositions(), Field.SIGN);
1659         } else {
1660             append(result, positivePrefix, delegate,
1661                    getPositivePrefixFieldPositions(), Field.SIGN);
1662         }
1663 
1664         if (useExponentialNotation) {
1665             int iFieldStart = result.length();
1666             int iFieldEnd = -1;
1667             int fFieldStart = -1;
1668 
1669             // Minimum integer digits are handled in exponential format by
1670             // adjusting the exponent.  For example, 0.01234 with 3 minimum
1671             // integer digits is "123.4E-4".
1672 
1673             // Maximum integer digits are interpreted as indicating the
1674             // repeating range.  This is useful for engineering notation, in
1675             // which the exponent is restricted to a multiple of 3.  For
1676             // example, 0.01234 with 3 maximum integer digits is "12.34e-3".
1677             // If maximum integer digits are > 1 and are larger than
1678             // minimum integer digits, then minimum integer digits are
1679             // ignored.
1680             int exponent = digitList.decimalAt;
1681             int repeat = maxIntDigits;
1682             int minimumIntegerDigits = minIntDigits;
1683             if (repeat > 1 && repeat > minIntDigits) {
1684                 // A repeating range is defined; adjust to it as follows.
1685                 // If repeat == 3, we have 6,5,4=>3; 3,2,1=>0; 0,-1,-2=>-3;
1686                 // -3,-4,-5=>-6, etc. This takes into account that the
1687                 // exponent we have here is off by one from what we expect;
1688                 // it is for the format 0.MMMMMx10^n.
1689                 if (exponent >= 1) {
1690                     exponent = ((exponent - 1) / repeat) * repeat;
1691                 } else {
1692                     // integer division rounds towards 0
1693                     exponent = ((exponent - repeat) / repeat) * repeat;
1694                 }
1695                 minimumIntegerDigits = 1;
1696             } else {
1697                 // No repeating range is defined; use minimum integer digits.
1698                 exponent -= minimumIntegerDigits;
1699             }
1700 
1701             // We now output a minimum number of digits, and more if there
1702             // are more digits, up to the maximum number of digits.  We
1703             // place the decimal point after the "integer" digits, which
1704             // are the first (decimalAt - exponent) digits.
1705             int minimumDigits = minIntDigits + minFraDigits;
1706             if (minimumDigits < 0) {    // overflow?
1707                 minimumDigits = Integer.MAX_VALUE;
1708             }
1709 
1710             // The number of integer digits is handled specially if the number
1711             // is zero, since then there may be no digits.
1712             int integerDigits = digitList.isZero() ? minimumIntegerDigits :
1713                     digitList.decimalAt - exponent;
1714             if (minimumDigits < integerDigits) {
1715                 minimumDigits = integerDigits;
1716             }
1717             int totalDigits = digitList.count;
1718             if (minimumDigits > totalDigits) {
1719                 totalDigits = minimumDigits;
1720             }
1721             boolean addedDecimalSeparator = false;
1722 
1723             for (int i=0; i<totalDigits; ++i) {
1724                 if (i == integerDigits) {
1725                     // Record field information for caller.
1726                     iFieldEnd = result.length();
1727 
1728                     result.append(decimal);
1729                     addedDecimalSeparator = true;
1730 
1731                     // Record field information for caller.
1732                     fFieldStart = result.length();
1733                 }
1734                 result.append((i < digitList.count) ?
1735                               (char)(digitList.digits[i] + zeroDelta) :
1736                               zero);
1737             }
1738 
1739             if (decimalSeparatorAlwaysShown && totalDigits == integerDigits) {
1740                 // Record field information for caller.
1741                 iFieldEnd = result.length();
1742 
1743                 result.append(decimal);
1744                 addedDecimalSeparator = true;
1745 
1746                 // Record field information for caller.
1747                 fFieldStart = result.length();
1748             }
1749 
1750             // Record field information
1751             if (iFieldEnd == -1) {
1752                 iFieldEnd = result.length();
1753             }
1754             delegate.formatted(INTEGER_FIELD, Field.INTEGER, Field.INTEGER,
1755                                iFieldStart, iFieldEnd, result);
1756             if (addedDecimalSeparator) {
1757                 delegate.formatted(Field.DECIMAL_SEPARATOR,
1758                                    Field.DECIMAL_SEPARATOR,
1759                                    iFieldEnd, fFieldStart, result);
1760             }
1761             if (fFieldStart == -1) {
1762                 fFieldStart = result.length();
1763             }
1764             delegate.formatted(FRACTION_FIELD, Field.FRACTION, Field.FRACTION,
1765                                fFieldStart, result.length(), result);
1766 
1767             // The exponent is output using the pattern-specified minimum
1768             // exponent digits.  There is no maximum limit to the exponent
1769             // digits, since truncating the exponent would result in an
1770             // unacceptable inaccuracy.
1771             int fieldStart = result.length();
1772 
1773             result.append(symbols.getExponentSeparator());
1774 
1775             delegate.formatted(Field.EXPONENT_SYMBOL, Field.EXPONENT_SYMBOL,
1776                                fieldStart, result.length(), result);
1777 
1778             // For zero values, we force the exponent to zero.  We
1779             // must do this here, and not earlier, because the value
1780             // is used to determine integer digit count above.
1781             if (digitList.isZero()) {
1782                 exponent = 0;
1783             }
1784 
1785             boolean negativeExponent = exponent < 0;
1786             if (negativeExponent) {
1787                 exponent = -exponent;
1788                 fieldStart = result.length();
1789                 result.append(symbols.getMinusSign());
1790                 delegate.formatted(Field.EXPONENT_SIGN, Field.EXPONENT_SIGN,
1791                                    fieldStart, result.length(), result);
1792             }
1793             digitList.set(negativeExponent, exponent);
1794 
1795             int eFieldStart = result.length();
1796 
1797             for (int i=digitList.decimalAt; i<minExponentDigits; ++i) {
1798                 result.append(zero);
1799             }
1800             for (int i=0; i<digitList.decimalAt; ++i) {
1801                 result.append((i < digitList.count) ?
1802                           (char)(digitList.digits[i] + zeroDelta) : zero);
1803             }
1804             delegate.formatted(Field.EXPONENT, Field.EXPONENT, eFieldStart,
1805                                result.length(), result);
1806         } else {
1807             int iFieldStart = result.length();
1808 
1809             // Output the integer portion.  Here ‘count‘ is the total
1810             // number of integer digits we will display, including both
1811             // leading zeros required to satisfy getMinimumIntegerDigits,
1812             // and actual digits present in the number.
1813             int count = minIntDigits;
1814             int digitIndex = 0; // Index into digitList.fDigits[]
1815             if (digitList.decimalAt > 0 && count < digitList.decimalAt) {
1816                 count = digitList.decimalAt;
1817             }
1818 
1819             // Handle the case where getMaximumIntegerDigits() is smaller
1820             // than the real number of integer digits.  If this is so, we
1821             // output the least significant max integer digits.  For example,
1822             // the value 1997 printed with 2 max integer digits is just "97".
1823             if (count > maxIntDigits) {
1824                 count = maxIntDigits;
1825                 digitIndex = digitList.decimalAt - count;
1826             }
1827 
1828             int sizeBeforeIntegerPart = result.length();
1829             for (int i=count-1; i>=0; --i) {
1830                 if (i < digitList.decimalAt && digitIndex < digitList.count) {
1831                     // Output a real digit
1832                     result.append((char)(digitList.digits[digitIndex++] + zeroDelta));
1833                 } else {
1834                     // Output a leading zero
1835                     result.append(zero);
1836                 }
1837 
1838                 // Output grouping separator if necessary.  Don‘t output a
1839                 // grouping separator if i==0 though; that‘s at the end of
1840                 // the integer part.
1841                 if (isGroupingUsed() && i>0 && (groupingSize != 0) &&
1842                     (i % groupingSize == 0)) {
1843                     int gStart = result.length();
1844                     result.append(grouping);
1845                     delegate.formatted(Field.GROUPING_SEPARATOR,
1846                                        Field.GROUPING_SEPARATOR, gStart,
1847                                        result.length(), result);
1848                 }
1849             }
1850 
1851             // Determine whether or not there are any printable fractional
1852             // digits.  If we‘ve used up the digits we know there aren‘t.
1853             boolean fractionPresent = (minFraDigits > 0) ||
1854                 (!isInteger && digitIndex < digitList.count);
1855 
1856             // If there is no fraction present, and we haven‘t printed any
1857             // integer digits, then print a zero.  Otherwise we won‘t print
1858             // _any_ digits, and we won‘t be able to parse this string.
1859             if (!fractionPresent && result.length() == sizeBeforeIntegerPart) {
1860                 result.append(zero);
1861             }
1862 
1863             delegate.formatted(INTEGER_FIELD, Field.INTEGER, Field.INTEGER,
1864                                iFieldStart, result.length(), result);
1865 
1866             // Output the decimal separator if we always do so.
1867             int sStart = result.length();
1868             if (decimalSeparatorAlwaysShown || fractionPresent) {
1869                 result.append(decimal);
1870             }
1871 
1872             if (sStart != result.length()) {
1873                 delegate.formatted(Field.DECIMAL_SEPARATOR,
1874                                    Field.DECIMAL_SEPARATOR,
1875                                    sStart, result.length(), result);
1876             }
1877             int fFieldStart = result.length();
1878 
1879             for (int i=0; i < maxFraDigits; ++i) {
1880                 // Here is where we escape from the loop.  We escape if we‘ve
1881                 // output the maximum fraction digits (specified in the for
1882                 // expression above).
1883                 // We also stop when we‘ve output the minimum digits and either:
1884                 // we have an integer, so there is no fractional stuff to
1885                 // display, or we‘re out of significant digits.
1886                 if (i >= minFraDigits &&
1887                     (isInteger || digitIndex >= digitList.count)) {
1888                     break;
1889                 }
1890 
1891                 // Output leading fractional zeros. These are zeros that come
1892                 // after the decimal but before any significant digits. These
1893                 // are only output if abs(number being formatted) < 1.0.
1894                 if (-1-i > (digitList.decimalAt-1)) {
1895                     result.append(zero);
1896                     continue;
1897                 }
1898 
1899                 // Output a digit, if we have any precision left, or a
1900                 // zero if we don‘t.  We don‘t want to output noise digits.
1901                 if (!isInteger && digitIndex < digitList.count) {
1902                     result.append((char)(digitList.digits[digitIndex++] + zeroDelta));
1903                 } else {
1904                     result.append(zero);
1905                 }
1906             }
1907 
1908             // Record field information for caller.
1909             delegate.formatted(FRACTION_FIELD, Field.FRACTION, Field.FRACTION,
1910                                fFieldStart, result.length(), result);
1911         }
1912 
1913         if (isNegative) {
1914             append(result, negativeSuffix, delegate,
1915                    getNegativeSuffixFieldPositions(), Field.SIGN);
1916         } else {
1917             append(result, positiveSuffix, delegate,
1918                    getPositiveSuffixFieldPositions(), Field.SIGN);
1919         }
1920 
1921         return result;
1922     }
1923 
1924     /**
1925      * Appends the String <code>string</code> to <code>result</code>.
1926      * <code>delegate</code> is notified of all  the
1927      * <code>FieldPosition</code>s in <code>positions</code>.
1928      * <p>
1929      * If one of the <code>FieldPosition</code>s in <code>positions</code>
1930      * identifies a <code>SIGN</code> attribute, it is mapped to
1931      * <code>signAttribute</code>. This is used
1932      * to map the <code>SIGN</code> attribute to the <code>EXPONENT</code>
1933      * attribute as necessary.
1934      * <p>
1935      * This is used by <code>subformat</code> to add the prefix/suffix.
1936      */
1937     private void append(StringBuffer result, String string,
1938                         FieldDelegate delegate,
1939                         FieldPosition[] positions,
1940                         Format.Field signAttribute) {
1941         int start = result.length();
1942 
1943         if (string.length() > 0) {
1944             result.append(string);
1945             for (int counter = 0, max = positions.length; counter < max;
1946                  counter++) {
1947                 FieldPosition fp = positions[counter];
1948                 Format.Field attribute = fp.getFieldAttribute();
1949 
1950                 if (attribute == Field.SIGN) {
1951                     attribute = signAttribute;
1952                 }
1953                 delegate.formatted(attribute, attribute,
1954                                    start + fp.getBeginIndex(),
1955                                    start + fp.getEndIndex(), result);
1956             }
1957         }
1958     }
1959 
1960     /**
1961      * Parses text from a string to produce a <code>Number</code>.
1962      * <p>
1963      * The method attempts to parse text starting at the index given by
1964      * <code>pos</code>.
1965      * If parsing succeeds, then the index of <code>pos</code> is updated
1966      * to the index after the last character used (parsing does not necessarily
1967      * use all characters up to the end of the string), and the parsed
1968      * number is returned. The updated <code>pos</code> can be used to
1969      * indicate the starting point for the next call to this method.
1970      * If an error occurs, then the index of <code>pos</code> is not
1971      * changed, the error index of <code>pos</code> is set to the index of
1972      * the character where the error occurred, and null is returned.
1973      * <p>
1974      * The subclass returned depends on the value of {@link #isParseBigDecimal}
1975      * as well as on the string being parsed.
1976      * <ul>
1977      *   <li>If <code>isParseBigDecimal()</code> is false (the default),
1978      *       most integer values are returned as <code>Long</code>
1979      *       objects, no matter how they are written: <code>"17"</code> and
1980      *       <code>"17.000"</code> both parse to <code>Long(17)</code>.
1981      *       Values that cannot fit into a <code>Long</code> are returned as
1982      *       <code>Double</code>s. This includes values with a fractional part,
1983      *       infinite values, <code>NaN</code>, and the value -0.0.
1984      *       <code>DecimalFormat</code> does <em>not</em> decide whether to
1985      *       return a <code>Double</code> or a <code>Long</code> based on the
1986      *       presence of a decimal separator in the source string. Doing so
1987      *       would prevent integers that overflow the mantissa of a double,
1988      *       such as <code>"-9,223,372,036,854,775,808.00"</code>, from being
1989      *       parsed accurately.
1990      *       <p>
1991      *       Callers may use the <code>Number</code> methods
1992      *       <code>doubleValue</code>, <code>longValue</code>, etc., to obtain
1993      *       the type they want.
1994      *   <li>If <code>isParseBigDecimal()</code> is true, values are returned
1995      *       as <code>BigDecimal</code> objects. The values are the ones
1996      *       constructed by {@link java.math.BigDecimal#BigDecimal(String)}
1997      *       for corresponding strings in locale-independent format. The
1998      *       special cases negative and positive infinity and NaN are returned
1999      *       as <code>Double</code> instances holding the values of the
2000      *       corresponding <code>Double</code> constants.
2001      * </ul>
2002      * <p>
2003      * <code>DecimalFormat</code> parses all Unicode characters that represent
2004      * decimal digits, as defined by <code>Character.digit()</code>. In
2005      * addition, <code>DecimalFormat</code> also recognizes as digits the ten
2006      * consecutive characters starting with the localized zero digit defined in
2007      * the <code>DecimalFormatSymbols</code> object.
2008      *
2009      * @param text the string to be parsed
2010      * @param pos  A <code>ParsePosition</code> object with index and error
2011      *             index information as described above.
2012      * @return     the parsed value, or <code>null</code> if the parse fails
2013      * @exception  NullPointerException if <code>text</code> or
2014      *             <code>pos</code> is null.
2015      */
2016     @Override
2017     public Number parse(String text, ParsePosition pos) {
2018         // special case NaN
2019         if (text.regionMatches(pos.index, symbols.getNaN(), 0, symbols.getNaN().length())) {
2020             pos.index = pos.index + symbols.getNaN().length();
2021             return Double.valueOf(Double.NaN);
2022         }
2023 
2024         boolean[] status = new boolean[STATUS_LENGTH];
2025         if (!subparse(text, pos, positivePrefix, negativePrefix, digitList, false, status)) {
2026             return null;
2027         }
2028 
2029         // special case INFINITY
2030         if (status[STATUS_INFINITE]) {
2031             if (status[STATUS_POSITIVE] == (multiplier >= 0)) {
2032                 return Double.valueOf(Double.POSITIVE_INFINITY);
2033             } else {
2034                 return Double.valueOf(Double.NEGATIVE_INFINITY);
2035             }
2036         }
2037 
2038         if (multiplier == 0) {
2039             if (digitList.isZero()) {
2040                 return Double.valueOf(Double.NaN);
2041             } else if (status[STATUS_POSITIVE]) {
2042                 return Double.valueOf(Double.POSITIVE_INFINITY);
2043             } else {
2044                 return Double.valueOf(Double.NEGATIVE_INFINITY);
2045             }
2046         }
2047 
2048         if (isParseBigDecimal()) {
2049             BigDecimal bigDecimalResult = digitList.getBigDecimal();
2050 
2051             if (multiplier != 1) {
2052                 try {
2053                     bigDecimalResult = bigDecimalResult.divide(getBigDecimalMultiplier());
2054                 }
2055                 catch (ArithmeticException e) {  // non-terminating decimal expansion
2056                     bigDecimalResult = bigDecimalResult.divide(getBigDecimalMultiplier(), roundingMode);
2057                 }
2058             }
2059 
2060             if (!status[STATUS_POSITIVE]) {
2061                 bigDecimalResult = bigDecimalResult.negate();
2062             }
2063             return bigDecimalResult;
2064         } else {
2065             boolean gotDouble = true;
2066             boolean gotLongMinimum = false;
2067             double  doubleResult = 0.0;
2068             long    longResult = 0;
2069 
2070             // Finally, have DigitList parse the digits into a value.
2071             if (digitList.fitsIntoLong(status[STATUS_POSITIVE], isParseIntegerOnly())) {
2072                 gotDouble = false;
2073                 longResult = digitList.getLong();
2074                 if (longResult < 0) {  // got Long.MIN_VALUE
2075                     gotLongMinimum = true;
2076                 }
2077             } else {
2078                 doubleResult = digitList.getDouble();
2079             }
2080 
2081             // Divide by multiplier. We have to be careful here not to do
2082             // unneeded conversions between double and long.
2083             if (multiplier != 1) {
2084                 if (gotDouble) {
2085                     doubleResult /= multiplier;
2086                 } else {
2087                     // Avoid converting to double if we can
2088                     if (longResult % multiplier == 0) {
2089                         longResult /= multiplier;
2090                     } else {
2091                         doubleResult = ((double)longResult) / multiplier;
2092                         gotDouble = true;
2093                     }
2094                 }
2095             }
2096 
2097             if (!status[STATUS_POSITIVE] && !gotLongMinimum) {
2098                 doubleResult = -doubleResult;
2099                 longResult = -longResult;
2100             }
2101 
2102             // At this point, if we divided the result by the multiplier, the
2103             // result may fit into a long.  We check for this case and return
2104             // a long if possible.
2105             // We must do this AFTER applying the negative (if appropriate)
2106             // in order to handle the case of LONG_MIN; otherwise, if we do
2107             // this with a positive value -LONG_MIN, the double is > 0, but
2108             // the long is < 0. We also must retain a double in the case of
2109             // -0.0, which will compare as == to a long 0 cast to a double
2110             // (bug 4162852).
2111             if (multiplier != 1 && gotDouble) {
2112                 longResult = (long)doubleResult;
2113                 gotDouble = ((doubleResult != (double)longResult) ||
2114                             (doubleResult == 0.0 && 1/doubleResult < 0.0)) &&
2115                             !isParseIntegerOnly();
2116             }
2117 
2118             // cast inside of ?: because of binary numeric promotion, JLS 15.25
2119             return gotDouble ? (Number)doubleResult : (Number)longResult;
2120         }
2121     }
2122 
2123     /**
2124      * Return a BigInteger multiplier.
2125      */
2126     private BigInteger getBigIntegerMultiplier() {
2127         if (bigIntegerMultiplier == null) {
2128             bigIntegerMultiplier = BigInteger.valueOf(multiplier);
2129         }
2130         return bigIntegerMultiplier;
2131     }
2132     private transient BigInteger bigIntegerMultiplier;
2133 
2134     /**
2135      * Return a BigDecimal multiplier.
2136      */
2137     private BigDecimal getBigDecimalMultiplier() {
2138         if (bigDecimalMultiplier == null) {
2139             bigDecimalMultiplier = new BigDecimal(multiplier);
2140         }
2141         return bigDecimalMultiplier;
2142     }
2143     private transient BigDecimal bigDecimalMultiplier;
2144 
2145     private static final int STATUS_INFINITE = 0;
2146     private static final int STATUS_POSITIVE = 1;
2147     private static final int STATUS_LENGTH   = 2;
2148 
2149     /**
2150      * Parse the given text into a number.  The text is parsed beginning at
2151      * parsePosition, until an unparseable character is seen.
2152      * @param text The string to parse.
2153      * @param parsePosition The position at which to being parsing.  Upon
2154      * return, the first unparseable character.
2155      * @param digits The DigitList to set to the parsed value.
2156      * @param isExponent If true, parse an exponent.  This means no
2157      * infinite values and integer only.
2158      * @param status Upon return contains boolean status flags indicating
2159      * whether the value was infinite and whether it was positive.
2160      */
2161     private final boolean subparse(String text, ParsePosition parsePosition,
2162                    String positivePrefix, String negativePrefix,
2163                    DigitList digits, boolean isExponent,
2164                    boolean status[]) {
2165         int position = parsePosition.index;
2166         int oldStart = parsePosition.index;
2167         int backup;
2168         boolean gotPositive, gotNegative;
2169 
2170         // check for positivePrefix; take longest
2171         gotPositive = text.regionMatches(position, positivePrefix, 0,
2172                                          positivePrefix.length());
2173         gotNegative = text.regionMatches(position, negativePrefix, 0,
2174                                          negativePrefix.length());
2175 
2176         if (gotPositive && gotNegative) {
2177             if (positivePrefix.length() > negativePrefix.length()) {
2178                 gotNegative = false;
2179             } else if (positivePrefix.length() < negativePrefix.length()) {
2180                 gotPositive = false;
2181             }
2182         }
2183 
2184         if (gotPositive) {
2185             position += positivePrefix.length();
2186         } else if (gotNegative) {
2187             position += negativePrefix.length();
2188         } else {
2189             parsePosition.errorIndex = position;
2190             return false;
2191         }
2192 
2193         // process digits or Inf, find decimal position
2194         status[STATUS_INFINITE] = false;
2195         if (!isExponent && text.regionMatches(position,symbols.getInfinity(),0,
2196                           symbols.getInfinity().length())) {
2197             position += symbols.getInfinity().length();
2198             status[STATUS_INFINITE] = true;
2199         } else {
2200             // We now have a string of digits, possibly with grouping symbols,
2201             // and decimal points.  We want to process these into a DigitList.
2202             // We don‘t want to put a bunch of leading zeros into the DigitList
2203             // though, so we keep track of the location of the decimal point,
2204             // put only significant digits into the DigitList, and adjust the
2205             // exponent as needed.
2206 
2207             digits.decimalAt = digits.count = 0;
2208             char zero = symbols.getZeroDigit();
2209             char decimal = isCurrencyFormat ?
2210                 symbols.getMonetaryDecimalSeparator() :
2211                 symbols.getDecimalSeparator();
2212             char grouping = symbols.getGroupingSeparator();
2213             String exponentString = symbols.getExponentSeparator();
2214             boolean sawDecimal = false;
2215             boolean sawExponent = false;
2216             boolean sawDigit = false;
2217             int exponent = 0; // Set to the exponent value, if any
2218 
2219             // We have to track digitCount ourselves, because digits.count will
2220             // pin when the maximum allowable digits is reached.
2221             int digitCount = 0;
2222 
2223             backup = -1;
2224             for (; position < text.length(); ++position) {
2225                 char ch = text.charAt(position);
2226 
2227                 /* We recognize all digit ranges, not only the Latin digit range
2228                  * ‘0‘..‘9‘.  We do so by using the Character.digit() method,
2229                  * which converts a valid Unicode digit to the range 0..9.
2230                  *
2231                  * The character ‘ch‘ may be a digit.  If so, place its value
2232                  * from 0 to 9 in ‘digit‘.  First try using the locale digit,
2233                  * which may or MAY NOT be a standard Unicode digit range.  If
2234                  * this fails, try using the standard Unicode digit ranges by
2235                  * calling Character.digit().  If this also fails, digit will
2236                  * have a value outside the range 0..9.
2237                  */
2238                 int digit = ch - zero;
2239                 if (digit < 0 || digit > 9) {
2240                     digit = Character.digit(ch, 10);
2241                 }
2242 
2243                 if (digit == 0) {
2244                     // Cancel out backup setting (see grouping handler below)
2245                     backup = -1; // Do this BEFORE continue statement below!!!
2246                     sawDigit = true;
2247 
2248                     // Handle leading zeros
2249                     if (digits.count == 0) {
2250                         // Ignore leading zeros in integer part of number.
2251                         if (!sawDecimal) {
2252                             continue;
2253                         }
2254 
2255                         // If we have seen the decimal, but no significant
2256                         // digits yet, then we account for leading zeros by
2257                         // decrementing the digits.decimalAt into negative
2258                         // values.
2259                         --digits.decimalAt;
2260                     } else {
2261                         ++digitCount;
2262                         digits.append((char)(digit + ‘0‘));
2263                     }
2264                 } else if (digit > 0 && digit <= 9) { // [sic] digit==0 handled above
2265                     sawDigit = true;
2266                     ++digitCount;
2267                     digits.append((char)(digit + ‘0‘));
2268 
2269                     // Cancel out backup setting (see grouping handler below)
2270                     backup = -1;
2271                 } else if (!isExponent && ch == decimal) {
2272                     // If we‘re only parsing integers, or if we ALREADY saw the
2273                     // decimal, then don‘t parse this one.
2274                     if (isParseIntegerOnly() || sawDecimal) {
2275                         break;
2276                     }
2277                     digits.decimalAt = digitCount; // Not digits.count!
2278                     sawDecimal = true;
2279                 } else if (!isExponent && ch == grouping && isGroupingUsed()) {
2280                     if (sawDecimal) {
2281                         break;
2282                     }
2283                     // Ignore grouping characters, if we are using them, but
2284                     // require that they be followed by a digit.  Otherwise
2285                     // we backup and reprocess them.
2286                     backup = position;
2287                 } else if (!isExponent && text.regionMatches(position, exponentString, 0, exponentString.length())
2288                              && !sawExponent) {
2289                     // Process the exponent by recursively calling this method.
2290                      ParsePosition pos = new ParsePosition(position + exponentString.length());
2291                     boolean[] stat = new boolean[STATUS_LENGTH];
2292                     DigitList exponentDigits = new DigitList();
2293 
2294                     if (subparse(text, pos, "", Character.toString(symbols.getMinusSign()), exponentDigits, true, stat) &&
2295                         exponentDigits.fitsIntoLong(stat[STATUS_POSITIVE], true)) {
2296                         position = pos.index; // Advance past the exponent
2297                         exponent = (int)exponentDigits.getLong();
2298                         if (!stat[STATUS_POSITIVE]) {
2299                             exponent = -exponent;
2300                         }
2301                         sawExponent = true;
2302                     }
2303                     break; // Whether we fail or succeed, we exit this loop
2304                 } else {
2305                     break;
2306                 }
2307             }
2308 
2309             if (backup != -1) {
2310                 position = backup;
2311             }
2312 
2313             // If there was no decimal point we have an integer
2314             if (!sawDecimal) {
2315                 digits.decimalAt = digitCount; // Not digits.count!
2316             }
2317 
2318             // Adjust for exponent, if any
2319             digits.decimalAt += exponent;
2320 
2321             // If none of the text string was recognized.  For example, parse
2322             // "x" with pattern "#0.00" (return index and error index both 0)
2323             // parse "$" with pattern "$#0.00". (return index 0 and error
2324             // index 1).
2325             if (!sawDigit && digitCount == 0) {
2326                 parsePosition.index = oldStart;
2327                 parsePosition.errorIndex = oldStart;
2328                 return false;
2329             }
2330         }
2331 
2332         // check for suffix
2333         if (!isExponent) {
2334             if (gotPositive) {
2335                 gotPositive = text.regionMatches(position,positiveSuffix,0,
2336                                                  positiveSuffix.length());
2337             }
2338             if (gotNegative) {
2339                 gotNegative = text.regionMatches(position,negativeSuffix,0,
2340                                                  negativeSuffix.length());
2341             }
2342 
2343         // if both match, take longest
2344         if (gotPositive && gotNegative) {
2345             if (positiveSuffix.length() > negativeSuffix.length()) {
2346                 gotNegative = false;
2347             } else if (positiveSuffix.length() < negativeSuffix.length()) {
2348                 gotPositive = false;
2349             }
2350         }
2351 
2352         // fail if neither or both
2353         if (gotPositive == gotNegative) {
2354             parsePosition.errorIndex = position;
2355             return false;
2356         }
2357 
2358         parsePosition.index = position +
2359             (gotPositive ? positiveSuffix.length() : negativeSuffix.length()); // mark success!
2360         } else {
2361             parsePosition.index = position;
2362         }
2363 
2364         status[STATUS_POSITIVE] = gotPositive;
2365         if (parsePosition.index == oldStart) {
2366             parsePosition.errorIndex = position;
2367             return false;
2368         }
2369         return true;
2370     }
2371 
2372     /**
2373      * Returns a copy of the decimal format symbols, which is generally not
2374      * changed by the programmer or user.
2375      * @return a copy of the desired DecimalFormatSymbols
2376      * @see java.text.DecimalFormatSymbols
2377      */
2378     public DecimalFormatSymbols getDecimalFormatSymbols() {
2379         try {
2380             // don‘t allow multiple references
2381             return (DecimalFormatSymbols) symbols.clone();
2382         } catch (Exception foo) {
2383             return null; // should never happen
2384         }
2385     }
2386 
2387 
2388     /**
2389      * Sets the decimal format symbols, which is generally not changed
2390      * by the programmer or user.
2391      * @param newSymbols desired DecimalFormatSymbols
2392      * @see java.text.DecimalFormatSymbols
2393      */
2394     public void setDecimalFormatSymbols(DecimalFormatSymbols newSymbols) {
2395         try {
2396             // don‘t allow multiple references
2397             symbols = (DecimalFormatSymbols) newSymbols.clone();
2398             expandAffixes();
2399             fastPathCheckNeeded = true;
2400         } catch (Exception foo) {
2401             // should never happen
2402         }
2403     }
2404 
2405     /**
2406      * Get the positive prefix.
2407      * <P>Examples: +123, $123, sFr123
2408      *
2409      * @return the positive prefix
2410      */
2411     public String getPositivePrefix () {
2412         return positivePrefix;
2413     }
2414 
2415     /**
2416      * Set the positive prefix.
2417      * <P>Examples: +123, $123, sFr123
2418      *
2419      * @param newValue the new positive prefix
2420      */
2421     public void setPositivePrefix (String newValue) {
2422         positivePrefix = newValue;
2423         posPrefixPattern = null;
2424         positivePrefixFieldPositions = null;
2425         fastPathCheckNeeded = true;
2426     }
2427 
2428     /**
2429      * Returns the FieldPositions of the fields in the prefix used for
2430      * positive numbers. This is not used if the user has explicitly set
2431      * a positive prefix via <code>setPositivePrefix</code>. This is
2432      * lazily created.
2433      *
2434      * @return FieldPositions in positive prefix
2435      */
2436     private FieldPosition[] getPositivePrefixFieldPositions() {
2437         if (positivePrefixFieldPositions == null) {
2438             if (posPrefixPattern != null) {
2439                 positivePrefixFieldPositions = expandAffix(posPrefixPattern);
2440             } else {
2441                 positivePrefixFieldPositions = EmptyFieldPositionArray;
2442             }
2443         }
2444         return positivePrefixFieldPositions;
2445     }
2446 
2447     /**
2448      * Get the negative prefix.
2449      * <P>Examples: -123, ($123) (with negative suffix), sFr-123
2450      *
2451      * @return the negative prefix
2452      */
2453     public String getNegativePrefix () {
2454         return negativePrefix;
2455     }
2456 
2457     /**
2458      * Set the negative prefix.
2459      * <P>Examples: -123, ($123) (with negative suffix), sFr-123
2460      *
2461      * @param newValue the new negative prefix
2462      */
2463     public void setNegativePrefix (String newValue) {
2464         negativePrefix = newValue;
2465         negPrefixPattern = null;
2466         fastPathCheckNeeded = true;
2467     }
2468 
2469     /**
2470      * Returns the FieldPositions of the fields in the prefix used for
2471      * negative numbers. This is not used if the user has explicitly set
2472      * a negative prefix via <code>setNegativePrefix</code>. This is
2473      * lazily created.
2474      *
2475      * @return FieldPositions in positive prefix
2476      */
2477     private FieldPosition[] getNegativePrefixFieldPositions() {
2478         if (negativePrefixFieldPositions == null) {
2479             if (negPrefixPattern != null) {
2480                 negativePrefixFieldPositions = expandAffix(negPrefixPattern);
2481             } else {
2482                 negativePrefixFieldPositions = EmptyFieldPositionArray;
2483             }
2484         }
2485         return negativePrefixFieldPositions;
2486     }
2487 
2488     /**
2489      * Get the positive suffix.
2490      * <P>Example: 123%
2491      *
2492      * @return the positive suffix
2493      */
2494     public String getPositiveSuffix () {
2495         return positiveSuffix;
2496     }
2497 
2498     /**
2499      * Set the positive suffix.
2500      * <P>Example: 123%
2501      *
2502      * @param newValue the new positive suffix
2503      */
2504     public void setPositiveSuffix (String newValue) {
2505         positiveSuffix = newValue;
2506         posSuffixPattern = null;
2507         fastPathCheckNeeded = true;
2508     }
2509 
2510     /**
2511      * Returns the FieldPositions of the fields in the suffix used for
2512      * positive numbers. This is not used if the user has explicitly set
2513      * a positive suffix via <code>setPositiveSuffix</code>. This is
2514      * lazily created.
2515      *
2516      * @return FieldPositions in positive prefix
2517      */
2518     private FieldPosition[] getPositiveSuffixFieldPositions() {
2519         if (positiveSuffixFieldPositions == null) {
2520             if (posSuffixPattern != null) {
2521                 positiveSuffixFieldPositions = expandAffix(posSuffixPattern);
2522             } else {
2523                 positiveSuffixFieldPositions = EmptyFieldPositionArray;
2524             }
2525         }
2526         return positiveSuffixFieldPositions;
2527     }
2528 
2529     /**
2530      * Get the negative suffix.
2531      * <P>Examples: -123%, ($123) (with positive suffixes)
2532      *
2533      * @return the negative suffix
2534      */
2535     public String getNegativeSuffix () {
2536         return negativeSuffix;
2537     }
2538 
2539     /**
2540      * Set the negative suffix.
2541      * <P>Examples: 123%
2542      *
2543      * @param newValue the new negative suffix
2544      */
2545     public void setNegativeSuffix (String newValue) {
2546         negativeSuffix = newValue;
2547         negSuffixPattern = null;
2548         fastPathCheckNeeded = true;
2549     }
2550 
2551     /**
2552      * Returns the FieldPositions of the fields in the suffix used for
2553      * negative numbers. This is not used if the user has explicitly set
2554      * a negative suffix via <code>setNegativeSuffix</code>. This is
2555      * lazily created.
2556      *
2557      * @return FieldPositions in positive prefix
2558      */
2559     private FieldPosition[] getNegativeSuffixFieldPositions() {
2560         if (negativeSuffixFieldPositions == null) {
2561             if (negSuffixPattern != null) {
2562                 negativeSuffixFieldPositions = expandAffix(negSuffixPattern);
2563             } else {
2564                 negativeSuffixFieldPositions = EmptyFieldPositionArray;
2565             }
2566         }
2567         return negativeSuffixFieldPositions;
2568     }
2569 
2570     /**
2571      * Gets the multiplier for use in percent, per mille, and similar
2572      * formats.
2573      *
2574      * @return the multiplier
2575      * @see #setMultiplier(int)
2576      */
2577     public int getMultiplier () {
2578         return multiplier;
2579     }
2580 
2581     /**
2582      * Sets the multiplier for use in percent, per mille, and similar
2583      * formats.
2584      * For a percent format, set the multiplier to 100 and the suffixes to
2585      * have ‘%‘ (for Arabic, use the Arabic percent sign).
2586      * For a per mille format, set the multiplier to 1000 and the suffixes to
2587      * have ‘&#92;u2030‘.
2588      *
2589      * <P>Example: with multiplier 100, 1.23 is formatted as "123", and
2590      * "123" is parsed into 1.23.
2591      *
2592      * @param newValue the new multiplier
2593      * @see #getMultiplier
2594      */
2595     public void setMultiplier (int newValue) {
2596         multiplier = newValue;
2597         bigDecimalMultiplier = null;
2598         bigIntegerMultiplier = null;
2599         fastPathCheckNeeded = true;
2600     }
2601 
2602     /**
2603      * {@inheritDoc}
2604      */
2605     @Override
2606     public void setGroupingUsed(boolean newValue) {
2607         super.setGroupingUsed(newValue);
2608         fastPathCheckNeeded = true;
2609     }
2610 
2611     /**
2612      * Return the grouping size. Grouping size is the number of digits between
2613      * grouping separators in the integer portion of a number.  For example,
2614      * in the number "123,456.78", the grouping size is 3.
2615      *
2616      * @return the grouping size
2617      * @see #setGroupingSize
2618      * @see java.text.NumberFormat#isGroupingUsed
2619      * @see java.text.DecimalFormatSymbols#getGroupingSeparator
2620      */
2621     public int getGroupingSize () {
2622         return groupingSize;
2623     }
2624 
2625     /**
2626      * Set the grouping size. Grouping size is the number of digits between
2627      * grouping separators in the integer portion of a number.  For example,
2628      * in the number "123,456.78", the grouping size is 3.
2629      * <br>
2630      * The value passed in is converted to a byte, which may lose information.
2631      *
2632      * @param newValue the new grouping size
2633      * @see #getGroupingSize
2634      * @see java.text.NumberFormat#setGroupingUsed
2635      * @see java.text.DecimalFormatSymbols#setGroupingSeparator
2636      */
2637     public void setGroupingSize (int newValue) {
2638         groupingSize = (byte)newValue;
2639         fastPathCheckNeeded = true;
2640     }
2641 
2642     /**
2643      * Allows you to get the behavior of the decimal separator with integers.
2644      * (The decimal separator will always appear with decimals.)
2645      * <P>Example: Decimal ON: 12345 &rarr; 12345.; OFF: 12345 &rarr; 12345
2646      *
2647      * @return {@code true} if the decimal separator is always shown;
2648      *         {@code false} otherwise
2649      */
2650     public boolean isDecimalSeparatorAlwaysShown() {
2651         return decimalSeparatorAlwaysShown;
2652     }
2653 
2654     /**
2655      * Allows you to set the behavior of the decimal separator with integers.
2656      * (The decimal separator will always appear with decimals.)
2657      * <P>Example: Decimal ON: 12345 &rarr; 12345.; OFF: 12345 &rarr; 12345
2658      *
2659      * @param newValue {@code true} if the decimal separator is always shown;
2660      *                 {@code false} otherwise
2661      */
2662     public void setDecimalSeparatorAlwaysShown(boolean newValue) {
2663         decimalSeparatorAlwaysShown = newValue;
2664         fastPathCheckNeeded = true;
2665     }
2666 
2667     /**
2668      * Returns whether the {@link #parse(java.lang.String, java.text.ParsePosition)}
2669      * method returns <code>BigDecimal</code>. The default value is false.
2670      *
2671      * @return {@code true} if the parse method returns BigDecimal;
2672      *         {@code false} otherwise
2673      * @see #setParseBigDecimal
2674      * @since 1.5
2675      */
2676     public boolean isParseBigDecimal() {
2677         return parseBigDecimal;
2678     }
2679 
2680     /**
2681      * Sets whether the {@link #parse(java.lang.String, java.text.ParsePosition)}
2682      * method returns <code>BigDecimal</code>.
2683      *
2684      * @param newValue {@code true} if the parse method returns BigDecimal;
2685      *                 {@code false} otherwise
2686      * @see #isParseBigDecimal
2687      * @since 1.5
2688      */
2689     public void setParseBigDecimal(boolean newValue) {
2690         parseBigDecimal = newValue;
2691     }
2692 
2693     /**
2694      * Standard override; no change in semantics.
2695      */
2696     @Override
2697     public Object clone() {
2698         DecimalFormat other = (DecimalFormat) super.clone();
2699         other.symbols = (DecimalFormatSymbols) symbols.clone();
2700         other.digitList = (DigitList) digitList.clone();
2701 
2702         // Fast-path is almost stateless algorithm. The only logical state is the
2703         // isFastPath flag. In addition fastPathCheckNeeded is a sentinel flag
2704         // that forces recalculation of all fast-path fields when set to true.
2705         //
2706         // There is thus no need to clone all the fast-path fields.
2707         // We just only need to set fastPathCheckNeeded to true when cloning,
2708         // and init fastPathData to null as if it were a truly new instance.
2709         // Every fast-path field will be recalculated (only once) at next usage of
2710         // fast-path algorithm.
2711         other.fastPathCheckNeeded = true;
2712         other.isFastPath = false;
2713         other.fastPathData = null;
2714 
2715         return other;
2716     }
2717 
2718     /**
2719      * Overrides equals
2720      */
2721     @Override
2722     public boolean equals(Object obj)
2723     {
2724         if (obj == null)
2725             return false;
2726         if (!super.equals(obj))
2727             return false; // super does class check
2728         DecimalFormat other = (DecimalFormat) obj;
2729         return ((posPrefixPattern == other.posPrefixPattern &&
2730                  positivePrefix.equals(other.positivePrefix))
2731                 || (posPrefixPattern != null &&
2732                     posPrefixPattern.equals(other.posPrefixPattern)))
2733             && ((posSuffixPattern == other.posSuffixPattern &&
2734                  positiveSuffix.equals(other.positiveSuffix))
2735                 || (posSuffixPattern != null &&
2736                     posSuffixPattern.equals(other.posSuffixPattern)))
2737             && ((negPrefixPattern == other.negPrefixPattern &&
2738                  negativePrefix.equals(other.negativePrefix))
2739                 || (negPrefixPattern != null &&
2740                     negPrefixPattern.equals(other.negPrefixPattern)))
2741             && ((negSuffixPattern == other.negSuffixPattern &&
2742                  negativeSuffix.equals(other.negativeSuffix))
2743                 || (negSuffixPattern != null &&
2744                     negSuffixPattern.equals(other.negSuffixPattern)))
2745             && multiplier == other.multiplier
2746             && groupingSize == other.groupingSize
2747             && decimalSeparatorAlwaysShown == other.decimalSeparatorAlwaysShown
2748             && parseBigDecimal == other.parseBigDecimal
2749             && useExponentialNotation == other.useExponentialNotation
2750             && (!useExponentialNotation ||
2751                 minExponentDigits == other.minExponentDigits)
2752             && maximumIntegerDigits == other.maximumIntegerDigits
2753             && minimumIntegerDigits == other.minimumIntegerDigits
2754             && maximumFractionDigits == other.maximumFractionDigits
2755             && minimumFractionDigits == other.minimumFractionDigits
2756             && roundingMode == other.roundingMode
2757             && symbols.equals(other.symbols);
2758     }
2759 
2760     /**
2761      * Overrides hashCode
2762      */
2763     @Override
2764     public int hashCode() {
2765         return super.hashCode() * 37 + positivePrefix.hashCode();
2766         // just enough fields for a reasonable distribution
2767     }
2768 
2769     /**
2770      * Synthesizes a pattern string that represents the current state
2771      * of this Format object.
2772      *
2773      * @return a pattern string
2774      * @see #applyPattern
2775      */
2776     public String toPattern() {
2777         return toPattern( false );
2778     }
2779 
2780     /**
2781      * Synthesizes a localized pattern string that represents the current
2782      * state of this Format object.
2783      *
2784      * @return a localized pattern string
2785      * @see #applyPattern
2786      */
2787     public String toLocalizedPattern() {
2788         return toPattern( true );
2789     }
2790 
2791     /**
2792      * Expand the affix pattern strings into the expanded affix strings.  If any
2793      * affix pattern string is null, do not expand it.  This method should be
2794      * called any time the symbols or the affix patterns change in order to keep
2795      * the expanded affix strings up to date.
2796      */
2797     private void expandAffixes() {
2798         // Reuse one StringBuffer for better performance
2799         StringBuffer buffer = new StringBuffer();
2800         if (posPrefixPattern != null) {
2801             positivePrefix = expandAffix(posPrefixPattern, buffer);
2802             positivePrefixFieldPositions = null;
2803         }
2804         if (posSuffixPattern != null) {
2805             positiveSuffix = expandAffix(posSuffixPattern, buffer);
2806             positiveSuffixFieldPositions = null;
2807         }
2808         if (negPrefixPattern != null) {
2809             negativePrefix = expandAffix(negPrefixPattern, buffer);
2810             negativePrefixFieldPositions = null;
2811         }
2812         if (negSuffixPattern != null) {
2813             negativeSuffix = expandAffix(negSuffixPattern, buffer);
2814             negativeSuffixFieldPositions = null;
2815         }
2816     }
2817 
2818     /**
2819      * Expand an affix pattern into an affix string.  All characters in the
2820      * pattern are literal unless prefixed by QUOTE.  The following characters
2821      * after QUOTE are recognized: PATTERN_PERCENT, PATTERN_PER_MILLE,
2822      * PATTERN_MINUS, and CURRENCY_SIGN.  If CURRENCY_SIGN is doubled (QUOTE +
2823      * CURRENCY_SIGN + CURRENCY_SIGN), it is interpreted as an ISO 4217
2824      * currency code.  Any other character after a QUOTE represents itself.
2825      * QUOTE must be followed by another character; QUOTE may not occur by
2826      * itself at the end of the pattern.
2827      *
2828      * @param pattern the non-null, possibly empty pattern
2829      * @param buffer a scratch StringBuffer; its contents will be lost
2830      * @return the expanded equivalent of pattern
2831      */
2832     private String expandAffix(String pattern, StringBuffer buffer) {
2833         buffer.setLength(0);
2834         for (int i=0; i<pattern.length(); ) {
2835             char c = pattern.charAt(i++);
2836             if (c == QUOTE) {
2837                 c = pattern.charAt(i++);
2838                 switch (c) {
2839                 case CURRENCY_SIGN:
2840                     if (i<pattern.length() &&
2841                         pattern.charAt(i) == CURRENCY_SIGN) {
2842                         ++i;
2843                         buffer.append(symbols.getInternationalCurrencySymbol());
2844                     } else {
2845                         buffer.append(symbols.getCurrencySymbol());
2846                     }
2847                     continue;
2848                 case PATTERN_PERCENT:
2849                     c = symbols.getPercent();
2850                     break;
2851                 case PATTERN_PER_MILLE:
2852                     c = symbols.getPerMill();
2853                     break;
2854                 case PATTERN_MINUS:
2855                     c = symbols.getMinusSign();
2856                     break;
2857                 }
2858             }
2859             buffer.append(c);
2860         }
2861         return buffer.toString();
2862     }
2863 
2864     /**
2865      * Expand an affix pattern into an array of FieldPositions describing
2866      * how the pattern would be expanded.
2867      * All characters in the
2868      * pattern are literal unless prefixed by QUOTE.  The following characters
2869      * after QUOTE are recognized: PATTERN_PERCENT, PATTERN_PER_MILLE,
2870      * PATTERN_MINUS, and CURRENCY_SIGN.  If CURRENCY_SIGN is doubled (QUOTE +
2871      * CURRENCY_SIGN + CURRENCY_SIGN), it is interpreted as an ISO 4217
2872      * currency code.  Any other character after a QUOTE represents itself.
2873      * QUOTE must be followed by another character; QUOTE may not occur by
2874      * itself at the end of the pattern.
2875      *
2876      * @param pattern the non-null, possibly empty pattern
2877      * @return FieldPosition array of the resulting fields.
2878      */
2879     private FieldPosition[] expandAffix(String pattern) {
2880         ArrayList<FieldPosition> positions = null;
2881         int stringIndex = 0;
2882         for (int i=0; i<pattern.length(); ) {
2883             char c = pattern.charAt(i++);
2884             if (c == QUOTE) {
2885                 int field = -1;
2886                 Format.Field fieldID = null;
2887                 c = pattern.charAt(i++);
2888                 switch (c) {
2889                 case CURRENCY_SIGN:
2890                     String string;
2891                     if (i<pattern.length() &&
2892                         pattern.charAt(i) == CURRENCY_SIGN) {
2893                         ++i;
2894                         string = symbols.getInternationalCurrencySymbol();
2895                     } else {
2896                         string = symbols.getCurrencySymbol();
2897                     }
2898                     if (string.length() > 0) {
2899                         if (positions == null) {
2900                             positions = new ArrayList<>(2);
2901                         }
2902                         FieldPosition fp = new FieldPosition(Field.CURRENCY);
2903                         fp.setBeginIndex(stringIndex);
2904                         fp.setEndIndex(stringIndex + string.length());
2905                         positions.add(fp);
2906                         stringIndex += string.length();
2907                     }
2908                     continue;
2909                 case PATTERN_PERCENT:
2910                     c = symbols.getPercent();
2911                     field = -1;
2912                     fieldID = Field.PERCENT;
2913                     break;
2914                 case PATTERN_PER_MILLE:
2915                     c = symbols.getPerMill();
2916                     field = -1;
2917                     fieldID = Field.PERMILLE;
2918                     break;
2919                 case PATTERN_MINUS:
2920                     c = symbols.getMinusSign();
2921                     field = -1;
2922                     fieldID = Field.SIGN;
2923                     break;
2924                 }
2925                 if (fieldID != null) {
2926                     if (positions == null) {
2927                         positions = new ArrayList<>(2);
2928                     }
2929                     FieldPosition fp = new FieldPosition(fieldID, field);
2930                     fp.setBeginIndex(stringIndex);
2931                     fp.setEndIndex(stringIndex + 1);
2932                     positions.add(fp);
2933                 }
2934             }
2935             stringIndex++;
2936         }
2937         if (positions != null) {
2938             return positions.toArray(EmptyFieldPositionArray);
2939         }
2940         return EmptyFieldPositionArray;
2941     }
2942 
2943     /**
2944      * Appends an affix pattern to the given StringBuffer, quoting special
2945      * characters as needed.  Uses the internal affix pattern, if that exists,
2946      * or the literal affix, if the internal affix pattern is null.  The
2947      * appended string will generate the same affix pattern (or literal affix)
2948      * when passed to toPattern().
2949      *
2950      * @param buffer the affix string is appended to this
2951      * @param affixPattern a pattern such as posPrefixPattern; may be null
2952      * @param expAffix a corresponding expanded affix, such as positivePrefix.
2953      * Ignored unless affixPattern is null.  If affixPattern is null, then
2954      * expAffix is appended as a literal affix.
2955      * @param localized true if the appended pattern should contain localized
2956      * pattern characters; otherwise, non-localized pattern chars are appended
2957      */
2958     private void appendAffix(StringBuffer buffer, String affixPattern,
2959                              String expAffix, boolean localized) {
2960         if (affixPattern == null) {
2961             appendAffix(buffer, expAffix, localized);
2962         } else {
2963             int i;
2964             for (int pos=0; pos<affixPattern.length(); pos=i) {
2965                 i = affixPattern.indexOf(QUOTE, pos);
2966                 if (i < 0) {
2967                     appendAffix(buffer, affixPattern.substring(pos), localized);
2968                     break;
2969                 }
2970                 if (i > pos) {
2971                     appendAffix(buffer, affixPattern.substring(pos, i), localized);
2972                 }
2973                 char c = affixPattern.charAt(++i);
2974                 ++i;
2975                 if (c == QUOTE) {
2976                     buffer.append(c);
2977                     // Fall through and append another QUOTE below
2978                 } else if (c == CURRENCY_SIGN &&
2979                            i<affixPattern.length() &&
2980                            affixPattern.charAt(i) == CURRENCY_SIGN) {
2981                     ++i;
2982                     buffer.append(c);
2983                     // Fall through and append another CURRENCY_SIGN below
2984                 } else if (localized) {
2985                     switch (c) {
2986                     case PATTERN_PERCENT:
2987                         c = symbols.getPercent();
2988                         break;
2989                     case PATTERN_PER_MILLE:
2990                         c = symbols.getPerMill();
2991                         break;
2992                     case PATTERN_MINUS:
2993                         c = symbols.getMinusSign();
2994                         break;
2995                     }
2996                 }
2997                 buffer.append(c);
2998             }
2999         }
3000     }
3001 
3002     /**
3003      * Append an affix to the given StringBuffer, using quotes if
3004      * there are special characters.  Single quotes themselves must be
3005      * escaped in either case.
3006      */
3007     private void appendAffix(StringBuffer buffer, String affix, boolean localized) {
3008         boolean needQuote;
3009         if (localized) {
3010             needQuote = affix.indexOf(symbols.getZeroDigit()) >= 0
3011                 || affix.indexOf(symbols.getGroupingSeparator()) >= 0
3012                 || affix.indexOf(symbols.getDecimalSeparator()) >= 0
3013                 || affix.indexOf(symbols.getPercent()) >= 0
3014                 || affix.indexOf(symbols.getPerMill()) >= 0
3015                 || affix.indexOf(symbols.getDigit()) >= 0
3016                 || affix.indexOf(symbols.getPatternSeparator()) >= 0
3017                 || affix.indexOf(symbols.getMinusSign()) >= 0
3018                 || affix.indexOf(CURRENCY_SIGN) >= 0;
3019         } else {
3020             needQuote = affix.indexOf(PATTERN_ZERO_DIGIT) >= 0
3021                 || affix.indexOf(PATTERN_GROUPING_SEPARATOR) >= 0
3022                 || affix.indexOf(PATTERN_DECIMAL_SEPARATOR) >= 0
3023                 || affix.indexOf(PATTERN_PERCENT) >= 0
3024                 || affix.indexOf(PATTERN_PER_MILLE) >= 0
3025                 || affix.indexOf(PATTERN_DIGIT) >= 0
3026                 || affix.indexOf(PATTERN_SEPARATOR) >= 0
3027                 || affix.indexOf(PATTERN_MINUS) >= 0
3028                 || affix.indexOf(CURRENCY_SIGN) >= 0;
3029         }
3030         if (needQuote) buffer.append(‘\‘‘);
3031         if (affix.indexOf(‘\‘‘) < 0) buffer.append(affix);
3032         else {
3033             for (int j=0; j<affix.length(); ++j) {
3034                 char c = affix.charAt(j);
3035                 buffer.append(c);
3036                 if (c == ‘\‘‘) buffer.append(c);
3037             }
3038         }
3039         if (needQuote) buffer.append(‘\‘‘);
3040     }
3041 
3042     /**
3043      * Does the real work of generating a pattern.  */
3044     private String toPattern(boolean localized) {
3045         StringBuffer result = new StringBuffer();
3046         for (int j = 1; j >= 0; --j) {
3047             if (j == 1)
3048                 appendAffix(result, posPrefixPattern, positivePrefix, localized);
3049             else appendAffix(result, negPrefixPattern, negativePrefix, localized);
3050             int i;
3051             int digitCount = useExponentialNotation
3052                         ? getMaximumIntegerDigits()
3053                         : Math.max(groupingSize, getMinimumIntegerDigits())+1;
3054             for (i = digitCount; i > 0; --i) {
3055                 if (i != digitCount && isGroupingUsed() && groupingSize != 0 &&
3056                     i % groupingSize == 0) {
3057                     result.append(localized ? symbols.getGroupingSeparator() :
3058                                   PATTERN_GROUPING_SEPARATOR);
3059                 }
3060                 result.append(i <= getMinimumIntegerDigits()
3061                     ? (localized ? symbols.getZeroDigit() : PATTERN_ZERO_DIGIT)
3062                     : (localized ? symbols.getDigit() : PATTERN_DIGIT));
3063             }
3064             if (getMaximumFractionDigits() > 0 || decimalSeparatorAlwaysShown)
3065                 result.append(localized ? symbols.getDecimalSeparator() :
3066                               PATTERN_DECIMAL_SEPARATOR);
3067             for (i = 0; i < getMaximumFractionDigits(); ++i) {
3068                 if (i < getMinimumFractionDigits()) {
3069                     result.append(localized ? symbols.getZeroDigit() :
3070                                   PATTERN_ZERO_DIGIT);
3071                 } else {
3072                     result.append(localized ? symbols.getDigit() :
3073                                   PATTERN_DIGIT);
3074                 }
3075             }
3076         if (useExponentialNotation)
3077         {
3078             result.append(localized ? symbols.getExponentSeparator() :
3079                   PATTERN_EXPONENT);
3080         for (i=0; i<minExponentDigits; ++i)
3081                     result.append(localized ? symbols.getZeroDigit() :
3082                                   PATTERN_ZERO_DIGIT);
3083         }
3084             if (j == 1) {
3085                 appendAffix(result, posSuffixPattern, positiveSuffix, localized);
3086                 if ((negSuffixPattern == posSuffixPattern && // n == p == null
3087                      negativeSuffix.equals(positiveSuffix))
3088                     || (negSuffixPattern != null &&
3089                         negSuffixPattern.equals(posSuffixPattern))) {
3090                     if ((negPrefixPattern != null && posPrefixPattern != null &&
3091                          negPrefixPattern.equals("‘-" + posPrefixPattern)) ||
3092                         (negPrefixPattern == posPrefixPattern && // n == p == null
3093                          negativePrefix.equals(symbols.getMinusSign() + positivePrefix)))
3094                         break;
3095                 }
3096                 result.append(localized ? symbols.getPatternSeparator() :
3097                               PATTERN_SEPARATOR);
3098             } else appendAffix(result, negSuffixPattern, negativeSuffix, localized);
3099         }
3100         return result.toString();
3101     }
3102 
3103     /**
3104      * Apply the given pattern to this Format object.  A pattern is a
3105      * short-hand specification for the various formatting properties.
3106      * These properties can also be changed individually through the
3107      * various setter methods.
3108      * <p>
3109      * There is no limit to integer digits set
3110      * by this routine, since that is the typical end-user desire;
3111      * use setMaximumInteger if you want to set a real value.
3112      * For negative numbers, use a second pattern, separated by a semicolon
3113      * <P>Example <code>"#,#00.0#"</code> &rarr; 1,234.56
3114      * <P>This means a minimum of 2 integer digits, 1 fraction digit, and
3115      * a maximum of 2 fraction digits.
3116      * <p>Example: <code>"#,#00.0#;(#,#00.0#)"</code> for negatives in
3117      * parentheses.
3118      * <p>In negative patterns, the minimum and maximum counts are ignored;
3119      * these are presumed to be set in the positive pattern.
3120      *
3121      * @param pattern a new pattern
3122      * @exception NullPointerException if <code>pattern</code> is null
3123      * @exception IllegalArgumentException if the given pattern is invalid.
3124      */
3125     public void applyPattern(String pattern) {
3126         applyPattern(pattern, false);
3127     }
3128 
3129     /**
3130      * Apply the given pattern to this Format object.  The pattern
3131      * is assumed to be in a localized notation. A pattern is a
3132      * short-hand specification for the various formatting properties.
3133      * These properties can also be changed individually through the
3134      * various setter methods.
3135      * <p>
3136      * There is no limit to integer digits set
3137      * by this routine, since that is the typical end-user desire;
3138      * use setMaximumInteger if you want to set a real value.
3139      * For negative numbers, use a second pattern, separated by a semicolon
3140      * <P>Example <code>"#,#00.0#"</code> &rarr; 1,234.56
3141      * <P>This means a minimum of 2 integer digits, 1 fraction digit, and
3142      * a maximum of 2 fraction digits.
3143      * <p>Example: <code>"#,#00.0#;(#,#00.0#)"</code> for negatives in
3144      * parentheses.
3145      * <p>In negative patterns, the minimum and maximum counts are ignored;
3146      * these are presumed to be set in the positive pattern.
3147      *
3148      * @param pattern a new pattern
3149      * @exception NullPointerException if <code>pattern</code> is null
3150      * @exception IllegalArgumentException if the given pattern is invalid.
3151      */
3152     public void applyLocalizedPattern(String pattern) {
3153         applyPattern(pattern, true);
3154     }
3155 
3156     /**
3157      * Does the real work of applying a pattern.
3158      */
3159     private void applyPattern(String pattern, boolean localized) {
3160         char zeroDigit         = PATTERN_ZERO_DIGIT;
3161         char groupingSeparator = PATTERN_GROUPING_SEPARATOR;
3162         char decimalSeparator  = PATTERN_DECIMAL_SEPARATOR;
3163         char percent           = PATTERN_PERCENT;
3164         char perMill           = PATTERN_PER_MILLE;
3165         char digit             = PATTERN_DIGIT;
3166         char separator         = PATTERN_SEPARATOR;
3167         String exponent          = PATTERN_EXPONENT;
3168         char minus             = PATTERN_MINUS;
3169         if (localized) {
3170             zeroDigit         = symbols.getZeroDigit();
3171             groupingSeparator = symbols.getGroupingSeparator();
3172             decimalSeparator  = symbols.getDecimalSeparator();
3173             percent           = symbols.getPercent();
3174             perMill           = symbols.getPerMill();
3175             digit             = symbols.getDigit();
3176             separator         = symbols.getPatternSeparator();
3177             exponent          = symbols.getExponentSeparator();
3178             minus             = symbols.getMinusSign();
3179         }
3180         boolean gotNegative = false;
3181         decimalSeparatorAlwaysShown = false;
3182         isCurrencyFormat = false;
3183         useExponentialNotation = false;
3184 
3185         // Two variables are used to record the subrange of the pattern
3186         // occupied by phase 1.  This is used during the processing of the
3187         // second pattern (the one representing negative numbers) to ensure
3188         // that no deviation exists in phase 1 between the two patterns.
3189         int phaseOneStart = 0;
3190         int phaseOneLength = 0;
3191 
3192         int start = 0;
3193         for (int j = 1; j >= 0 && start < pattern.length(); --j) {
3194             boolean inQuote = false;
3195             StringBuffer prefix = new StringBuffer();
3196             StringBuffer suffix = new StringBuffer();
3197             int decimalPos = -1;
3198             int multiplier = 1;
3199             int digitLeftCount = 0, zeroDigitCount = 0, digitRightCount = 0;
3200             byte groupingCount = -1;
3201 
3202             // The phase ranges from 0 to 2.  Phase 0 is the prefix.  Phase 1 is
3203             // the section of the pattern with digits, decimal separator,
3204             // grouping characters.  Phase 2 is the suffix.  In phases 0 and 2,
3205             // percent, per mille, and currency symbols are recognized and
3206             // translated.  The separation of the characters into phases is
3207             // strictly enforced; if phase 1 characters are to appear in the
3208             // suffix, for example, they must be quoted.
3209             int phase = 0;
3210 
3211             // The affix is either the prefix or the suffix.
3212             StringBuffer affix = prefix;
3213 
3214             for (int pos = start; pos < pattern.length(); ++pos) {
3215                 char ch = pattern.charAt(pos);
3216                 switch (phase) {
3217                 case 0:
3218                 case 2:
3219                     // Process the prefix / suffix characters
3220                     if (inQuote) {
3221                         // A quote within quotes indicates either the closing
3222                         // quote or two quotes, which is a quote literal. That
3223                         // is, we have the second quote in ‘do‘ or ‘don‘‘t‘.
3224                         if (ch == QUOTE) {
3225                             if ((pos+1) < pattern.length() &&
3226                                 pattern.charAt(pos+1) == QUOTE) {
3227                                 ++pos;
3228                                 affix.append("‘‘"); // ‘don‘‘t‘
3229                             } else {
3230                                 inQuote = false; // ‘do‘
3231                             }
3232                             continue;
3233                         }
3234                     } else {
3235                         // Process unquoted characters seen in prefix or suffix
3236                         // phase.
3237                         if (ch == digit ||
3238                             ch == zeroDigit ||
3239                             ch == groupingSeparator ||
3240                             ch == decimalSeparator) {
3241                             phase = 1;
3242                             if (j == 1) {
3243                                 phaseOneStart = pos;
3244                             }
3245                             --pos; // Reprocess this character
3246                             continue;
3247                         } else if (ch == CURRENCY_SIGN) {
3248                             // Use lookahead to determine if the currency sign
3249                             // is doubled or not.
3250                             boolean doubled = (pos + 1) < pattern.length() &&
3251                                 pattern.charAt(pos + 1) == CURRENCY_SIGN;
3252                             if (doubled) { // Skip over the doubled character
3253                              ++pos;
3254                             }
3255                             isCurrencyFormat = true;
3256                             affix.append(doubled ? "‘\u00A4\u00A4" : "‘\u00A4");
3257                             continue;
3258                         } else if (ch == QUOTE) {
3259                             // A quote outside quotes indicates either the
3260                             // opening quote or two quotes, which is a quote
3261                             // literal. That is, we have the first quote in ‘do‘
3262                             // or o‘‘clock.
3263                             if (ch == QUOTE) {
3264                                 if ((pos+1) < pattern.length() &&
3265                                     pattern.charAt(pos+1) == QUOTE) {
3266                                     ++pos;
3267                                     affix.append("‘‘"); // o‘‘clock
3268                                 } else {
3269                                     inQuote = true; // ‘do‘
3270                                 }
3271                                 continue;
3272                             }
3273                         } else if (ch == separator) {
3274                             // Don‘t allow separators before we see digit
3275                             // characters of phase 1, and don‘t allow separators
3276                             // in the second pattern (j == 0).
3277                             if (phase == 0 || j == 0) {
3278                                 throw new IllegalArgumentException("Unquoted special character ‘" +
3279                                     ch + "‘ in pattern \"" + pattern + ‘"‘);
3280                             }
3281                             start = pos + 1;
3282                             pos = pattern.length();
3283                             continue;
3284                         }
3285 
3286                         // Next handle characters which are appended directly.
3287                         else if (ch == percent) {
3288                             if (multiplier != 1) {
3289                                 throw new IllegalArgumentException("Too many percent/per mille characters in pattern \"" +
3290                                     pattern + ‘"‘);
3291                             }
3292                             multiplier = 100;
3293                             affix.append("‘%");
3294                             continue;
3295                         } else if (ch == perMill) {
3296                             if (multiplier != 1) {
3297                                 throw new IllegalArgumentException("Too many percent/per mille characters in pattern \"" +
3298                                     pattern + ‘"‘);
3299                             }
3300                             multiplier = 1000;
3301                             affix.append("‘\u2030");
3302                             continue;
3303                         } else if (ch == minus) {
3304                             affix.append("‘-");
3305                             continue;
3306                         }
3307                     }
3308                     // Note that if we are within quotes, or if this is an
3309                     // unquoted, non-special character, then we usually fall
3310                     // through to here.
3311                     affix.append(ch);
3312                     break;
3313 
3314                 case 1:
3315                     // Phase one must be identical in the two sub-patterns. We
3316                     // enforce this by doing a direct comparison. While
3317                     // processing the first sub-pattern, we just record its
3318                     // length. While processing the second, we compare
3319                     // characters.
3320                     if (j == 1) {
3321                         ++phaseOneLength;
3322                     } else {
3323                         if (--phaseOneLength == 0) {
3324                             phase = 2;
3325                             affix = suffix;
3326                         }
3327                         continue;
3328                     }
3329 
3330                     // Process the digits, decimal, and grouping characters. We
3331                     // record five pieces of information. We expect the digits
3332                     // to occur in the pattern ####0000.####, and we record the
3333                     // number of left digits, zero (central) digits, and right
3334                     // digits. The position of the last grouping character is
3335                     // recorded (should be somewhere within the first two blocks
3336                     // of characters), as is the position of the decimal point,
3337                     // if any (should be in the zero digits). If there is no
3338                     // decimal point, then there should be no right digits.
3339                     if (ch == digit) {
3340                         if (zeroDigitCount > 0) {
3341                             ++digitRightCount;
3342                         } else {
3343                             ++digitLeftCount;
3344                         }
3345                         if (groupingCount >= 0 && decimalPos < 0) {
3346                             ++groupingCount;
3347                         }
3348                     } else if (ch == zeroDigit) {
3349                         if (digitRightCount > 0) {
3350                             throw new IllegalArgumentException("Unexpected ‘0‘ in pattern \"" +
3351                                 pattern + ‘"‘);
3352                         }
3353                         ++zeroDigitCount;
3354                         if (groupingCount >= 0 && decimalPos < 0) {
3355                             ++groupingCount;
3356                         }
3357                     } else if (ch == groupingSeparator) {
3358                         groupingCount = 0;
3359                     } else if (ch == decimalSeparator) {
3360                         if (decimalPos >= 0) {
3361                             throw new IllegalArgumentException("Multiple decimal separators in pattern \"" +
3362                                 pattern + ‘"‘);
3363                         }
3364                         decimalPos = digitLeftCount + zeroDigitCount + digitRightCount;
3365                     } else if (pattern.regionMatches(pos, exponent, 0, exponent.length())){
3366                         if (useExponentialNotation) {
3367                             throw new IllegalArgumentException("Multiple exponential " +
3368                                 "symbols in pattern \"" + pattern + ‘"‘);
3369                         }
3370                         useExponentialNotation = true;
3371                         minExponentDigits = 0;
3372 
3373                         // Use lookahead to parse out the exponential part
3374                         // of the pattern, then jump into phase 2.
3375                         pos = pos+exponent.length();
3376                          while (pos < pattern.length() &&
3377                                pattern.charAt(pos) == zeroDigit) {
3378                             ++minExponentDigits;
3379                             ++phaseOneLength;
3380                             ++pos;
3381                         }
3382 
3383                         if ((digitLeftCount + zeroDigitCount) < 1 ||
3384                             minExponentDigits < 1) {
3385                             throw new IllegalArgumentException("Malformed exponential " +
3386                                 "pattern \"" + pattern + ‘"‘);
3387                         }
3388 
3389                         // Transition to phase 2
3390                         phase = 2;
3391                         affix = suffix;
3392                         --pos;
3393                         continue;
3394                     } else {
3395                         phase = 2;
3396                         affix = suffix;
3397                         --pos;
3398                         --phaseOneLength;
3399                         continue;
3400                     }
3401                     break;
3402                 }
3403             }
3404 
3405             // Handle patterns with no ‘0‘ pattern character. These patterns
3406             // are legal, but must be interpreted.  "##.###" -> "#0.###".
3407             // ".###" -> ".0##".
3408             /* We allow patterns of the form "####" to produce a zeroDigitCount
3409              * of zero (got that?); although this seems like it might make it
3410              * possible for format() to produce empty strings, format() checks
3411              * for this condition and outputs a zero digit in this situation.
3412              * Having a zeroDigitCount of zero yields a minimum integer digits
3413              * of zero, which allows proper round-trip patterns.  That is, we
3414              * don‘t want "#" to become "#0" when toPattern() is called (even
3415              * though that‘s what it really is, semantically).
3416              */
3417             if (zeroDigitCount == 0 && digitLeftCount > 0 && decimalPos >= 0) {
3418                 // Handle "###.###" and "###." and ".###"
3419                 int n = decimalPos;
3420                 if (n == 0) { // Handle ".###"
3421                     ++n;
3422                 }
3423                 digitRightCount = digitLeftCount - n;
3424                 digitLeftCount = n - 1;
3425                 zeroDigitCount = 1;
3426             }
3427 
3428             // Do syntax checking on the digits.
3429             if ((decimalPos < 0 && digitRightCount > 0) ||
3430                 (decimalPos >= 0 && (decimalPos < digitLeftCount ||
3431                  decimalPos > (digitLeftCount + zeroDigitCount))) ||
3432                  groupingCount == 0 || inQuote) {
3433                 throw new IllegalArgumentException("Malformed pattern \"" +
3434                     pattern + ‘"‘);
3435             }
3436 
3437             if (j == 1) {
3438                 posPrefixPattern = prefix.toString();
3439                 posSuffixPattern = suffix.toString();
3440                 negPrefixPattern = posPrefixPattern;   // assume these for now
3441                 negSuffixPattern = posSuffixPattern;
3442                 int digitTotalCount = digitLeftCount + zeroDigitCount + digitRightCount;
3443                 /* The effectiveDecimalPos is the position the decimal is at or
3444                  * would be at if there is no decimal. Note that if decimalPos<0,
3445                  * then digitTotalCount == digitLeftCount + zeroDigitCount.
3446                  */
3447                 int effectiveDecimalPos = decimalPos >= 0 ?
3448                     decimalPos : digitTotalCount;
3449                 setMinimumIntegerDigits(effectiveDecimalPos - digitLeftCount);
3450                 setMaximumIntegerDigits(useExponentialNotation ?
3451                     digitLeftCount + getMinimumIntegerDigits() :
3452                     MAXIMUM_INTEGER_DIGITS);
3453                 setMaximumFractionDigits(decimalPos >= 0 ?
3454                     (digitTotalCount - decimalPos) : 0);
3455                 setMinimumFractionDigits(decimalPos >= 0 ?
3456                     (digitLeftCount + zeroDigitCount - decimalPos) : 0);
3457                 setGroupingUsed(groupingCount > 0);
3458                 this.groupingSize = (groupingCount > 0) ? groupingCount : 0;
3459                 this.multiplier = multiplier;
3460                 setDecimalSeparatorAlwaysShown(decimalPos == 0 ||
3461                     decimalPos == digitTotalCount);
3462             } else {
3463                 negPrefixPattern = prefix.toString();
3464                 negSuffixPattern = suffix.toString();
3465                 gotNegative = true;
3466             }
3467         }
3468 
3469         if (pattern.length() == 0) {
3470             posPrefixPattern = posSuffixPattern = "";
3471             setMinimumIntegerDigits(0);
3472             setMaximumIntegerDigits(MAXIMUM_INTEGER_DIGITS);
3473             setMinimumFractionDigits(0);
3474             setMaximumFractionDigits(MAXIMUM_FRACTION_DIGITS);
3475         }
3476 
3477         // If there was no negative pattern, or if the negative pattern is
3478         // identical to the positive pattern, then prepend the minus sign to
3479         // the positive pattern to form the negative pattern.
3480         if (!gotNegative ||
3481             (negPrefixPattern.equals(posPrefixPattern)
3482              && negSuffixPattern.equals(posSuffixPattern))) {
3483             negSuffixPattern = posSuffixPattern;
3484             negPrefixPattern = "‘-" + posPrefixPattern;
3485         }
3486 
3487         expandAffixes();
3488     }
3489 
3490     /**
3491      * Sets the maximum number of digits allowed in the integer portion of a
3492      * number.
3493      * For formatting numbers other than <code>BigInteger</code> and
3494      * <code>BigDecimal</code> objects, the lower of <code>newValue</code> and
3495      * 309 is used. Negative input values are replaced with 0.
3496      * @see NumberFormat#setMaximumIntegerDigits
3497      */
3498     @Override
3499     public void setMaximumIntegerDigits(int newValue) {
3500         maximumIntegerDigits = Math.min(Math.max(0, newValue), MAXIMUM_INTEGER_DIGITS);
3501         super.setMaximumIntegerDigits((maximumIntegerDigits > DOUBLE_INTEGER_DIGITS) ?
3502             DOUBLE_INTEGER_DIGITS : maximumIntegerDigits);
3503         if (minimumIntegerDigits > maximumIntegerDigits) {
3504             minimumIntegerDigits = maximumIntegerDigits;
3505             super.setMinimumIntegerDigits((minimumIntegerDigits > DOUBLE_INTEGER_DIGITS) ?
3506                 DOUBLE_INTEGER_DIGITS : minimumIntegerDigits);
3507         }
3508         fastPathCheckNeeded = true;
3509     }
3510 
3511     /**
3512      * Sets the minimum number of digits allowed in the integer portion of a
3513      * number.
3514      * For formatting numbers other than <code>BigInteger</code> and
3515      * <code>BigDecimal</code> objects, the lower of <code>newValue</code> and
3516      * 309 is used. Negative input values are replaced with 0.
3517      * @see NumberFormat#setMinimumIntegerDigits
3518      */
3519     @Override
3520     public void setMinimumIntegerDigits(int newValue) {
3521         minimumIntegerDigits = Math.min(Math.max(0, newValue), MAXIMUM_INTEGER_DIGITS);
3522         super.setMinimumIntegerDigits((minimumIntegerDigits > DOUBLE_INTEGER_DIGITS) ?
3523             DOUBLE_INTEGER_DIGITS : minimumIntegerDigits);
3524         if (minimumIntegerDigits > maximumIntegerDigits) {
3525             maximumIntegerDigits = minimumIntegerDigits;
3526             super.setMaximumIntegerDigits((maximumIntegerDigits > DOUBLE_INTEGER_DIGITS) ?
3527                 DOUBLE_INTEGER_DIGITS : maximumIntegerDigits);
3528         }
3529         fastPathCheckNeeded = true;
3530     }
3531 
3532     /**
3533      * Sets the maximum number of digits allowed in the fraction portion of a
3534      * number.
3535      * For formatting numbers other than <code>BigInteger</code> and
3536      * <code>BigDecimal</code> objects, the lower of <code>newValue</code> and
3537      * 340 is used. Negative input values are replaced with 0.
3538      * @see NumberFormat#setMaximumFractionDigits
3539      */
3540     @Override
3541     public void setMaximumFractionDigits(int newValue) {
3542         maximumFractionDigits = Math.min(Math.max(0, newValue), MAXIMUM_FRACTION_DIGITS);
3543         super.setMaximumFractionDigits((maximumFractionDigits > DOUBLE_FRACTION_DIGITS) ?
3544             DOUBLE_FRACTION_DIGITS : maximumFractionDigits);
3545         if (minimumFractionDigits > maximumFractionDigits) {
3546             minimumFractionDigits = maximumFractionDigits;
3547             super.setMinimumFractionDigits((minimumFractionDigits > DOUBLE_FRACTION_DIGITS) ?
3548                 DOUBLE_FRACTION_DIGITS : minimumFractionDigits);
3549         }
3550         fastPathCheckNeeded = true;
3551     }
3552 
3553     /**
3554      * Sets the minimum number of digits allowed in the fraction portion of a
3555      * number.
3556      * For formatting numbers other than <code>BigInteger</code> and
3557      * <code>BigDecimal</code> objects, the lower of <code>newValue</code> and
3558      * 340 is used. Negative input values are replaced with 0.
3559      * @see NumberFormat#setMinimumFractionDigits
3560      */
3561     @Override
3562     public void setMinimumFractionDigits(int newValue) {
3563         minimumFractionDigits = Math.min(Math.max(0, newValue), MAXIMUM_FRACTION_DIGITS);
3564         super.setMinimumFractionDigits((minimumFractionDigits > DOUBLE_FRACTION_DIGITS) ?
3565             DOUBLE_FRACTION_DIGITS : minimumFractionDigits);
3566         if (minimumFractionDigits > maximumFractionDigits) {
3567             maximumFractionDigits = minimumFractionDigits;
3568             super.setMaximumFractionDigits((maximumFractionDigits > DOUBLE_FRACTION_DIGITS) ?
3569                 DOUBLE_FRACTION_DIGITS : maximumFractionDigits);
3570         }
3571         fastPathCheckNeeded = true;
3572     }
3573 
3574     /**
3575      * Gets the maximum number of digits allowed in the integer portion of a
3576      * number.
3577      * For formatting numbers other than <code>BigInteger</code> and
3578      * <code>BigDecimal</code> objects, the lower of the return value and
3579      * 309 is used.
3580      * @see #setMaximumIntegerDigits
3581      */
3582     @Override
3583     public int getMaximumIntegerDigits() {
3584         return maximumIntegerDigits;
3585     }
3586 
3587     /**
3588      * Gets the minimum number of digits allowed in the integer portion of a
3589      * number.
3590      * For formatting numbers other than <code>BigInteger</code> and
3591      * <code>BigDecimal</code> objects, the lower of the return value and
3592      * 309 is used.
3593      * @see #setMinimumIntegerDigits
3594      */
3595     @Override
3596     public int getMinimumIntegerDigits() {
3597         return minimumIntegerDigits;
3598     }
3599 
3600     /**
3601      * Gets the maximum number of digits allowed in the fraction portion of a
3602      * number.
3603      * For formatting numbers other than <code>BigInteger</code> and
3604      * <code>BigDecimal</code> objects, the lower of the return value and
3605      * 340 is used.
3606      * @see #setMaximumFractionDigits
3607      */
3608     @Override
3609     public int getMaximumFractionDigits() {
3610         return maximumFractionDigits;
3611     }
3612 
3613     /**
3614      * Gets the minimum number of digits allowed in the fraction portion of a
3615      * number.
3616      * For formatting numbers other than <code>BigInteger</code> and
3617      * <code>BigDecimal</code> objects, the lower of the return value and
3618      * 340 is used.
3619      * @see #setMinimumFractionDigits
3620      */
3621     @Override
3622     public int getMinimumFractionDigits() {
3623         return minimumFractionDigits;
3624     }
3625 
3626     /**
3627      * Gets the currency used by this decimal format when formatting
3628      * currency values.
3629      * The currency is obtained by calling
3630      * {@link DecimalFormatSymbols#getCurrency DecimalFormatSymbols.getCurrency}
3631      * on this number format‘s symbols.
3632      *
3633      * @return the currency used by this decimal format, or <code>null</code>
3634      * @since 1.4
3635      */
3636     @Override
3637     public Currency getCurrency() {
3638         return symbols.getCurrency();
3639     }
3640 
3641     /**
3642      * Sets the currency used by this number format when formatting
3643      * currency values. This does not update the minimum or maximum
3644      * number of fraction digits used by the number format.
3645      * The currency is set by calling
3646      * {@link DecimalFormatSymbols#setCurrency DecimalFormatSymbols.setCurrency}
3647      * on this number format‘s symbols.
3648      *
3649      * @param currency the new currency to be used by this decimal format
3650      * @exception NullPointerException if <code>currency</code> is null
3651      * @since 1.4
3652      */
3653     @Override
3654     public void setCurrency(Currency currency) {
3655         if (currency != symbols.getCurrency()) {
3656             symbols.setCurrency(currency);
3657             if (isCurrencyFormat) {
3658                 expandAffixes();
3659             }
3660         }
3661         fastPathCheckNeeded = true;
3662     }
3663 
3664     /**
3665      * Gets the {@link java.math.RoundingMode} used in this DecimalFormat.
3666      *
3667      * @return The <code>RoundingMode</code> used for this DecimalFormat.
3668      * @see #setRoundingMode(RoundingMode)
3669      * @since 1.6
3670      */
3671     @Override
3672     public RoundingMode getRoundingMode() {
3673         return roundingMode;
3674     }
3675 
3676     /**
3677      * Sets the {@link java.math.RoundingMode} used in this DecimalFormat.
3678      *
3679      * @param roundingMode The <code>RoundingMode</code> to be used
3680      * @see #getRoundingMode()
3681      * @exception NullPointerException if <code>roundingMode</code> is null.
3682      * @since 1.6
3683      */
3684     @Override
3685     public void setRoundingMode(RoundingMode roundingMode) {
3686         if (roundingMode == null) {
3687             throw new NullPointerException();
3688         }
3689 
3690         this.roundingMode = roundingMode;
3691         digitList.setRoundingMode(roundingMode);
3692         fastPathCheckNeeded = true;
3693     }
3694 
3695     /**
3696      * Reads the default serializable fields from the stream and performs
3697      * validations and adjustments for older serialized versions. The
3698      * validations and adjustments are:
3699      * <ol>
3700      * <li>
3701      * Verify that the superclass‘s digit count fields correctly reflect
3702      * the limits imposed on formatting numbers other than
3703      * <code>BigInteger</code> and <code>BigDecimal</code> objects. These
3704      * limits are stored in the superclass for serialization compatibility
3705      * with older versions, while the limits for <code>BigInteger</code> and
3706      * <code>BigDecimal</code> objects are kept in this class.
3707      * If, in the superclass, the minimum or maximum integer digit count is
3708      * larger than <code>DOUBLE_INTEGER_DIGITS</code> or if the minimum or
3709      * maximum fraction digit count is larger than
3710      * <code>DOUBLE_FRACTION_DIGITS</code>, then the stream data is invalid
3711      * and this method throws an <code>InvalidObjectException</code>.
3712      * <li>
3713      * If <code>serialVersionOnStream</code> is less than 4, initialize
3714      * <code>roundingMode</code> to {@link java.math.RoundingMode#HALF_EVEN
3715      * RoundingMode.HALF_EVEN}.  This field is new with version 4.
3716      * <li>
3717      * If <code>serialVersionOnStream</code> is less than 3, then call
3718      * the setters for the minimum and maximum integer and fraction digits with
3719      * the values of the corresponding superclass getters to initialize the
3720      * fields in this class. The fields in this class are new with version 3.
3721      * <li>
3722      * If <code>serialVersionOnStream</code> is less than 1, indicating that
3723      * the stream was written by JDK 1.1, initialize
3724      * <code>useExponentialNotation</code>
3725      * to false, since it was not present in JDK 1.1.
3726      * <li>
3727      * Set <code>serialVersionOnStream</code> to the maximum allowed value so
3728      * that default serialization will work properly if this object is streamed
3729      * out again.
3730      * </ol>
3731      *
3732      * <p>Stream versions older than 2 will not have the affix pattern variables
3733      * <code>posPrefixPattern</code> etc.  As a result, they will be initialized
3734      * to <code>null</code>, which means the affix strings will be taken as
3735      * literal values.  This is exactly what we want, since that corresponds to
3736      * the pre-version-2 behavior.
3737      */
3738     private void readObject(ObjectInputStream stream)
3739          throws IOException, ClassNotFoundException
3740     {
3741         stream.defaultReadObject();
3742         digitList = new DigitList();
3743 
3744         // We force complete fast-path reinitialization when the instance is
3745         // deserialized. See clone() comment on fastPathCheckNeeded.
3746         fastPathCheckNeeded = true;
3747         isFastPath = false;
3748         fastPathData = null;
3749 
3750         if (serialVersionOnStream < 4) {
3751             setRoundingMode(RoundingMode.HALF_EVEN);
3752         } else {
3753             setRoundingMode(getRoundingMode());
3754         }
3755 
3756         // We only need to check the maximum counts because NumberFormat
3757         // .readObject has already ensured that the maximum is greater than the
3758         // minimum count.
3759         if (super.getMaximumIntegerDigits() > DOUBLE_INTEGER_DIGITS ||
3760             super.getMaximumFractionDigits() > DOUBLE_FRACTION_DIGITS) {
3761             throw new InvalidObjectException("Digit count out of range");
3762         }
3763         if (serialVersionOnStream < 3) {
3764             setMaximumIntegerDigits(super.getMaximumIntegerDigits());
3765             setMinimumIntegerDigits(super.getMinimumIntegerDigits());
3766             setMaximumFractionDigits(super.getMaximumFractionDigits());
3767             setMinimumFractionDigits(super.getMinimumFractionDigits());
3768         }
3769         if (serialVersionOnStream < 1) {
3770             // Didn‘t have exponential fields
3771             useExponentialNotation = false;
3772         }
3773         serialVersionOnStream = currentSerialVersion;
3774     }
3775 
3776     //----------------------------------------------------------------------
3777     // INSTANCE VARIABLES
3778     //----------------------------------------------------------------------
3779 
3780     private transient DigitList digitList = new DigitList();
3781 
3782     /**
3783      * The symbol used as a prefix when formatting positive numbers, e.g. "+".
3784      *
3785      * @serial
3786      * @see #getPositivePrefix
3787      */
3788     private String  positivePrefix = "";
3789 
3790     /**
3791      * The symbol used as a suffix when formatting positive numbers.
3792      * This is often an empty string.
3793      *
3794      * @serial
3795      * @see #getPositiveSuffix
3796      */
3797     private String  positiveSuffix = "";
3798 
3799     /**
3800      * The symbol used as a prefix when formatting negative numbers, e.g. "-".
3801      *
3802      * @serial
3803      * @see #getNegativePrefix
3804      */
3805     private String  negativePrefix = "-";
3806 
3807     /**
3808      * The symbol used as a suffix when formatting negative numbers.
3809      * This is often an empty string.
3810      *
3811      * @serial
3812      * @see #getNegativeSuffix
3813      */
3814     private String  negativeSuffix = "";
3815 
3816     /**
3817      * The prefix pattern for non-negative numbers.  This variable corresponds
3818      * to <code>positivePrefix</code>.
3819      *
3820      * <p>This pattern is expanded by the method <code>expandAffix()</code> to
3821      * <code>positivePrefix</code> to update the latter to reflect changes in
3822      * <code>symbols</code>.  If this variable is <code>null</code> then
3823      * <code>positivePrefix</code> is taken as a literal value that does not
3824      * change when <code>symbols</code> changes.  This variable is always
3825      * <code>null</code> for <code>DecimalFormat</code> objects older than
3826      * stream version 2 restored from stream.
3827      *
3828      * @serial
3829      * @since 1.3
3830      */
3831     private String posPrefixPattern;
3832 
3833     /**
3834      * The suffix pattern for non-negative numbers.  This variable corresponds
3835      * to <code>positiveSuffix</code>.  This variable is analogous to
3836      * <code>posPrefixPattern</code>; see that variable for further
3837      * documentation.
3838      *
3839      * @serial
3840      * @since 1.3
3841      */
3842     private String posSuffixPattern;
3843 
3844     /**
3845      * The prefix pattern for negative numbers.  This variable corresponds
3846      * to <code>negativePrefix</code>.  This variable is analogous to
3847      * <code>posPrefixPattern</code>; see that variable for further
3848      * documentation.
3849      *
3850      * @serial
3851      * @since 1.3
3852      */
3853     private String negPrefixPattern;
3854 
3855     /**
3856      * The suffix pattern for negative numbers.  This variable corresponds
3857      * to <code>negativeSuffix</code>.  This variable is analogous to
3858      * <code>posPrefixPattern</code>; see that variable for further
3859      * documentation.
3860      *
3861      * @serial
3862      * @since 1.3
3863      */
3864     private String negSuffixPattern;
3865 
3866     /**
3867      * The multiplier for use in percent, per mille, etc.
3868      *
3869      * @serial
3870      * @see #getMultiplier
3871      */
3872     private int     multiplier = 1;
3873 
3874     /**
3875      * The number of digits between grouping separators in the integer
3876      * portion of a number.  Must be greater than 0 if
3877      * <code>NumberFormat.groupingUsed</code> is true.
3878      *
3879      * @serial
3880      * @see #getGroupingSize
3881      * @see java.text.NumberFormat#isGroupingUsed
3882      */
3883     private byte    groupingSize = 3;  // invariant, > 0 if useThousands
3884 
3885     /**
3886      * If true, forces the decimal separator to always appear in a formatted
3887      * number, even if the fractional part of the number is zero.
3888      *
3889      * @serial
3890      * @see #isDecimalSeparatorAlwaysShown
3891      */
3892     private boolean decimalSeparatorAlwaysShown = false;
3893 
3894     /**
3895      * If true, parse returns BigDecimal wherever possible.
3896      *
3897      * @serial
3898      * @see #isParseBigDecimal
3899      * @since 1.5
3900      */
3901     private boolean parseBigDecimal = false;
3902 
3903 
3904     /**
3905      * True if this object represents a currency format.  This determines
3906      * whether the monetary decimal separator is used instead of the normal one.
3907      */
3908     private transient boolean isCurrencyFormat = false;
3909 
3910     /**
3911      * The <code>DecimalFormatSymbols</code> object used by this format.
3912      * It contains the symbols used to format numbers, e.g. the grouping separator,
3913      * decimal separator, and so on.
3914      *
3915      * @serial
3916      * @see #setDecimalFormatSymbols
3917      * @see java.text.DecimalFormatSymbols
3918      */
3919     private DecimalFormatSymbols symbols = null; // LIU new DecimalFormatSymbols();
3920 
3921     /**
3922      * True to force the use of exponential (i.e. scientific) notation when formatting
3923      * numbers.
3924      *
3925      * @serial
3926      * @since 1.2
3927      */
3928     private boolean useExponentialNotation;  // Newly persistent in the Java 2 platform v.1.2
3929 
3930     /**
3931      * FieldPositions describing the positive prefix String. This is
3932      * lazily created. Use <code>getPositivePrefixFieldPositions</code>
3933      * when needed.
3934      */
3935     private transient FieldPosition[] positivePrefixFieldPositions;
3936 
3937     /**
3938      * FieldPositions describing the positive suffix String. This is
3939      * lazily created. Use <code>getPositiveSuffixFieldPositions</code>
3940      * when needed.
3941      */
3942     private transient FieldPosition[] positiveSuffixFieldPositions;
3943 
3944     /**
3945      * FieldPositions describing the negative prefix String. This is
3946      * lazily created. Use <code>getNegativePrefixFieldPositions</code>
3947      * when needed.
3948      */
3949     private transient FieldPosition[] negativePrefixFieldPositions;
3950 
3951     /**
3952      * FieldPositions describing the negative suffix String. This is
3953      * lazily created. Use <code>getNegativeSuffixFieldPositions</code>
3954      * when needed.
3955      */
3956     private transient FieldPosition[] negativeSuffixFieldPositions;
3957 
3958     /**
3959      * The minimum number of digits used to display the exponent when a number is
3960      * formatted in exponential notation.  This field is ignored if
3961      * <code>useExponentialNotation</code> is not true.
3962      *
3963      * @serial
3964      * @since 1.2
3965      */
3966     private byte    minExponentDigits;       // Newly persistent in the Java 2 platform v.1.2
3967 
3968     /**
3969      * The maximum number of digits allowed in the integer portion of a
3970      * <code>BigInteger</code> or <code>BigDecimal</code> number.
3971      * <code>maximumIntegerDigits</code> must be greater than or equal to
3972      * <code>minimumIntegerDigits</code>.
3973      *
3974      * @serial
3975      * @see #getMaximumIntegerDigits
3976      * @since 1.5
3977      */
3978     private int    maximumIntegerDigits = super.getMaximumIntegerDigits();
3979 
3980     /**
3981      * The minimum number of digits allowed in the integer portion of a
3982      * <code>BigInteger</code> or <code>BigDecimal</code> number.
3983      * <code>minimumIntegerDigits</code> must be less than or equal to
3984      * <code>maximumIntegerDigits</code>.
3985      *
3986      * @serial
3987      * @see #getMinimumIntegerDigits
3988      * @since 1.5
3989      */
3990     private int    minimumIntegerDigits = super.getMinimumIntegerDigits();
3991 
3992     /**
3993      * The maximum number of digits allowed in the fractional portion of a
3994      * <code>BigInteger</code> or <code>BigDecimal</code> number.
3995      * <code>maximumFractionDigits</code> must be greater than or equal to
3996      * <code>minimumFractionDigits</code>.
3997      *
3998      * @serial
3999      * @see #getMaximumFractionDigits
4000      * @since 1.5
4001      */
4002     private int    maximumFractionDigits = super.getMaximumFractionDigits();
4003 
4004     /**
4005      * The minimum number of digits allowed in the fractional portion of a
4006      * <code>BigInteger</code> or <code>BigDecimal</code> number.
4007      * <code>minimumFractionDigits</code> must be less than or equal to
4008      * <code>maximumFractionDigits</code>.
4009      *
4010      * @serial
4011      * @see #getMinimumFractionDigits
4012      * @since 1.5
4013      */
4014     private int    minimumFractionDigits = super.getMinimumFractionDigits();
4015 
4016     /**
4017      * The {@link java.math.RoundingMode} used in this DecimalFormat.
4018      *
4019      * @serial
4020      * @since 1.6
4021      */
4022     private RoundingMode roundingMode = RoundingMode.HALF_EVEN;
4023 
4024     // ------ DecimalFormat fields for fast-path for double algorithm  ------
4025 
4026     /**
4027      * Helper inner utility class for storing the data used in the fast-path
4028      * algorithm. Almost all fields related to fast-path are encapsulated in
4029      * this class.
4030      *
4031      * Any {@code DecimalFormat} instance has a {@code fastPathData}
4032      * reference field that is null unless both the properties of the instance
4033      * are such that the instance is in the "fast-path" state, and a format call
4034      * has been done at least once while in this state.
4035      *
4036      * Almost all fields are related to the "fast-path" state only and don‘t
4037      * change until one of the instance properties is changed.
4038      *
4039      * {@code firstUsedIndex} and {@code lastFreeIndex} are the only
4040      * two fields that are used and modified while inside a call to
4041      * {@code fastDoubleFormat}.
4042      *
4043      */
4044     private static class FastPathData {
4045         // --- Temporary fields used in fast-path, shared by several methods.
4046 
4047         /** The first unused index at the end of the formatted result. */
4048         int lastFreeIndex;
4049 
4050         /** The first used index at the beginning of the formatted result */
4051         int firstUsedIndex;
4052 
4053         // --- State fields related to fast-path status. Changes due to a
4054         //     property change only. Set by checkAndSetFastPathStatus() only.
4055 
4056         /** Difference between locale zero and default zero representation. */
4057         int  zeroDelta;
4058 
4059         /** Locale char for grouping separator. */
4060         char groupingChar;
4061 
4062         /**  Fixed index position of last integral digit of formatted result */
4063         int integralLastIndex;
4064 
4065         /**  Fixed index position of first fractional digit of formatted result */
4066         int fractionalFirstIndex;
4067 
4068         /** Fractional constants depending on decimal|currency state */
4069         double fractionalScaleFactor;
4070         int fractionalMaxIntBound;
4071 
4072 
4073         /** The char array buffer that will contain the formatted result */
4074         char[] fastPathContainer;
4075 
4076         /** Suffixes recorded as char array for efficiency. */
4077         char[] charsPositivePrefix;
4078         char[] charsNegativePrefix;
4079         char[] charsPositiveSuffix;
4080         char[] charsNegativeSuffix;
4081         boolean positiveAffixesRequired = true;
4082         boolean negativeAffixesRequired = true;
4083     }
4084 
4085     /** The format fast-path status of the instance. Logical state. */
4086     private transient boolean isFastPath = false;
4087 
4088     /** Flag stating need of check and reinit fast-path status on next format call. */
4089     private transient boolean fastPathCheckNeeded = true;
4090 
4091     /** DecimalFormat reference to its FastPathData */
4092     private transient FastPathData fastPathData;
4093 
4094 
4095     //----------------------------------------------------------------------
4096 
4097     static final int currentSerialVersion = 4;
4098 
4099     /**
4100      * The internal serial version which says which version was written.
4101      * Possible values are:
4102      * <ul>
4103      * <li><b>0</b> (default): versions before the Java 2 platform v1.2
4104      * <li><b>1</b>: version for 1.2, which includes the two new fields
4105      *      <code>useExponentialNotation</code> and
4106      *      <code>minExponentDigits</code>.
4107      * <li><b>2</b>: version for 1.3 and later, which adds four new fields:
4108      *      <code>posPrefixPattern</code>, <code>posSuffixPattern</code>,
4109      *      <code>negPrefixPattern</code>, and <code>negSuffixPattern</code>.
4110      * <li><b>3</b>: version for 1.5 and later, which adds five new fields:
4111      *      <code>maximumIntegerDigits</code>,
4112      *      <code>minimumIntegerDigits</code>,
4113      *      <code>maximumFractionDigits</code>,
4114      *      <code>minimumFractionDigits</code>, and
4115      *      <code>parseBigDecimal</code>.
4116      * <li><b>4</b>: version for 1.6 and later, which adds one new field:
4117      *      <code>roundingMode</code>.
4118      * </ul>
4119      * @since 1.2
4120      * @serial
4121      */
4122     private int serialVersionOnStream = currentSerialVersion;
4123 
4124     //----------------------------------------------------------------------
4125     // CONSTANTS
4126     //----------------------------------------------------------------------
4127 
4128     // ------ Fast-Path for double Constants ------
4129 
4130     /** Maximum valid integer value for applying fast-path algorithm */
4131     private static final double MAX_INT_AS_DOUBLE = (double) Integer.MAX_VALUE;
4132 
4133     /**
4134      * The digit arrays used in the fast-path methods for collecting digits.
4135      * Using 3 constants arrays of chars ensures a very fast collection of digits
4136      */
4137     private static class DigitArrays {
4138         static final char[] DigitOnes1000 = new char[1000];
4139         static final char[] DigitTens1000 = new char[1000];
4140         static final char[] DigitHundreds1000 = new char[1000];
4141 
4142         // initialize on demand holder class idiom for arrays of digits
4143         static {
4144             int tenIndex = 0;
4145             int hundredIndex = 0;
4146             char digitOne = ‘0‘;
4147             char digitTen = ‘0‘;
4148             char digitHundred = ‘0‘;
4149             for (int i = 0;  i < 1000; i++ ) {
4150 
4151                 DigitOnes1000[i] = digitOne;
4152                 if (digitOne == ‘9‘)
4153                     digitOne = ‘0‘;
4154                 else
4155                     digitOne++;
4156 
4157                 DigitTens1000[i] = digitTen;
4158                 if (i == (tenIndex + 9)) {
4159                     tenIndex += 10;
4160                     if (digitTen == ‘9‘)
4161                         digitTen = ‘0‘;
4162                     else
4163                         digitTen++;
4164                 }
4165 
4166                 DigitHundreds1000[i] = digitHundred;
4167                 if (i == (hundredIndex + 99)) {
4168                     digitHundred++;
4169                     hundredIndex += 100;
4170                 }
4171             }
4172         }
4173     }
4174     // ------ Fast-Path for double Constants end ------
4175 
4176     // Constants for characters used in programmatic (unlocalized) patterns.
4177     private static final char       PATTERN_ZERO_DIGIT         = ‘0‘;
4178     private static final char       PATTERN_GROUPING_SEPARATOR = ‘,‘;
4179     private static final char       PATTERN_DECIMAL_SEPARATOR  = ‘.‘;
4180     private static final char       PATTERN_PER_MILLE          = ‘\u2030‘;
4181     private static final char       PATTERN_PERCENT            = ‘%‘;
4182     private static final char       PATTERN_DIGIT              = ‘#‘;
4183     private static final char       PATTERN_SEPARATOR          = ‘;‘;
4184     private static final String     PATTERN_EXPONENT           = "E";
4185     private static final char       PATTERN_MINUS              = ‘-‘;
4186 
4187     /**
4188      * The CURRENCY_SIGN is the standard Unicode symbol for currency.  It
4189      * is used in patterns and substituted with either the currency symbol,
4190      * or if it is doubled, with the international currency symbol.  If the
4191      * CURRENCY_SIGN is seen in a pattern, then the decimal separator is
4192      * replaced with the monetary decimal separator.
4193      *
4194      * The CURRENCY_SIGN is not localized.
4195      */
4196     private static final char       CURRENCY_SIGN = ‘\u00A4‘;
4197 
4198     private static final char       QUOTE = ‘\‘‘;
4199 
4200     private static FieldPosition[] EmptyFieldPositionArray = new FieldPosition[0];
4201 
4202     // Upper limit on integer and fraction digits for a Java double
4203     static final int DOUBLE_INTEGER_DIGITS  = 309;
4204     static final int DOUBLE_FRACTION_DIGITS = 340;
4205 
4206     // Upper limit on integer and fraction digits for BigDecimal and BigInteger
4207     static final int MAXIMUM_INTEGER_DIGITS  = Integer.MAX_VALUE;
4208     static final int MAXIMUM_FRACTION_DIGITS = Integer.MAX_VALUE;
4209 
4210     // Proclaim JDK 1.1 serial compatibility.
4211     static final long serialVersionUID = 864413376551465018L;
4212 }

 

DecimalFormat详解

标签:iterator   tde   elf   parse   next   ctf   第一个   har   item   

原文地址:https://www.cnblogs.com/XT-xutao/p/9882404.html

(0)
(0)
   
举报
评论 一句话评论(0
登录后才能评论!
© 2014 mamicode.com 版权所有  联系我们:gaon5@hotmail.com
迷上了代码!