This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
handy.h: Add comments
[perl5.git] / handy.h
CommitLineData
a0d0e21e 1/* handy.h
a687059c 2 *
1129b882 3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1999, 2000,
da5d8dbb 4 * 2001, 2002, 2004, 2005, 2006, 2007, 2008, 2012 by Larry Wall and others
a687059c 5 *
6e21c824
LW
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
8d063cd8 8 *
8d063cd8
LW
9 */
10
4650c663
KW
11/* IMPORTANT NOTE: Everything whose name begins with an underscore is for
12 * internal core Perl use only. */
13
6a5bc5ac
KW
14#ifndef PERL_HANDY_H_ /* Guard against nested #inclusion */
15#define PERL_HANDY_H_
9d745869 16
24792b8d
NC
17#ifndef PERL_CORE
18# define Null(type) ((type)NULL)
954c1994
GS
19
20/*
ccfc67b7 21=head1 Handy Values
954c1994 22
3609ea0d 23=for apidoc AmU||Nullch
72d33970
FC
24Null character pointer. (No longer available when C<PERL_CORE> is
25defined.)
2307c6d0 26
954c1994 27=for apidoc AmU||Nullsv
72d33970 28Null SV pointer. (No longer available when C<PERL_CORE> is defined.)
954c1994
GS
29
30=cut
31*/
32
24792b8d
NC
33# define Nullch Null(char*)
34# define Nullfp Null(PerlIO*)
35# define Nullsv Null(SV*)
36#endif
8d063cd8 37
641d3f0b
PP
38#ifdef TRUE
39#undef TRUE
40#endif
41#ifdef FALSE
42#undef FALSE
43#endif
44#define TRUE (1)
45#define FALSE (0)
46
cf3f0ffb
DM
47/* The MUTABLE_*() macros cast pointers to the types shown, in such a way
48 * (compiler permitting) that casting away const-ness will give a warning;
49 * e.g.:
50 *
51 * const SV *sv = ...;
52 * AV *av1 = (AV*)sv; <== BAD: the const has been silently cast away
53 * AV *av2 = MUTABLE_AV(sv); <== GOOD: it may warn
54 */
55
b1bc3f34 56#if defined(__GNUC__) && !defined(PERL_GCC_BRACE_GROUPS_FORBIDDEN)
f065df04 57# define MUTABLE_PTR(p) ({ void *_p = (p); _p; })
b1bc3f34
NC
58#else
59# define MUTABLE_PTR(p) ((void *) (p))
60#endif
61
a062e10d 62#define MUTABLE_AV(p) ((AV *)MUTABLE_PTR(p))
ea726b52 63#define MUTABLE_CV(p) ((CV *)MUTABLE_PTR(p))
159b6efe 64#define MUTABLE_GV(p) ((GV *)MUTABLE_PTR(p))
dbebbdb4 65#define MUTABLE_HV(p) ((HV *)MUTABLE_PTR(p))
a45c7426 66#define MUTABLE_IO(p) ((IO *)MUTABLE_PTR(p))
b1bc3f34 67#define MUTABLE_SV(p) ((SV *)MUTABLE_PTR(p))
27d4fb96 68
f789f6a4 69#if defined(I_STDBOOL) && !defined(PERL_BOOL_AS_CHAR)
bd31be4b
NC
70# include <stdbool.h>
71# ifndef HAS_BOOL
72# define HAS_BOOL 1
73# endif
74#endif
75
8e84507e 76/* bool is built-in for g++-2.6.3 and later, which might be used
c1d22f6b
GS
77 for extensions. <_G_config.h> defines _G_HAVE_BOOL, but we can't
78 be sure _G_config.h will be included before this file. _G_config.h
8e84507e 79 also defines _G_HAVE_BOOL for both gcc and g++, but only g++
c1d22f6b
GS
80 actually has bool. Hence, _G_HAVE_BOOL is pretty useless for us.
81 g++ can be identified by __GNUG__.
82 Andy Dougherty February 2000
5d94fbed 83*/
3609ea0d 84#ifdef __GNUG__ /* GNU g++ has bool built-in */
f789f6a4 85# ifndef PERL_BOOL_AS_CHAR
5d94fbed 86# ifndef HAS_BOOL
c1d22f6b 87# define HAS_BOOL 1
5d94fbed 88# endif
f789f6a4 89# endif
5d94fbed
AD
90#endif
91
92#ifndef HAS_BOOL
f789f6a4
FC
93# ifdef bool
94# undef bool
95# endif
70d5cb32 96# define bool char
c1d22f6b 97# define HAS_BOOL 1
a687059c 98#endif
0d3e774c 99
236415e2
KW
100/* cast-to-bool. A simple (bool) cast may not do the right thing: if bool is
101 * defined as char for example, then the cast from int is
102 * implementation-defined (bool)!!(cbool) in a ternary triggers a bug in xlc on
103 * AIX */
18f5643b 104#define cBOOL(cbool) ((cbool) ? (bool)1 : (bool)0)
f2338a2e 105
46c6c7e2 106/* Try to figure out __func__ or __FUNCTION__ equivalent, if any.
e352bcff
JH
107 * XXX Should really be a Configure probe, with HAS__FUNCTION__
108 * and FUNCTION__ as results.
109 * XXX Similarly, a Configure probe for __FILE__ and __LINE__ is needed. */
46c6c7e2
JH
110#if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L) || (defined(__SUNPRO_C)) /* C99 or close enough. */
111# define FUNCTION__ __func__
07798b17
AC
112#elif (defined(USING_MSVC6)) || /* MSVC6 has neither __func__ nor __FUNCTION and no good workarounds, either. */ \
113 (defined(__DECC_VER)) /* Tru64 or VMS, and strict C89 being used, but not modern enough cc (in Tur64, -c99 not known, only -std1). */
114# define FUNCTION__ ""
46c6c7e2 115#else
07798b17 116# define FUNCTION__ __FUNCTION__ /* Common extension. */
46c6c7e2
JH
117#endif
118
27d4fb96
PP
119/* XXX A note on the perl source internal type system. The
120 original intent was that I32 be *exactly* 32 bits.
121
122 Currently, we only guarantee that I32 is *at least* 32 bits.
123 Specifically, if int is 64 bits, then so is I32. (This is the case
124 for the Cray.) This has the advantage of meshing nicely with
125 standard library calls (where we pass an I32 and the library is
126 expecting an int), but the disadvantage that an I32 is not 32 bits.
127 Andy Dougherty August 1996
24fef2a7 128
dc45a647
MB
129 There is no guarantee that there is *any* integral type with
130 exactly 32 bits. It is perfectly legal for a system to have
131 sizeof(short) == sizeof(int) == sizeof(long) == 8.
693762b4 132
dc45a647
MB
133 Similarly, there is no guarantee that I16 and U16 have exactly 16
134 bits.
693762b4 135
8e84507e
NIS
136 For dealing with issues that may arise from various 32/64-bit
137 systems, we will ask Configure to check out
8175356b 138
3609ea0d
JH
139 SHORTSIZE == sizeof(short)
140 INTSIZE == sizeof(int)
141 LONGSIZE == sizeof(long)
dc45a647 142 LONGLONGSIZE == sizeof(long long) (if HAS_LONG_LONG)
3609ea0d 143 PTRSIZE == sizeof(void *)
dc45a647
MB
144 DOUBLESIZE == sizeof(double)
145 LONG_DOUBLESIZE == sizeof(long double) (if HAS_LONG_DOUBLE).
8175356b 146
27d4fb96
PP
147*/
148
69512466
JH
149#ifdef I_INTTYPES /* e.g. Linux has int64_t without <inttypes.h> */
150# include <inttypes.h>
dd0eed91
JH
151# ifdef INT32_MIN_BROKEN
152# undef INT32_MIN
153# define INT32_MIN (-2147483647-1)
154# endif
155# ifdef INT64_MIN_BROKEN
156# undef INT64_MIN
157# define INT64_MIN (-9223372036854775807LL-1)
158# endif
69512466
JH
159#endif
160
8175356b
JH
161typedef I8TYPE I8;
162typedef U8TYPE U8;
163typedef I16TYPE I16;
164typedef U16TYPE U16;
165typedef I32TYPE I32;
166typedef U32TYPE U32;
16d89be8
JH
167
168#ifdef HAS_QUAD
8175356b
JH
169typedef I64TYPE I64;
170typedef U64TYPE U64;
16d89be8 171#endif
8175356b 172
9c39b21c
JH
173/* INT64_C/UINT64_C are C99 from <stdint.h> (so they will not be
174 * available in strict C89 mode), but they are nice, so let's define
175 * them if necessary. */
0d3bd0a9 176#if defined(HAS_QUAD)
9c39b21c
JH
177# undef PeRl_INT64_C
178# undef PeRl_UINT64_C
179/* Prefer the native integer types (int and long) over long long
180 * (which is not C89) and Win32-specific __int64. */
181# if QUADKIND == QUAD_IS_INT && INTSIZE == 8
182# define PeRl_INT64_C(c) (c)
183# define PeRl_UINT64_C(c) CAT2(c,U)
184# endif
185# if QUADKIND == QUAD_IS_LONG && LONGSIZE == 8
186# define PeRl_INT64_C(c) CAT2(c,L)
187# define PeRl_UINT64_C(c) CAT2(c,UL)
188# endif
189# if QUADKIND == QUAD_IS_LONG_LONG && defined(HAS_LONG_LONG)
190# define PeRl_INT64_C(c) CAT2(c,LL)
191# define PeRl_UINT64_C(c) CAT2(c,ULL)
192# endif
193# if QUADKIND == QUAD_IS___INT64
194# define PeRl_INT64_C(c) CAT2(c,I64)
195# define PeRl_UINT64_C(c) CAT2(c,UI64)
196# endif
197# ifndef PeRl_INT64_C
702c92eb
JH
198# define PeRl_INT64_C(c) ((I64)(c)) /* last resort */
199# define PeRl_UINT64_C(c) ((U64)(c))
9c39b21c
JH
200# endif
201/* In OS X the INT64_C/UINT64_C are defined with LL/ULL, which will
202 * not fly with C89-pedantic gcc, so let's undefine them first so that
203 * we can redefine them with our native integer preferring versions. */
8eafc6eb 204# if defined(PERL_DARWIN) && defined(PERL_GCC_PEDANTIC)
9c39b21c
JH
205# undef INT64_C
206# undef UINT64_C
207# endif
208# ifndef INT64_C
209# define INT64_C(c) PeRl_INT64_C(c)
210# endif
211# ifndef UINT64_C
212# define UINT64_C(c) PeRl_UINT64_C(c)
213# endif
e8c95190 214#endif
dc750f2e 215
d8668976 216#if defined(UINT8_MAX) && defined(INT16_MAX) && defined(INT32_MAX)
5ff3f7a4 217
5ff3f7a4
GS
218/* I8_MAX and I8_MIN constants are not defined, as I8 is an ambiguous type.
219 Please search CHAR_MAX in perl.h for further details. */
220#define U8_MAX UINT8_MAX
221#define U8_MIN UINT8_MIN
222
5ff3f7a4
GS
223#define I16_MAX INT16_MAX
224#define I16_MIN INT16_MIN
225#define U16_MAX UINT16_MAX
226#define U16_MIN UINT16_MIN
227
5ff3f7a4
GS
228#define I32_MAX INT32_MAX
229#define I32_MIN INT32_MIN
0e983133
GS
230#ifndef UINT32_MAX_BROKEN /* e.g. HP-UX with gcc messes this up */
231# define U32_MAX UINT32_MAX
232#else
233# define U32_MAX 4294967295U
234#endif
5ff3f7a4
GS
235#define U32_MIN UINT32_MIN
236
237#else
238
5c9fa16e
KA
239/* I8_MAX and I8_MIN constants are not defined, as I8 is an ambiguous type.
240 Please search CHAR_MAX in perl.h for further details. */
27d4fb96
PP
241#define U8_MAX PERL_UCHAR_MAX
242#define U8_MIN PERL_UCHAR_MIN
79072805 243
27d4fb96
PP
244#define I16_MAX PERL_SHORT_MAX
245#define I16_MIN PERL_SHORT_MIN
246#define U16_MAX PERL_USHORT_MAX
247#define U16_MIN PERL_USHORT_MIN
79072805 248
c4f23d77 249#if LONGSIZE > 4
27d4fb96
PP
250# define I32_MAX PERL_INT_MAX
251# define I32_MIN PERL_INT_MIN
252# define U32_MAX PERL_UINT_MAX
253# define U32_MIN PERL_UINT_MIN
79072805 254#else
27d4fb96
PP
255# define I32_MAX PERL_LONG_MAX
256# define I32_MIN PERL_LONG_MIN
257# define U32_MAX PERL_ULONG_MAX
258# define U32_MIN PERL_ULONG_MIN
79072805
LW
259#endif
260
5ff3f7a4
GS
261#endif
262
58a9a5d5 263/* log(2) is pretty close to 0.30103, just in case anyone is grepping for it */
fc36a67e
PP
264#define BIT_DIGITS(N) (((N)*146)/485 + 1) /* log2(10) =~ 146/485 */
265#define TYPE_DIGITS(T) BIT_DIGITS(sizeof(T) * 8)
266#define TYPE_CHARS(T) (TYPE_DIGITS(T) + 2) /* sign, NUL */
267
88794300 268/* Unused by core; should be deprecated */
ff68c719 269#define Ctl(ch) ((ch) & 037)
8d063cd8 270
98fce2a4
KW
271#if defined(PERL_CORE) || defined(PERL_EXT)
272# ifndef MIN
273# define MIN(a,b) ((a) < (b) ? (a) : (b))
274# endif
275# ifndef MAX
276# define MAX(a,b) ((a) > (b) ? (a) : (b))
277# endif
278#endif
279
4650c663
KW
280/* This is a helper macro to avoid preprocessor issues, replaced by nothing
281 * unless under DEBUGGING, where it expands to an assert of its argument,
282 * followed by a comma (hence the comma operator). If we just used a straight
3e94db23
JH
283 * assert(), we would get a comma with nothing before it when not DEBUGGING.
284 *
285 * We also use empty definition under Coverity since the __ASSERT__
286 * checks often check for things that Really Cannot Happen, and Coverity
287 * detects that and gets all excited. */
288
289#if defined(DEBUGGING) && !defined(__COVERITY__)
0f092d08
KW
290# define __ASSERT_(statement) assert(statement),
291#else
292# define __ASSERT_(statement)
293#endif
294
3fe05580
MHM
295/*
296=head1 SV-Body Allocation
297
298=for apidoc Ama|SV*|newSVpvs|const char* s
0c395ea5 299Like C<newSVpvn>, but takes a C<NUL>-terminated literal string instead of a
30a15352 300string/length pair.
3fe05580 301
84bafc02 302=for apidoc Ama|SV*|newSVpvs_flags|const char* s|U32 flags
0c395ea5 303Like C<newSVpvn_flags>, but takes a C<NUL>-terminated literal string instead of
30a15352 304a string/length pair.
84bafc02 305
3fe05580 306=for apidoc Ama|SV*|newSVpvs_share|const char* s
0c395ea5 307Like C<newSVpvn_share>, but takes a C<NUL>-terminated literal string instead of
30a15352 308a string/length pair and omits the hash parameter.
3fe05580 309
9dcc53ea 310=for apidoc Am|void|sv_catpvs_flags|SV* sv|const char* s|I32 flags
0c395ea5 311Like C<sv_catpvn_flags>, but takes a C<NUL>-terminated literal string instead
30a15352 312of a string/length pair.
9dcc53ea
Z
313
314=for apidoc Am|void|sv_catpvs_nomg|SV* sv|const char* s
0c395ea5
KW
315Like C<sv_catpvn_nomg>, but takes a C<NUL>-terminated literal string instead of
316a string/length pair.
9dcc53ea 317
7b4e7469 318=for apidoc Am|void|sv_catpvs|SV* sv|const char* s
0c395ea5
KW
319Like C<sv_catpvn>, but takes a C<NUL>-terminated literal string instead of a
320string/length pair.
3fe05580 321
9dcc53ea 322=for apidoc Am|void|sv_catpvs_mg|SV* sv|const char* s
0c395ea5 323Like C<sv_catpvn_mg>, but takes a C<NUL>-terminated literal string instead of a
9dcc53ea
Z
324string/length pair.
325
7b4e7469 326=for apidoc Am|void|sv_setpvs|SV* sv|const char* s
0c395ea5
KW
327Like C<sv_setpvn>, but takes a C<NUL>-terminated literal string instead of a
328string/length pair.
3fe05580 329
9dcc53ea 330=for apidoc Am|void|sv_setpvs_mg|SV* sv|const char* s
0c395ea5 331Like C<sv_setpvn_mg>, but takes a C<NUL>-terminated literal string instead of a
9dcc53ea
Z
332string/length pair.
333
334=for apidoc Am|SV *|sv_setref_pvs|const char* s
0c395ea5
KW
335Like C<sv_setref_pvn>, but takes a C<NUL>-terminated literal string instead of
336a string/length pair.
9dcc53ea 337
3fe05580
MHM
338=head1 Memory Management
339
340=for apidoc Ama|char*|savepvs|const char* s
0c395ea5 341Like C<savepvn>, but takes a C<NUL>-terminated literal string instead of a
30a15352 342string/length pair.
3fe05580 343
9dcc53ea
Z
344=for apidoc Ama|char*|savesharedpvs|const char* s
345A version of C<savepvs()> which allocates the duplicate string in memory
346which is shared between threads.
347
3fe05580
MHM
348=head1 GV Functions
349
350=for apidoc Am|HV*|gv_stashpvs|const char* name|I32 create
0c395ea5
KW
351Like C<gv_stashpvn>, but takes a C<NUL>-terminated literal string instead of a
352string/length pair.
3fe05580
MHM
353
354=head1 Hash Manipulation Functions
355
356=for apidoc Am|SV**|hv_fetchs|HV* tb|const char* key|I32 lval
0c395ea5
KW
357Like C<hv_fetch>, but takes a C<NUL>-terminated literal string instead of a
358string/length pair.
3fe05580
MHM
359
360=for apidoc Am|SV**|hv_stores|HV* tb|const char* key|NULLOK SV* val
0c395ea5
KW
361Like C<hv_store>, but takes a C<NUL>-terminated literal string instead of a
362string/length pair
3fe05580
MHM
363and omits the hash parameter.
364
510966aa
Z
365=head1 Lexer interface
366
367=for apidoc Amx|void|lex_stuff_pvs|const char *pv|U32 flags
368
0c395ea5
KW
369Like L</lex_stuff_pvn>, but takes a C<NUL>-terminated literal string instead of
370a string/length pair.
510966aa 371
3fe05580
MHM
372=cut
373*/
374
926bb54c
KW
375/* concatenating with "" ensures that only literal strings are accepted as
376 * argument */
26bfd924 377#define STR_WITH_LEN(s) ("" s ""), (sizeof(s)-1)
2efa8cc7 378
926bb54c
KW
379/* note that STR_WITH_LEN() can't be used as argument to macros or functions
380 * that under some configurations might be macros, which means that it requires
381 * the full Perl_xxx(aTHX_ ...) form for any API calls where it's used.
ba3a79e7
GA
382 */
383
384/* STR_WITH_LEN() shortcuts */
385#define newSVpvs(str) Perl_newSVpvn(aTHX_ STR_WITH_LEN(str))
84bafc02
NC
386#define newSVpvs_flags(str,flags) \
387 Perl_newSVpvn_flags(aTHX_ STR_WITH_LEN(str), flags)
ba3a79e7 388#define newSVpvs_share(str) Perl_newSVpvn_share(aTHX_ STR_WITH_LEN(str), 0)
9dcc53ea
Z
389#define sv_catpvs_flags(sv, str, flags) \
390 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), flags)
391#define sv_catpvs_nomg(sv, str) \
392 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), 0)
393#define sv_catpvs(sv, str) \
394 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), SV_GMAGIC)
395#define sv_catpvs_mg(sv, str) \
396 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), SV_GMAGIC|SV_SMAGIC)
3fe05580 397#define sv_setpvs(sv, str) Perl_sv_setpvn(aTHX_ sv, STR_WITH_LEN(str))
9dcc53ea
Z
398#define sv_setpvs_mg(sv, str) Perl_sv_setpvn_mg(aTHX_ sv, STR_WITH_LEN(str))
399#define sv_setref_pvs(rv, classname, str) \
400 Perl_sv_setref_pvn(aTHX_ rv, classname, STR_WITH_LEN(str))
ba3a79e7 401#define savepvs(str) Perl_savepvn(aTHX_ STR_WITH_LEN(str))
9dcc53ea
Z
402#define savesharedpvs(str) Perl_savesharedpvn(aTHX_ STR_WITH_LEN(str))
403#define gv_stashpvs(str, create) \
404 Perl_gv_stashpvn(aTHX_ STR_WITH_LEN(str), create)
405#define gv_fetchpvs(namebeg, add, sv_type) \
406 Perl_gv_fetchpvn_flags(aTHX_ STR_WITH_LEN(namebeg), add, sv_type)
407#define gv_fetchpvn(namebeg, len, add, sv_type) \
408 Perl_gv_fetchpvn_flags(aTHX_ namebeg, len, add, sv_type)
409#define sv_catxmlpvs(dsv, str, utf8) \
410 Perl_sv_catxmlpvn(aTHX_ dsv, STR_WITH_LEN(str), utf8)
4ac46235 411
ba3a79e7 412
510966aa
Z
413#define lex_stuff_pvs(pv,flags) Perl_lex_stuff_pvn(aTHX_ STR_WITH_LEN(pv), flags)
414
b96d8cd9
NC
415#define get_cvs(str, flags) \
416 Perl_get_cvn_flags(aTHX_ STR_WITH_LEN(str), (flags))
5c1737d1 417
954c1994 418/*
ccfc67b7
JH
419=head1 Miscellaneous Functions
420
954c1994 421=for apidoc Am|bool|strNE|char* s1|char* s2
dc6b0978
KW
422Test two C<NUL>-terminated strings to see if they are different. Returns true
423or false.
954c1994
GS
424
425=for apidoc Am|bool|strEQ|char* s1|char* s2
dc6b0978
KW
426Test two C<NUL>-terminated strings to see if they are equal. Returns true or
427false.
954c1994
GS
428
429=for apidoc Am|bool|strLT|char* s1|char* s2
dc6b0978
KW
430Test two C<NUL>-terminated strings to see if the first, C<s1>, is less than the
431second, C<s2>. Returns true or false.
954c1994
GS
432
433=for apidoc Am|bool|strLE|char* s1|char* s2
dc6b0978
KW
434Test two C<NUL>-terminated strings to see if the first, C<s1>, is less than or
435equal to the second, C<s2>. Returns true or false.
954c1994
GS
436
437=for apidoc Am|bool|strGT|char* s1|char* s2
dc6b0978
KW
438Test two C<NUL>-terminated strings to see if the first, C<s1>, is greater than
439the second, C<s2>. Returns true or false.
954c1994
GS
440
441=for apidoc Am|bool|strGE|char* s1|char* s2
dc6b0978
KW
442Test two C<NUL>-terminated strings to see if the first, C<s1>, is greater than
443or equal to the second, C<s2>. Returns true or false.
954c1994
GS
444
445=for apidoc Am|bool|strnNE|char* s1|char* s2|STRLEN len
dc6b0978
KW
446Test two C<NUL>-terminated strings to see if they are different. The C<len>
447parameter indicates the number of bytes to compare. Returns true or false. (A
954c1994
GS
448wrapper for C<strncmp>).
449
450=for apidoc Am|bool|strnEQ|char* s1|char* s2|STRLEN len
dc6b0978
KW
451Test two C<NUL>-terminated strings to see if they are equal. The C<len>
452parameter indicates the number of bytes to compare. Returns true or false. (A
453wrapper for C<strncmp>).
954c1994 454
bd18bd40
KW
455=for apidoc Am|bool|memEQ|char* s1|char* s2|STRLEN len
456Test two buffers (which may contain embedded C<NUL> characters, to see if they
457are equal. The C<len> parameter indicates the number of bytes to compare.
458Returns zero if equal, or non-zero if non-equal.
459
460=for apidoc Am|bool|memNE|char* s1|char* s2|STRLEN len
461Test two buffers (which may contain embedded C<NUL> characters, to see if they
462are not equal. The C<len> parameter indicates the number of bytes to compare.
463Returns zero if non-equal, or non-zero if equal.
464
954c1994 465=cut
fc169e00
KW
466
467New macros should use the following conventions for their names (which are
468based on the underlying C library functions):
469
470 (mem | str n? ) (EQ | NE | LT | GT | GE | (( BEGIN | END ) P? )) l? s?
471
472 Each has two main parameters, string-like operands that are compared
473 against each other, as specified by the macro name. Some macros may
474 additionally have one or potentially even two length parameters. If a length
475 parameter applies to both string parameters, it will be positioned third;
476 otherwise any length parameter immediately follows the string parameter it
477 applies to.
478
479 If the prefix to the name is 'str', the string parameter is a pointer to a C
480 language string. Such a string does not contain embedded NUL bytes; its
481 length may be unknown, but can be calculated by C<strlen()>, since it is
482 terminated by a NUL, which isn't included in its length.
483
484 The optional 'n' following 'str' means that that there is a third parameter,
485 giving the maximum number of bytes to look at in each string. Even if both
486 strings are longer than the length parameter, those extra bytes will be
487 unexamined.
488
489 The 's' suffix means that the 2nd byte string parameter is a literal C
490 double-quoted string. Its length will automatically be calculated by the
491 macro, so no length parameter will ever be needed for it.
492
493 If the prefix is 'mem', the string parameters don't have to be C strings;
494 they may contain embedded NUL bytes, do not necessarily have a terminating
495 NUL, and their lengths can be known only through other means, which in
496 practice are additional parameter(s) passed to the function. All 'mem'
497 functions have at least one length parameter. Barring any 'l' or 's' suffix,
498 there is a single length parameter, in position 3, which applies to both
499 string parameters. The 's' suffix means, as described above, that the 2nd
500 string is a literal double-quoted C string (hence its length is calculated by
501 the macro, and the length parameter to the function applies just to the first
502 string parameter, and hence is positioned just after it). An 'l' suffix
503 means that the 2nd string parameter has its own length parameter, and the
504 signature will look like memFOOl(s1, l1, s2, l2).
505
506 BEGIN (and END) are for testing if the 2nd string is an initial (or final)
507 substring of the 1st string. 'P' if present indicates that the substring
508 must be a "proper" one in tha mathematical sense that the first one must be
509 strictly larger than the 2nd.
510
954c1994
GS
511*/
512
62946e08 513
75400963
KW
514#define strNE(s1,s2) (strcmp(s1,s2) != 0)
515#define strEQ(s1,s2) (strcmp(s1,s2) == 0)
8d063cd8
LW
516#define strLT(s1,s2) (strcmp(s1,s2) < 0)
517#define strLE(s1,s2) (strcmp(s1,s2) <= 0)
518#define strGT(s1,s2) (strcmp(s1,s2) > 0)
519#define strGE(s1,s2) (strcmp(s1,s2) >= 0)
62946e08 520
75400963
KW
521#define strnNE(s1,s2,l) (strncmp(s1,s2,l) != 0)
522#define strnEQ(s1,s2,l) (strncmp(s1,s2,l) == 0)
378cc40b 523
75400963
KW
524#define memNE(s1,s2,l) (memcmp(s1,s2,l) != 0)
525#define memEQ(s1,s2,l) (memcmp(s1,s2,l) == 0)
36477c24 526
085b7534 527/* memEQ and memNE where second comparand is a string constant */
568a785a 528#define memEQs(s1, l, s2) \
777fa2cb 529 (((sizeof(s2)-1) == (l)) && memEQ((s1), ("" s2 ""), (sizeof(s2)-1)))
5f50c6c9 530#define memNEs(s1, l, s2) (! memEQs(s1, l, s2))
568a785a 531
fdbb9a7c
KW
532/* Keep these private until we decide it was a good idea */
533#if defined(PERL_CORE) || defined(PERL_EXT) || defined(PERL_EXT_POSIX)
534
535#define strBEGINs(s1,s2) (strncmp(s1,"" s2 "", sizeof(s2)-1) == 0)
536
bdb7e3f0
KW
537#define memBEGINs(s1, l, s2) \
538 ( (l) >= sizeof(s2) - 1 \
539 && memEQ(s1, "" s2 "", sizeof(s2)-1))
de627158
KW
540#define memBEGINPs(s1, l, s2) \
541 ( (l) > sizeof(s2) - 1 \
542 && memEQ(s1, "" s2 "", sizeof(s2)-1))
bdb7e3f0
KW
543#define memENDs(s1, l, s2) \
544 ( (l) >= sizeof(s2) - 1 \
545 && memEQ(s1 + (l) - (sizeof(s2) - 1), "" s2 "", sizeof(s2)-1))
b80f8424
KW
546#define memENDPs(s1, l, s2) \
547 ( (l) > sizeof(s2) \
548 && memEQ(s1 + (l) - (sizeof(s2) - 1), "" s2 "", sizeof(s2)-1))
fdbb9a7c 549#endif /* End of making macros private */
bdb7e3f0 550
062b6850
KW
551#define memLT(s1,s2,l) (memcmp(s1,s2,l) < 0)
552#define memLE(s1,s2,l) (memcmp(s1,s2,l) <= 0)
553#define memGT(s1,s2,l) (memcmp(s1,s2,l) > 0)
554#define memGE(s1,s2,l) (memcmp(s1,s2,l) >= 0)
555
bbce6d69
PP
556/*
557 * Character classes.
558 *
559 * Unfortunately, the introduction of locales means that we
560 * can't trust isupper(), etc. to tell the truth. And when
561 * it comes to /\w+/ with tainting enabled, we *must* be able
562 * to trust our character classes.
563 *
564 * Therefore, the default tests in the text of Perl will be
565 * independent of locale. Any code that wants to depend on
566 * the current locale will use the tests that begin with "lc".
567 */
568
2304df62
AD
569#ifdef HAS_SETLOCALE /* XXX Is there a better test for this? */
570# ifndef CTYPE256
571# define CTYPE256
572# endif
573#endif
574
954c1994 575/*
ccfc67b7 576
dcccc8ff 577=head1 Character classification
243effed
KW
578This section is about functions (really macros) that classify characters
579into types, such as punctuation versus alphabetic, etc. Most of these are
580analogous to regular expression character classes. (See
581L<perlrecharclass/POSIX Character Classes>.) There are several variants for
582each class. (Not all macros have all variants; each item below lists the
583ones valid for it.) None are affected by C<use bytes>, and only the ones
584with C<LC> in the name are affected by the current locale.
585
586The base function, e.g., C<isALPHA()>, takes an octet (either a C<char> or a
587C<U8>) as input and returns a boolean as to whether or not the character
6aff1f14
KW
588represented by that octet is (or on non-ASCII platforms, corresponds to) an
589ASCII character in the named class based on platform, Unicode, and Perl rules.
590If the input is a number that doesn't fit in an octet, FALSE is returned.
243effed 591
c98722a4 592Variant C<isI<FOO>_A> (e.g., C<isALPHA_A()>) is identical to the base function
550da823
KW
593with no suffix C<"_A">. This variant is used to emphasize by its name that
594only ASCII-range characters can return TRUE.
4b9734bf 595
d60679e1 596Variant C<isI<FOO>_L1> imposes the Latin-1 (or EBCDIC equivalent) character set
4b9734bf
KW
597onto the platform. That is, the code points that are ASCII are unaffected,
598since ASCII is a subset of Latin-1. But the non-ASCII code points are treated
599as if they are Latin-1 characters. For example, C<isWORDCHAR_L1()> will return
600true when called with the code point 0xDF, which is a word character in both
4650c663 601ASCII and EBCDIC (though it represents different characters in each).
243effed 602
c98722a4 603Variant C<isI<FOO>_uvchr> is like the C<isI<FOO>_L1> variant, but accepts any UV code
243effed
KW
604point as input. If the code point is larger than 255, Unicode rules are used
605to determine if it is in the character class. For example,
d0da05db 606C<isWORDCHAR_uvchr(0x100)> returns TRUE, since 0x100 is LATIN CAPITAL LETTER A
6aff1f14 607WITH MACRON in Unicode, and is a word character.
243effed 608
c98722a4 609Variant C<isI<FOO>_utf8_safe> is like C<isI<FOO>_uvchr>, but is used for UTF-8
da8c1a98
KW
610encoded strings. Each call classifies one character, even if the string
611contains many. This variant takes two parameters. The first, C<p>, is a
612pointer to the first byte of the character to be classified. (Recall that it
613may take more than one byte to represent a character in UTF-8 strings.) The
614second parameter, C<e>, points to anywhere in the string beyond the first
615character, up to one byte past the end of the entire string. The suffix
616C<_safe> in the function's name indicates that it will not attempt to read
617beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is true (this
618is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the input
619character is malformed in some way, the program may croak, or the function may
620return FALSE, at the discretion of the implementation, and subject to change in
621future releases.
622
c98722a4 623Variant C<isI<FOO>_utf8> is like C<isI<FOO>_utf8_safe>, but takes just a single
da8c1a98 624parameter, C<p>, which has the same meaning as the corresponding parameter does
c98722a4 625in C<isI<FOO>_utf8_safe>. The function therefore can't check if it is reading
34aeb2e9 626beyond the end of the string. Starting in Perl v5.30, it will take a second
c98722a4 627parameter, becoming a synonym for C<isI<FOO>_utf8_safe>. At that time every
34aeb2e9 628program that uses it will have to be changed to successfully compile. In the
c98722a4 629meantime, the first runtime call to C<isI<FOO>_utf8> from each call point in the
34aeb2e9 630program will raise a deprecation warning, enabled by default. You can convert
c98722a4 631your program now to use C<isI<FOO>_utf8_safe>, and avoid the warnings, and get an
34aeb2e9
KW
632extra measure of protection, or you can wait until v5.30, when you'll be forced
633to add the C<e> parameter.
243effed 634
c98722a4 635Variant C<isI<FOO>_LC> is like the C<isI<FOO>_A> and C<isI<FOO>_L1> variants, but the
31f05a37
KW
636result is based on the current locale, which is what C<LC> in the name stands
637for. If Perl can determine that the current locale is a UTF-8 locale, it uses
638the published Unicode rules; otherwise, it uses the C library function that
639gives the named classification. For example, C<isDIGIT_LC()> when not in a
640UTF-8 locale returns the result of calling C<isdigit()>. FALSE is always
1a83413c
KW
641returned if the input won't fit into an octet. On some platforms where the C
642library function is known to be defective, Perl changes its result to follow
643the POSIX standard's rules.
243effed 644
c98722a4
KW
645Variant C<isI<FOO>_LC_uvchr> is like C<isI<FOO>_LC>, but is defined on any UV. It
646returns the same as C<isI<FOO>_LC> for input code points less than 256, and
243effed
KW
647returns the hard-coded, not-affected-by-locale, Unicode results for larger ones.
648
c98722a4 649Variant C<isI<FOO>_LC_utf8_safe> is like C<isI<FOO>_LC_uvchr>, but is used for UTF-8
da8c1a98
KW
650encoded strings. Each call classifies one character, even if the string
651contains many. This variant takes two parameters. The first, C<p>, is a
652pointer to the first byte of the character to be classified. (Recall that it
653may take more than one byte to represent a character in UTF-8 strings.) The
654second parameter, C<e>, points to anywhere in the string beyond the first
655character, up to one byte past the end of the entire string. The suffix
656C<_safe> in the function's name indicates that it will not attempt to read
657beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is true (this
658is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the input
659character is malformed in some way, the program may croak, or the function may
660return FALSE, at the discretion of the implementation, and subject to change in
661future releases.
662
c98722a4 663Variant C<isI<FOO>_LC_utf8> is like C<isI<FOO>_LC_utf8_safe>, but takes just a single
da8c1a98 664parameter, C<p>, which has the same meaning as the corresponding parameter does
c98722a4 665in C<isI<FOO>_LC_utf8_safe>. The function therefore can't check if it is reading
34aeb2e9 666beyond the end of the string. Starting in Perl v5.30, it will take a second
c98722a4 667parameter, becoming a synonym for C<isI<FOO>_LC_utf8_safe>. At that time every
34aeb2e9 668program that uses it will have to be changed to successfully compile. In the
c98722a4 669meantime, the first runtime call to C<isI<FOO>_LC_utf8> from each call point in
34aeb2e9 670the program will raise a deprecation warning, enabled by default. You can
c98722a4 671convert your program now to use C<isI<FOO>_LC_utf8_safe>, and avoid the warnings,
34aeb2e9
KW
672and get an extra measure of protection, or you can wait until v5.30, when
673you'll be forced to add the C<e> parameter.
ccfc67b7 674
954c1994 675=for apidoc Am|bool|isALPHA|char ch
2787a470 676Returns a boolean indicating whether the specified character is an
6aff1f14 677alphabetic character, analogous to C<m/[[:alpha:]]/>.
dcccc8ff
KW
678See the L<top of this section|/Character classification> for an explanation of
679variants
da8c1a98
KW
680C<isALPHA_A>, C<isALPHA_L1>, C<isALPHA_uvchr>, C<isALPHA_utf8_safe>,
681C<isALPHA_LC>, C<isALPHA_LC_uvchr>, and C<isALPHA_LC_utf8_safe>.
8a58bdcf 682
15861f94
KW
683=for apidoc Am|bool|isALPHANUMERIC|char ch
684Returns a boolean indicating whether the specified character is a either an
6aff1f14 685alphabetic character or decimal digit, analogous to C<m/[[:alnum:]]/>.
dcccc8ff
KW
686See the L<top of this section|/Character classification> for an explanation of
687variants
d0da05db 688C<isALPHANUMERIC_A>, C<isALPHANUMERIC_L1>, C<isALPHANUMERIC_uvchr>,
da8c1a98
KW
689C<isALPHANUMERIC_utf8_safe>, C<isALPHANUMERIC_LC>, C<isALPHANUMERIC_LC_uvchr>,
690and C<isALPHANUMERIC_LC_utf8_safe>.
15861f94 691
8a58bdcf
KW
692=for apidoc Am|bool|isASCII|char ch
693Returns a boolean indicating whether the specified character is one of the 128
243effed 694characters in the ASCII character set, analogous to C<m/[[:ascii:]]/>.
e5ad6aba 695On non-ASCII platforms, it returns TRUE iff this
8a58bdcf
KW
696character corresponds to an ASCII character. Variants C<isASCII_A()> and
697C<isASCII_L1()> are identical to C<isASCII()>.
dcccc8ff
KW
698See the L<top of this section|/Character classification> for an explanation of
699variants
da8c1a98
KW
700C<isASCII_uvchr>, C<isASCII_utf8_safe>, C<isASCII_LC>, C<isASCII_LC_uvchr>, and
701C<isASCII_LC_utf8_safe>. Note, however, that some platforms do not have the C
243effed
KW
702library routine C<isascii()>. In these cases, the variants whose names contain
703C<LC> are the same as the corresponding ones without.
704
d98532ea
KW
705Also note, that because all ASCII characters are UTF-8 invariant (meaning they
706have the exact same representation (always a single byte) whether encoded in
707UTF-8 or not), C<isASCII> will give the correct results when called with any
da8c1a98
KW
708byte in any string encoded or not in UTF-8. And similarly C<isASCII_utf8_safe>
709will work properly on any string encoded or not in UTF-8.
d98532ea 710
243effed
KW
711=for apidoc Am|bool|isBLANK|char ch
712Returns a boolean indicating whether the specified character is a
6aff1f14 713character considered to be a blank, analogous to C<m/[[:blank:]]/>.
dcccc8ff
KW
714See the L<top of this section|/Character classification> for an explanation of
715variants
da8c1a98
KW
716C<isBLANK_A>, C<isBLANK_L1>, C<isBLANK_uvchr>, C<isBLANK_utf8_safe>,
717C<isBLANK_LC>, C<isBLANK_LC_uvchr>, and C<isBLANK_LC_utf8_safe>. Note,
718however, that some platforms do not have the C library routine
719C<isblank()>. In these cases, the variants whose names contain C<LC> are
720the same as the corresponding ones without.
243effed
KW
721
722=for apidoc Am|bool|isCNTRL|char ch
723Returns a boolean indicating whether the specified character is a
6aff1f14 724control character, analogous to C<m/[[:cntrl:]]/>.
dcccc8ff
KW
725See the L<top of this section|/Character classification> for an explanation of
726variants
da8c1a98
KW
727C<isCNTRL_A>, C<isCNTRL_L1>, C<isCNTRL_uvchr>, C<isCNTRL_utf8_safe>,
728C<isCNTRL_LC>, C<isCNTRL_LC_uvchr>, and C<isCNTRL_LC_utf8_safe> On EBCDIC
729platforms, you almost always want to use the C<isCNTRL_L1> variant.
954c1994 730
954c1994 731=for apidoc Am|bool|isDIGIT|char ch
2787a470 732Returns a boolean indicating whether the specified character is a
6aff1f14 733digit, analogous to C<m/[[:digit:]]/>.
8a58bdcf 734Variants C<isDIGIT_A> and C<isDIGIT_L1> are identical to C<isDIGIT>.
dcccc8ff
KW
735See the L<top of this section|/Character classification> for an explanation of
736variants
da8c1a98
KW
737C<isDIGIT_uvchr>, C<isDIGIT_utf8_safe>, C<isDIGIT_LC>, C<isDIGIT_LC_uvchr>, and
738C<isDIGIT_LC_utf8_safe>.
243effed
KW
739
740=for apidoc Am|bool|isGRAPH|char ch
741Returns a boolean indicating whether the specified character is a
6aff1f14 742graphic character, analogous to C<m/[[:graph:]]/>.
dcccc8ff 743See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
744variants C<isGRAPH_A>, C<isGRAPH_L1>, C<isGRAPH_uvchr>, C<isGRAPH_utf8_safe>,
745C<isGRAPH_LC>, C<isGRAPH_LC_uvchr>, and C<isGRAPH_LC_utf8_safe>.
954c1994 746
0c82b6df 747=for apidoc Am|bool|isLOWER|char ch
2787a470 748Returns a boolean indicating whether the specified character is a
6aff1f14 749lowercase character, analogous to C<m/[[:lower:]]/>.
dcccc8ff
KW
750See the L<top of this section|/Character classification> for an explanation of
751variants
da8c1a98
KW
752C<isLOWER_A>, C<isLOWER_L1>, C<isLOWER_uvchr>, C<isLOWER_utf8_safe>,
753C<isLOWER_LC>, C<isLOWER_LC_uvchr>, and C<isLOWER_LC_utf8_safe>.
0c82b6df 754
c99e91e9 755=for apidoc Am|bool|isOCTAL|char ch
2787a470 756Returns a boolean indicating whether the specified character is an
6aff1f14 757octal digit, [0-7].
243effed
KW
758The only two variants are C<isOCTAL_A> and C<isOCTAL_L1>; each is identical to
759C<isOCTAL>.
760
761=for apidoc Am|bool|isPUNCT|char ch
762Returns a boolean indicating whether the specified character is a
6aff1f14
KW
763punctuation character, analogous to C<m/[[:punct:]]/>.
764Note that the definition of what is punctuation isn't as
243effed
KW
765straightforward as one might desire. See L<perlrecharclass/POSIX Character
766Classes> for details.
dcccc8ff 767See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
768variants C<isPUNCT_A>, C<isPUNCT_L1>, C<isPUNCT_uvchr>, C<isPUNCT_utf8_safe>,
769C<isPUNCT_LC>, C<isPUNCT_LC_uvchr>, and C<isPUNCT_LC_utf8_safe>.
c99e91e9 770
0c82b6df 771=for apidoc Am|bool|isSPACE|char ch
2787a470 772Returns a boolean indicating whether the specified character is a
6aff1f14 773whitespace character. This is analogous
398d098a 774to what C<m/\s/> matches in a regular expression. Starting in Perl 5.18
779cf272 775this also matches what C<m/[[:space:]]/> does. Prior to 5.18, only the
398d098a
KW
776locale forms of this macro (the ones with C<LC> in their names) matched
777precisely what C<m/[[:space:]]/> does. In those releases, the only difference,
778in the non-locale variants, was that C<isSPACE()> did not match a vertical tab.
779(See L</isPSXSPC> for a macro that matches a vertical tab in all releases.)
dcccc8ff
KW
780See the L<top of this section|/Character classification> for an explanation of
781variants
da8c1a98
KW
782C<isSPACE_A>, C<isSPACE_L1>, C<isSPACE_uvchr>, C<isSPACE_utf8_safe>,
783C<isSPACE_LC>, C<isSPACE_LC_uvchr>, and C<isSPACE_LC_utf8_safe>.
0c82b6df 784
398d098a
KW
785=for apidoc Am|bool|isPSXSPC|char ch
786(short for Posix Space)
779cf272
KW
787Starting in 5.18, this is identical in all its forms to the
788corresponding C<isSPACE()> macros.
398d098a
KW
789The locale forms of this macro are identical to their corresponding
790C<isSPACE()> forms in all Perl releases. In releases prior to 5.18, the
791non-locale forms differ from their C<isSPACE()> forms only in that the
792C<isSPACE()> forms don't match a Vertical Tab, and the C<isPSXSPC()> forms do.
793Otherwise they are identical. Thus this macro is analogous to what
794C<m/[[:space:]]/> matches in a regular expression.
dcccc8ff 795See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
796variants C<isPSXSPC_A>, C<isPSXSPC_L1>, C<isPSXSPC_uvchr>, C<isPSXSPC_utf8_safe>,
797C<isPSXSPC_LC>, C<isPSXSPC_LC_uvchr>, and C<isPSXSPC_LC_utf8_safe>.
398d098a 798
954c1994 799=for apidoc Am|bool|isUPPER|char ch
2787a470 800Returns a boolean indicating whether the specified character is an
6aff1f14 801uppercase character, analogous to C<m/[[:upper:]]/>.
dcccc8ff 802See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
803variants C<isUPPER_A>, C<isUPPER_L1>, C<isUPPER_uvchr>, C<isUPPER_utf8_safe>,
804C<isUPPER_LC>, C<isUPPER_LC_uvchr>, and C<isUPPER_LC_utf8_safe>.
954c1994 805
243effed 806=for apidoc Am|bool|isPRINT|char ch
8eea39dd 807Returns a boolean indicating whether the specified character is a
6aff1f14 808printable character, analogous to C<m/[[:print:]]/>.
dcccc8ff
KW
809See the L<top of this section|/Character classification> for an explanation of
810variants
da8c1a98
KW
811C<isPRINT_A>, C<isPRINT_L1>, C<isPRINT_uvchr>, C<isPRINT_utf8_safe>,
812C<isPRINT_LC>, C<isPRINT_LC_uvchr>, and C<isPRINT_LC_utf8_safe>.
243effed
KW
813
814=for apidoc Am|bool|isWORDCHAR|char ch
815Returns a boolean indicating whether the specified character is a character
816that is a word character, analogous to what C<m/\w/> and C<m/[[:word:]]/> match
817in a regular expression. A word character is an alphabetic character, a
818decimal digit, a connecting punctuation character (such as an underscore), or
819a "mark" character that attaches to one of those (like some sort of accent).
820C<isALNUM()> is a synonym provided for backward compatibility, even though a
821word character includes more than the standard C language meaning of
822alphanumeric.
dcccc8ff 823See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
824variants C<isWORDCHAR_A>, C<isWORDCHAR_L1>, C<isWORDCHAR_uvchr>, and
825C<isWORDCHAR_utf8_safe>. C<isWORDCHAR_LC>, C<isWORDCHAR_LC_uvchr>, and
826C<isWORDCHAR_LC_utf8_safe> are also as described there, but additionally
827include the platform's native underscore.
8a58bdcf
KW
828
829=for apidoc Am|bool|isXDIGIT|char ch
830Returns a boolean indicating whether the specified character is a hexadecimal
243effed
KW
831digit. In the ASCII range these are C<[0-9A-Fa-f]>. Variants C<isXDIGIT_A()>
832and C<isXDIGIT_L1()> are identical to C<isXDIGIT()>.
dcccc8ff
KW
833See the L<top of this section|/Character classification> for an explanation of
834variants
da8c1a98
KW
835C<isXDIGIT_uvchr>, C<isXDIGIT_utf8_safe>, C<isXDIGIT_LC>, C<isXDIGIT_LC_uvchr>,
836and C<isXDIGIT_LC_utf8_safe>.
243effed 837
3c3ecf18
KW
838=for apidoc Am|bool|isIDFIRST|char ch
839Returns a boolean indicating whether the specified character can be the first
840character of an identifier. This is very close to, but not quite the same as
841the official Unicode property C<XID_Start>. The difference is that this
842returns true only if the input character also matches L</isWORDCHAR>.
dcccc8ff
KW
843See the L<top of this section|/Character classification> for an explanation of
844variants
da8c1a98
KW
845C<isIDFIRST_A>, C<isIDFIRST_L1>, C<isIDFIRST_uvchr>, C<isIDFIRST_utf8_safe>,
846C<isIDFIRST_LC>, C<isIDFIRST_LC_uvchr>, and C<isIDFIRST_LC_utf8_safe>.
3c3ecf18
KW
847
848=for apidoc Am|bool|isIDCONT|char ch
849Returns a boolean indicating whether the specified character can be the
850second or succeeding character of an identifier. This is very close to, but
851not quite the same as the official Unicode property C<XID_Continue>. The
852difference is that this returns true only if the input character also matches
dcccc8ff
KW
853L</isWORDCHAR>. See the L<top of this section|/Character classification> for
854an
d0da05db 855explanation of variants C<isIDCONT_A>, C<isIDCONT_L1>, C<isIDCONT_uvchr>,
da8c1a98
KW
856C<isIDCONT_utf8_safe>, C<isIDCONT_LC>, C<isIDCONT_LC_uvchr>, and
857C<isIDCONT_LC_utf8_safe>.
3c3ecf18 858
243effed 859=head1 Miscellaneous Functions
8eea39dd 860
95a59cab 861=for apidoc Am|U8|READ_XDIGIT|char str*
243effed 862Returns the value of an ASCII-range hex digit and advances the string pointer.
95a59cab
YO
863Behaviour is only well defined when isXDIGIT(*str) is true.
864
e7c1e6c1 865=head1 Character case changing
21da7284
KW
866Perl uses "full" Unicode case mappings. This means that converting a single
867character to another case may result in a sequence of more than one character.
868For example, the uppercase of C<E<223>> (LATIN SMALL LETTER SHARP S) is the two
869character sequence C<SS>. This presents some complications The lowercase of
870all characters in the range 0..255 is a single character, and thus
871C<L</toLOWER_L1>> is furnished. But, C<toUPPER_L1> can't exist, as it couldn't
872return a valid result for all legal inputs. Instead C<L</toUPPER_uvchr>> has
873an API that does allow every possible legal result to be returned.) Likewise
874no other function that is crippled by not being able to give the correct
875results for the full range of possible inputs has been implemented here.
e7c1e6c1 876
1f607577
KW
877=for apidoc Am|U8|toUPPER|U8 ch
878Converts the specified character to uppercase. If the input is anything but an
879ASCII lowercase character, that input character itself is returned. Variant
c753c8d3 880C<toUPPER_A> is equivalent.
954c1994 881
d0da05db
KW
882=for apidoc Am|UV|toUPPER_uvchr|UV cp|U8* s|STRLEN* lenp
883Converts the code point C<cp> to its uppercase version, and
884stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
885point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
886that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
887bytes since the uppercase version may be longer than the original character.
888
889The first code point of the uppercased version is returned
21da7284
KW
890(but note, as explained at L<the top of this section|/Character case
891changing>, that there may be more.)
1f607577 892
a239b1e2
KW
893=for apidoc Am|UV|toUPPER_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
894Converts the first UTF-8 encoded character in the sequence starting at C<p> and
895extending no further than S<C<e - 1>> to its uppercase version, and
1f607577
KW
896stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
897that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
898bytes since the uppercase version may be longer than the original character.
899
900The first code point of the uppercased version is returned
21da7284
KW
901(but note, as explained at L<the top of this section|/Character case
902changing>, that there may be more).
1f607577 903
a239b1e2
KW
904The suffix C<_safe> in the function's name indicates that it will not attempt
905to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
906true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
907input character is malformed in some way, the program may croak, or the
908function may return the REPLACEMENT CHARACTER, at the discretion of the
909implementation, and subject to change in future releases.
910
911=for apidoc Am|UV|toUPPER_utf8|U8* p|U8* s|STRLEN* lenp
912This is like C<L</toUPPER_utf8_safe>>, but doesn't have the C<e>
913parameter The function therefore can't check if it is reading
607313a1
KW
914beyond the end of the string. Starting in Perl v5.30, it will take the C<e>
915parameter, becoming a synonym for C<toUPPER_utf8_safe>. At that time every
916program that uses it will have to be changed to successfully compile. In the
917meantime, the first runtime call to C<toUPPER_utf8> from each call point in the
918program will raise a deprecation warning, enabled by default. You can convert
919your program now to use C<toUPPER_utf8_safe>, and avoid the warnings, and get an
920extra measure of protection, or you can wait until v5.30, when you'll be forced
921to add the C<e> parameter.
1f607577 922
25200305
KW
923=for apidoc Am|U8|toFOLD|U8 ch
924Converts the specified character to foldcase. If the input is anything but an
925ASCII uppercase character, that input character itself is returned. Variant
926C<toFOLD_A> is equivalent. (There is no equivalent C<to_FOLD_L1> for the full
d0da05db 927Latin1 range, as the full generality of L</toFOLD_uvchr> is needed there.)
25200305 928
d0da05db
KW
929=for apidoc Am|UV|toFOLD_uvchr|UV cp|U8* s|STRLEN* lenp
930Converts the code point C<cp> to its foldcase version, and
931stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
932point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
933that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
934bytes since the foldcase version may be longer than the original character.
935
936The first code point of the foldcased version is returned
21da7284
KW
937(but note, as explained at L<the top of this section|/Character case
938changing>, that there may be more).
1f607577 939
a239b1e2
KW
940=for apidoc Am|UV|toFOLD_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
941Converts the first UTF-8 encoded character in the sequence starting at C<p> and
942extending no further than S<C<e - 1>> to its foldcase version, and
1f607577
KW
943stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
944that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
945bytes since the foldcase version may be longer than the original character.
946
947The first code point of the foldcased version is returned
21da7284
KW
948(but note, as explained at L<the top of this section|/Character case
949changing>, that there may be more).
1f607577 950
a239b1e2
KW
951The suffix C<_safe> in the function's name indicates that it will not attempt
952to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
953true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
954input character is malformed in some way, the program may croak, or the
955function may return the REPLACEMENT CHARACTER, at the discretion of the
956implementation, and subject to change in future releases.
957
958=for apidoc Am|UV|toFOLD_utf8|U8* p|U8* s|STRLEN* lenp
959This is like C<L</toFOLD_utf8_safe>>, but doesn't have the C<e>
960parameter The function therefore can't check if it is reading
607313a1
KW
961beyond the end of the string. Starting in Perl v5.30, it will take the C<e>
962parameter, becoming a synonym for C<toFOLD_utf8_safe>. At that time every
963program that uses it will have to be changed to successfully compile. In the
964meantime, the first runtime call to C<toFOLD_utf8> from each call point in the
965program will raise a deprecation warning, enabled by default. You can convert
966your program now to use C<toFOLD_utf8_safe>, and avoid the warnings, and get an
967extra measure of protection, or you can wait until v5.30, when you'll be forced
968to add the C<e> parameter.
1f607577
KW
969
970=for apidoc Am|U8|toLOWER|U8 ch
971Converts the specified character to lowercase. If the input is anything but an
972ASCII uppercase character, that input character itself is returned. Variant
c753c8d3 973C<toLOWER_A> is equivalent.
954c1994 974
1f607577 975=for apidoc Am|U8|toLOWER_L1|U8 ch
b7d90381
KW
976Converts the specified Latin1 character to lowercase. The results are
977undefined if the input doesn't fit in a byte.
1f607577
KW
978
979=for apidoc Am|U8|toLOWER_LC|U8 ch
980Converts the specified character to lowercase using the current locale's rules,
981if possible; otherwise returns the input character itself.
982
d0da05db
KW
983=for apidoc Am|UV|toLOWER_uvchr|UV cp|U8* s|STRLEN* lenp
984Converts the code point C<cp> to its lowercase version, and
985stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
986point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
987that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
988bytes since the lowercase version may be longer than the original character.
989
990The first code point of the lowercased version is returned
21da7284
KW
991(but note, as explained at L<the top of this section|/Character case
992changing>, that there may be more).
1f607577 993
a239b1e2
KW
994
995=for apidoc Am|UV|toLOWER_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
996Converts the first UTF-8 encoded character in the sequence starting at C<p> and
997extending no further than S<C<e - 1>> to its lowercase version, and
1f607577
KW
998stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
999that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
1000bytes since the lowercase version may be longer than the original character.
1001
1002The first code point of the lowercased version is returned
21da7284
KW
1003(but note, as explained at L<the top of this section|/Character case
1004changing>, that there may be more).
1f607577 1005
a239b1e2
KW
1006The suffix C<_safe> in the function's name indicates that it will not attempt
1007to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
1008true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
1009input character is malformed in some way, the program may croak, or the
1010function may return the REPLACEMENT CHARACTER, at the discretion of the
1011implementation, and subject to change in future releases.
1012
1013=for apidoc Am|UV|toLOWER_utf8|U8* p|U8* s|STRLEN* lenp
1014This is like C<L</toLOWER_utf8_safe>>, but doesn't have the C<e>
1015parameter The function therefore can't check if it is reading
607313a1
KW
1016beyond the end of the string. Starting in Perl v5.30, it will take the C<e>
1017parameter, becoming a synonym for C<toLOWER_utf8_safe>. At that time every
1018program that uses it will have to be changed to successfully compile. In the
1019meantime, the first runtime call to C<toLOWER_utf8> from each call point in the
1020program will raise a deprecation warning, enabled by default. You can convert
1021your program now to use C<toLOWER_utf8_safe>, and avoid the warnings, and get an
1022extra measure of protection, or you can wait until v5.30, when you'll be forced
1023to add the C<e> parameter.
1f607577 1024
25200305
KW
1025=for apidoc Am|U8|toTITLE|U8 ch
1026Converts the specified character to titlecase. If the input is anything but an
1027ASCII lowercase character, that input character itself is returned. Variant
b7d90381 1028C<toTITLE_A> is equivalent. (There is no C<toTITLE_L1> for the full Latin1
d0da05db 1029range, as the full generality of L</toTITLE_uvchr> is needed there. Titlecase is
b7d90381 1030not a concept used in locale handling, so there is no functionality for that.)
25200305 1031
d0da05db
KW
1032=for apidoc Am|UV|toTITLE_uvchr|UV cp|U8* s|STRLEN* lenp
1033Converts the code point C<cp> to its titlecase version, and
1034stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
1035point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
1036that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
1037bytes since the titlecase version may be longer than the original character.
1038
1039The first code point of the titlecased version is returned
21da7284
KW
1040(but note, as explained at L<the top of this section|/Character case
1041changing>, that there may be more).
1f607577 1042
a239b1e2
KW
1043=for apidoc Am|UV|toTITLE_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
1044Converts the first UTF-8 encoded character in the sequence starting at C<p> and
1045extending no further than S<C<e - 1>> to its titlecase version, and
1f607577
KW
1046stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
1047that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
1048bytes since the titlecase version may be longer than the original character.
1049
1050The first code point of the titlecased version is returned
21da7284
KW
1051(but note, as explained at L<the top of this section|/Character case
1052changing>, that there may be more).
1f607577 1053
a239b1e2
KW
1054The suffix C<_safe> in the function's name indicates that it will not attempt
1055to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
1056true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
1057input character is malformed in some way, the program may croak, or the
1058function may return the REPLACEMENT CHARACTER, at the discretion of the
1059implementation, and subject to change in future releases.
1060
1061=for apidoc Am|UV|toTITLE_utf8|U8* p|U8* s|STRLEN* lenp
1062This is like C<L</toLOWER_utf8_safe>>, but doesn't have the C<e>
1063parameter The function therefore can't check if it is reading
607313a1
KW
1064beyond the end of the string. Starting in Perl v5.30, it will take the C<e>
1065parameter, becoming a synonym for C<toTITLE_utf8_safe>. At that time every
1066program that uses it will have to be changed to successfully compile. In the
1067meantime, the first runtime call to C<toTITLE_utf8> from each call point in the
1068program will raise a deprecation warning, enabled by default. You can convert
1069your program now to use C<toTITLE_utf8_safe>, and avoid the warnings, and get an
1070extra measure of protection, or you can wait until v5.30, when you'll be forced
1071to add the C<e> parameter.
1f607577 1072
954c1994 1073=cut
353c9b6f 1074
d0da05db 1075XXX Still undocumented isVERTWS_uvchr and _utf8; it's unclear what their names
1e222e4f
KW
1076really should be. Also toUPPER_LC and toFOLD_LC, which are subject to change,
1077and aren't general purpose as they don't work on U+DF, and assert against that.
243effed 1078
8a58bdcf 1079Note that these macros are repeated in Devel::PPPort, so should also be
62fa66b6
KW
1080patched there. The file as of this writing is cpan/Devel-PPPort/parts/inc/misc
1081
954c1994
GS
1082*/
1083
7c062697
KW
1084/* Specify the widest unsigned type on the platform. Use U64TYPE because U64
1085 * is known only in the perl core, and this macro can be called from outside
1086 * that */
1087#ifdef HAS_QUAD
1088# define WIDEST_UTYPE U64TYPE
1089#else
1090# define WIDEST_UTYPE U32
1091#endif
1092
3912bc88
KW
1093/* FITS_IN_8_BITS(c) returns true if c doesn't have a bit set other than in
1094 * the lower 8. It is designed to be hopefully bomb-proof, making sure that no
1095 * bits of information are lost even on a 64-bit machine, but to get the
1096 * compiler to optimize it out if possible. This is because Configure makes
1097 * sure that the machine has an 8-bit byte, so if c is stored in a byte, the
1098 * sizeof() guarantees that this evaluates to a constant true at compile time.
7e75d1a1
JH
1099 *
1100 * For Coverity, be always true, because otherwise Coverity thinks
1101 * it finds several expressions that are always true, independent
1102 * of operands. Well, they are, but that is kind of the point.
220c71bf 1103 */
7e75d1a1 1104#ifndef __COVERITY__
6c5b02ac
KW
1105 /* The '| 0' part ensures a compiler error if c is not integer (like e.g., a
1106 * pointer) */
1107#define FITS_IN_8_BITS(c) ( (sizeof(c) == 1) \
ace3ad0f 1108 || !(((WIDEST_UTYPE)((c) | 0)) & ~0xFF))
7e75d1a1
JH
1109#else
1110#define FITS_IN_8_BITS(c) (1)
1111#endif
cf301eb7 1112
41f43cc2 1113#ifdef EBCDIC
b6340bd0 1114# ifndef _ALL_SOURCE
0852beac
KW
1115 /* The native libc isascii() et.al. functions return the wrong results
1116 * on at least z/OS unless this is defined. */
b6340bd0
KW
1117# error _ALL_SOURCE should probably be defined
1118# endif
41f43cc2 1119#else
0852beac
KW
1120 /* There is a simple definition of ASCII for ASCII platforms. But the
1121 * EBCDIC one isn't so simple, so is defined using table look-up like the
9c903d59 1122 * other macros below.
3f3c579d
KW
1123 *
1124 * The cast here is used instead of '(c) >= 0', because some compilers emit
1125 * a warning that that test is always true when the parameter is an
1126 * unsigned type. khw supposes that it could be written as
1127 * && ((c) == '\0' || (c) > 0)
1128 * to avoid the message, but the cast will likely avoid extra branches even
1129 * with stupid compilers.
1130 *
1131 * The '| 0' part ensures a compiler error if c is not integer (like e.g.,
1132 * a pointer) */
9c903d59 1133# define isASCII(c) ((WIDEST_UTYPE)((c) | 0) < 128)
41f43cc2
KW
1134#endif
1135
38694112
KW
1136/* Take the eight possible bit patterns of the lower 3 bits and you get the
1137 * lower 3 bits of the 8 octal digits, in both ASCII and EBCDIC, so those bits
1138 * can be ignored. If the rest match '0', we have an octal */
1139#define isOCTAL_A(c) (((WIDEST_UTYPE)((c) | 0) & ~7) == '0')
c2da0b36 1140
9fb1bf9d 1141#ifdef H_PERL /* If have access to perl.h, lookup in its table */
f4cdb42c 1142
a500dc72
KW
1143/* Character class numbers. For internal core Perl use only. The ones less
1144 * than 32 are used in PL_charclass[] and the ones up through the one that
1145 * corresponds to <_HIGHEST_REGCOMP_DOT_H_SYNC> are used by regcomp.h and
1146 * related files. PL_charclass ones use names used in l1_char_class_tab.h but
1147 * their actual definitions are here. If that file has a name not used here,
1148 * it won't compile.
1709d539
KW
1149 *
1150 * The first group of these is ordered in what I (khw) estimate to be the
31c7f561
KW
1151 * frequency of their use. This gives a slight edge to exiting a loop earlier
1152 * (in reginclass() in regexec.c) */
1709d539
KW
1153# define _CC_WORDCHAR 0 /* \w and [:word:] */
1154# define _CC_DIGIT 1 /* \d and [:digit:] */
1155# define _CC_ALPHA 2 /* [:alpha:] */
1156# define _CC_LOWER 3 /* [:lower:] */
1157# define _CC_UPPER 4 /* [:upper:] */
1158# define _CC_PUNCT 5 /* [:punct:] */
1159# define _CC_PRINT 6 /* [:print:] */
15861f94 1160# define _CC_ALPHANUMERIC 7 /* [:alnum:] */
1709d539 1161# define _CC_GRAPH 8 /* [:graph:] */
359b005e 1162# define _CC_CASED 9 /* [:lower:] or [:upper:] under /i */
1709d539 1163
b0d691b2 1164#define _FIRST_NON_SWASH_CC 10
1709d539
KW
1165/* The character classes above are implemented with swashes. The second group
1166 * (just below) contains the ones implemented without. These are also sorted
1167 * in rough order of the frequency of their use, except that \v should be last,
1168 * as it isn't a real Posix character class, and some (small) inefficiencies in
1169 * regular expression handling would be introduced by putting it in the middle
1170 * of those that are. Also, cntrl and ascii come after the others as it may be
1171 * useful to group these which have no members that match above Latin1, (or
1172 * above ASCII in the latter case) */
1173
779cf272 1174# define _CC_SPACE 10 /* \s, [:space:] */
34aeb2e9
KW
1175# define _CC_PSXSPC _CC_SPACE /* XXX Temporary, can be removed
1176 when the deprecated isFOO_utf8()
1177 functions are removed */
b0d691b2
KW
1178# define _CC_BLANK 11 /* [:blank:] */
1179# define _CC_XDIGIT 12 /* [:xdigit:] */
779cf272
KW
1180# define _CC_CNTRL 13 /* [:cntrl:] */
1181# define _CC_ASCII 14 /* [:ascii:] */
1182# define _CC_VERTSPACE 15 /* \v */
1709d539 1183
a0947d7b
KW
1184# define _HIGHEST_REGCOMP_DOT_H_SYNC _CC_VERTSPACE
1185
1709d539 1186/* The members of the third group below do not need to be coordinated with data
3ffc8c70 1187 * structures in regcomp.[ch] and regexec.c. */
779cf272
KW
1188# define _CC_IDFIRST 16
1189# define _CC_CHARNAME_CONT 17
1190# define _CC_NONLATIN1_FOLD 18
1191# define _CC_NONLATIN1_SIMPLE_FOLD 19
1192# define _CC_QUOTEMETA 20
1193# define _CC_NON_FINAL_FOLD 21
1194# define _CC_IS_IN_SOME_FOLD 22
1195# define _CC_MNEMONIC_CNTRL 23
073c22b3 1196
34aeb2e9
KW
1197# define _CC_IDCONT 24 /* XXX Temporary, can be removed when the deprecated
1198 isFOO_utf8() functions are removed */
1199
073c22b3
KW
1200/* This next group is only used on EBCDIC platforms, so theoretically could be
1201 * shared with something entirely different that's only on ASCII platforms */
5d5376e2
KW
1202# define _CC_UTF8_START_BYTE_IS_FOR_AT_LEAST_SURROGATE 28
1203# define _CC_UTF8_IS_START 29
1204# define _CC_UTF8_IS_DOWNGRADEABLE_START 30
1205# define _CC_UTF8_IS_CONTINUATION 31
1206/* Unused: 24-27
f4cdb42c
KW
1207 * If more bits are needed, one could add a second word for non-64bit
1208 * QUAD_IS_INT systems, using some #ifdefs to distinguish between having a 2nd
37ede926
KW
1209 * word or not. The IS_IN_SOME_FOLD bit is the most easily expendable, as it
1210 * is used only for optimization (as of this writing), and differs in the
1211 * Latin1 range from the ALPHA bit only in two relatively unimportant
a500dc72 1212 * characters: the masculine and feminine ordinal indicators, so removing it
073c22b3
KW
1213 * would just cause /i regexes which match them to run less efficiently.
1214 * Similarly the EBCDIC-only bits are used just for speed, and could be
1215 * replaced by other means */
96ac0975 1216
3a371f2f
KW
1217#if defined(PERL_CORE) || defined(PERL_EXT)
1218/* An enum version of the character class numbers, to help compilers
1219 * optimize */
1220typedef enum {
3a371f2f 1221 _CC_ENUM_ALPHA = _CC_ALPHA,
e8d596e0
KW
1222 _CC_ENUM_ALPHANUMERIC = _CC_ALPHANUMERIC,
1223 _CC_ENUM_ASCII = _CC_ASCII,
1224 _CC_ENUM_BLANK = _CC_BLANK,
b0d691b2 1225 _CC_ENUM_CASED = _CC_CASED,
e8d596e0 1226 _CC_ENUM_CNTRL = _CC_CNTRL,
3a371f2f
KW
1227 _CC_ENUM_DIGIT = _CC_DIGIT,
1228 _CC_ENUM_GRAPH = _CC_GRAPH,
1229 _CC_ENUM_LOWER = _CC_LOWER,
1230 _CC_ENUM_PRINT = _CC_PRINT,
1231 _CC_ENUM_PUNCT = _CC_PUNCT,
e8d596e0 1232 _CC_ENUM_SPACE = _CC_SPACE,
3a371f2f 1233 _CC_ENUM_UPPER = _CC_UPPER,
e8d596e0 1234 _CC_ENUM_VERTSPACE = _CC_VERTSPACE,
3a371f2f 1235 _CC_ENUM_WORDCHAR = _CC_WORDCHAR,
e8d596e0 1236 _CC_ENUM_XDIGIT = _CC_XDIGIT
3a371f2f
KW
1237} _char_class_number;
1238#endif
1239
63c61c3f 1240#define POSIX_SWASH_COUNT _FIRST_NON_SWASH_CC
86f72d56 1241#define POSIX_CC_COUNT (_HIGHEST_REGCOMP_DOT_H_SYNC + 1)
63c61c3f 1242
b7d90381
KW
1243#if defined(PERL_IN_UTF8_C) \
1244 || defined(PERL_IN_REGCOMP_C) \
1245 || defined(PERL_IN_REGEXEC_C)
922e8cb4 1246# if _CC_WORDCHAR != 0 || _CC_DIGIT != 1 || _CC_ALPHA != 2 || _CC_LOWER != 3 \
b0d691b2
KW
1247 || _CC_UPPER != 4 || _CC_PUNCT != 5 || _CC_PRINT != 6 \
1248 || _CC_ALPHANUMERIC != 7 || _CC_GRAPH != 8 || _CC_CASED != 9
922e8cb4
KW
1249 #error Need to adjust order of swash_property_names[]
1250# endif
1251
b44daeeb
KW
1252/* This is declared static in each of the few files that this is #defined for
1253 * to keep them from being publicly accessible. Hence there is a small amount
1254 * of wasted space */
1255
922e8cb4
KW
1256static const char* const swash_property_names[] = {
1257 "XPosixWord",
1258 "XPosixDigit",
1259 "XPosixAlpha",
1260 "XPosixLower",
1261 "XPosixUpper",
1262 "XPosixPunct",
1263 "XPosixPrint",
1264 "XPosixAlnum",
b0d691b2
KW
1265 "XPosixGraph",
1266 "Cased"
922e8cb4
KW
1267};
1268#endif
1269
6635f04f 1270START_EXTERN_C
96ac0975
NC
1271# ifdef DOINIT
1272EXTCONST U32 PL_charclass[] = {
1273# include "l1_char_class_tab.h"
1274};
1275
1276# else /* ! DOINIT */
1277EXTCONST U32 PL_charclass[];
1278# endif
6635f04f 1279END_EXTERN_C
96ac0975 1280
265c1f46 1281 /* The 1U keeps Solaris from griping when shifting sets the uppermost bit */
430b7c70 1282# define _CC_mask(classnum) (1U << (classnum))
4650c663
KW
1283
1284 /* For internal core Perl use only: the base macro for defining macros like
1285 * isALPHA */
ff7ecfc3 1286# define _generic_isCC(c, classnum) cBOOL(FITS_IN_8_BITS(c) \
f4cd282c 1287 && (PL_charclass[(U8) (c)] & _CC_mask(classnum)))
4eeeb416 1288
f4cdb42c
KW
1289 /* The mask for the _A versions of the macros; it just adds in the bit for
1290 * ASCII. */
1291# define _CC_mask_A(classnum) (_CC_mask(classnum) | _CC_mask(_CC_ASCII))
1292
4650c663
KW
1293 /* For internal core Perl use only: the base macro for defining macros like
1294 * isALPHA_A. The foo_A version makes sure that both the desired bit and
1295 * the ASCII bit are present */
b7d90381
KW
1296# define _generic_isCC_A(c, classnum) (FITS_IN_8_BITS(c) \
1297 && ((PL_charclass[(U8) (c)] & _CC_mask_A(classnum)) \
1298 == _CC_mask_A(classnum)))
f4cdb42c 1299
f4cdb42c 1300# define isALPHA_A(c) _generic_isCC_A(c, _CC_ALPHA)
15861f94 1301# define isALPHANUMERIC_A(c) _generic_isCC_A(c, _CC_ALPHANUMERIC)
f4cdb42c
KW
1302# define isBLANK_A(c) _generic_isCC_A(c, _CC_BLANK)
1303# define isCNTRL_A(c) _generic_isCC_A(c, _CC_CNTRL)
96ca48da 1304# define isDIGIT_A(c) _generic_isCC(c, _CC_DIGIT) /* No non-ASCII digits */
f4cdb42c
KW
1305# define isGRAPH_A(c) _generic_isCC_A(c, _CC_GRAPH)
1306# define isLOWER_A(c) _generic_isCC_A(c, _CC_LOWER)
1307# define isPRINT_A(c) _generic_isCC_A(c, _CC_PRINT)
f4cdb42c
KW
1308# define isPUNCT_A(c) _generic_isCC_A(c, _CC_PUNCT)
1309# define isSPACE_A(c) _generic_isCC_A(c, _CC_SPACE)
1310# define isUPPER_A(c) _generic_isCC_A(c, _CC_UPPER)
1311# define isWORDCHAR_A(c) _generic_isCC_A(c, _CC_WORDCHAR)
b7d90381
KW
1312# define isXDIGIT_A(c) _generic_isCC(c, _CC_XDIGIT) /* No non-ASCII xdigits
1313 */
d95f8b6a 1314# define isIDFIRST_A(c) _generic_isCC_A(c, _CC_IDFIRST)
3ded5eb0
KW
1315# define isALPHA_L1(c) _generic_isCC(c, _CC_ALPHA)
1316# define isALPHANUMERIC_L1(c) _generic_isCC(c, _CC_ALPHANUMERIC)
1317# define isBLANK_L1(c) _generic_isCC(c, _CC_BLANK)
1318
1319 /* continuation character for legal NAME in \N{NAME} */
1320# define isCHARNAME_CONT(c) _generic_isCC(c, _CC_CHARNAME_CONT)
1321
1322# define isCNTRL_L1(c) _generic_isCC(c, _CC_CNTRL)
1323# define isGRAPH_L1(c) _generic_isCC(c, _CC_GRAPH)
1324# define isLOWER_L1(c) _generic_isCC(c, _CC_LOWER)
1325# define isPRINT_L1(c) _generic_isCC(c, _CC_PRINT)
b7d90381 1326# define isPSXSPC_L1(c) isSPACE_L1(c)
3ded5eb0
KW
1327# define isPUNCT_L1(c) _generic_isCC(c, _CC_PUNCT)
1328# define isSPACE_L1(c) _generic_isCC(c, _CC_SPACE)
1329# define isUPPER_L1(c) _generic_isCC(c, _CC_UPPER)
1330# define isWORDCHAR_L1(c) _generic_isCC(c, _CC_WORDCHAR)
1331# define isIDFIRST_L1(c) _generic_isCC(c, _CC_IDFIRST)
f4cdb42c 1332
0852beac
KW
1333# ifdef EBCDIC
1334# define isASCII(c) _generic_isCC(c, _CC_ASCII)
1335# endif
1336
f12c0118
KW
1337 /* Participates in a single-character fold with a character above 255 */
1338# define _HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) ((! cBOOL(FITS_IN_8_BITS(c))) || (PL_charclass[(U8) (c)] & _CC_mask(_CC_NONLATIN1_SIMPLE_FOLD)))
1339
1340 /* Like the above, but also can be part of a multi-char fold */
f4cd282c 1341# define _HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) ((! cBOOL(FITS_IN_8_BITS(c))) || (PL_charclass[(U8) (c)] & _CC_mask(_CC_NONLATIN1_FOLD)))
430b7c70 1342
4eeeb416 1343# define _isQUOTEMETA(c) _generic_isCC(c, _CC_QUOTEMETA)
26faadbd 1344# define _IS_NON_FINAL_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
b7d90381 1345 _generic_isCC(c, _CC_NON_FINAL_FOLD)
37ede926 1346# define _IS_IN_SOME_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
b7d90381 1347 _generic_isCC(c, _CC_IS_IN_SOME_FOLD)
8e35b056
KW
1348# define _IS_MNEMONIC_CNTRL_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
1349 _generic_isCC(c, _CC_MNEMONIC_CNTRL)
687c8d01 1350#else /* else we don't have perl.h H_PERL */
3ded5eb0
KW
1351
1352 /* If we don't have perl.h, we are compiling a utility program. Below we
1353 * hard-code various macro definitions that wouldn't otherwise be available
fc273927 1354 * to it. Most are coded based on first principles. These are written to
74665a89 1355 * avoid EBCDIC vs. ASCII #ifdef's as much as possible. */
0852beac
KW
1356# define isDIGIT_A(c) ((c) <= '9' && (c) >= '0')
1357# define isBLANK_A(c) ((c) == ' ' || (c) == '\t')
74665a89
KW
1358# define isSPACE_A(c) (isBLANK_A(c) \
1359 || (c) == '\n' \
1360 || (c) == '\r' \
1361 || (c) == '\v' \
0852beac 1362 || (c) == '\f')
74665a89
KW
1363 /* On EBCDIC, there are gaps between 'i' and 'j'; 'r' and 's'. Same for
1364 * uppercase. The tests for those aren't necessary on ASCII, but hurt only
1365 * performance (if optimization isn't on), and allow the same code to be
1366 * used for both platform types */
1367# define isLOWER_A(c) ((c) >= 'a' && (c) <= 'z' \
1368 && ( (c) <= 'i' \
1369 || ((c) >= 'j' && (c) <= 'r') \
1370 || (c) >= 's'))
1371# define isUPPER_A(c) ((c) >= 'A' && (c) <= 'Z' \
1372 && ( (c) <= 'I' \
1373 || ((c) >= 'J' && (c) <= 'R') \
1374 || (c) >= 'S'))
a4d7a999
KW
1375# define isALPHA_A(c) (isUPPER_A(c) || isLOWER_A(c))
1376# define isALPHANUMERIC_A(c) (isALPHA_A(c) || isDIGIT_A(c))
3ded5eb0 1377# define isWORDCHAR_A(c) (isALPHANUMERIC_A(c) || (c) == '_')
0852beac 1378# define isIDFIRST_A(c) (isALPHA_A(c) || (c) == '_')
74665a89
KW
1379# define isXDIGIT_A(c) (isDIGIT_A(c) \
1380 || ((c) >= 'a' && (c) <= 'f') \
a4d7a999 1381 || ((c) <= 'F' && (c) >= 'A'))
74665a89
KW
1382# define isPUNCT_A(c) ((c) == '-' || (c) == '!' || (c) == '"' \
1383 || (c) == '#' || (c) == '$' || (c) == '%' \
1384 || (c) == '&' || (c) == '\'' || (c) == '(' \
1385 || (c) == ')' || (c) == '*' || (c) == '+' \
1386 || (c) == ',' || (c) == '.' || (c) == '/' \
1387 || (c) == ':' || (c) == ';' || (c) == '<' \
1388 || (c) == '=' || (c) == '>' || (c) == '?' \
1389 || (c) == '@' || (c) == '[' || (c) == '\\' \
1390 || (c) == ']' || (c) == '^' || (c) == '_' \
1391 || (c) == '`' || (c) == '{' || (c) == '|' \
1392 || (c) == '}' || (c) == '~')
1393# define isGRAPH_A(c) (isALPHANUMERIC_A(c) || isPUNCT_A(c))
1394# define isPRINT_A(c) (isGRAPH_A(c) || (c) == ' ')
3ded5eb0 1395
0852beac 1396# ifdef EBCDIC
74665a89
KW
1397 /* The below is accurate for the 3 EBCDIC code pages traditionally
1398 * supported by perl. The only difference between them in the controls
1399 * is the position of \n, and that is represented symbolically below */
1400# define isCNTRL_A(c) ((c) == '\0' || (c) == '\a' || (c) == '\b' \
1401 || (c) == '\f' || (c) == '\n' || (c) == '\r' \
1402 || (c) == '\t' || (c) == '\v' \
1403 || ((c) <= 3 && (c) >= 1) /* SOH, STX, ETX */ \
1404 || (c) == 7 /* U+7F DEL */ \
1405 || ((c) <= 0x13 && (c) >= 0x0E) /* SO, SI */ \
1406 /* DLE, DC[1-3] */ \
1407 || (c) == 0x18 /* U+18 CAN */ \
1408 || (c) == 0x19 /* U+19 EOM */ \
1409 || ((c) <= 0x1F && (c) >= 0x1C) /* [FGRU]S */ \
1410 || (c) == 0x26 /* U+17 ETB */ \
1411 || (c) == 0x27 /* U+1B ESC */ \
1412 || (c) == 0x2D /* U+05 ENQ */ \
1413 || (c) == 0x2E /* U+06 ACK */ \
1414 || (c) == 0x32 /* U+16 SYN */ \
1415 || (c) == 0x37 /* U+04 EOT */ \
1416 || (c) == 0x3C /* U+14 DC4 */ \
1417 || (c) == 0x3D /* U+15 NAK */ \
1418 || (c) == 0x3F)/* U+1A SUB */
0852beac 1419# define isASCII(c) (isCNTRL_A(c) || isPRINT_A(c))
74665a89
KW
1420# else /* isASCII is already defined for ASCII platforms, so can use that to
1421 define isCNTRL */
1422# define isCNTRL_A(c) (isASCII(c) && ! isPRINT_A(c))
0852beac
KW
1423# endif
1424
3ffc8c70 1425 /* The _L1 macros may be unnecessary for the utilities; I (khw) added them
caa94d35
KW
1426 * during debugging, and it seems best to keep them. We may be called
1427 * without NATIVE_TO_LATIN1 being defined. On ASCII platforms, it doesn't
1428 * do anything anyway, so make it not a problem */
1429# if ! defined(EBCDIC) && ! defined(NATIVE_TO_LATIN1)
1430# define NATIVE_TO_LATIN1(ch) (ch)
1431# endif
3ded5eb0
KW
1432# define isALPHA_L1(c) (isUPPER_L1(c) || isLOWER_L1(c))
1433# define isALPHANUMERIC_L1(c) (isALPHA_L1(c) || isDIGIT_A(c))
1434# define isBLANK_L1(c) (isBLANK_A(c) \
1435 || (FITS_IN_8_BITS(c) \
1436 && NATIVE_TO_LATIN1((U8) c) == 0xA0))
1437# define isCNTRL_L1(c) (FITS_IN_8_BITS(c) && (! isPRINT_L1(c)))
1438# define isGRAPH_L1(c) (isPRINT_L1(c) && (! isBLANK_L1(c)))
1439# define isLOWER_L1(c) (isLOWER_A(c) \
1440 || (FITS_IN_8_BITS(c) \
ae683a5f 1441 && (( NATIVE_TO_LATIN1((U8) c) >= 0xDF \
3ded5eb0
KW
1442 && NATIVE_TO_LATIN1((U8) c) != 0xF7) \
1443 || NATIVE_TO_LATIN1((U8) c) == 0xAA \
1444 || NATIVE_TO_LATIN1((U8) c) == 0xBA \
1445 || NATIVE_TO_LATIN1((U8) c) == 0xB5)))
1446# define isPRINT_L1(c) (isPRINT_A(c) \
1447 || (FITS_IN_8_BITS(c) \
1448 && NATIVE_TO_LATIN1((U8) c) >= 0xA0))
3ded5eb0
KW
1449# define isPUNCT_L1(c) (isPUNCT_A(c) \
1450 || (FITS_IN_8_BITS(c) \
ae683a5f 1451 && ( NATIVE_TO_LATIN1((U8) c) == 0xA1 \
3ded5eb0
KW
1452 || NATIVE_TO_LATIN1((U8) c) == 0xA7 \
1453 || NATIVE_TO_LATIN1((U8) c) == 0xAB \
1454 || NATIVE_TO_LATIN1((U8) c) == 0xB6 \
1455 || NATIVE_TO_LATIN1((U8) c) == 0xB7 \
1456 || NATIVE_TO_LATIN1((U8) c) == 0xBB \
1457 || NATIVE_TO_LATIN1((U8) c) == 0xBF)))
1458# define isSPACE_L1(c) (isSPACE_A(c) \
1459 || (FITS_IN_8_BITS(c) \
ae683a5f 1460 && ( NATIVE_TO_LATIN1((U8) c) == 0x85 \
3ded5eb0
KW
1461 || NATIVE_TO_LATIN1((U8) c) == 0xA0)))
1462# define isUPPER_L1(c) (isUPPER_A(c) \
1463 || (FITS_IN_8_BITS(c) \
ae683a5f 1464 && ( NATIVE_TO_LATIN1((U8) c) >= 0xC0 \
3ded5eb0
KW
1465 && NATIVE_TO_LATIN1((U8) c) <= 0xDE \
1466 && NATIVE_TO_LATIN1((U8) c) != 0xD7)))
1467# define isWORDCHAR_L1(c) (isIDFIRST_L1(c) || isDIGIT_A(c))
1468# define isIDFIRST_L1(c) (isALPHA_L1(c) || NATIVE_TO_LATIN1(c) == '_')
1469# define isCHARNAME_CONT(c) (isWORDCHAR_L1(c) \
1470 || isBLANK_L1(c) \
1471 || (c) == '-' \
1472 || (c) == '(' \
1473 || (c) == ')')
1474 /* The following are not fully accurate in the above-ASCII range. I (khw)
1475 * don't think it's necessary to be so for the purposes where this gets
1476 * compiled */
1477# define _isQUOTEMETA(c) (FITS_IN_8_BITS(c) && ! isWORDCHAR_L1(c))
1478# define _IS_IN_SOME_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) isALPHA_L1(c)
1479
1480 /* And these aren't accurate at all. They are useful only for above
1481 * Latin1, which utilities and bootstrapping don't deal with */
1482# define _IS_NON_FINAL_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) 0
6838b41e 1483# define _HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) 0
3ded5eb0
KW
1484# define _HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) 0
1485
1486 /* Many of the macros later in this file are defined in terms of these. By
1487 * implementing them with a function, which converts the class number into
1488 * a call to the desired macro, all of the later ones work. However, that
1489 * function won't be actually defined when building a utility program (no
1490 * perl.h), and so a compiler error will be generated if one is attempted
1491 * to be used. And the above-Latin1 code points require Unicode tables to
1492 * be present, something unlikely to be the case when bootstrapping */
1493# define _generic_isCC(c, classnum) \
1494 (FITS_IN_8_BITS(c) && S_bootstrap_ctype((U8) (c), (classnum), TRUE))
1495# define _generic_isCC_A(c, classnum) \
1496 (FITS_IN_8_BITS(c) && S_bootstrap_ctype((U8) (c), (classnum), FALSE))
687c8d01 1497#endif /* End of no perl.h H_PERL */
8a58bdcf 1498
e66b99e9
KW
1499#define isALPHANUMERIC(c) isALPHANUMERIC_A(c)
1500#define isALPHA(c) isALPHA_A(c)
0852beac
KW
1501#define isASCII_A(c) isASCII(c)
1502#define isASCII_L1(c) isASCII(c)
e66b99e9
KW
1503#define isBLANK(c) isBLANK_A(c)
1504#define isCNTRL(c) isCNTRL_A(c)
1505#define isDIGIT(c) isDIGIT_A(c)
1506#define isGRAPH(c) isGRAPH_A(c)
1507#define isIDFIRST(c) isIDFIRST_A(c)
1508#define isLOWER(c) isLOWER_A(c)
1509#define isPRINT(c) isPRINT_A(c)
779cf272 1510#define isPSXSPC_A(c) isSPACE_A(c)
e66b99e9 1511#define isPSXSPC(c) isPSXSPC_A(c)
779cf272 1512#define isPSXSPC_L1(c) isSPACE_L1(c)
e66b99e9
KW
1513#define isPUNCT(c) isPUNCT_A(c)
1514#define isSPACE(c) isSPACE_A(c)
1515#define isUPPER(c) isUPPER_A(c)
1516#define isWORDCHAR(c) isWORDCHAR_A(c)
1517#define isXDIGIT(c) isXDIGIT_A(c)
1518
1519/* ASCII casing. These could also be written as
1520 #define toLOWER(c) (isASCII(c) ? toLOWER_LATIN1(c) : (c))
1521 #define toUPPER(c) (isASCII(c) ? toUPPER_LATIN1_MOD(c) : (c))
1522 which uses table lookup and mask instead of subtraction. (This would
1523 work because the _MOD does not apply in the ASCII range) */
68067e4e
DM
1524#define toLOWER(c) (isUPPER(c) ? (U8)((c) + ('a' - 'A')) : (c))
1525#define toUPPER(c) (isLOWER(c) ? (U8)((c) - ('a' - 'A')) : (c))
bbce6d69 1526
25200305
KW
1527/* In the ASCII range, these are equivalent to what they're here defined to be.
1528 * But by creating these definitions, other code doesn't have to be aware of
1529 * this detail */
1530#define toFOLD(c) toLOWER(c)
25200305
KW
1531#define toTITLE(c) toUPPER(c)
1532
c753c8d3
KW
1533#define toLOWER_A(c) toLOWER(c)
1534#define toUPPER_A(c) toUPPER(c)
25200305
KW
1535#define toFOLD_A(c) toFOLD(c)
1536#define toTITLE_A(c) toTITLE(c)
1a0901db 1537
4650c663 1538/* Use table lookup for speed; returns the input itself if is out-of-range */
b2bf251f 1539#define toLOWER_LATIN1(c) ((! FITS_IN_8_BITS(c)) \
8e7c6e7d 1540 ? (c) \
f4cd282c 1541 : PL_latin1_lc[ (U8) (c) ])
c753c8d3
KW
1542#define toLOWER_L1(c) toLOWER_LATIN1(c) /* Synonym for consistency */
1543
1a0901db 1544/* Modified uc. Is correct uc except for three non-ascii chars which are
4650c663
KW
1545 * all mapped to one of them, and these need special handling; returns the
1546 * input itself if is out-of-range */
b2bf251f 1547#define toUPPER_LATIN1_MOD(c) ((! FITS_IN_8_BITS(c)) \
8e7c6e7d 1548 ? (c) \
f4cd282c 1549 : PL_mod_latin1_uc[ (U8) (c) ])
31f05a37 1550#define IN_UTF8_CTYPE_LOCALE PL_in_utf8_CTYPE_locale
84061b6a 1551
beab9ebe
KW
1552/* Use foo_LC_uvchr() instead of these for beyond the Latin1 range */
1553
1554/* For internal core Perl use only: the base macro for defining macros like
1555 * isALPHA_LC, which uses the current LC_CTYPE locale. 'c' is the code point
31f05a37
KW
1556 * (0-255) to check. In a UTF-8 locale, the result is the same as calling
1557 * isFOO_L1(); the 'utf8_locale_classnum' parameter is something like
1558 * _CC_UPPER, which gives the class number for doing this. For non-UTF-8
1559 * locales, the code to actually do the test this is passed in 'non_utf8'. If
1560 * 'c' is above 255, 0 is returned. For accessing the full range of possible
1561 * code points under locale rules, use the macros based on _generic_LC_uvchr
1562 * instead of this. */
beab9ebe
KW
1563#define _generic_LC_base(c, utf8_locale_classnum, non_utf8) \
1564 (! FITS_IN_8_BITS(c) \
1565 ? 0 \
31f05a37
KW
1566 : IN_UTF8_CTYPE_LOCALE \
1567 ? cBOOL(PL_charclass[(U8) (c)] & _CC_mask(utf8_locale_classnum)) \
beab9ebe
KW
1568 : cBOOL(non_utf8))
1569
1570/* For internal core Perl use only: a helper macro for defining macros like
1571 * isALPHA_LC. 'c' is the code point (0-255) to check. The function name to
1572 * actually do this test is passed in 'non_utf8_func', which is called on 'c',
1573 * casting 'c' to the macro _LC_CAST, which should not be parenthesized. See
1574 * _generic_LC_base for more info */
1575#define _generic_LC(c, utf8_locale_classnum, non_utf8_func) \
1576 _generic_LC_base(c,utf8_locale_classnum, \
1577 non_utf8_func( (_LC_CAST) (c)))
1578
1579/* For internal core Perl use only: like _generic_LC, but also returns TRUE if
1580 * 'c' is the platform's native underscore character */
1581#define _generic_LC_underscore(c,utf8_locale_classnum,non_utf8_func) \
1582 _generic_LC_base(c, utf8_locale_classnum, \
1583 (non_utf8_func( (_LC_CAST) (c)) \
1584 || (char)(c) == '_'))
1585
1586/* These next three are also for internal core Perl use only: case-change
1587 * helper macros */
1588#define _generic_toLOWER_LC(c, function, cast) (! FITS_IN_8_BITS(c) \
1589 ? (c) \
31f05a37
KW
1590 : (IN_UTF8_CTYPE_LOCALE) \
1591 ? PL_latin1_lc[ (U8) (c) ] \
d2a8cf7f 1592 : (cast)function((cast)(c)))
beab9ebe 1593
31f05a37
KW
1594/* Note that the result can be larger than a byte in a UTF-8 locale. It
1595 * returns a single value, so can't adequately return the upper case of LATIN
1596 * SMALL LETTER SHARP S in a UTF-8 locale (which should be a string of two
1597 * values "SS"); instead it asserts against that under DEBUGGING, and
1598 * otherwise returns its input */
beab9ebe
KW
1599#define _generic_toUPPER_LC(c, function, cast) \
1600 (! FITS_IN_8_BITS(c) \
1601 ? (c) \
31f05a37 1602 : ((! IN_UTF8_CTYPE_LOCALE) \
b7d90381 1603 ? (cast)function((cast)(c)) \
31f05a37
KW
1604 : ((((U8)(c)) == MICRO_SIGN) \
1605 ? GREEK_CAPITAL_LETTER_MU \
1606 : ((((U8)(c)) == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS) \
1607 ? LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS \
1608 : ((((U8)(c)) == LATIN_SMALL_LETTER_SHARP_S) \
1609 ? (__ASSERT_(0) (c)) \
1610 : PL_mod_latin1_uc[ (U8) (c) ])))))
1611
1612/* Note that the result can be larger than a byte in a UTF-8 locale. It
1613 * returns a single value, so can't adequately return the fold case of LATIN
1614 * SMALL LETTER SHARP S in a UTF-8 locale (which should be a string of two
1615 * values "ss"); instead it asserts against that under DEBUGGING, and
1616 * otherwise returns its input */
beab9ebe 1617#define _generic_toFOLD_LC(c, function, cast) \
4d7db1b9 1618 ((UNLIKELY((c) == MICRO_SIGN) && IN_UTF8_CTYPE_LOCALE) \
31f05a37 1619 ? GREEK_SMALL_LETTER_MU \
4d7db1b9
KW
1620 : (__ASSERT_(! IN_UTF8_CTYPE_LOCALE \
1621 || (c) != LATIN_SMALL_LETTER_SHARP_S) \
1622 _generic_toLOWER_LC(c, function, cast)))
beab9ebe 1623
84061b6a 1624/* Use the libc versions for these if available. */
f05550c0 1625#if defined(HAS_ISASCII)
84061b6a
KW
1626# define isASCII_LC(c) (FITS_IN_8_BITS(c) && isascii( (U8) (c)))
1627#else
1628# define isASCII_LC(c) isASCII(c)
1629#endif
1630
f05550c0 1631#if defined(HAS_ISBLANK)
84061b6a 1632# define isBLANK_LC(c) _generic_LC(c, _CC_BLANK, isblank)
95f34b6f 1633#else /* Unlike isASCII, varies if in a UTF-8 locale */
278b2b56 1634# define isBLANK_LC(c) ((IN_UTF8_CTYPE_LOCALE) ? isBLANK_L1(c) : isBLANK(c))
84061b6a
KW
1635#endif
1636
f05550c0 1637#define _LC_CAST U8
bbce6d69 1638
f05550c0 1639#ifdef WIN32
375f5f06
KW
1640 /* The Windows functions don't bother to follow the POSIX standard, which
1641 * for example says that something can't both be a printable and a control.
1642 * But Windows treats the \t control as a printable, and does such things
1643 * as making superscripts into both digits and punctuation. This tames
1644 * these flaws by assuming that the definitions of both controls and space
1645 * are correct, and then making sure that other definitions don't have
1646 * weirdnesses, by making sure that isalnum() isn't also ispunct(), etc.
1647 * Not all possible weirdnesses are checked for, just the ones that were
1648 * detected on actual Microsoft code pages */
1649
b7d90381
KW
1650# define isCNTRL_LC(c) _generic_LC(c, _CC_CNTRL, iscntrl)
1651# define isSPACE_LC(c) _generic_LC(c, _CC_SPACE, isspace)
1652
1653# define isALPHA_LC(c) (_generic_LC(c, _CC_ALPHA, isalpha) \
1654 && isALPHANUMERIC_LC(c))
1655# define isALPHANUMERIC_LC(c) (_generic_LC(c, _CC_ALPHANUMERIC, isalnum) && \
1656 ! isPUNCT_LC(c))
1657# define isDIGIT_LC(c) (_generic_LC(c, _CC_DIGIT, isdigit) && \
1658 isALPHANUMERIC_LC(c))
1659# define isGRAPH_LC(c) (_generic_LC(c, _CC_GRAPH, isgraph) && isPRINT_LC(c))
1660# define isIDFIRST_LC(c) (((c) == '_') \
1661 || (_generic_LC(c, _CC_IDFIRST, isalpha) && ! isPUNCT_LC(c)))
1662# define isLOWER_LC(c) (_generic_LC(c, _CC_LOWER, islower) && isALPHA_LC(c))
1663# define isPRINT_LC(c) (_generic_LC(c, _CC_PRINT, isprint) && ! isCNTRL_LC(c))
1664# define isPUNCT_LC(c) (_generic_LC(c, _CC_PUNCT, ispunct) && ! isCNTRL_LC(c))
1665# define isUPPER_LC(c) (_generic_LC(c, _CC_UPPER, isupper) && isALPHA_LC(c))
f05550c0 1666# define isWORDCHAR_LC(c) (((c) == '_') || isALPHANUMERIC_LC(c))
b7d90381
KW
1667# define isXDIGIT_LC(c) (_generic_LC(c, _CC_XDIGIT, isxdigit) \
1668 && isALPHANUMERIC_LC(c))
f05550c0
BF
1669
1670# define toLOWER_LC(c) _generic_toLOWER_LC((c), tolower, U8)
1671# define toUPPER_LC(c) _generic_toUPPER_LC((c), toupper, U8)
1672# define toFOLD_LC(c) _generic_toFOLD_LC((c), tolower, U8)
1673
1674#elif defined(CTYPE256) || (!defined(isascii) && !defined(HAS_ISASCII))
4650c663 1675 /* For most other platforms */
beab9ebe 1676
f05550c0
BF
1677# define isALPHA_LC(c) _generic_LC(c, _CC_ALPHA, isalpha)
1678# define isALPHANUMERIC_LC(c) _generic_LC(c, _CC_ALPHANUMERIC, isalnum)
1679# define isCNTRL_LC(c) _generic_LC(c, _CC_CNTRL, iscntrl)
1680# define isDIGIT_LC(c) _generic_LC(c, _CC_DIGIT, isdigit)
1681# define isGRAPH_LC(c) _generic_LC(c, _CC_GRAPH, isgraph)
1682# define isIDFIRST_LC(c) _generic_LC_underscore(c, _CC_IDFIRST, isalpha)
1683# define isLOWER_LC(c) _generic_LC(c, _CC_LOWER, islower)
1684# define isPRINT_LC(c) _generic_LC(c, _CC_PRINT, isprint)
1685# define isPUNCT_LC(c) _generic_LC(c, _CC_PUNCT, ispunct)
1686# define isSPACE_LC(c) _generic_LC(c, _CC_SPACE, isspace)
1687# define isUPPER_LC(c) _generic_LC(c, _CC_UPPER, isupper)
1688# define isWORDCHAR_LC(c) _generic_LC_underscore(c, _CC_WORDCHAR, isalnum)
1689# define isXDIGIT_LC(c) _generic_LC(c, _CC_XDIGIT, isxdigit)
1690
1691
1692# define toLOWER_LC(c) _generic_toLOWER_LC((c), tolower, U8)
1693# define toUPPER_LC(c) _generic_toUPPER_LC((c), toupper, U8)
1694# define toFOLD_LC(c) _generic_toFOLD_LC((c), tolower, U8)
1695
1696#else /* The final fallback position */
1697
b7d90381
KW
1698# define isALPHA_LC(c) (isascii(c) && isalpha(c))
1699# define isALPHANUMERIC_LC(c) (isascii(c) && isalnum(c))
1700# define isCNTRL_LC(c) (isascii(c) && iscntrl(c))
1701# define isDIGIT_LC(c) (isascii(c) && isdigit(c))
1702# define isGRAPH_LC(c) (isascii(c) && isgraph(c))
f05550c0 1703# define isIDFIRST_LC(c) (isascii(c) && (isalpha(c) || (c) == '_'))
b7d90381
KW
1704# define isLOWER_LC(c) (isascii(c) && islower(c))
1705# define isPRINT_LC(c) (isascii(c) && isprint(c))
1706# define isPUNCT_LC(c) (isascii(c) && ispunct(c))
1707# define isSPACE_LC(c) (isascii(c) && isspace(c))
1708# define isUPPER_LC(c) (isascii(c) && isupper(c))
f05550c0 1709# define isWORDCHAR_LC(c) (isascii(c) && (isalnum(c) || (c) == '_'))
b7d90381 1710# define isXDIGIT_LC(c) (isascii(c) && isxdigit(c))
f05550c0
BF
1711
1712# define toLOWER_LC(c) (isascii(c) ? tolower(c) : (c))
1713# define toUPPER_LC(c) (isascii(c) ? toupper(c) : (c))
1714# define toFOLD_LC(c) (isascii(c) ? tolower(c) : (c))
bbce6d69 1715
f05550c0 1716#endif
55204971 1717
eba68aa0
KW
1718#define isIDCONT(c) isWORDCHAR(c)
1719#define isIDCONT_A(c) isWORDCHAR_A(c)
1720#define isIDCONT_L1(c) isWORDCHAR_L1(c)
1721#define isIDCONT_LC(c) isWORDCHAR_LC(c)
13380643 1722#define isPSXSPC_LC(c) isSPACE_LC(c)
aaa51d5e 1723
4650c663 1724/* For internal core Perl use only: the base macros for defining macros like
d0da05db
KW
1725 * isALPHA_uvchr. 'c' is the code point to check. 'classnum' is the POSIX class
1726 * number defined earlier in this file. _generic_uvchr() is used for POSIX
4650c663
KW
1727 * classes where there is a macro or function 'above_latin1' that takes the
1728 * single argument 'c' and returns the desired value. These exist for those
1729 * classes which have simple definitions, avoiding the overhead of a hash
d0da05db 1730 * lookup or inversion list binary search. _generic_swash_uvchr() can be used
4650c663 1731 * for classes where that overhead is faster than a direct lookup.
d0da05db 1732 * _generic_uvchr() won't compile if 'c' isn't unsigned, as it won't match the
4650c663
KW
1733 * 'above_latin1' prototype. _generic_isCC() macro does bounds checking, so
1734 * have duplicate checks here, so could create versions of the macros that
1735 * don't, but experiments show that gcc optimizes them out anyway. */
66c17564
KW
1736
1737/* Note that all ignore 'use bytes' */
1e222e4f
KW
1738#define _generic_uvchr(classnum, above_latin1, c) ((c) < 256 \
1739 ? _generic_isCC(c, classnum) \
cd500f2f 1740 : above_latin1(c))
1e222e4f
KW
1741#define _generic_swash_uvchr(classnum, c) ((c) < 256 \
1742 ? _generic_isCC(c, classnum) \
922e8cb4 1743 : _is_uni_FOO(classnum, c))
d0da05db
KW
1744#define isALPHA_uvchr(c) _generic_swash_uvchr(_CC_ALPHA, c)
1745#define isALPHANUMERIC_uvchr(c) _generic_swash_uvchr(_CC_ALPHANUMERIC, c)
1746#define isASCII_uvchr(c) isASCII(c)
1747#define isBLANK_uvchr(c) _generic_uvchr(_CC_BLANK, is_HORIZWS_cp_high, c)
1748#define isCNTRL_uvchr(c) isCNTRL_L1(c) /* All controls are in Latin1 */
1749#define isDIGIT_uvchr(c) _generic_swash_uvchr(_CC_DIGIT, c)
1750#define isGRAPH_uvchr(c) _generic_swash_uvchr(_CC_GRAPH, c)
1e222e4f
KW
1751#define isIDCONT_uvchr(c) \
1752 _generic_uvchr(_CC_WORDCHAR, _is_uni_perl_idcont, c)
1753#define isIDFIRST_uvchr(c) \
1754 _generic_uvchr(_CC_IDFIRST, _is_uni_perl_idstart, c)
d0da05db
KW
1755#define isLOWER_uvchr(c) _generic_swash_uvchr(_CC_LOWER, c)
1756#define isPRINT_uvchr(c) _generic_swash_uvchr(_CC_PRINT, c)
1757
1758#define isPUNCT_uvchr(c) _generic_swash_uvchr(_CC_PUNCT, c)
1759#define isSPACE_uvchr(c) _generic_uvchr(_CC_SPACE, is_XPERLSPACE_cp_high, c)
1760#define isPSXSPC_uvchr(c) isSPACE_uvchr(c)
1761
1762#define isUPPER_uvchr(c) _generic_swash_uvchr(_CC_UPPER, c)
1763#define isVERTWS_uvchr(c) _generic_uvchr(_CC_VERTSPACE, is_VERTWS_cp_high, c)
1764#define isWORDCHAR_uvchr(c) _generic_swash_uvchr(_CC_WORDCHAR, c)
1765#define isXDIGIT_uvchr(c) _generic_uvchr(_CC_XDIGIT, is_XDIGIT_cp_high, c)
1766
1767#define toFOLD_uvchr(c,s,l) to_uni_fold(c,s,l)
1768#define toLOWER_uvchr(c,s,l) to_uni_lower(c,s,l)
1769#define toTITLE_uvchr(c,s,l) to_uni_title(c,s,l)
1770#define toUPPER_uvchr(c,s,l) to_uni_upper(c,s,l)
1771
1772/* For backwards compatibility, even though '_uni' should mean official Unicode
1773 * code points, in Perl it means native for those below 256 */
1774#define isALPHA_uni(c) isALPHA_uvchr(c)
1775#define isALPHANUMERIC_uni(c) isALPHANUMERIC_uvchr(c)
1776#define isASCII_uni(c) isASCII_uvchr(c)
1777#define isBLANK_uni(c) isBLANK_uvchr(c)
1778#define isCNTRL_uni(c) isCNTRL_uvchr(c)
1779#define isDIGIT_uni(c) isDIGIT_uvchr(c)
1780#define isGRAPH_uni(c) isGRAPH_uvchr(c)
1781#define isIDCONT_uni(c) isIDCONT_uvchr(c)
1782#define isIDFIRST_uni(c) isIDFIRST_uvchr(c)
1783#define isLOWER_uni(c) isLOWER_uvchr(c)
1784#define isPRINT_uni(c) isPRINT_uvchr(c)
1785#define isPUNCT_uni(c) isPUNCT_uvchr(c)
1786#define isSPACE_uni(c) isSPACE_uvchr(c)
1787#define isPSXSPC_uni(c) isPSXSPC_uvchr(c)
1788#define isUPPER_uni(c) isUPPER_uvchr(c)
1789#define isVERTWS_uni(c) isVERTWS_uvchr(c)
1790#define isWORDCHAR_uni(c) isWORDCHAR_uvchr(c)
1791#define isXDIGIT_uni(c) isXDIGIT_uvchr(c)
1792#define toFOLD_uni(c,s,l) toFOLD_uvchr(c,s,l)
1793#define toLOWER_uni(c,s,l) toLOWER_uvchr(c,s,l)
1794#define toTITLE_uni(c,s,l) toTITLE_uvchr(c,s,l)
1795#define toUPPER_uni(c,s,l) toUPPER_uvchr(c,s,l)
a0ed51b3 1796
4650c663
KW
1797/* For internal core Perl use only: the base macros for defining macros like
1798 * isALPHA_LC_uvchr. These are like isALPHA_LC, but the input can be any code
d0da05db 1799 * point, not just 0-255. Like _generic_uvchr, there are two versions, one for
4650c663 1800 * simple class definitions; the other for more complex. These are like
d0da05db 1801 * _generic_uvchr, so see it for more info. */
cd500f2f
KW
1802#define _generic_LC_uvchr(latin1, above_latin1, c) \
1803 (c < 256 ? latin1(c) : above_latin1(c))
1804#define _generic_LC_swash_uvchr(latin1, classnum, c) \
1805 (c < 256 ? latin1(c) : _is_uni_FOO(classnum, c))
1806
1807#define isALPHA_LC_uvchr(c) _generic_LC_swash_uvchr(isALPHA_LC, _CC_ALPHA, c)
1808#define isALPHANUMERIC_LC_uvchr(c) _generic_LC_swash_uvchr(isALPHANUMERIC_LC, \
1809 _CC_ALPHANUMERIC, c)
b7d90381
KW
1810#define isASCII_LC_uvchr(c) isASCII_LC(c)
1811#define isBLANK_LC_uvchr(c) _generic_LC_uvchr(isBLANK_LC, \
1812 is_HORIZWS_cp_high, c)
feeab5a9 1813#define isCNTRL_LC_uvchr(c) (c < 256 ? isCNTRL_LC(c) : 0)
cd500f2f
KW
1814#define isDIGIT_LC_uvchr(c) _generic_LC_swash_uvchr(isDIGIT_LC, _CC_DIGIT, c)
1815#define isGRAPH_LC_uvchr(c) _generic_LC_swash_uvchr(isGRAPH_LC, _CC_GRAPH, c)
b7d90381 1816#define isIDCONT_LC_uvchr(c) _generic_LC_uvchr(isIDCONT_LC, \
eba68aa0 1817 _is_uni_perl_idcont, c)
b7d90381 1818#define isIDFIRST_LC_uvchr(c) _generic_LC_uvchr(isIDFIRST_LC, \
cd500f2f
KW
1819 _is_uni_perl_idstart, c)
1820#define isLOWER_LC_uvchr(c) _generic_LC_swash_uvchr(isLOWER_LC, _CC_LOWER, c)
1821#define isPRINT_LC_uvchr(c) _generic_LC_swash_uvchr(isPRINT_LC, _CC_PRINT, c)
b7d90381 1822#define isPSXSPC_LC_uvchr(c) isSPACE_LC_uvchr(c)
cd500f2f 1823#define isPUNCT_LC_uvchr(c) _generic_LC_swash_uvchr(isPUNCT_LC, _CC_PUNCT, c)
b7d90381 1824#define isSPACE_LC_uvchr(c) _generic_LC_uvchr(isSPACE_LC, \
509fb054 1825 is_XPERLSPACE_cp_high, c)
cd500f2f 1826#define isUPPER_LC_uvchr(c) _generic_LC_swash_uvchr(isUPPER_LC, _CC_UPPER, c)
b7d90381 1827#define isWORDCHAR_LC_uvchr(c) _generic_LC_swash_uvchr(isWORDCHAR_LC, \
cd500f2f 1828 _CC_WORDCHAR, c)
b7d90381
KW
1829#define isXDIGIT_LC_uvchr(c) _generic_LC_uvchr(isXDIGIT_LC, \
1830 is_XDIGIT_cp_high, c)
e712593e 1831
b7d90381 1832#define isBLANK_LC_uni(c) isBLANK_LC_uvchr(UNI_TO_NATIVE(c))
aaa51d5e 1833
4650c663
KW
1834/* For internal core Perl use only: the base macros for defining macros like
1835 * isALPHA_utf8. These are like the earlier defined macros, but take an input
1836 * UTF-8 encoded string 'p'. If the input is in the Latin1 range, use
1837 * the Latin1 macro 'classnum' on 'p'. Otherwise use the value given by the
1838 * 'utf8' parameter. This relies on the fact that ASCII characters have the
1839 * same representation whether utf8 or not. Note that it assumes that the utf8
1840 * has been validated, and ignores 'use bytes' */
34aeb2e9
KW
1841#define _base_generic_utf8(enum_name, name, p, use_locale ) \
1842 _is_utf8_FOO(CAT2(_CC_, enum_name), \
1843 (const U8 *) p, \
1844 "is" STRINGIFY(name) "_utf8", \
1845 "is" STRINGIFY(name) "_utf8_safe", \
1846 1, use_locale, __FILE__,__LINE__)
1847
1848#define _generic_utf8(name, p) _base_generic_utf8(name, name, p, 0)
da8c1a98
KW
1849
1850/* The "_safe" macros make sure that we don't attempt to read beyond 'e', but
1851 * they don't otherwise go out of their way to look for malformed UTF-8. If
1852 * they can return accurate results without knowing if the input is otherwise
1853 * malformed, they do so. For example isASCII is accurate in spite of any
1854 * non-length malformations because it looks only at a single byte. Likewise
1855 * isDIGIT looks just at the first byte for code points 0-255, as all UTF-8
1856 * variant ones return FALSE. But, if the input has to be well-formed in order
1857 * for the results to be accurate, the macros will test and if malformed will
1858 * call a routine to die
1859 *
1860 * Except for toke.c, the macros do assume that e > p, asserting that on
1861 * DEBUGGING builds. Much code that calls these depends on this being true,
1862 * for other reasons. toke.c is treated specially as using the regular
1863 * assertion breaks it in many ways. All strings that these operate on there
1864 * are supposed to have an extra NUL character at the end, so that *e = \0. A
1865 * bunch of code in toke.c assumes that this is true, so the assertion allows
1866 * for that */
1867#ifdef PERL_IN_TOKE_C
1868# define _utf8_safe_assert(p,e) ((e) > (p) || ((e) == (p) && *(p) == '\0'))
1869#else
1870# define _utf8_safe_assert(p,e) ((e) > (p))
1871#endif
1872
1873#define _generic_utf8_safe(classnum, p, e, above_latin1) \
1874 (__ASSERT_(_utf8_safe_assert(p, e)) \
1875 (UTF8_IS_INVARIANT(*(p))) \
1876 ? _generic_isCC(*(p), classnum) \
1877 : (UTF8_IS_DOWNGRADEABLE_START(*(p)) \
1878 ? ((LIKELY((e) - (p) > 1 && UTF8_IS_CONTINUATION(*((p)+1)))) \
1879 ? _generic_isCC(EIGHT_BIT_UTF8_TO_NATIVE(*(p), *((p)+1 )), \
1880 classnum) \
1881 : (_force_out_malformed_utf8_message( \
1882 (U8 *) (p), (U8 *) (e), 0, 1), 0)) \
1883 : above_latin1))
b7d90381
KW
1884/* Like the above, but calls 'above_latin1(p)' to get the utf8 value.
1885 * 'above_latin1' can be a macro */
da8c1a98
KW
1886#define _generic_func_utf8_safe(classnum, above_latin1, p, e) \
1887 _generic_utf8_safe(classnum, p, e, above_latin1(p, e))
1888#define _generic_non_swash_utf8_safe(classnum, above_latin1, p, e) \
1889 _generic_utf8_safe(classnum, p, e, \
1890 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
1891 ? (_force_out_malformed_utf8_message( \
1892 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
1893 : above_latin1(p)))
4650c663 1894/* Like the above, but passes classnum to _isFOO_utf8(), instead of having an
cd500f2f 1895 * 'above_latin1' parameter */
da8c1a98
KW
1896#define _generic_swash_utf8_safe(classnum, p, e) \
1897_generic_utf8_safe(classnum, p, e, _is_utf8_FOO_with_len(classnum, p, e))
922e8cb4 1898
cc8ab7c0 1899/* Like the above, but should be used only when it is known that there are no
ff7ecfc3
KW
1900 * characters in the upper-Latin1 range (128-255 on ASCII platforms) which the
1901 * class is TRUE for. Hence it can skip the tests for this range.
1902 * 'above_latin1' should include its arguments */
da8c1a98
KW
1903#define _generic_utf8_safe_no_upper_latin1(classnum, p, e, above_latin1) \
1904 (__ASSERT_(_utf8_safe_assert(p, e)) \
1905 (UTF8_IS_INVARIANT(*(p))) \
1906 ? _generic_isCC(*(p), classnum) \
1907 : (UTF8_IS_DOWNGRADEABLE_START(*(p))) \
1908 ? 0 /* Note that doesn't check validity for latin1 */ \
1909 : above_latin1)
1910
84238efa
KW
1911/* NOTE that some of these macros have very similar ones in regcharclass.h.
1912 * For example, there is (at the time of this writing) an 'is_SPACE_utf8()'
1913 * there, differing in name only by an underscore from the one here
1914 * 'isSPACE_utf8(). The difference is that the ones here are probably more
1915 * efficient and smaller, using an O(1) array lookup for Latin1-range code
1916 * points; the regcharclass.h ones are implemented as a series of
1917 * "if-else-if-else ..." */
1918
34aeb2e9
KW
1919#define isALPHA_utf8(p) _generic_utf8(ALPHA, p)
1920#define isALPHANUMERIC_utf8(p) _generic_utf8(ALPHANUMERIC, p)
1921#define isASCII_utf8(p) _generic_utf8(ASCII, p)
1922#define isBLANK_utf8(p) _generic_utf8(BLANK, p)
1923#define isCNTRL_utf8(p) _generic_utf8(CNTRL, p)
1924#define isDIGIT_utf8(p) _generic_utf8(DIGIT, p)
1925#define isGRAPH_utf8(p) _generic_utf8(GRAPH, p)
1926#define isIDCONT_utf8(p) _generic_utf8(IDCONT, p)
1927#define isIDFIRST_utf8(p) _generic_utf8(IDFIRST, p)
1928#define isLOWER_utf8(p) _generic_utf8(LOWER, p)
1929#define isPRINT_utf8(p) _generic_utf8(PRINT, p)
1930#define isPSXSPC_utf8(p) _generic_utf8(PSXSPC, p)
1931#define isPUNCT_utf8(p) _generic_utf8(PUNCT, p)
1932#define isSPACE_utf8(p) _generic_utf8(SPACE, p)
1933#define isUPPER_utf8(p) _generic_utf8(UPPER, p)
1934#define isVERTWS_utf8(p) _generic_utf8(VERTSPACE, p)
1935#define isWORDCHAR_utf8(p) _generic_utf8(WORDCHAR, p)
1936#define isXDIGIT_utf8(p) _generic_utf8(XDIGIT, p)
e8fa43e2 1937
da8c1a98
KW
1938#define isALPHA_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_ALPHA, p, e)
1939#define isALPHANUMERIC_utf8_safe(p, e) \
1940 _generic_swash_utf8_safe(_CC_ALPHANUMERIC, p, e)
1941#define isASCII_utf8_safe(p, e) \
1942 /* Because ASCII is invariant under utf8, the non-utf8 macro \
1943 * works */ \
1944 (__ASSERT_(_utf8_safe_assert(p, e)) isASCII(*(p)))
1945#define isBLANK_utf8_safe(p, e) \
1946 _generic_non_swash_utf8_safe(_CC_BLANK, is_HORIZWS_high, p, e)
1947
e8fa43e2
KW
1948#ifdef EBCDIC
1949 /* Because all controls are UTF-8 invariants in EBCDIC, we can use this
1950 * more efficient macro instead of the more general one */
da8c1a98 1951# define isCNTRL_utf8_safe(p, e) \
56d02b8c 1952 (__ASSERT_(_utf8_safe_assert(p, e)) isCNTRL_L1(*(p)))
e8fa43e2 1953#else
da8c1a98 1954# define isCNTRL_utf8_safe(p, e) _generic_utf8_safe(_CC_CNTRL, p, e, 0)
e8fa43e2
KW
1955#endif
1956
da8c1a98
KW
1957#define isDIGIT_utf8_safe(p, e) \
1958 _generic_utf8_safe_no_upper_latin1(_CC_DIGIT, p, e, \
1959 _is_utf8_FOO_with_len(_CC_DIGIT, p, e))
1960#define isGRAPH_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_GRAPH, p, e)
1961#define isIDCONT_utf8_safe(p, e) _generic_func_utf8_safe(_CC_WORDCHAR, \
1962 _is_utf8_perl_idcont_with_len, p, e)
e5dcd934 1963
c11ff943
KW
1964/* To prevent S_scan_word in toke.c from hanging, we have to make sure that
1965 * IDFIRST is an alnum. See
c8362b00 1966 * http://rt.perl.org/rt3/Ticket/Display.html?id=74022 for more detail than you
f91dcd13
KW
1967 * ever wanted to know about. (In the ASCII range, there isn't a difference.)
1968 * This used to be not the XID version, but we decided to go with the more
1969 * modern Unicode definition */
da8c1a98
KW
1970#define isIDFIRST_utf8_safe(p, e) \
1971 _generic_func_utf8_safe(_CC_IDFIRST, \
1972 _is_utf8_perl_idstart_with_len, (U8 *) (p), (U8 *) (e))
1973
1974#define isLOWER_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_LOWER, p, e)
1975#define isPRINT_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_PRINT, p, e)
1976#define isPSXSPC_utf8_safe(p, e) isSPACE_utf8_safe(p, e)
1977#define isPUNCT_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_PUNCT, p, e)
1978#define isSPACE_utf8_safe(p, e) \
1979 _generic_non_swash_utf8_safe(_CC_SPACE, is_XPERLSPACE_high, p, e)
1980#define isUPPER_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_UPPER, p, e)
1981#define isVERTWS_utf8_safe(p, e) \
1982 _generic_non_swash_utf8_safe(_CC_VERTSPACE, is_VERTWS_high, p, e)
1983#define isWORDCHAR_utf8_safe(p, e) \
1984 _generic_swash_utf8_safe(_CC_WORDCHAR, p, e)
1985#define isXDIGIT_utf8_safe(p, e) \
1986 _generic_utf8_safe_no_upper_latin1(_CC_XDIGIT, p, e, \
1987 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
1988 ? (_force_out_malformed_utf8_message( \
1989 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
1990 : is_XDIGIT_high(p)))
a0ed51b3 1991
7d2b61dd 1992#define toFOLD_utf8(p,s,l) to_utf8_fold(p,s,l)
509fb054
KW
1993#define toLOWER_utf8(p,s,l) to_utf8_lower(p,s,l)
1994#define toTITLE_utf8(p,s,l) to_utf8_title(p,s,l)
1995#define toUPPER_utf8(p,s,l) to_utf8_upper(p,s,l)
2e8adce6 1996
567b353c 1997/* For internal core use only, subject to change */
607313a1
KW
1998#define _toFOLD_utf8_flags(p,e,s,l,f) _to_utf8_fold_flags (p,e,s,l,f, "", 0)
1999#define _toLOWER_utf8_flags(p,e,s,l,f) _to_utf8_lower_flags(p,e,s,l,f, "", 0)
2000#define _toTITLE_utf8_flags(p,e,s,l,f) _to_utf8_title_flags(p,e,s,l,f, "", 0)
2001#define _toUPPER_utf8_flags(p,e,s,l,f) _to_utf8_upper_flags(p,e,s,l,f, "", 0)
a1a5ec35
KW
2002
2003#define toFOLD_utf8_safe(p,e,s,l) _toFOLD_utf8_flags(p,e,s,l, FOLD_FLAGS_FULL)
2004#define toLOWER_utf8_safe(p,e,s,l) _toLOWER_utf8_flags(p,e,s,l, 0)
2005#define toTITLE_utf8_safe(p,e,s,l) _toTITLE_utf8_flags(p,e,s,l, 0)
2006#define toUPPER_utf8_safe(p,e,s,l) _toUPPER_utf8_flags(p,e,s,l, 0)
567b353c 2007
4650c663
KW
2008/* For internal core Perl use only: the base macros for defining macros like
2009 * isALPHA_LC_utf8. These are like _generic_utf8, but if the first code point
2010 * in 'p' is within the 0-255 range, it uses locale rules from the passed-in
2011 * 'macro' parameter */
34aeb2e9
KW
2012#define _generic_LC_utf8(name, p) _base_generic_utf8(name, name, p, 1)
2013
2014#define isALPHA_LC_utf8(p) _generic_LC_utf8(ALPHA, p)
2015#define isALPHANUMERIC_LC_utf8(p) _generic_LC_utf8(ALPHANUMERIC, p)
2016#define isASCII_LC_utf8(p) _generic_LC_utf8(ASCII, p)
2017#define isBLANK_LC_utf8(p) _generic_LC_utf8(BLANK, p)
2018#define isCNTRL_LC_utf8(p) _generic_LC_utf8(CNTRL, p)
2019#define isDIGIT_LC_utf8(p) _generic_LC_utf8(DIGIT, p)
2020#define isGRAPH_LC_utf8(p) _generic_LC_utf8(GRAPH, p)
2021#define isIDCONT_LC_utf8(p) _generic_LC_utf8(IDCONT, p)
2022#define isIDFIRST_LC_utf8(p) _generic_LC_utf8(IDFIRST, p)
2023#define isLOWER_LC_utf8(p) _generic_LC_utf8(LOWER, p)
2024#define isPRINT_LC_utf8(p) _generic_LC_utf8(PRINT, p)
2025#define isPSXSPC_LC_utf8(p) _generic_LC_utf8(PSXSPC, p)
2026#define isPUNCT_LC_utf8(p) _generic_LC_utf8(PUNCT, p)
2027#define isSPACE_LC_utf8(p) _generic_LC_utf8(SPACE, p)
2028#define isUPPER_LC_utf8(p) _generic_LC_utf8(UPPER, p)
2029#define isWORDCHAR_LC_utf8(p) _generic_LC_utf8(WORDCHAR, p)
2030#define isXDIGIT_LC_utf8(p) _generic_LC_utf8(XDIGIT, p)
2031
da8c1a98
KW
2032/* For internal core Perl use only: the base macros for defining macros like
2033 * isALPHA_LC_utf8_safe. These are like _generic_utf8, but if the first code
2034 * point in 'p' is within the 0-255 range, it uses locale rules from the
2035 * passed-in 'macro' parameter */
2036#define _generic_LC_utf8_safe(macro, p, e, above_latin1) \
2037 (__ASSERT_(_utf8_safe_assert(p, e)) \
2038 (UTF8_IS_INVARIANT(*(p))) \
2039 ? macro(*(p)) \
2040 : (UTF8_IS_DOWNGRADEABLE_START(*(p)) \
2041 ? ((LIKELY((e) - (p) > 1 && UTF8_IS_CONTINUATION(*((p)+1)))) \
2042 ? macro(EIGHT_BIT_UTF8_TO_NATIVE(*(p), *((p)+1))) \
2043 : (_force_out_malformed_utf8_message( \
2044 (U8 *) (p), (U8 *) (e), 0, 1), 0)) \
2045 : above_latin1))
2046
2047#define _generic_LC_swash_utf8_safe(macro, classnum, p, e) \
2048 _generic_LC_utf8_safe(macro, p, e, \
2049 _is_utf8_FOO_with_len(classnum, p, e))
2050
2051#define _generic_LC_func_utf8_safe(macro, above_latin1, p, e) \
2052 _generic_LC_utf8_safe(macro, p, e, above_latin1(p, e))
2053
2054#define _generic_LC_non_swash_utf8_safe(classnum, above_latin1, p, e) \
2055 _generic_LC_utf8_safe(classnum, p, e, \
2056 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
2057 ? (_force_out_malformed_utf8_message( \
2058 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
2059 : above_latin1(p)))
2060
2061#define isALPHANUMERIC_LC_utf8_safe(p, e) \
2062 _generic_LC_swash_utf8_safe(isALPHANUMERIC_LC, \
2063 _CC_ALPHANUMERIC, p, e)
2064#define isALPHA_LC_utf8_safe(p, e) \
2065 _generic_LC_swash_utf8_safe(isALPHA_LC, _CC_ALPHA, p, e)
2066#define isASCII_LC_utf8_safe(p, e) \
2067 (__ASSERT_(_utf8_safe_assert(p, e)) isASCII_LC(*(p)))
2068#define isBLANK_LC_utf8_safe(p, e) \
2069 _generic_LC_non_swash_utf8_safe(isBLANK_LC, is_HORIZWS_high, p, e)
2070#define isCNTRL_LC_utf8_safe(p, e) \
2071 _generic_LC_utf8_safe(isCNTRL_LC, p, e, 0)
2072#define isDIGIT_LC_utf8_safe(p, e) \
2073 _generic_LC_swash_utf8_safe(isDIGIT_LC, _CC_DIGIT, p, e)
2074#define isGRAPH_LC_utf8_safe(p, e) \
2075 _generic_LC_swash_utf8_safe(isGRAPH_LC, _CC_GRAPH, p, e)
2076#define isIDCONT_LC_utf8_safe(p, e) \
2077 _generic_LC_func_utf8_safe(isIDCONT_LC, \
2078 _is_utf8_perl_idcont_with_len, p, e)
2079#define isIDFIRST_LC_utf8_safe(p, e) \
2080 _generic_LC_func_utf8_safe(isIDFIRST_LC, \
2081 _is_utf8_perl_idstart_with_len, p, e)
2082#define isLOWER_LC_utf8_safe(p, e) \
2083 _generic_LC_swash_utf8_safe(isLOWER_LC, _CC_LOWER, p, e)
2084#define isPRINT_LC_utf8_safe(p, e) \
2085 _generic_LC_swash_utf8_safe(isPRINT_LC, _CC_PRINT, p, e)
2086#define isPSXSPC_LC_utf8_safe(p, e) isSPACE_LC_utf8_safe(p, e)
2087#define isPUNCT_LC_utf8_safe(p, e) \
2088 _generic_LC_swash_utf8_safe(isPUNCT_LC, _CC_PUNCT, p, e)
2089#define isSPACE_LC_utf8_safe(p, e) \
2090 _generic_LC_non_swash_utf8_safe(isSPACE_LC, is_XPERLSPACE_high, p, e)
2091#define isUPPER_LC_utf8_safe(p, e) \
2092 _generic_LC_swash_utf8_safe(isUPPER_LC, _CC_UPPER, p, e)
2093#define isWORDCHAR_LC_utf8_safe(p, e) \
2094 _generic_LC_swash_utf8_safe(isWORDCHAR_LC, _CC_WORDCHAR, p, e)
2095#define isXDIGIT_LC_utf8_safe(p, e) \
2096 _generic_LC_non_swash_utf8_safe(isXDIGIT_LC, is_XDIGIT_high, p, e)
aaa51d5e 2097
fbc19f27
KW
2098/* Macros for backwards compatibility and for completeness when the ASCII and
2099 * Latin1 values are identical */
b7d90381
KW
2100#define isALPHAU(c) isALPHA_L1(c)
2101#define isDIGIT_L1(c) isDIGIT_A(c)
2102#define isOCTAL(c) isOCTAL_A(c)
2103#define isOCTAL_L1(c) isOCTAL_A(c)
2104#define isXDIGIT_L1(c) isXDIGIT_A(c)
2105#define isALNUM(c) isWORDCHAR(c)
2106#define isALNUMU(c) isWORDCHAR_L1(c)
2107#define isALNUM_LC(c) isWORDCHAR_LC(c)
2108#define isALNUM_uni(c) isWORDCHAR_uni(c)
2e28f0b9 2109#define isALNUM_LC_uvchr(c) isWORDCHAR_LC_uvchr(c)
b7d90381
KW
2110#define isALNUM_utf8(p) isWORDCHAR_utf8(p)
2111#define isALNUM_LC_utf8(p) isWORDCHAR_LC_utf8(p)
2112#define isALNUMC_A(c) isALPHANUMERIC_A(c) /* Mnemonic: "C's alnum" */
2113#define isALNUMC_L1(c) isALPHANUMERIC_L1(c)
2114#define isALNUMC(c) isALPHANUMERIC(c)
2115#define isALNUMC_LC(c) isALPHANUMERIC_LC(c)
2116#define isALNUMC_uni(c) isALPHANUMERIC_uni(c)
15861f94 2117#define isALNUMC_LC_uvchr(c) isALPHANUMERIC_LC_uvchr(c)
b7d90381 2118#define isALNUMC_utf8(p) isALPHANUMERIC_utf8(p)
15861f94 2119#define isALNUMC_LC_utf8(p) isALPHANUMERIC_LC_utf8(p)
fbc19f27 2120
2bd1cbf6
KW
2121/* On EBCDIC platforms, CTRL-@ is 0, CTRL-A is 1, etc, just like on ASCII,
2122 * except that they don't necessarily mean the same characters, e.g. CTRL-D is
2123 * 4 on both systems, but that is EOT on ASCII; ST on EBCDIC.
2124 * '?' is special-cased on EBCDIC to APC, which is the control there that is
2125 * the outlier from the block that contains the other controls, just like
2126 * toCTRL('?') on ASCII yields DEL, the control that is the outlier from the C0
2127 * block. If it weren't special cased, it would yield a non-control.
88794300
KW
2128 * The conversion works both ways, so toCTRL('D') is 4, and toCTRL(4) is D,
2129 * etc. */
2bd1cbf6 2130#ifndef EBCDIC
75763b3a 2131# define toCTRL(c) (__ASSERT_(FITS_IN_8_BITS(c)) toUPPER(((U8)(c))) ^ 64)
2bd1cbf6 2132#else
75763b3a
KW
2133# define toCTRL(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
2134 ((isPRINT_A(c)) \
2135 ? (UNLIKELY((c) == '?') \
2136 ? QUESTION_MARK_CTRL \
2137 : (NATIVE_TO_LATIN1(toUPPER((U8) (c))) ^ 64)) \
2138 : (UNLIKELY((c) == QUESTION_MARK_CTRL) \
2139 ? '?' \
2140 : (LATIN1_TO_NATIVE(((U8) (c)) ^ 64)))))
2bd1cbf6 2141#endif
bbce6d69 2142
dea28490
JJ
2143/* Line numbers are unsigned, 32 bits. */
2144typedef U32 line_t;
e5dcd934 2145#define NOLINE ((line_t) 4294967295UL) /* = FFFFFFFF */
378cc40b 2146
91152fc1
DG
2147/* Helpful alias for version prescan */
2148#define is_LAX_VERSION(a,b) \
2149 (a != Perl_prescan_version(aTHX_ a, FALSE, b, NULL, NULL, NULL, NULL))
2150
2151#define is_STRICT_VERSION(a,b) \
2152 (a != Perl_prescan_version(aTHX_ a, TRUE, b, NULL, NULL, NULL, NULL))
2153
2154#define BADVERSION(a,b,c) \
2155 if (b) { \
2156 *b = c; \
2157 } \
2158 return a;
8c52afec 2159
cb27eebd
KW
2160/* Converts a character known to represent a hexadecimal digit (0-9, A-F, or
2161 * a-f) to its numeric value. READ_XDIGIT's argument is a string pointer,
2162 * which is advanced. The input is validated only by an assert() in DEBUGGING
2163 * builds. In both ASCII and EBCDIC the last 4 bits of the digits are 0-9; and
2164 * the last 4 bits of A-F and a-f are 1-6, so adding 9 yields 10-15 */
2165#define XDIGIT_VALUE(c) (__ASSERT_(isXDIGIT(c)) (0xf & (isDIGIT(c) \
2166 ? (c) \
2167 : ((c) + 9))))
902008b8
KW
2168#define READ_XDIGIT(s) (__ASSERT_(isXDIGIT(*s)) (0xf & (isDIGIT(*(s)) \
2169 ? (*(s)++) \
2170 : (*(s)++ + 9))))
95a59cab 2171
cb27eebd
KW
2172/* Converts a character known to represent an octal digit (0-7) to its numeric
2173 * value. The input is validated only by an assert() in DEBUGGING builds. In
2174 * both ASCII and EBCDIC the last 3 bits of the octal digits range from 0-7. */
2175#define OCTAL_VALUE(c) (__ASSERT_(isOCTAL(c)) (7 & (c)))
2176
305b8651
KW
2177/* Efficiently returns a boolean as to if two native characters are equivalent
2178 * case-insenstively. At least one of the characters must be one of [A-Za-z];
2179 * the ALPHA in the name is to remind you of that. This is asserted() in
2180 * DEBUGGING builds. Because [A-Za-z] are invariant under UTF-8, this macro
2181 * works (on valid input) for both non- and UTF-8-encoded bytes.
2182 *
2183 * When one of the inputs is a compile-time constant and gets folded by the
2184 * compiler, this reduces to an AND and a TEST. On both EBCDIC and ASCII
2185 * machines, 'A' and 'a' differ by a single bit; the same with the upper and
2186 * lower case of all other ASCII-range alphabetics. On ASCII platforms, they
96ca48da
KW
2187 * are 32 apart; on EBCDIC, they are 64. At compile time, this uses an
2188 * exclusive 'or' to find that bit and then inverts it to form a mask, with
2189 * just a single 0, in the bit position where the upper- and lowercase differ.
2190 * */
305b8651
KW
2191#define isALPHA_FOLD_EQ(c1, c2) \
2192 (__ASSERT_(isALPHA_A(c1) || isALPHA_A(c2)) \
2193 ((c1) & ~('A' ^ 'a')) == ((c2) & ~('A' ^ 'a')))
2194#define isALPHA_FOLD_NE(c1, c2) (! isALPHA_FOLD_EQ((c1), (c2)))
2195
8e84507e 2196/*
ccfc67b7
JH
2197=head1 Memory Management
2198
a02a5408 2199=for apidoc Am|void|Newx|void* ptr|int nitems|type
954c1994
GS
2200The XSUB-writer's interface to the C C<malloc> function.
2201
596f7718 2202Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2203
c5008215
JC
2204In 5.9.3, Newx() and friends replace the older New() API, and drops
2205the first parameter, I<x>, a debug aid which allowed callers to identify
37b8b4c9 2206themselves. This aid has been superseded by a new build option,
d10b4965 2207PERL_MEM_LOG (see L<perlhacktips/PERL_MEM_LOG>). The older API is still
c5008215
JC
2208there for use in XS modules supporting older perls.
2209
a02a5408 2210=for apidoc Am|void|Newxc|void* ptr|int nitems|type|cast
954c1994 2211The XSUB-writer's interface to the C C<malloc> function, with
fbe13c60 2212cast. See also C<L</Newx>>.
954c1994 2213
596f7718 2214Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2215
a02a5408 2216=for apidoc Am|void|Newxz|void* ptr|int nitems|type
954c1994 2217The XSUB-writer's interface to the C C<malloc> function. The allocated
fbe13c60 2218memory is zeroed with C<memzero>. See also C<L</Newx>>.
a02a5408 2219
596f7718 2220Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2221
954c1994
GS
2222=for apidoc Am|void|Renew|void* ptr|int nitems|type
2223The XSUB-writer's interface to the C C<realloc> function.
2224
596f7718 2225Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2226
954c1994
GS
2227=for apidoc Am|void|Renewc|void* ptr|int nitems|type|cast
2228The XSUB-writer's interface to the C C<realloc> function, with
2229cast.
2230
596f7718 2231Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2232
49b8b560 2233=for apidoc Am|void|Safefree|void* ptr
954c1994
GS
2234The XSUB-writer's interface to the C C<free> function.
2235
596f7718 2236This should B<ONLY> be used on memory obtained using L</"Newx"> and friends.
0d7b2759 2237
954c1994
GS
2238=for apidoc Am|void|Move|void* src|void* dest|int nitems|type
2239The XSUB-writer's interface to the C C<memmove> function. The C<src> is the
926bb54c 2240source, C<dest> is the destination, C<nitems> is the number of items, and
fbe13c60 2241C<type> is the type. Can do overlapping moves. See also C<L</Copy>>.
954c1994 2242
e90e2364 2243=for apidoc Am|void *|MoveD|void* src|void* dest|int nitems|type
796b6530 2244Like C<Move> but returns C<dest>. Useful
72d33970 2245for encouraging compilers to tail-call
e90e2364
NC
2246optimise.
2247
954c1994
GS
2248=for apidoc Am|void|Copy|void* src|void* dest|int nitems|type
2249The XSUB-writer's interface to the C C<memcpy> function. The C<src> is the
926bb54c 2250source, C<dest> is the destination, C<nitems> is the number of items, and
fbe13c60 2251C<type> is the type. May fail on overlapping copies. See also C<L</Move>>.
954c1994 2252
e90e2364
NC
2253=for apidoc Am|void *|CopyD|void* src|void* dest|int nitems|type
2254
796b6530 2255Like C<Copy> but returns C<dest>. Useful
72d33970 2256for encouraging compilers to tail-call
e90e2364
NC
2257optimise.
2258
954c1994
GS
2259=for apidoc Am|void|Zero|void* dest|int nitems|type
2260
2261The XSUB-writer's interface to the C C<memzero> function. The C<dest> is the
2262destination, C<nitems> is the number of items, and C<type> is the type.
2263
e90e2364
NC
2264=for apidoc Am|void *|ZeroD|void* dest|int nitems|type
2265
72d33970
FC
2266Like C<Zero> but returns dest. Useful
2267for encouraging compilers to tail-call
e90e2364
NC
2268optimise.
2269
da5d8dbb 2270=for apidoc Am|void|StructCopy|type *src|type *dest|type
4375e838 2271This is an architecture-independent macro to copy one structure to another.
954c1994 2272
7e337ee0
JH
2273=for apidoc Am|void|PoisonWith|void* dest|int nitems|type|U8 byte
2274
2275Fill up memory with a byte pattern (a byte repeated over and over
2276again) that hopefully catches attempts to access uninitialized memory.
2277
2278=for apidoc Am|void|PoisonNew|void* dest|int nitems|type
2279
2280PoisonWith(0xAB) for catching access to allocated but uninitialized memory.
2281
1c12ffb4 2282=for apidoc Am|void|PoisonFree|void* dest|int nitems|type
7e337ee0
JH
2283
2284PoisonWith(0xEF) for catching access to freed memory.
2285
9965345d
JH
2286=for apidoc Am|void|Poison|void* dest|int nitems|type
2287
7e337ee0 2288PoisonWith(0xEF) for catching access to freed memory.
9965345d
JH
2289
2290=cut */
954c1994 2291
561b68a9
SH
2292/* Maintained for backwards-compatibility only. Use newSV() instead. */
2293#ifndef PERL_CORE
ff06c60c 2294#define NEWSV(x,len) newSV(len)
561b68a9 2295#endif
ff06c60c 2296
19a94d75
JH
2297#define MEM_SIZE_MAX ((MEM_SIZE)~0)
2298
e6bdf523 2299
27d5b266 2300#ifdef PERL_MALLOC_WRAP
e6bdf523
DM
2301
2302/* This expression will be constant-folded at compile time. It checks
2303 * whether or not the type of the count n is so small (e.g. U8 or U16, or
2304 * U32 on 64-bit systems) that there's no way a wrap-around could occur.
2305 * As well as avoiding the need for a run-time check in some cases, it's
2306 * designed to avoid compiler warnings like:
2307 * comparison is always false due to limited range of data type
73e8ff00
DM
2308 * It's mathematically equivalent to
2309 * max(n) * sizeof(t) > MEM_SIZE_MAX
e6bdf523
DM
2310 */
2311
2312# define _MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) \
004073ba 2313 (8 * sizeof(n) + sizeof(t) > sizeof(MEM_SIZE))
e6bdf523 2314
88f9f128 2315/* This is written in a slightly odd way to avoid various spurious
d98e5cde
DM
2316 * compiler warnings. We *want* to write the expression as
2317 * _MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) && (n > C)
2318 * (for some compile-time constant C), but even when the LHS
2319 * constant-folds to false at compile-time, g++ insists on emitting
2320 * warnings about the RHS (e.g. "comparison is always false"), so instead
2321 * we write it as
e6bdf523 2322 *
d98e5cde 2323 * (cond ? n : X) > C
88f9f128 2324 *
d98e5cde
DM
2325 * where X is a constant with X > C always false. Choosing a value for X
2326 * is tricky. If 0, some compilers will complain about 0 > C always being
2327 * false; if 1, Coverity complains when n happens to be the constant value
2328 * '1', that cond ? 1 : 1 has the same value on both branches; so use C
2329 * for X and hope that nothing else whines.
e6bdf523
DM
2330 */
2331
2332# define _MEM_WRAP_WILL_WRAP(n,t) \
88f9f128
DM
2333 ((_MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) ? (MEM_SIZE)(n) : \
2334 MEM_SIZE_MAX/sizeof(t)) > MEM_SIZE_MAX/sizeof(t))
e6bdf523
DM
2335
2336# define MEM_WRAP_CHECK(n,t) \
2337 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2338 && (croak_memory_wrap(),0))
2339
2340# define MEM_WRAP_CHECK_1(n,t,a) \
2341 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2342 && (Perl_croak_nocontext("%s",(a)),0))
2343
8b44ba4c 2344#define MEM_WRAP_CHECK_(n,t) MEM_WRAP_CHECK(n,t),
27d5b266 2345
d1decf2b 2346#define PERL_STRLEN_ROUNDUP(n) ((void)(((n) > MEM_SIZE_MAX - 2 * PERL_STRLEN_ROUNDUP_QUANTUM) ? (croak_memory_wrap(),0):0),((n-1+PERL_STRLEN_ROUNDUP_QUANTUM)&~((MEM_SIZE)PERL_STRLEN_ROUNDUP_QUANTUM-1)))
27d5b266
JH
2347#else
2348
410319be
NC
2349#define MEM_WRAP_CHECK(n,t)
2350#define MEM_WRAP_CHECK_1(n,t,a)
2351#define MEM_WRAP_CHECK_2(n,t,a,b)
8b44ba4c
NC
2352#define MEM_WRAP_CHECK_(n,t)
2353
6fff79ce 2354#define PERL_STRLEN_ROUNDUP(n) (((n-1+PERL_STRLEN_ROUNDUP_QUANTUM)&~((MEM_SIZE)PERL_STRLEN_ROUNDUP_QUANTUM-1)))
27d5b266 2355
1936d2a7 2356#endif
8b44ba4c 2357
fe4f188c 2358#ifdef PERL_MEM_LOG
46c6c7e2 2359/*
9f653bb5 2360 * If PERL_MEM_LOG is defined, all Newx()s, Renew()s, and Safefree()s
46c6c7e2
JH
2361 * go through functions, which are handy for debugging breakpoints, but
2362 * which more importantly get the immediate calling environment (file and
e352bcff
JH
2363 * line number, and C function name if available) passed in. This info can
2364 * then be used for logging the calls, for which one gets a sample
73d1d973 2365 * implementation unless -DPERL_MEM_LOG_NOIMPL is also defined.
3609ea0d 2366 *
46c6c7e2 2367 * Known problems:
94e892a6 2368 * - not all memory allocs get logged, only those
46c6c7e2 2369 * that go through Newx() and derivatives (while all
94e892a6 2370 * Safefrees do get logged)
46c6c7e2
JH
2371 * - __FILE__ and __LINE__ do not work everywhere
2372 * - __func__ or __FUNCTION__ even less so
2373 * - I think more goes on after the perlio frees but
2374 * the thing is that STDERR gets closed (as do all
2375 * the file descriptors)
2376 * - no deeper calling stack than the caller of the Newx()
2377 * or the kind, but do I look like a C reflection/introspection
2378 * utility to you?
2379 * - the function prototypes for the logging functions
2380 * probably should maybe be somewhere else than handy.h
2381 * - one could consider inlining (macrofying) the logging
2382 * for speed, but I am too lazy
2383 * - one could imagine recording the allocations in a hash,
2384 * (keyed by the allocation address?), and maintain that
2385 * through reallocs and frees, but how to do that without
2386 * any News() happening...?
73d1d973 2387 * - lots of -Ddefines to get useful/controllable output
b953482e 2388 * - lots of ENV reads
46c6c7e2
JH
2389 */
2390
0b0ab801 2391# ifdef PERL_CORE
73d1d973 2392# ifndef PERL_MEM_LOG_NOIMPL
0b0ab801
MHM
2393enum mem_log_type {
2394 MLT_ALLOC,
2395 MLT_REALLOC,
d7a2c63c
MHM
2396 MLT_FREE,
2397 MLT_NEW_SV,
2398 MLT_DEL_SV
0b0ab801
MHM
2399};
2400# endif
12754f92 2401# if defined(PERL_IN_SV_C) /* those are only used in sv.c */
d7a2c63c
MHM
2402void Perl_mem_log_new_sv(const SV *sv, const char *filename, const int linenumber, const char *funcname);
2403void Perl_mem_log_del_sv(const SV *sv, const char *filename, const int linenumber, const char *funcname);
12754f92 2404# endif
0b0ab801
MHM
2405# endif
2406
fe4f188c
JH
2407#endif
2408
2409#ifdef PERL_MEM_LOG
d1401ee9
MHM
2410#define MEM_LOG_ALLOC(n,t,a) Perl_mem_log_alloc(n,sizeof(t),STRINGIFY(t),a,__FILE__,__LINE__,FUNCTION__)
2411#define MEM_LOG_REALLOC(n,t,v,a) Perl_mem_log_realloc(n,sizeof(t),STRINGIFY(t),v,a,__FILE__,__LINE__,FUNCTION__)
46c6c7e2 2412#define MEM_LOG_FREE(a) Perl_mem_log_free(a,__FILE__,__LINE__,FUNCTION__)
fe4f188c
JH
2413#endif
2414
2415#ifndef MEM_LOG_ALLOC
2416#define MEM_LOG_ALLOC(n,t,a) (a)
2417#endif
2418#ifndef MEM_LOG_REALLOC
2419#define MEM_LOG_REALLOC(n,t,v,a) (a)
2420#endif
2421#ifndef MEM_LOG_FREE
2422#define MEM_LOG_FREE(a) (a)
2423#endif
2424
d1401ee9
MHM
2425#define Newx(v,n,t) (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_ALLOC(n,t,safemalloc((MEM_SIZE)((n)*sizeof(t))))))
2426#define Newxc(v,n,t,c) (v = (MEM_WRAP_CHECK_(n,t) (c*)MEM_LOG_ALLOC(n,t,safemalloc((MEM_SIZE)((n)*sizeof(t))))))
2427#define Newxz(v,n,t) (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_ALLOC(n,t,safecalloc((n),sizeof(t)))))
a6f6820f
NC
2428
2429#ifndef PERL_CORE
a02a5408
JC
2430/* pre 5.9.x compatibility */
2431#define New(x,v,n,t) Newx(v,n,t)
2432#define Newc(x,v,n,t,c) Newxc(v,n,t,c)
4541904d 2433#define Newz(x,v,n,t) Newxz(v,n,t)
a6f6820f 2434#endif
a02a5408 2435
ff68c719 2436#define Renew(v,n,t) \
d1401ee9 2437 (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_REALLOC(n,t,v,saferealloc((Malloc_t)(v),(MEM_SIZE)((n)*sizeof(t))))))
ff68c719 2438#define Renewc(v,n,t,c) \
d1401ee9 2439 (v = (MEM_WRAP_CHECK_(n,t) (c*)MEM_LOG_REALLOC(n,t,v,saferealloc((Malloc_t)(v),(MEM_SIZE)((n)*sizeof(t))))))
94010e71
NC
2440
2441#ifdef PERL_POISON
2442#define Safefree(d) \
06c0cc96 2443 ((d) ? (void)(safefree(MEM_LOG_FREE((Malloc_t)(d))), Poison(&(d), 1, Malloc_t)) : (void) 0)
94010e71 2444#else
fe4f188c 2445#define Safefree(d) safefree(MEM_LOG_FREE((Malloc_t)(d)))
94010e71 2446#endif
55497cff 2447
45908e4d 2448#define perl_assert_ptr(p) assert( ((void*)(p)) != 0 )
55497cff 2449
45908e4d
YO
2450
2451#define Move(s,d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), perl_assert_ptr(s), (void)memmove((char*)(d),(const char*)(s), (n) * sizeof(t)))
2452#define Copy(s,d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), perl_assert_ptr(s), (void)memcpy((char*)(d),(const char*)(s), (n) * sizeof(t)))
2453#define Zero(d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), (void)memzero((char*)(d), (n) * sizeof(t)))
2454
2455#define MoveD(s,d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), perl_assert_ptr(s), memmove((char*)(d),(const char*)(s), (n) * sizeof(t)))
2456#define CopyD(s,d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), perl_assert_ptr(s), memcpy((char*)(d),(const char*)(s), (n) * sizeof(t)))
45908e4d 2457#define ZeroD(d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), memzero((char*)(d), (n) * sizeof(t)))
e90e2364 2458
7e337ee0
JH
2459#define PoisonWith(d,n,t,b) (MEM_WRAP_CHECK_(n,t) (void)memset((char*)(d), (U8)(b), (n) * sizeof(t)))
2460#define PoisonNew(d,n,t) PoisonWith(d,n,t,0xAB)
2461#define PoisonFree(d,n,t) PoisonWith(d,n,t,0xEF)
2462#define Poison(d,n,t) PoisonFree(d,n,t)
27d5b266 2463
caa674f3
DD
2464#ifdef PERL_POISON
2465# define PERL_POISON_EXPR(x) x
2466#else
2467# define PERL_POISON_EXPR(x)
2468#endif
2469
ff68c719 2470#define StructCopy(s,d,t) (*((t*)(d)) = *((t*)(s)))
2cc61e15 2471
c3caa5c3
JH
2472/* C_ARRAY_LENGTH is the number of elements in the C array (so you
2473 * want your zero-based indices to be less than but not equal to).
2474 *
2475 * C_ARRAY_END is one past the last: half-open/half-closed range,
2476 * not last-inclusive range. */
622913ab 2477#define C_ARRAY_LENGTH(a) (sizeof(a)/sizeof((a)[0]))
c3caa5c3 2478#define C_ARRAY_END(a) ((a) + C_ARRAY_LENGTH(a))
622913ab 2479
2cc61e15
DD
2480#ifdef NEED_VA_COPY
2481# ifdef va_copy
2482# define Perl_va_copy(s, d) va_copy(d, s)
07798b17
AC
2483# elif defined(__va_copy)
2484# define Perl_va_copy(s, d) __va_copy(d, s)
2cc61e15 2485# else
07798b17 2486# define Perl_va_copy(s, d) Copy(s, d, 1, va_list)
2cc61e15
DD
2487# endif
2488#endif
2489
472d47bc
SB
2490/* convenience debug macros */
2491#ifdef USE_ITHREADS
2492#define pTHX_FORMAT "Perl interpreter: 0x%p"
2493#define pTHX__FORMAT ", Perl interpreter: 0x%p"
f54cb97a
AL
2494#define pTHX_VALUE_ (void *)my_perl,
2495#define pTHX_VALUE (void *)my_perl
2496#define pTHX__VALUE_ ,(void *)my_perl,
2497#define pTHX__VALUE ,(void *)my_perl
472d47bc 2498#else
3609ea0d 2499#define pTHX_FORMAT
472d47bc 2500#define pTHX__FORMAT
3609ea0d 2501#define pTHX_VALUE_
472d47bc 2502#define pTHX_VALUE
3609ea0d 2503#define pTHX__VALUE_
472d47bc
SB
2504#define pTHX__VALUE
2505#endif /* USE_ITHREADS */
3609ea0d 2506
2acdbac1
NC
2507/* Perl_deprecate was not part of the public API, and did not have a deprecate()
2508 shortcut macro defined without -DPERL_CORE. Neither codesearch.google.com nor
2509 CPAN::Unpack show any users outside the core. */
2510#ifdef PERL_CORE
dc6e8de0
A
2511# define deprecate(s) Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2512 "Use of " s " is deprecated")
c9680906
A
2513# define deprecate_disappears_in(when,message) \
2514 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2515 message ", and will disappear in Perl " when)
ac641426
A
2516# define deprecate_fatal_in(when,message) \
2517 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2518 message ". Its use will be fatal in Perl " when)
2acdbac1
NC
2519#endif
2520
dfff4baf
BF
2521/* Internal macros to deal with gids and uids */
2522#ifdef PERL_CORE
2523
2524# if Uid_t_size > IVSIZE
2525# define sv_setuid(sv, uid) sv_setnv((sv), (NV)(uid))
2526# define SvUID(sv) SvNV(sv)
07798b17
AC
2527# elif Uid_t_sign <= 0
2528# define sv_setuid(sv, uid) sv_setiv((sv), (IV)(uid))
2529# define SvUID(sv) SvIV(sv)
dfff4baf 2530# else
07798b17
AC
2531# define sv_setuid(sv, uid) sv_setuv((sv), (UV)(uid))
2532# define SvUID(sv) SvUV(sv)
dfff4baf
BF
2533# endif /* Uid_t_size */
2534
2535# if Gid_t_size > IVSIZE
2536# define sv_setgid(sv, gid) sv_setnv((sv), (NV)(gid))
2537# define SvGID(sv) SvNV(sv)
07798b17
AC
2538# elif Gid_t_sign <= 0
2539# define sv_setgid(sv, gid) sv_setiv((sv), (IV)(gid))
2540# define SvGID(sv) SvIV(sv)
dfff4baf 2541# else
07798b17
AC
2542# define sv_setgid(sv, gid) sv_setuv((sv), (UV)(gid))
2543# define SvGID(sv) SvUV(sv)
dfff4baf
BF
2544# endif /* Gid_t_size */
2545
2546#endif
2547
6a5bc5ac 2548#endif /* PERL_HANDY_H_ */
9d745869 2549
e9a8c099 2550/*
14d04a33 2551 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 2552 */