blob: 27bd86784895f83479fb43cfeef805e0e34bddcf [file] [log] [blame]
Victor Chang73229502020-09-17 13:39:19 +01001// © 2018 and later: Unicode, Inc. and others.
2// License & terms of use: http://www.unicode.org/copyright.html
3//
4// From the double-conversion library. Original license:
5//
6// Copyright 2012 the V8 project authors. All rights reserved.
7// Redistribution and use in source and binary forms, with or without
8// modification, are permitted provided that the following conditions are
9// met:
10//
11// * Redistributions of source code must retain the above copyright
12// notice, this list of conditions and the following disclaimer.
13// * Redistributions in binary form must reproduce the above
14// copyright notice, this list of conditions and the following
15// disclaimer in the documentation and/or other materials provided
16// with the distribution.
17// * Neither the name of Google Inc. nor the names of its
18// contributors may be used to endorse or promote products derived
19// from this software without specific prior written permission.
20//
21// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
24// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
25// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
26// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
27// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
28// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
29// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
30// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
31// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
32
33// ICU PATCH: ifdef around UCONFIG_NO_FORMATTING
34#include "unicode/utypes.h"
35#if !UCONFIG_NO_FORMATTING
36
37#ifndef DOUBLE_CONVERSION_DOUBLE_TO_STRING_H_
38#define DOUBLE_CONVERSION_DOUBLE_TO_STRING_H_
39
40// ICU PATCH: Customize header file paths for ICU.
41
42#include "double-conversion-utils.h"
43
44// ICU PATCH: Wrap in ICU namespace
45U_NAMESPACE_BEGIN
46
47namespace double_conversion {
48
49class DoubleToStringConverter {
50 public:
51#if 0 // not needed for ICU
52 // When calling ToFixed with a double > 10^kMaxFixedDigitsBeforePoint
53 // or a requested_digits parameter > kMaxFixedDigitsAfterPoint then the
54 // function returns false.
55 static const int kMaxFixedDigitsBeforePoint = 60;
56 static const int kMaxFixedDigitsAfterPoint = 60;
57
58 // When calling ToExponential with a requested_digits
59 // parameter > kMaxExponentialDigits then the function returns false.
60 static const int kMaxExponentialDigits = 120;
61
62 // When calling ToPrecision with a requested_digits
63 // parameter < kMinPrecisionDigits or requested_digits > kMaxPrecisionDigits
64 // then the function returns false.
65 static const int kMinPrecisionDigits = 1;
66 static const int kMaxPrecisionDigits = 120;
67
68 enum Flags {
69 NO_FLAGS = 0,
70 EMIT_POSITIVE_EXPONENT_SIGN = 1,
71 EMIT_TRAILING_DECIMAL_POINT = 2,
72 EMIT_TRAILING_ZERO_AFTER_POINT = 4,
73 UNIQUE_ZERO = 8
74 };
75
76 // Flags should be a bit-or combination of the possible Flags-enum.
77 // - NO_FLAGS: no special flags.
78 // - EMIT_POSITIVE_EXPONENT_SIGN: when the number is converted into exponent
79 // form, emits a '+' for positive exponents. Example: 1.2e+2.
80 // - EMIT_TRAILING_DECIMAL_POINT: when the input number is an integer and is
81 // converted into decimal format then a trailing decimal point is appended.
82 // Example: 2345.0 is converted to "2345.".
83 // - EMIT_TRAILING_ZERO_AFTER_POINT: in addition to a trailing decimal point
84 // emits a trailing '0'-character. This flag requires the
85 // EXMIT_TRAILING_DECIMAL_POINT flag.
86 // Example: 2345.0 is converted to "2345.0".
87 // - UNIQUE_ZERO: "-0.0" is converted to "0.0".
88 //
89 // Infinity symbol and nan_symbol provide the string representation for these
90 // special values. If the string is NULL and the special value is encountered
91 // then the conversion functions return false.
92 //
93 // The exponent_character is used in exponential representations. It is
94 // usually 'e' or 'E'.
95 //
96 // When converting to the shortest representation the converter will
97 // represent input numbers in decimal format if they are in the interval
98 // [10^decimal_in_shortest_low; 10^decimal_in_shortest_high[
99 // (lower boundary included, greater boundary excluded).
100 // Example: with decimal_in_shortest_low = -6 and
101 // decimal_in_shortest_high = 21:
102 // ToShortest(0.000001) -> "0.000001"
103 // ToShortest(0.0000001) -> "1e-7"
104 // ToShortest(111111111111111111111.0) -> "111111111111111110000"
105 // ToShortest(100000000000000000000.0) -> "100000000000000000000"
106 // ToShortest(1111111111111111111111.0) -> "1.1111111111111111e+21"
107 //
108 // When converting to precision mode the converter may add
109 // max_leading_padding_zeroes before returning the number in exponential
110 // format.
111 // Example with max_leading_padding_zeroes_in_precision_mode = 6.
112 // ToPrecision(0.0000012345, 2) -> "0.0000012"
113 // ToPrecision(0.00000012345, 2) -> "1.2e-7"
114 // Similarily the converter may add up to
115 // max_trailing_padding_zeroes_in_precision_mode in precision mode to avoid
116 // returning an exponential representation. A zero added by the
117 // EMIT_TRAILING_ZERO_AFTER_POINT flag is counted for this limit.
118 // Examples for max_trailing_padding_zeroes_in_precision_mode = 1:
119 // ToPrecision(230.0, 2) -> "230"
120 // ToPrecision(230.0, 2) -> "230." with EMIT_TRAILING_DECIMAL_POINT.
121 // ToPrecision(230.0, 2) -> "2.3e2" with EMIT_TRAILING_ZERO_AFTER_POINT.
122 //
123 // The min_exponent_width is used for exponential representations.
124 // The converter adds leading '0's to the exponent until the exponent
125 // is at least min_exponent_width digits long.
126 // The min_exponent_width is clamped to 5.
127 // As such, the exponent may never have more than 5 digits in total.
128 DoubleToStringConverter(int flags,
129 const char* infinity_symbol,
130 const char* nan_symbol,
131 char exponent_character,
132 int decimal_in_shortest_low,
133 int decimal_in_shortest_high,
134 int max_leading_padding_zeroes_in_precision_mode,
135 int max_trailing_padding_zeroes_in_precision_mode,
136 int min_exponent_width = 0)
137 : flags_(flags),
138 infinity_symbol_(infinity_symbol),
139 nan_symbol_(nan_symbol),
140 exponent_character_(exponent_character),
141 decimal_in_shortest_low_(decimal_in_shortest_low),
142 decimal_in_shortest_high_(decimal_in_shortest_high),
143 max_leading_padding_zeroes_in_precision_mode_(
144 max_leading_padding_zeroes_in_precision_mode),
145 max_trailing_padding_zeroes_in_precision_mode_(
146 max_trailing_padding_zeroes_in_precision_mode),
147 min_exponent_width_(min_exponent_width) {
148 // When 'trailing zero after the point' is set, then 'trailing point'
149 // must be set too.
150 DOUBLE_CONVERSION_ASSERT(((flags & EMIT_TRAILING_DECIMAL_POINT) != 0) ||
151 !((flags & EMIT_TRAILING_ZERO_AFTER_POINT) != 0));
152 }
153
154 // Returns a converter following the EcmaScript specification.
155 static const DoubleToStringConverter& EcmaScriptConverter();
156
157 // Computes the shortest string of digits that correctly represent the input
158 // number. Depending on decimal_in_shortest_low and decimal_in_shortest_high
159 // (see constructor) it then either returns a decimal representation, or an
160 // exponential representation.
161 // Example with decimal_in_shortest_low = -6,
162 // decimal_in_shortest_high = 21,
163 // EMIT_POSITIVE_EXPONENT_SIGN activated, and
164 // EMIT_TRAILING_DECIMAL_POINT deactived:
165 // ToShortest(0.000001) -> "0.000001"
166 // ToShortest(0.0000001) -> "1e-7"
167 // ToShortest(111111111111111111111.0) -> "111111111111111110000"
168 // ToShortest(100000000000000000000.0) -> "100000000000000000000"
169 // ToShortest(1111111111111111111111.0) -> "1.1111111111111111e+21"
170 //
171 // Note: the conversion may round the output if the returned string
172 // is accurate enough to uniquely identify the input-number.
173 // For example the most precise representation of the double 9e59 equals
174 // "899999999999999918767229449717619953810131273674690656206848", but
175 // the converter will return the shorter (but still correct) "9e59".
176 //
177 // Returns true if the conversion succeeds. The conversion always succeeds
178 // except when the input value is special and no infinity_symbol or
179 // nan_symbol has been given to the constructor.
180 bool ToShortest(double value, StringBuilder* result_builder) const {
181 return ToShortestIeeeNumber(value, result_builder, SHORTEST);
182 }
183
184 // Same as ToShortest, but for single-precision floats.
185 bool ToShortestSingle(float value, StringBuilder* result_builder) const {
186 return ToShortestIeeeNumber(value, result_builder, SHORTEST_SINGLE);
187 }
188
189
190 // Computes a decimal representation with a fixed number of digits after the
191 // decimal point. The last emitted digit is rounded.
192 //
193 // Examples:
194 // ToFixed(3.12, 1) -> "3.1"
195 // ToFixed(3.1415, 3) -> "3.142"
196 // ToFixed(1234.56789, 4) -> "1234.5679"
197 // ToFixed(1.23, 5) -> "1.23000"
198 // ToFixed(0.1, 4) -> "0.1000"
199 // ToFixed(1e30, 2) -> "1000000000000000019884624838656.00"
200 // ToFixed(0.1, 30) -> "0.100000000000000005551115123126"
201 // ToFixed(0.1, 17) -> "0.10000000000000001"
202 //
203 // If requested_digits equals 0, then the tail of the result depends on
204 // the EMIT_TRAILING_DECIMAL_POINT and EMIT_TRAILING_ZERO_AFTER_POINT.
205 // Examples, for requested_digits == 0,
206 // let EMIT_TRAILING_DECIMAL_POINT and EMIT_TRAILING_ZERO_AFTER_POINT be
207 // - false and false: then 123.45 -> 123
208 // 0.678 -> 1
209 // - true and false: then 123.45 -> 123.
210 // 0.678 -> 1.
211 // - true and true: then 123.45 -> 123.0
212 // 0.678 -> 1.0
213 //
214 // Returns true if the conversion succeeds. The conversion always succeeds
215 // except for the following cases:
216 // - the input value is special and no infinity_symbol or nan_symbol has
217 // been provided to the constructor,
218 // - 'value' > 10^kMaxFixedDigitsBeforePoint, or
219 // - 'requested_digits' > kMaxFixedDigitsAfterPoint.
220 // The last two conditions imply that the result will never contain more than
221 // 1 + kMaxFixedDigitsBeforePoint + 1 + kMaxFixedDigitsAfterPoint characters
222 // (one additional character for the sign, and one for the decimal point).
223 bool ToFixed(double value,
224 int requested_digits,
225 StringBuilder* result_builder) const;
226
227 // Computes a representation in exponential format with requested_digits
228 // after the decimal point. The last emitted digit is rounded.
229 // If requested_digits equals -1, then the shortest exponential representation
230 // is computed.
231 //
232 // Examples with EMIT_POSITIVE_EXPONENT_SIGN deactivated, and
233 // exponent_character set to 'e'.
234 // ToExponential(3.12, 1) -> "3.1e0"
235 // ToExponential(5.0, 3) -> "5.000e0"
236 // ToExponential(0.001, 2) -> "1.00e-3"
237 // ToExponential(3.1415, -1) -> "3.1415e0"
238 // ToExponential(3.1415, 4) -> "3.1415e0"
239 // ToExponential(3.1415, 3) -> "3.142e0"
240 // ToExponential(123456789000000, 3) -> "1.235e14"
241 // ToExponential(1000000000000000019884624838656.0, -1) -> "1e30"
242 // ToExponential(1000000000000000019884624838656.0, 32) ->
243 // "1.00000000000000001988462483865600e30"
244 // ToExponential(1234, 0) -> "1e3"
245 //
246 // Returns true if the conversion succeeds. The conversion always succeeds
247 // except for the following cases:
248 // - the input value is special and no infinity_symbol or nan_symbol has
249 // been provided to the constructor,
250 // - 'requested_digits' > kMaxExponentialDigits.
251 // The last condition implies that the result will never contain more than
252 // kMaxExponentialDigits + 8 characters (the sign, the digit before the
253 // decimal point, the decimal point, the exponent character, the
254 // exponent's sign, and at most 3 exponent digits).
255 bool ToExponential(double value,
256 int requested_digits,
257 StringBuilder* result_builder) const;
258
259 // Computes 'precision' leading digits of the given 'value' and returns them
260 // either in exponential or decimal format, depending on
261 // max_{leading|trailing}_padding_zeroes_in_precision_mode (given to the
262 // constructor).
263 // The last computed digit is rounded.
264 //
265 // Example with max_leading_padding_zeroes_in_precision_mode = 6.
266 // ToPrecision(0.0000012345, 2) -> "0.0000012"
267 // ToPrecision(0.00000012345, 2) -> "1.2e-7"
268 // Similarily the converter may add up to
269 // max_trailing_padding_zeroes_in_precision_mode in precision mode to avoid
270 // returning an exponential representation. A zero added by the
271 // EMIT_TRAILING_ZERO_AFTER_POINT flag is counted for this limit.
272 // Examples for max_trailing_padding_zeroes_in_precision_mode = 1:
273 // ToPrecision(230.0, 2) -> "230"
274 // ToPrecision(230.0, 2) -> "230." with EMIT_TRAILING_DECIMAL_POINT.
275 // ToPrecision(230.0, 2) -> "2.3e2" with EMIT_TRAILING_ZERO_AFTER_POINT.
276 // Examples for max_trailing_padding_zeroes_in_precision_mode = 3, and no
277 // EMIT_TRAILING_ZERO_AFTER_POINT:
278 // ToPrecision(123450.0, 6) -> "123450"
279 // ToPrecision(123450.0, 5) -> "123450"
280 // ToPrecision(123450.0, 4) -> "123500"
281 // ToPrecision(123450.0, 3) -> "123000"
282 // ToPrecision(123450.0, 2) -> "1.2e5"
283 //
284 // Returns true if the conversion succeeds. The conversion always succeeds
285 // except for the following cases:
286 // - the input value is special and no infinity_symbol or nan_symbol has
287 // been provided to the constructor,
288 // - precision < kMinPericisionDigits
289 // - precision > kMaxPrecisionDigits
290 // The last condition implies that the result will never contain more than
291 // kMaxPrecisionDigits + 7 characters (the sign, the decimal point, the
292 // exponent character, the exponent's sign, and at most 3 exponent digits).
293 bool ToPrecision(double value,
294 int precision,
295 StringBuilder* result_builder) const;
296#endif // not needed for ICU
297
298 enum DtoaMode {
299 // Produce the shortest correct representation.
300 // For example the output of 0.299999999999999988897 is (the less accurate
301 // but correct) 0.3.
302 SHORTEST,
303 // Same as SHORTEST, but for single-precision floats.
304 SHORTEST_SINGLE,
305 // Produce a fixed number of digits after the decimal point.
306 // For instance fixed(0.1, 4) becomes 0.1000
307 // If the input number is big, the output will be big.
308 FIXED,
309 // Fixed number of digits (independent of the decimal point).
310 PRECISION
311 };
312
313 // The maximal number of digits that are needed to emit a double in base 10.
314 // A higher precision can be achieved by using more digits, but the shortest
315 // accurate representation of any double will never use more digits than
316 // kBase10MaximalLength.
317 // Note that DoubleToAscii null-terminates its input. So the given buffer
318 // should be at least kBase10MaximalLength + 1 characters long.
319 static const int kBase10MaximalLength = 17;
320
321 // Converts the given double 'v' to digit characters. 'v' must not be NaN,
322 // +Infinity, or -Infinity. In SHORTEST_SINGLE-mode this restriction also
323 // applies to 'v' after it has been casted to a single-precision float. That
324 // is, in this mode static_cast<float>(v) must not be NaN, +Infinity or
325 // -Infinity.
326 //
327 // The result should be interpreted as buffer * 10^(point-length).
328 //
329 // The digits are written to the buffer in the platform's charset, which is
330 // often UTF-8 (with ASCII-range digits) but may be another charset, such
331 // as EBCDIC.
332 //
333 // The output depends on the given mode:
334 // - SHORTEST: produce the least amount of digits for which the internal
335 // identity requirement is still satisfied. If the digits are printed
336 // (together with the correct exponent) then reading this number will give
337 // 'v' again. The buffer will choose the representation that is closest to
338 // 'v'. If there are two at the same distance, than the one farther away
339 // from 0 is chosen (halfway cases - ending with 5 - are rounded up).
340 // In this mode the 'requested_digits' parameter is ignored.
341 // - SHORTEST_SINGLE: same as SHORTEST but with single-precision.
342 // - FIXED: produces digits necessary to print a given number with
343 // 'requested_digits' digits after the decimal point. The produced digits
344 // might be too short in which case the caller has to fill the remainder
345 // with '0's.
346 // Example: toFixed(0.001, 5) is allowed to return buffer="1", point=-2.
347 // Halfway cases are rounded towards +/-Infinity (away from 0). The call
348 // toFixed(0.15, 2) thus returns buffer="2", point=0.
349 // The returned buffer may contain digits that would be truncated from the
350 // shortest representation of the input.
351 // - PRECISION: produces 'requested_digits' where the first digit is not '0'.
352 // Even though the length of produced digits usually equals
353 // 'requested_digits', the function is allowed to return fewer digits, in
354 // which case the caller has to fill the missing digits with '0's.
355 // Halfway cases are again rounded away from 0.
356 // DoubleToAscii expects the given buffer to be big enough to hold all
357 // digits and a terminating null-character. In SHORTEST-mode it expects a
358 // buffer of at least kBase10MaximalLength + 1. In all other modes the
359 // requested_digits parameter and the padding-zeroes limit the size of the
360 // output. Don't forget the decimal point, the exponent character and the
361 // terminating null-character when computing the maximal output size.
362 // The given length is only used in debug mode to ensure the buffer is big
363 // enough.
364 // ICU PATCH: Export this as U_I18N_API for unit tests.
365 static void U_I18N_API DoubleToAscii(double v,
366 DtoaMode mode,
367 int requested_digits,
368 char* buffer,
369 int buffer_length,
370 bool* sign,
371 int* length,
372 int* point);
373
374#if 0 // not needed for ICU
375 private:
376 // Implementation for ToShortest and ToShortestSingle.
377 bool ToShortestIeeeNumber(double value,
378 StringBuilder* result_builder,
379 DtoaMode mode) const;
380
381 // If the value is a special value (NaN or Infinity) constructs the
382 // corresponding string using the configured infinity/nan-symbol.
383 // If either of them is NULL or the value is not special then the
384 // function returns false.
385 bool HandleSpecialValues(double value, StringBuilder* result_builder) const;
386 // Constructs an exponential representation (i.e. 1.234e56).
387 // The given exponent assumes a decimal point after the first decimal digit.
388 void CreateExponentialRepresentation(const char* decimal_digits,
389 int length,
390 int exponent,
391 StringBuilder* result_builder) const;
392 // Creates a decimal representation (i.e 1234.5678).
393 void CreateDecimalRepresentation(const char* decimal_digits,
394 int length,
395 int decimal_point,
396 int digits_after_point,
397 StringBuilder* result_builder) const;
398
399 const int flags_;
400 const char* const infinity_symbol_;
401 const char* const nan_symbol_;
402 const char exponent_character_;
403 const int decimal_in_shortest_low_;
404 const int decimal_in_shortest_high_;
405 const int max_leading_padding_zeroes_in_precision_mode_;
406 const int max_trailing_padding_zeroes_in_precision_mode_;
407 const int min_exponent_width_;
408#endif // not needed for ICU
409
410 DOUBLE_CONVERSION_DISALLOW_IMPLICIT_CONSTRUCTORS(DoubleToStringConverter);
411};
412
413} // namespace double_conversion
414
415// ICU PATCH: Close ICU namespace
416U_NAMESPACE_END
417
418#endif // DOUBLE_CONVERSION_DOUBLE_TO_STRING_H_
419#endif // ICU PATCH: close #if !UCONFIG_NO_FORMATTING