cpython/Objects/unicodectype.c

427 lines
8.4 KiB
C
Raw Normal View History

/*
Unicode character type helpers.
Written by Marc-Andre Lemburg (mal@lemburg.com).
Modified for Python 2.0 by Fredrik Lundh (fredrik@pythonware.com)
Copyright (c) Corporation for National Research Initiatives.
*/
#include "Python.h"
#include "unicodeobject.h"
#define ALPHA_MASK 0x01
#define DECIMAL_MASK 0x02
#define DIGIT_MASK 0x04
#define LOWER_MASK 0x08
#define LINEBREAK_MASK 0x10
#define SPACE_MASK 0x20
#define TITLE_MASK 0x40
#define UPPER_MASK 0x80
typedef struct {
const unsigned short flags;
const Py_UNICODE upper;
const Py_UNICODE lower;
const Py_UNICODE title;
const unsigned char decimal;
const unsigned char digit;
} _PyUnicode_TypeRecord;
#include "unicodetype_db.h"
static const _PyUnicode_TypeRecord *
gettyperecord(Py_UNICODE code)
{
int index;
#ifdef Py_UNICODE_WIDE
2002-10-18 13:11:54 -03:00
if (code >= 0x110000)
index = 0;
else
#endif
{
index = index1[(code>>SHIFT)];
index = index2[(index<<SHIFT)+(code&((1<<SHIFT)-1))];
}
return &_PyUnicode_TypeRecords[index];
}
/* Returns 1 for Unicode characters having the category 'Zl' or type
'B', 0 otherwise. */
int _PyUnicode_IsLinebreak(Py_UNICODE ch)
{
const _PyUnicode_TypeRecord *ctype = gettyperecord(ch);
return (ctype->flags & LINEBREAK_MASK) != 0;
}
/* Returns the titlecase Unicode characters corresponding to ch or just
ch if no titlecase mapping is known. */
Py_UNICODE _PyUnicode_ToTitlecase(register Py_UNICODE ch)
{
const _PyUnicode_TypeRecord *ctype = gettyperecord(ch);
int delta;
if (ctype->title)
delta = ctype->title;
else
delta = ctype->upper;
if (delta >= 32768)
delta -= 65536;
return ch + delta;
}
/* Returns 1 for Unicode characters having the category 'Lt', 0
otherwise. */
int _PyUnicode_IsTitlecase(Py_UNICODE ch)
{
const _PyUnicode_TypeRecord *ctype = gettyperecord(ch);
return (ctype->flags & TITLE_MASK) != 0;
}
/* Returns the integer decimal (0-9) for Unicode characters having
this property, -1 otherwise. */
int _PyUnicode_ToDecimalDigit(Py_UNICODE ch)
{
const _PyUnicode_TypeRecord *ctype = gettyperecord(ch);
return (ctype->flags & DECIMAL_MASK) ? ctype->decimal : -1;
}
int _PyUnicode_IsDecimalDigit(Py_UNICODE ch)
{
if (_PyUnicode_ToDecimalDigit(ch) < 0)
return 0;
return 1;
}
/* Returns the integer digit (0-9) for Unicode characters having
this property, -1 otherwise. */
int _PyUnicode_ToDigit(Py_UNICODE ch)
{
const _PyUnicode_TypeRecord *ctype = gettyperecord(ch);
return (ctype->flags & DIGIT_MASK) ? ctype->digit : -1;
}
int _PyUnicode_IsDigit(Py_UNICODE ch)
{
if (_PyUnicode_ToDigit(ch) < 0)
return 0;
return 1;
}
/* Returns the numeric value as double for Unicode characters having
this property, -1.0 otherwise. */
/* TODO: replace with unicodetype_db.h table */
double _PyUnicode_ToNumeric(Py_UNICODE ch)
{
switch (ch) {
case 0x3007:
return (double) 0;
case 0x09F4:
case 0x215F:
case 0x2160:
case 0x2170:
case 0x3021:
case 0x3280:
return (double) 1;
case 0x00BD:
return (double) 1 / 2;
case 0x2153:
return (double) 1 / 3;
case 0x00BC:
return (double) 1 / 4;
case 0x2155:
return (double) 1 / 5;
case 0x2159:
return (double) 1 / 6;
case 0x215B:
return (double) 1 / 8;
case 0x0BF0:
case 0x1372:
case 0x2169:
case 0x2179:
case 0x2469:
case 0x247D:
case 0x2491:
case 0x277F:
case 0x2789:
case 0x2793:
case 0x3038:
case 0x3289:
return (double) 10;
case 0x0BF1:
case 0x137B:
case 0x216D:
case 0x217D:
return (double) 100;
case 0x0BF2:
case 0x216F:
case 0x217F:
case 0x2180:
return (double) 1000;
case 0x137C:
case 0x2182:
return (double) 10000;
case 0x216A:
case 0x217A:
case 0x246A:
case 0x247E:
case 0x2492:
return (double) 11;
case 0x216B:
case 0x217B:
case 0x246B:
case 0x247F:
case 0x2493:
return (double) 12;
case 0x246C:
case 0x2480:
case 0x2494:
return (double) 13;
case 0x246D:
case 0x2481:
case 0x2495:
return (double) 14;
case 0x246E:
case 0x2482:
case 0x2496:
return (double) 15;
case 0x09F9:
case 0x246F:
case 0x2483:
case 0x2497:
return (double) 16;
case 0x16EE:
case 0x2470:
case 0x2484:
case 0x2498:
return (double) 17;
case 0x16EF:
case 0x2471:
case 0x2485:
case 0x2499:
return (double) 18;
case 0x16F0:
case 0x2472:
case 0x2486:
case 0x249A:
return (double) 19;
case 0x09F5:
case 0x2161:
case 0x2171:
case 0x3022:
case 0x3281:
return (double) 2;
case 0x2154:
return (double) 2 / 3;
case 0x2156:
return (double) 2 / 5;
case 0x1373:
case 0x2473:
case 0x2487:
case 0x249B:
case 0x3039:
return (double) 20;
case 0x09F6:
case 0x2162:
case 0x2172:
case 0x3023:
case 0x3282:
return (double) 3;
case 0x00BE:
return (double) 3 / 4;
case 0x2157:
return (double) 3 / 5;
case 0x215C:
return (double) 3 / 8;
case 0x1374:
case 0x303A:
return (double) 30;
case 0x09F7:
case 0x2163:
case 0x2173:
case 0x3024:
case 0x3283:
return (double) 4;
case 0x2158:
return (double) 4 / 5;
case 0x1375:
return (double) 40;
case 0x2164:
case 0x2174:
case 0x3025:
case 0x3284:
return (double) 5;
case 0x215A:
return (double) 5 / 6;
case 0x215D:
return (double) 5 / 8;
case 0x1376:
case 0x216C:
case 0x217C:
return (double) 50;
case 0x216E:
case 0x217E:
return (double) 500;
case 0x2181:
return (double) 5000;
case 0x2165:
case 0x2175:
case 0x3026:
case 0x3285:
return (double) 6;
case 0x1377:
return (double) 60;
case 0x2166:
case 0x2176:
case 0x3027:
case 0x3286:
return (double) 7;
case 0x215E:
return (double) 7 / 8;
case 0x1378:
return (double) 70;
case 0x2167:
case 0x2177:
case 0x3028:
case 0x3287:
return (double) 8;
case 0x1379:
return (double) 80;
case 0x2168:
case 0x2178:
case 0x3029:
case 0x3288:
return (double) 9;
case 0x137A:
return (double) 90;
default:
return (double) _PyUnicode_ToDigit(ch);
}
}
int _PyUnicode_IsNumeric(Py_UNICODE ch)
{
if (_PyUnicode_ToNumeric(ch) < 0.0)
return 0;
return 1;
}
#ifndef WANT_WCTYPE_FUNCTIONS
/* Returns 1 for Unicode characters having the bidirectional type
'WS', 'B' or 'S' or the category 'Zs', 0 otherwise. */
int _PyUnicode_IsWhitespace(Py_UNICODE ch)
{
const _PyUnicode_TypeRecord *ctype = gettyperecord(ch);
return (ctype->flags & SPACE_MASK) != 0;
}
/* Returns 1 for Unicode characters having the category 'Ll', 0
otherwise. */
int _PyUnicode_IsLowercase(Py_UNICODE ch)
{
const _PyUnicode_TypeRecord *ctype = gettyperecord(ch);
return (ctype->flags & LOWER_MASK) != 0;
}
/* Returns 1 for Unicode characters having the category 'Lu', 0
otherwise. */
int _PyUnicode_IsUppercase(Py_UNICODE ch)
{
const _PyUnicode_TypeRecord *ctype = gettyperecord(ch);
return (ctype->flags & UPPER_MASK) != 0;
}
/* Returns the uppercase Unicode characters corresponding to ch or just
ch if no uppercase mapping is known. */
Py_UNICODE _PyUnicode_ToUppercase(Py_UNICODE ch)
{
const _PyUnicode_TypeRecord *ctype = gettyperecord(ch);
int delta = ctype->upper;
if (delta >= 32768)
delta -= 65536;
return ch + delta;
}
/* Returns the lowercase Unicode characters corresponding to ch or just
ch if no lowercase mapping is known. */
Py_UNICODE _PyUnicode_ToLowercase(Py_UNICODE ch)
{
const _PyUnicode_TypeRecord *ctype = gettyperecord(ch);
int delta = ctype->lower;
if (delta >= 32768)
delta -= 65536;
return ch + delta;
}
/* Returns 1 for Unicode characters having the category 'Ll', 'Lu', 'Lt',
'Lo' or 'Lm', 0 otherwise. */
int _PyUnicode_IsAlpha(Py_UNICODE ch)
{
const _PyUnicode_TypeRecord *ctype = gettyperecord(ch);
return (ctype->flags & ALPHA_MASK) != 0;
}
#else
/* Export the interfaces using the wchar_t type for portability
reasons: */
int _PyUnicode_IsWhitespace(Py_UNICODE ch)
{
return iswspace(ch);
}
int _PyUnicode_IsLowercase(Py_UNICODE ch)
{
return iswlower(ch);
}
int _PyUnicode_IsUppercase(Py_UNICODE ch)
{
return iswupper(ch);
}
Py_UNICODE _PyUnicode_ToLowercase(Py_UNICODE ch)
{
return towlower(ch);
}
Py_UNICODE _PyUnicode_ToUppercase(Py_UNICODE ch)
{
return towupper(ch);
}
int _PyUnicode_IsAlpha(Py_UNICODE ch)
{
return iswalpha(ch);
}
#endif