This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Check for Unicode replacement character in t/porting/podcheck.t
[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
85e6fe83 17#if !defined(__STDC__)
378cc40b
LW
18#ifdef NULL
19#undef NULL
20#endif
a687059c 21# define NULL 0
85e6fe83
LW
22#endif
23
24792b8d
NC
24#ifndef PERL_CORE
25# define Null(type) ((type)NULL)
954c1994
GS
26
27/*
ccfc67b7 28=head1 Handy Values
954c1994 29
3609ea0d 30=for apidoc AmU||Nullch
72d33970
FC
31Null character pointer. (No longer available when C<PERL_CORE> is
32defined.)
2307c6d0 33
954c1994 34=for apidoc AmU||Nullsv
72d33970 35Null SV pointer. (No longer available when C<PERL_CORE> is defined.)
954c1994
GS
36
37=cut
38*/
39
24792b8d
NC
40# define Nullch Null(char*)
41# define Nullfp Null(PerlIO*)
42# define Nullsv Null(SV*)
43#endif
8d063cd8 44
641d3f0b
PP
45#ifdef TRUE
46#undef TRUE
47#endif
48#ifdef FALSE
49#undef FALSE
50#endif
51#define TRUE (1)
52#define FALSE (0)
53
cf3f0ffb
DM
54/* The MUTABLE_*() macros cast pointers to the types shown, in such a way
55 * (compiler permitting) that casting away const-ness will give a warning;
56 * e.g.:
57 *
58 * const SV *sv = ...;
59 * AV *av1 = (AV*)sv; <== BAD: the const has been silently cast away
60 * AV *av2 = MUTABLE_AV(sv); <== GOOD: it may warn
61 */
62
b1bc3f34 63#if defined(__GNUC__) && !defined(PERL_GCC_BRACE_GROUPS_FORBIDDEN)
f065df04 64# define MUTABLE_PTR(p) ({ void *_p = (p); _p; })
b1bc3f34
NC
65#else
66# define MUTABLE_PTR(p) ((void *) (p))
67#endif
68
a062e10d 69#define MUTABLE_AV(p) ((AV *)MUTABLE_PTR(p))
ea726b52 70#define MUTABLE_CV(p) ((CV *)MUTABLE_PTR(p))
159b6efe 71#define MUTABLE_GV(p) ((GV *)MUTABLE_PTR(p))
dbebbdb4 72#define MUTABLE_HV(p) ((HV *)MUTABLE_PTR(p))
a45c7426 73#define MUTABLE_IO(p) ((IO *)MUTABLE_PTR(p))
b1bc3f34 74#define MUTABLE_SV(p) ((SV *)MUTABLE_PTR(p))
27d4fb96 75
f789f6a4 76#if defined(I_STDBOOL) && !defined(PERL_BOOL_AS_CHAR)
bd31be4b
NC
77# include <stdbool.h>
78# ifndef HAS_BOOL
79# define HAS_BOOL 1
80# endif
81#endif
82
8e84507e 83/* bool is built-in for g++-2.6.3 and later, which might be used
c1d22f6b
GS
84 for extensions. <_G_config.h> defines _G_HAVE_BOOL, but we can't
85 be sure _G_config.h will be included before this file. _G_config.h
8e84507e 86 also defines _G_HAVE_BOOL for both gcc and g++, but only g++
c1d22f6b
GS
87 actually has bool. Hence, _G_HAVE_BOOL is pretty useless for us.
88 g++ can be identified by __GNUG__.
89 Andy Dougherty February 2000
5d94fbed 90*/
3609ea0d 91#ifdef __GNUG__ /* GNU g++ has bool built-in */
f789f6a4 92# ifndef PERL_BOOL_AS_CHAR
5d94fbed 93# ifndef HAS_BOOL
c1d22f6b 94# define HAS_BOOL 1
5d94fbed 95# endif
f789f6a4 96# endif
5d94fbed
AD
97#endif
98
99#ifndef HAS_BOOL
f789f6a4
FC
100# ifdef bool
101# undef bool
102# endif
70d5cb32 103# define bool char
c1d22f6b 104# define HAS_BOOL 1
a687059c 105#endif
0d3e774c 106
236415e2
KW
107/* cast-to-bool. A simple (bool) cast may not do the right thing: if bool is
108 * defined as char for example, then the cast from int is
109 * implementation-defined (bool)!!(cbool) in a ternary triggers a bug in xlc on
110 * AIX */
18f5643b 111#define cBOOL(cbool) ((cbool) ? (bool)1 : (bool)0)
f2338a2e 112
46c6c7e2 113/* Try to figure out __func__ or __FUNCTION__ equivalent, if any.
e352bcff
JH
114 * XXX Should really be a Configure probe, with HAS__FUNCTION__
115 * and FUNCTION__ as results.
116 * XXX Similarly, a Configure probe for __FILE__ and __LINE__ is needed. */
46c6c7e2
JH
117#if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L) || (defined(__SUNPRO_C)) /* C99 or close enough. */
118# define FUNCTION__ __func__
119#else
9ea40801 120# if (defined(USING_MSVC6)) || /* MSVC6 has neither __func__ nor __FUNCTION and no good workarounds, either. */ \
be73ebb0 121 (defined(__DECC_VER)) /* Tru64 or VMS, and strict C89 being used, but not modern enough cc (in Tur64, -c99 not known, only -std1). */
46c6c7e2
JH
122# define FUNCTION__ ""
123# else
124# define FUNCTION__ __FUNCTION__ /* Common extension. */
125# endif
126#endif
127
27d4fb96
PP
128/* XXX A note on the perl source internal type system. The
129 original intent was that I32 be *exactly* 32 bits.
130
131 Currently, we only guarantee that I32 is *at least* 32 bits.
132 Specifically, if int is 64 bits, then so is I32. (This is the case
133 for the Cray.) This has the advantage of meshing nicely with
134 standard library calls (where we pass an I32 and the library is
135 expecting an int), but the disadvantage that an I32 is not 32 bits.
136 Andy Dougherty August 1996
24fef2a7 137
dc45a647
MB
138 There is no guarantee that there is *any* integral type with
139 exactly 32 bits. It is perfectly legal for a system to have
140 sizeof(short) == sizeof(int) == sizeof(long) == 8.
693762b4 141
dc45a647
MB
142 Similarly, there is no guarantee that I16 and U16 have exactly 16
143 bits.
693762b4 144
8e84507e
NIS
145 For dealing with issues that may arise from various 32/64-bit
146 systems, we will ask Configure to check out
8175356b 147
3609ea0d
JH
148 SHORTSIZE == sizeof(short)
149 INTSIZE == sizeof(int)
150 LONGSIZE == sizeof(long)
dc45a647 151 LONGLONGSIZE == sizeof(long long) (if HAS_LONG_LONG)
3609ea0d 152 PTRSIZE == sizeof(void *)
dc45a647
MB
153 DOUBLESIZE == sizeof(double)
154 LONG_DOUBLESIZE == sizeof(long double) (if HAS_LONG_DOUBLE).
8175356b 155
27d4fb96
PP
156*/
157
69512466
JH
158#ifdef I_INTTYPES /* e.g. Linux has int64_t without <inttypes.h> */
159# include <inttypes.h>
dd0eed91
JH
160# ifdef INT32_MIN_BROKEN
161# undef INT32_MIN
162# define INT32_MIN (-2147483647-1)
163# endif
164# ifdef INT64_MIN_BROKEN
165# undef INT64_MIN
166# define INT64_MIN (-9223372036854775807LL-1)
167# endif
69512466
JH
168#endif
169
8175356b
JH
170typedef I8TYPE I8;
171typedef U8TYPE U8;
172typedef I16TYPE I16;
173typedef U16TYPE U16;
174typedef I32TYPE I32;
175typedef U32TYPE U32;
16d89be8
JH
176
177#ifdef HAS_QUAD
8175356b
JH
178typedef I64TYPE I64;
179typedef U64TYPE U64;
16d89be8 180#endif
8175356b 181
9c39b21c
JH
182/* INT64_C/UINT64_C are C99 from <stdint.h> (so they will not be
183 * available in strict C89 mode), but they are nice, so let's define
184 * them if necessary. */
0d3bd0a9 185#if defined(HAS_QUAD)
9c39b21c
JH
186# undef PeRl_INT64_C
187# undef PeRl_UINT64_C
188/* Prefer the native integer types (int and long) over long long
189 * (which is not C89) and Win32-specific __int64. */
190# if QUADKIND == QUAD_IS_INT && INTSIZE == 8
191# define PeRl_INT64_C(c) (c)
192# define PeRl_UINT64_C(c) CAT2(c,U)
193# endif
194# if QUADKIND == QUAD_IS_LONG && LONGSIZE == 8
195# define PeRl_INT64_C(c) CAT2(c,L)
196# define PeRl_UINT64_C(c) CAT2(c,UL)
197# endif
198# if QUADKIND == QUAD_IS_LONG_LONG && defined(HAS_LONG_LONG)
199# define PeRl_INT64_C(c) CAT2(c,LL)
200# define PeRl_UINT64_C(c) CAT2(c,ULL)
201# endif
202# if QUADKIND == QUAD_IS___INT64
203# define PeRl_INT64_C(c) CAT2(c,I64)
204# define PeRl_UINT64_C(c) CAT2(c,UI64)
205# endif
206# ifndef PeRl_INT64_C
702c92eb
JH
207# define PeRl_INT64_C(c) ((I64)(c)) /* last resort */
208# define PeRl_UINT64_C(c) ((U64)(c))
9c39b21c
JH
209# endif
210/* In OS X the INT64_C/UINT64_C are defined with LL/ULL, which will
211 * not fly with C89-pedantic gcc, so let's undefine them first so that
212 * we can redefine them with our native integer preferring versions. */
8eafc6eb 213# if defined(PERL_DARWIN) && defined(PERL_GCC_PEDANTIC)
9c39b21c
JH
214# undef INT64_C
215# undef UINT64_C
216# endif
217# ifndef INT64_C
218# define INT64_C(c) PeRl_INT64_C(c)
219# endif
220# ifndef UINT64_C
221# define UINT64_C(c) PeRl_UINT64_C(c)
222# endif
e8c95190 223#endif
dc750f2e 224
d8668976 225#if defined(UINT8_MAX) && defined(INT16_MAX) && defined(INT32_MAX)
5ff3f7a4 226
5ff3f7a4
GS
227/* I8_MAX and I8_MIN constants are not defined, as I8 is an ambiguous type.
228 Please search CHAR_MAX in perl.h for further details. */
229#define U8_MAX UINT8_MAX
230#define U8_MIN UINT8_MIN
231
5ff3f7a4
GS
232#define I16_MAX INT16_MAX
233#define I16_MIN INT16_MIN
234#define U16_MAX UINT16_MAX
235#define U16_MIN UINT16_MIN
236
5ff3f7a4
GS
237#define I32_MAX INT32_MAX
238#define I32_MIN INT32_MIN
0e983133
GS
239#ifndef UINT32_MAX_BROKEN /* e.g. HP-UX with gcc messes this up */
240# define U32_MAX UINT32_MAX
241#else
242# define U32_MAX 4294967295U
243#endif
5ff3f7a4
GS
244#define U32_MIN UINT32_MIN
245
246#else
247
5c9fa16e
KA
248/* I8_MAX and I8_MIN constants are not defined, as I8 is an ambiguous type.
249 Please search CHAR_MAX in perl.h for further details. */
27d4fb96
PP
250#define U8_MAX PERL_UCHAR_MAX
251#define U8_MIN PERL_UCHAR_MIN
79072805 252
27d4fb96
PP
253#define I16_MAX PERL_SHORT_MAX
254#define I16_MIN PERL_SHORT_MIN
255#define U16_MAX PERL_USHORT_MAX
256#define U16_MIN PERL_USHORT_MIN
79072805 257
c4f23d77 258#if LONGSIZE > 4
27d4fb96
PP
259# define I32_MAX PERL_INT_MAX
260# define I32_MIN PERL_INT_MIN
261# define U32_MAX PERL_UINT_MAX
262# define U32_MIN PERL_UINT_MIN
79072805 263#else
27d4fb96
PP
264# define I32_MAX PERL_LONG_MAX
265# define I32_MIN PERL_LONG_MIN
266# define U32_MAX PERL_ULONG_MAX
267# define U32_MIN PERL_ULONG_MIN
79072805
LW
268#endif
269
5ff3f7a4
GS
270#endif
271
58a9a5d5 272/* log(2) is pretty close to 0.30103, just in case anyone is grepping for it */
fc36a67e
PP
273#define BIT_DIGITS(N) (((N)*146)/485 + 1) /* log2(10) =~ 146/485 */
274#define TYPE_DIGITS(T) BIT_DIGITS(sizeof(T) * 8)
275#define TYPE_CHARS(T) (TYPE_DIGITS(T) + 2) /* sign, NUL */
276
88794300 277/* Unused by core; should be deprecated */
ff68c719 278#define Ctl(ch) ((ch) & 037)
8d063cd8 279
98fce2a4
KW
280#if defined(PERL_CORE) || defined(PERL_EXT)
281# ifndef MIN
282# define MIN(a,b) ((a) < (b) ? (a) : (b))
283# endif
284# ifndef MAX
285# define MAX(a,b) ((a) > (b) ? (a) : (b))
286# endif
287#endif
288
4650c663
KW
289/* This is a helper macro to avoid preprocessor issues, replaced by nothing
290 * unless under DEBUGGING, where it expands to an assert of its argument,
291 * followed by a comma (hence the comma operator). If we just used a straight
3e94db23
JH
292 * assert(), we would get a comma with nothing before it when not DEBUGGING.
293 *
294 * We also use empty definition under Coverity since the __ASSERT__
295 * checks often check for things that Really Cannot Happen, and Coverity
296 * detects that and gets all excited. */
297
298#if defined(DEBUGGING) && !defined(__COVERITY__)
0f092d08
KW
299# define __ASSERT_(statement) assert(statement),
300#else
301# define __ASSERT_(statement)
302#endif
303
3fe05580
MHM
304/*
305=head1 SV-Body Allocation
306
307=for apidoc Ama|SV*|newSVpvs|const char* s
0c395ea5 308Like C<newSVpvn>, but takes a C<NUL>-terminated literal string instead of a
30a15352 309string/length pair.
3fe05580 310
84bafc02 311=for apidoc Ama|SV*|newSVpvs_flags|const char* s|U32 flags
0c395ea5 312Like C<newSVpvn_flags>, but takes a C<NUL>-terminated literal string instead of
30a15352 313a string/length pair.
84bafc02 314
3fe05580 315=for apidoc Ama|SV*|newSVpvs_share|const char* s
0c395ea5 316Like C<newSVpvn_share>, but takes a C<NUL>-terminated literal string instead of
30a15352 317a string/length pair and omits the hash parameter.
3fe05580 318
9dcc53ea 319=for apidoc Am|void|sv_catpvs_flags|SV* sv|const char* s|I32 flags
0c395ea5 320Like C<sv_catpvn_flags>, but takes a C<NUL>-terminated literal string instead
30a15352 321of a string/length pair.
9dcc53ea
Z
322
323=for apidoc Am|void|sv_catpvs_nomg|SV* sv|const char* s
0c395ea5
KW
324Like C<sv_catpvn_nomg>, but takes a C<NUL>-terminated literal string instead of
325a string/length pair.
9dcc53ea 326
7b4e7469 327=for apidoc Am|void|sv_catpvs|SV* sv|const char* s
0c395ea5
KW
328Like C<sv_catpvn>, but takes a C<NUL>-terminated literal string instead of a
329string/length pair.
3fe05580 330
9dcc53ea 331=for apidoc Am|void|sv_catpvs_mg|SV* sv|const char* s
0c395ea5 332Like C<sv_catpvn_mg>, but takes a C<NUL>-terminated literal string instead of a
9dcc53ea
Z
333string/length pair.
334
7b4e7469 335=for apidoc Am|void|sv_setpvs|SV* sv|const char* s
0c395ea5
KW
336Like C<sv_setpvn>, but takes a C<NUL>-terminated literal string instead of a
337string/length pair.
3fe05580 338
9dcc53ea 339=for apidoc Am|void|sv_setpvs_mg|SV* sv|const char* s
0c395ea5 340Like C<sv_setpvn_mg>, but takes a C<NUL>-terminated literal string instead of a
9dcc53ea
Z
341string/length pair.
342
343=for apidoc Am|SV *|sv_setref_pvs|const char* s
0c395ea5
KW
344Like C<sv_setref_pvn>, but takes a C<NUL>-terminated literal string instead of
345a string/length pair.
9dcc53ea 346
3fe05580
MHM
347=head1 Memory Management
348
349=for apidoc Ama|char*|savepvs|const char* s
0c395ea5 350Like C<savepvn>, but takes a C<NUL>-terminated literal string instead of a
30a15352 351string/length pair.
3fe05580 352
9dcc53ea
Z
353=for apidoc Ama|char*|savesharedpvs|const char* s
354A version of C<savepvs()> which allocates the duplicate string in memory
355which is shared between threads.
356
3fe05580
MHM
357=head1 GV Functions
358
359=for apidoc Am|HV*|gv_stashpvs|const char* name|I32 create
0c395ea5
KW
360Like C<gv_stashpvn>, but takes a C<NUL>-terminated literal string instead of a
361string/length pair.
3fe05580
MHM
362
363=head1 Hash Manipulation Functions
364
365=for apidoc Am|SV**|hv_fetchs|HV* tb|const char* key|I32 lval
0c395ea5
KW
366Like C<hv_fetch>, but takes a C<NUL>-terminated literal string instead of a
367string/length pair.
3fe05580
MHM
368
369=for apidoc Am|SV**|hv_stores|HV* tb|const char* key|NULLOK SV* val
0c395ea5
KW
370Like C<hv_store>, but takes a C<NUL>-terminated literal string instead of a
371string/length pair
3fe05580
MHM
372and omits the hash parameter.
373
510966aa
Z
374=head1 Lexer interface
375
376=for apidoc Amx|void|lex_stuff_pvs|const char *pv|U32 flags
377
0c395ea5
KW
378Like L</lex_stuff_pvn>, but takes a C<NUL>-terminated literal string instead of
379a string/length pair.
510966aa 380
3fe05580
MHM
381=cut
382*/
383
926bb54c
KW
384/* concatenating with "" ensures that only literal strings are accepted as
385 * argument */
26bfd924 386#define STR_WITH_LEN(s) ("" s ""), (sizeof(s)-1)
2efa8cc7 387
926bb54c
KW
388/* note that STR_WITH_LEN() can't be used as argument to macros or functions
389 * that under some configurations might be macros, which means that it requires
390 * the full Perl_xxx(aTHX_ ...) form for any API calls where it's used.
ba3a79e7
GA
391 */
392
393/* STR_WITH_LEN() shortcuts */
394#define newSVpvs(str) Perl_newSVpvn(aTHX_ STR_WITH_LEN(str))
84bafc02
NC
395#define newSVpvs_flags(str,flags) \
396 Perl_newSVpvn_flags(aTHX_ STR_WITH_LEN(str), flags)
ba3a79e7 397#define newSVpvs_share(str) Perl_newSVpvn_share(aTHX_ STR_WITH_LEN(str), 0)
9dcc53ea
Z
398#define sv_catpvs_flags(sv, str, flags) \
399 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), flags)
400#define sv_catpvs_nomg(sv, str) \
401 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), 0)
402#define sv_catpvs(sv, str) \
403 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), SV_GMAGIC)
404#define sv_catpvs_mg(sv, str) \
405 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), SV_GMAGIC|SV_SMAGIC)
3fe05580 406#define sv_setpvs(sv, str) Perl_sv_setpvn(aTHX_ sv, STR_WITH_LEN(str))
9dcc53ea
Z
407#define sv_setpvs_mg(sv, str) Perl_sv_setpvn_mg(aTHX_ sv, STR_WITH_LEN(str))
408#define sv_setref_pvs(rv, classname, str) \
409 Perl_sv_setref_pvn(aTHX_ rv, classname, STR_WITH_LEN(str))
ba3a79e7 410#define savepvs(str) Perl_savepvn(aTHX_ STR_WITH_LEN(str))
9dcc53ea
Z
411#define savesharedpvs(str) Perl_savesharedpvn(aTHX_ STR_WITH_LEN(str))
412#define gv_stashpvs(str, create) \
413 Perl_gv_stashpvn(aTHX_ STR_WITH_LEN(str), create)
414#define gv_fetchpvs(namebeg, add, sv_type) \
415 Perl_gv_fetchpvn_flags(aTHX_ STR_WITH_LEN(namebeg), add, sv_type)
416#define gv_fetchpvn(namebeg, len, add, sv_type) \
417 Perl_gv_fetchpvn_flags(aTHX_ namebeg, len, add, sv_type)
418#define sv_catxmlpvs(dsv, str, utf8) \
419 Perl_sv_catxmlpvn(aTHX_ dsv, STR_WITH_LEN(str), utf8)
4ac46235 420
ba3a79e7 421
510966aa
Z
422#define lex_stuff_pvs(pv,flags) Perl_lex_stuff_pvn(aTHX_ STR_WITH_LEN(pv), flags)
423
b96d8cd9
NC
424#define get_cvs(str, flags) \
425 Perl_get_cvn_flags(aTHX_ STR_WITH_LEN(str), (flags))
5c1737d1 426
954c1994 427/*
ccfc67b7
JH
428=head1 Miscellaneous Functions
429
954c1994 430=for apidoc Am|bool|strNE|char* s1|char* s2
dc6b0978
KW
431Test two C<NUL>-terminated strings to see if they are different. Returns true
432or false.
954c1994
GS
433
434=for apidoc Am|bool|strEQ|char* s1|char* s2
dc6b0978
KW
435Test two C<NUL>-terminated strings to see if they are equal. Returns true or
436false.
954c1994
GS
437
438=for apidoc Am|bool|strLT|char* s1|char* s2
dc6b0978
KW
439Test two C<NUL>-terminated strings to see if the first, C<s1>, is less than the
440second, C<s2>. Returns true or false.
954c1994
GS
441
442=for apidoc Am|bool|strLE|char* s1|char* s2
dc6b0978
KW
443Test two C<NUL>-terminated strings to see if the first, C<s1>, is less than or
444equal to the second, C<s2>. Returns true or false.
954c1994
GS
445
446=for apidoc Am|bool|strGT|char* s1|char* s2
dc6b0978
KW
447Test two C<NUL>-terminated strings to see if the first, C<s1>, is greater than
448the second, C<s2>. Returns true or false.
954c1994
GS
449
450=for apidoc Am|bool|strGE|char* s1|char* s2
dc6b0978
KW
451Test two C<NUL>-terminated strings to see if the first, C<s1>, is greater than
452or equal to the second, C<s2>. Returns true or false.
954c1994
GS
453
454=for apidoc Am|bool|strnNE|char* s1|char* s2|STRLEN len
dc6b0978
KW
455Test two C<NUL>-terminated strings to see if they are different. The C<len>
456parameter indicates the number of bytes to compare. Returns true or false. (A
954c1994
GS
457wrapper for C<strncmp>).
458
459=for apidoc Am|bool|strnEQ|char* s1|char* s2|STRLEN len
dc6b0978
KW
460Test two C<NUL>-terminated strings to see if they are equal. The C<len>
461parameter indicates the number of bytes to compare. Returns true or false. (A
462wrapper for C<strncmp>).
954c1994 463
bd18bd40
KW
464=for apidoc Am|bool|memEQ|char* s1|char* s2|STRLEN len
465Test two buffers (which may contain embedded C<NUL> characters, to see if they
466are equal. The C<len> parameter indicates the number of bytes to compare.
467Returns zero if equal, or non-zero if non-equal.
468
469=for apidoc Am|bool|memNE|char* s1|char* s2|STRLEN len
470Test two buffers (which may contain embedded C<NUL> characters, to see if they
471are not equal. The C<len> parameter indicates the number of bytes to compare.
472Returns zero if non-equal, or non-zero if equal.
473
954c1994
GS
474=cut
475*/
476
62946e08 477
8d063cd8
LW
478#define strNE(s1,s2) (strcmp(s1,s2))
479#define strEQ(s1,s2) (!strcmp(s1,s2))
480#define strLT(s1,s2) (strcmp(s1,s2) < 0)
481#define strLE(s1,s2) (strcmp(s1,s2) <= 0)
482#define strGT(s1,s2) (strcmp(s1,s2) > 0)
483#define strGE(s1,s2) (strcmp(s1,s2) >= 0)
62946e08 484
8d063cd8
LW
485#define strnNE(s1,s2,l) (strncmp(s1,s2,l))
486#define strnEQ(s1,s2,l) (!strncmp(s1,s2,l))
378cc40b 487
8f716499
KW
488/* These names are controversial, so guarding against their being used in more
489 * places than they already are. strBEGs and StrStartsWith are potential
490 * candidates */
491#if defined(PERL_IN_DOIO_C) || defined(PERL_IN_GV_C) || defined(PERL_IN_HV_C) || defined(PERL_IN_LOCALE_C) || defined(PERL_IN_PERL_C) || defined(PERL_IN_TOKE_C) || defined(PERL_EXT)
62946e08
YO
492#define strNEs(s1,s2) (strncmp(s1,"" s2 "", sizeof(s2)-1))
493#define strEQs(s1,s2) (!strncmp(s1,"" s2 "", sizeof(s2)-1))
8f716499 494#endif
62946e08 495
36477c24
PP
496#ifdef HAS_MEMCMP
497# define memNE(s1,s2,l) (memcmp(s1,s2,l))
498# define memEQ(s1,s2,l) (!memcmp(s1,s2,l))
499#else
500# define memNE(s1,s2,l) (bcmp(s1,s2,l))
501# define memEQ(s1,s2,l) (!bcmp(s1,s2,l))
502#endif
503
085b7534 504/* memEQ and memNE where second comparand is a string constant */
568a785a 505#define memEQs(s1, l, s2) \
777fa2cb 506 (((sizeof(s2)-1) == (l)) && memEQ((s1), ("" s2 ""), (sizeof(s2)-1)))
568a785a
NC
507#define memNEs(s1, l, s2) !memEQs(s1, l, s2)
508
777fa2cb
YO
509/* memEQ and memNE where second comparand is a string constant
510 * and we can assume the length of s1 is at least that of the string */
511#define _memEQs(s1, s2) \
512 (memEQ((s1), ("" s2 ""), (sizeof(s2)-1)))
513#define _memNEs(s1, s2) (memNE((s1),("" s2 ""),(sizeof(s2)-1)))
514
062b6850
KW
515#define memLT(s1,s2,l) (memcmp(s1,s2,l) < 0)
516#define memLE(s1,s2,l) (memcmp(s1,s2,l) <= 0)
517#define memGT(s1,s2,l) (memcmp(s1,s2,l) > 0)
518#define memGE(s1,s2,l) (memcmp(s1,s2,l) >= 0)
519
bbce6d69
PP
520/*
521 * Character classes.
522 *
523 * Unfortunately, the introduction of locales means that we
524 * can't trust isupper(), etc. to tell the truth. And when
525 * it comes to /\w+/ with tainting enabled, we *must* be able
526 * to trust our character classes.
527 *
528 * Therefore, the default tests in the text of Perl will be
529 * independent of locale. Any code that wants to depend on
530 * the current locale will use the tests that begin with "lc".
531 */
532
2304df62
AD
533#ifdef HAS_SETLOCALE /* XXX Is there a better test for this? */
534# ifndef CTYPE256
535# define CTYPE256
536# endif
537#endif
538
954c1994 539/*
ccfc67b7 540
dcccc8ff 541=head1 Character classification
243effed
KW
542This section is about functions (really macros) that classify characters
543into types, such as punctuation versus alphabetic, etc. Most of these are
544analogous to regular expression character classes. (See
545L<perlrecharclass/POSIX Character Classes>.) There are several variants for
546each class. (Not all macros have all variants; each item below lists the
547ones valid for it.) None are affected by C<use bytes>, and only the ones
548with C<LC> in the name are affected by the current locale.
549
550The base function, e.g., C<isALPHA()>, takes an octet (either a C<char> or a
551C<U8>) as input and returns a boolean as to whether or not the character
6aff1f14
KW
552represented by that octet is (or on non-ASCII platforms, corresponds to) an
553ASCII character in the named class based on platform, Unicode, and Perl rules.
554If the input is a number that doesn't fit in an octet, FALSE is returned.
243effed 555
c98722a4 556Variant C<isI<FOO>_A> (e.g., C<isALPHA_A()>) is identical to the base function
550da823
KW
557with no suffix C<"_A">. This variant is used to emphasize by its name that
558only ASCII-range characters can return TRUE.
4b9734bf 559
d60679e1 560Variant C<isI<FOO>_L1> imposes the Latin-1 (or EBCDIC equivalent) character set
4b9734bf
KW
561onto the platform. That is, the code points that are ASCII are unaffected,
562since ASCII is a subset of Latin-1. But the non-ASCII code points are treated
563as if they are Latin-1 characters. For example, C<isWORDCHAR_L1()> will return
564true when called with the code point 0xDF, which is a word character in both
4650c663 565ASCII and EBCDIC (though it represents different characters in each).
243effed 566
c98722a4 567Variant C<isI<FOO>_uvchr> is like the C<isI<FOO>_L1> variant, but accepts any UV code
243effed
KW
568point as input. If the code point is larger than 255, Unicode rules are used
569to determine if it is in the character class. For example,
d0da05db 570C<isWORDCHAR_uvchr(0x100)> returns TRUE, since 0x100 is LATIN CAPITAL LETTER A
6aff1f14 571WITH MACRON in Unicode, and is a word character.
243effed 572
c98722a4 573Variant C<isI<FOO>_utf8_safe> is like C<isI<FOO>_uvchr>, but is used for UTF-8
da8c1a98
KW
574encoded strings. Each call classifies one character, even if the string
575contains many. This variant takes two parameters. The first, C<p>, is a
576pointer to the first byte of the character to be classified. (Recall that it
577may take more than one byte to represent a character in UTF-8 strings.) The
578second parameter, C<e>, points to anywhere in the string beyond the first
579character, up to one byte past the end of the entire string. The suffix
580C<_safe> in the function's name indicates that it will not attempt to read
581beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is true (this
582is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the input
583character is malformed in some way, the program may croak, or the function may
584return FALSE, at the discretion of the implementation, and subject to change in
585future releases.
586
c98722a4 587Variant C<isI<FOO>_utf8> is like C<isI<FOO>_utf8_safe>, but takes just a single
da8c1a98 588parameter, C<p>, which has the same meaning as the corresponding parameter does
c98722a4 589in C<isI<FOO>_utf8_safe>. The function therefore can't check if it is reading
34aeb2e9 590beyond the end of the string. Starting in Perl v5.30, it will take a second
c98722a4 591parameter, becoming a synonym for C<isI<FOO>_utf8_safe>. At that time every
34aeb2e9 592program that uses it will have to be changed to successfully compile. In the
c98722a4 593meantime, the first runtime call to C<isI<FOO>_utf8> from each call point in the
34aeb2e9 594program will raise a deprecation warning, enabled by default. You can convert
c98722a4 595your program now to use C<isI<FOO>_utf8_safe>, and avoid the warnings, and get an
34aeb2e9
KW
596extra measure of protection, or you can wait until v5.30, when you'll be forced
597to add the C<e> parameter.
243effed 598
c98722a4 599Variant C<isI<FOO>_LC> is like the C<isI<FOO>_A> and C<isI<FOO>_L1> variants, but the
31f05a37
KW
600result is based on the current locale, which is what C<LC> in the name stands
601for. If Perl can determine that the current locale is a UTF-8 locale, it uses
602the published Unicode rules; otherwise, it uses the C library function that
603gives the named classification. For example, C<isDIGIT_LC()> when not in a
604UTF-8 locale returns the result of calling C<isdigit()>. FALSE is always
1a83413c
KW
605returned if the input won't fit into an octet. On some platforms where the C
606library function is known to be defective, Perl changes its result to follow
607the POSIX standard's rules.
243effed 608
c98722a4
KW
609Variant C<isI<FOO>_LC_uvchr> is like C<isI<FOO>_LC>, but is defined on any UV. It
610returns the same as C<isI<FOO>_LC> for input code points less than 256, and
243effed
KW
611returns the hard-coded, not-affected-by-locale, Unicode results for larger ones.
612
c98722a4 613Variant C<isI<FOO>_LC_utf8_safe> is like C<isI<FOO>_LC_uvchr>, but is used for UTF-8
da8c1a98
KW
614encoded strings. Each call classifies one character, even if the string
615contains many. This variant takes two parameters. The first, C<p>, is a
616pointer to the first byte of the character to be classified. (Recall that it
617may take more than one byte to represent a character in UTF-8 strings.) The
618second parameter, C<e>, points to anywhere in the string beyond the first
619character, up to one byte past the end of the entire string. The suffix
620C<_safe> in the function's name indicates that it will not attempt to read
621beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is true (this
622is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the input
623character is malformed in some way, the program may croak, or the function may
624return FALSE, at the discretion of the implementation, and subject to change in
625future releases.
626
c98722a4 627Variant C<isI<FOO>_LC_utf8> is like C<isI<FOO>_LC_utf8_safe>, but takes just a single
da8c1a98 628parameter, C<p>, which has the same meaning as the corresponding parameter does
c98722a4 629in C<isI<FOO>_LC_utf8_safe>. The function therefore can't check if it is reading
34aeb2e9 630beyond the end of the string. Starting in Perl v5.30, it will take a second
c98722a4 631parameter, becoming a synonym for C<isI<FOO>_LC_utf8_safe>. At that time every
34aeb2e9 632program that uses it will have to be changed to successfully compile. In the
c98722a4 633meantime, the first runtime call to C<isI<FOO>_LC_utf8> from each call point in
34aeb2e9 634the program will raise a deprecation warning, enabled by default. You can
c98722a4 635convert your program now to use C<isI<FOO>_LC_utf8_safe>, and avoid the warnings,
34aeb2e9
KW
636and get an extra measure of protection, or you can wait until v5.30, when
637you'll be forced to add the C<e> parameter.
ccfc67b7 638
954c1994 639=for apidoc Am|bool|isALPHA|char ch
2787a470 640Returns a boolean indicating whether the specified character is an
6aff1f14 641alphabetic character, analogous to C<m/[[:alpha:]]/>.
dcccc8ff
KW
642See the L<top of this section|/Character classification> for an explanation of
643variants
da8c1a98
KW
644C<isALPHA_A>, C<isALPHA_L1>, C<isALPHA_uvchr>, C<isALPHA_utf8_safe>,
645C<isALPHA_LC>, C<isALPHA_LC_uvchr>, and C<isALPHA_LC_utf8_safe>.
8a58bdcf 646
15861f94
KW
647=for apidoc Am|bool|isALPHANUMERIC|char ch
648Returns a boolean indicating whether the specified character is a either an
6aff1f14 649alphabetic character or decimal digit, analogous to C<m/[[:alnum:]]/>.
dcccc8ff
KW
650See the L<top of this section|/Character classification> for an explanation of
651variants
d0da05db 652C<isALPHANUMERIC_A>, C<isALPHANUMERIC_L1>, C<isALPHANUMERIC_uvchr>,
da8c1a98
KW
653C<isALPHANUMERIC_utf8_safe>, C<isALPHANUMERIC_LC>, C<isALPHANUMERIC_LC_uvchr>,
654and C<isALPHANUMERIC_LC_utf8_safe>.
15861f94 655
8a58bdcf
KW
656=for apidoc Am|bool|isASCII|char ch
657Returns a boolean indicating whether the specified character is one of the 128
243effed 658characters in the ASCII character set, analogous to C<m/[[:ascii:]]/>.
e5ad6aba 659On non-ASCII platforms, it returns TRUE iff this
8a58bdcf
KW
660character corresponds to an ASCII character. Variants C<isASCII_A()> and
661C<isASCII_L1()> are identical to C<isASCII()>.
dcccc8ff
KW
662See the L<top of this section|/Character classification> for an explanation of
663variants
da8c1a98
KW
664C<isASCII_uvchr>, C<isASCII_utf8_safe>, C<isASCII_LC>, C<isASCII_LC_uvchr>, and
665C<isASCII_LC_utf8_safe>. Note, however, that some platforms do not have the C
243effed
KW
666library routine C<isascii()>. In these cases, the variants whose names contain
667C<LC> are the same as the corresponding ones without.
668
d98532ea
KW
669Also note, that because all ASCII characters are UTF-8 invariant (meaning they
670have the exact same representation (always a single byte) whether encoded in
671UTF-8 or not), C<isASCII> will give the correct results when called with any
da8c1a98
KW
672byte in any string encoded or not in UTF-8. And similarly C<isASCII_utf8_safe>
673will work properly on any string encoded or not in UTF-8.
d98532ea 674
243effed
KW
675=for apidoc Am|bool|isBLANK|char ch
676Returns a boolean indicating whether the specified character is a
6aff1f14 677character considered to be a blank, analogous to C<m/[[:blank:]]/>.
dcccc8ff
KW
678See the L<top of this section|/Character classification> for an explanation of
679variants
da8c1a98
KW
680C<isBLANK_A>, C<isBLANK_L1>, C<isBLANK_uvchr>, C<isBLANK_utf8_safe>,
681C<isBLANK_LC>, C<isBLANK_LC_uvchr>, and C<isBLANK_LC_utf8_safe>. Note,
682however, that some platforms do not have the C library routine
683C<isblank()>. In these cases, the variants whose names contain C<LC> are
684the same as the corresponding ones without.
243effed
KW
685
686=for apidoc Am|bool|isCNTRL|char ch
687Returns a boolean indicating whether the specified character is a
6aff1f14 688control character, analogous to C<m/[[:cntrl:]]/>.
dcccc8ff
KW
689See the L<top of this section|/Character classification> for an explanation of
690variants
da8c1a98
KW
691C<isCNTRL_A>, C<isCNTRL_L1>, C<isCNTRL_uvchr>, C<isCNTRL_utf8_safe>,
692C<isCNTRL_LC>, C<isCNTRL_LC_uvchr>, and C<isCNTRL_LC_utf8_safe> On EBCDIC
693platforms, you almost always want to use the C<isCNTRL_L1> variant.
954c1994 694
954c1994 695=for apidoc Am|bool|isDIGIT|char ch
2787a470 696Returns a boolean indicating whether the specified character is a
6aff1f14 697digit, analogous to C<m/[[:digit:]]/>.
8a58bdcf 698Variants C<isDIGIT_A> and C<isDIGIT_L1> are identical to C<isDIGIT>.
dcccc8ff
KW
699See the L<top of this section|/Character classification> for an explanation of
700variants
da8c1a98
KW
701C<isDIGIT_uvchr>, C<isDIGIT_utf8_safe>, C<isDIGIT_LC>, C<isDIGIT_LC_uvchr>, and
702C<isDIGIT_LC_utf8_safe>.
243effed
KW
703
704=for apidoc Am|bool|isGRAPH|char ch
705Returns a boolean indicating whether the specified character is a
6aff1f14 706graphic character, analogous to C<m/[[:graph:]]/>.
dcccc8ff 707See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
708variants C<isGRAPH_A>, C<isGRAPH_L1>, C<isGRAPH_uvchr>, C<isGRAPH_utf8_safe>,
709C<isGRAPH_LC>, C<isGRAPH_LC_uvchr>, and C<isGRAPH_LC_utf8_safe>.
954c1994 710
0c82b6df 711=for apidoc Am|bool|isLOWER|char ch
2787a470 712Returns a boolean indicating whether the specified character is a
6aff1f14 713lowercase character, analogous to C<m/[[:lower:]]/>.
dcccc8ff
KW
714See the L<top of this section|/Character classification> for an explanation of
715variants
da8c1a98
KW
716C<isLOWER_A>, C<isLOWER_L1>, C<isLOWER_uvchr>, C<isLOWER_utf8_safe>,
717C<isLOWER_LC>, C<isLOWER_LC_uvchr>, and C<isLOWER_LC_utf8_safe>.
0c82b6df 718
c99e91e9 719=for apidoc Am|bool|isOCTAL|char ch
2787a470 720Returns a boolean indicating whether the specified character is an
6aff1f14 721octal digit, [0-7].
243effed
KW
722The only two variants are C<isOCTAL_A> and C<isOCTAL_L1>; each is identical to
723C<isOCTAL>.
724
725=for apidoc Am|bool|isPUNCT|char ch
726Returns a boolean indicating whether the specified character is a
6aff1f14
KW
727punctuation character, analogous to C<m/[[:punct:]]/>.
728Note that the definition of what is punctuation isn't as
243effed
KW
729straightforward as one might desire. See L<perlrecharclass/POSIX Character
730Classes> for details.
dcccc8ff 731See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
732variants C<isPUNCT_A>, C<isPUNCT_L1>, C<isPUNCT_uvchr>, C<isPUNCT_utf8_safe>,
733C<isPUNCT_LC>, C<isPUNCT_LC_uvchr>, and C<isPUNCT_LC_utf8_safe>.
c99e91e9 734
0c82b6df 735=for apidoc Am|bool|isSPACE|char ch
2787a470 736Returns a boolean indicating whether the specified character is a
6aff1f14 737whitespace character. This is analogous
398d098a 738to what C<m/\s/> matches in a regular expression. Starting in Perl 5.18
779cf272 739this also matches what C<m/[[:space:]]/> does. Prior to 5.18, only the
398d098a
KW
740locale forms of this macro (the ones with C<LC> in their names) matched
741precisely what C<m/[[:space:]]/> does. In those releases, the only difference,
742in the non-locale variants, was that C<isSPACE()> did not match a vertical tab.
743(See L</isPSXSPC> for a macro that matches a vertical tab in all releases.)
dcccc8ff
KW
744See the L<top of this section|/Character classification> for an explanation of
745variants
da8c1a98
KW
746C<isSPACE_A>, C<isSPACE_L1>, C<isSPACE_uvchr>, C<isSPACE_utf8_safe>,
747C<isSPACE_LC>, C<isSPACE_LC_uvchr>, and C<isSPACE_LC_utf8_safe>.
0c82b6df 748
398d098a
KW
749=for apidoc Am|bool|isPSXSPC|char ch
750(short for Posix Space)
779cf272
KW
751Starting in 5.18, this is identical in all its forms to the
752corresponding C<isSPACE()> macros.
398d098a
KW
753The locale forms of this macro are identical to their corresponding
754C<isSPACE()> forms in all Perl releases. In releases prior to 5.18, the
755non-locale forms differ from their C<isSPACE()> forms only in that the
756C<isSPACE()> forms don't match a Vertical Tab, and the C<isPSXSPC()> forms do.
757Otherwise they are identical. Thus this macro is analogous to what
758C<m/[[:space:]]/> matches in a regular expression.
dcccc8ff 759See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
760variants C<isPSXSPC_A>, C<isPSXSPC_L1>, C<isPSXSPC_uvchr>, C<isPSXSPC_utf8_safe>,
761C<isPSXSPC_LC>, C<isPSXSPC_LC_uvchr>, and C<isPSXSPC_LC_utf8_safe>.
398d098a 762
954c1994 763=for apidoc Am|bool|isUPPER|char ch
2787a470 764Returns a boolean indicating whether the specified character is an
6aff1f14 765uppercase character, analogous to C<m/[[:upper:]]/>.
dcccc8ff 766See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
767variants C<isUPPER_A>, C<isUPPER_L1>, C<isUPPER_uvchr>, C<isUPPER_utf8_safe>,
768C<isUPPER_LC>, C<isUPPER_LC_uvchr>, and C<isUPPER_LC_utf8_safe>.
954c1994 769
243effed 770=for apidoc Am|bool|isPRINT|char ch
8eea39dd 771Returns a boolean indicating whether the specified character is a
6aff1f14 772printable character, analogous to C<m/[[:print:]]/>.
dcccc8ff
KW
773See the L<top of this section|/Character classification> for an explanation of
774variants
da8c1a98
KW
775C<isPRINT_A>, C<isPRINT_L1>, C<isPRINT_uvchr>, C<isPRINT_utf8_safe>,
776C<isPRINT_LC>, C<isPRINT_LC_uvchr>, and C<isPRINT_LC_utf8_safe>.
243effed
KW
777
778=for apidoc Am|bool|isWORDCHAR|char ch
779Returns a boolean indicating whether the specified character is a character
780that is a word character, analogous to what C<m/\w/> and C<m/[[:word:]]/> match
781in a regular expression. A word character is an alphabetic character, a
782decimal digit, a connecting punctuation character (such as an underscore), or
783a "mark" character that attaches to one of those (like some sort of accent).
784C<isALNUM()> is a synonym provided for backward compatibility, even though a
785word character includes more than the standard C language meaning of
786alphanumeric.
dcccc8ff 787See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
788variants C<isWORDCHAR_A>, C<isWORDCHAR_L1>, C<isWORDCHAR_uvchr>, and
789C<isWORDCHAR_utf8_safe>. C<isWORDCHAR_LC>, C<isWORDCHAR_LC_uvchr>, and
790C<isWORDCHAR_LC_utf8_safe> are also as described there, but additionally
791include the platform's native underscore.
8a58bdcf
KW
792
793=for apidoc Am|bool|isXDIGIT|char ch
794Returns a boolean indicating whether the specified character is a hexadecimal
243effed
KW
795digit. In the ASCII range these are C<[0-9A-Fa-f]>. Variants C<isXDIGIT_A()>
796and C<isXDIGIT_L1()> are identical to C<isXDIGIT()>.
dcccc8ff
KW
797See the L<top of this section|/Character classification> for an explanation of
798variants
da8c1a98
KW
799C<isXDIGIT_uvchr>, C<isXDIGIT_utf8_safe>, C<isXDIGIT_LC>, C<isXDIGIT_LC_uvchr>,
800and C<isXDIGIT_LC_utf8_safe>.
243effed 801
3c3ecf18
KW
802=for apidoc Am|bool|isIDFIRST|char ch
803Returns a boolean indicating whether the specified character can be the first
804character of an identifier. This is very close to, but not quite the same as
805the official Unicode property C<XID_Start>. The difference is that this
806returns true only if the input character also matches L</isWORDCHAR>.
dcccc8ff
KW
807See the L<top of this section|/Character classification> for an explanation of
808variants
da8c1a98
KW
809C<isIDFIRST_A>, C<isIDFIRST_L1>, C<isIDFIRST_uvchr>, C<isIDFIRST_utf8_safe>,
810C<isIDFIRST_LC>, C<isIDFIRST_LC_uvchr>, and C<isIDFIRST_LC_utf8_safe>.
3c3ecf18
KW
811
812=for apidoc Am|bool|isIDCONT|char ch
813Returns a boolean indicating whether the specified character can be the
814second or succeeding character of an identifier. This is very close to, but
815not quite the same as the official Unicode property C<XID_Continue>. The
816difference is that this returns true only if the input character also matches
dcccc8ff
KW
817L</isWORDCHAR>. See the L<top of this section|/Character classification> for
818an
d0da05db 819explanation of variants C<isIDCONT_A>, C<isIDCONT_L1>, C<isIDCONT_uvchr>,
da8c1a98
KW
820C<isIDCONT_utf8_safe>, C<isIDCONT_LC>, C<isIDCONT_LC_uvchr>, and
821C<isIDCONT_LC_utf8_safe>.
3c3ecf18 822
243effed 823=head1 Miscellaneous Functions
8eea39dd 824
95a59cab 825=for apidoc Am|U8|READ_XDIGIT|char str*
243effed 826Returns the value of an ASCII-range hex digit and advances the string pointer.
95a59cab
YO
827Behaviour is only well defined when isXDIGIT(*str) is true.
828
e7c1e6c1 829=head1 Character case changing
21da7284
KW
830Perl uses "full" Unicode case mappings. This means that converting a single
831character to another case may result in a sequence of more than one character.
832For example, the uppercase of C<E<223>> (LATIN SMALL LETTER SHARP S) is the two
833character sequence C<SS>. This presents some complications The lowercase of
834all characters in the range 0..255 is a single character, and thus
835C<L</toLOWER_L1>> is furnished. But, C<toUPPER_L1> can't exist, as it couldn't
836return a valid result for all legal inputs. Instead C<L</toUPPER_uvchr>> has
837an API that does allow every possible legal result to be returned.) Likewise
838no other function that is crippled by not being able to give the correct
839results for the full range of possible inputs has been implemented here.
e7c1e6c1 840
1f607577
KW
841=for apidoc Am|U8|toUPPER|U8 ch
842Converts the specified character to uppercase. If the input is anything but an
843ASCII lowercase character, that input character itself is returned. Variant
c753c8d3 844C<toUPPER_A> is equivalent.
954c1994 845
d0da05db
KW
846=for apidoc Am|UV|toUPPER_uvchr|UV cp|U8* s|STRLEN* lenp
847Converts the code point C<cp> to its uppercase version, and
848stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
849point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
850that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
851bytes since the uppercase version may be longer than the original character.
852
853The first code point of the uppercased version is returned
21da7284
KW
854(but note, as explained at L<the top of this section|/Character case
855changing>, that there may be more.)
1f607577 856
a239b1e2
KW
857=for apidoc Am|UV|toUPPER_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
858Converts the first UTF-8 encoded character in the sequence starting at C<p> and
859extending no further than S<C<e - 1>> to its uppercase version, and
1f607577
KW
860stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
861that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
862bytes since the uppercase version may be longer than the original character.
863
864The first code point of the uppercased version is returned
21da7284
KW
865(but note, as explained at L<the top of this section|/Character case
866changing>, that there may be more).
1f607577 867
a239b1e2
KW
868The suffix C<_safe> in the function's name indicates that it will not attempt
869to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
870true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
871input character is malformed in some way, the program may croak, or the
872function may return the REPLACEMENT CHARACTER, at the discretion of the
873implementation, and subject to change in future releases.
874
875=for apidoc Am|UV|toUPPER_utf8|U8* p|U8* s|STRLEN* lenp
876This is like C<L</toUPPER_utf8_safe>>, but doesn't have the C<e>
877parameter The function therefore can't check if it is reading
607313a1
KW
878beyond the end of the string. Starting in Perl v5.30, it will take the C<e>
879parameter, becoming a synonym for C<toUPPER_utf8_safe>. At that time every
880program that uses it will have to be changed to successfully compile. In the
881meantime, the first runtime call to C<toUPPER_utf8> from each call point in the
882program will raise a deprecation warning, enabled by default. You can convert
883your program now to use C<toUPPER_utf8_safe>, and avoid the warnings, and get an
884extra measure of protection, or you can wait until v5.30, when you'll be forced
885to add the C<e> parameter.
1f607577 886
25200305
KW
887=for apidoc Am|U8|toFOLD|U8 ch
888Converts the specified character to foldcase. If the input is anything but an
889ASCII uppercase character, that input character itself is returned. Variant
890C<toFOLD_A> is equivalent. (There is no equivalent C<to_FOLD_L1> for the full
d0da05db 891Latin1 range, as the full generality of L</toFOLD_uvchr> is needed there.)
25200305 892
d0da05db
KW
893=for apidoc Am|UV|toFOLD_uvchr|UV cp|U8* s|STRLEN* lenp
894Converts the code point C<cp> to its foldcase version, and
895stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
896point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
897that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
898bytes since the foldcase version may be longer than the original character.
899
900The first code point of the foldcased 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
904=for apidoc Am|UV|toFOLD_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
905Converts the first UTF-8 encoded character in the sequence starting at C<p> and
906extending no further than S<C<e - 1>> to its foldcase version, and
1f607577
KW
907stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
908that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
909bytes since the foldcase version may be longer than the original character.
910
911The first code point of the foldcased version is returned
21da7284
KW
912(but note, as explained at L<the top of this section|/Character case
913changing>, that there may be more).
1f607577 914
a239b1e2
KW
915The suffix C<_safe> in the function's name indicates that it will not attempt
916to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
917true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
918input character is malformed in some way, the program may croak, or the
919function may return the REPLACEMENT CHARACTER, at the discretion of the
920implementation, and subject to change in future releases.
921
922=for apidoc Am|UV|toFOLD_utf8|U8* p|U8* s|STRLEN* lenp
923This is like C<L</toFOLD_utf8_safe>>, but doesn't have the C<e>
924parameter The function therefore can't check if it is reading
607313a1
KW
925beyond the end of the string. Starting in Perl v5.30, it will take the C<e>
926parameter, becoming a synonym for C<toFOLD_utf8_safe>. At that time every
927program that uses it will have to be changed to successfully compile. In the
928meantime, the first runtime call to C<toFOLD_utf8> from each call point in the
929program will raise a deprecation warning, enabled by default. You can convert
930your program now to use C<toFOLD_utf8_safe>, and avoid the warnings, and get an
931extra measure of protection, or you can wait until v5.30, when you'll be forced
932to add the C<e> parameter.
1f607577
KW
933
934=for apidoc Am|U8|toLOWER|U8 ch
935Converts the specified character to lowercase. If the input is anything but an
936ASCII uppercase character, that input character itself is returned. Variant
c753c8d3 937C<toLOWER_A> is equivalent.
954c1994 938
1f607577 939=for apidoc Am|U8|toLOWER_L1|U8 ch
b7d90381
KW
940Converts the specified Latin1 character to lowercase. The results are
941undefined if the input doesn't fit in a byte.
1f607577
KW
942
943=for apidoc Am|U8|toLOWER_LC|U8 ch
944Converts the specified character to lowercase using the current locale's rules,
945if possible; otherwise returns the input character itself.
946
d0da05db
KW
947=for apidoc Am|UV|toLOWER_uvchr|UV cp|U8* s|STRLEN* lenp
948Converts the code point C<cp> to its lowercase version, and
949stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
950point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
951that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
952bytes since the lowercase version may be longer than the original character.
953
954The first code point of the lowercased version is returned
21da7284
KW
955(but note, as explained at L<the top of this section|/Character case
956changing>, that there may be more).
1f607577 957
a239b1e2
KW
958
959=for apidoc Am|UV|toLOWER_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
960Converts the first UTF-8 encoded character in the sequence starting at C<p> and
961extending no further than S<C<e - 1>> to its lowercase version, and
1f607577
KW
962stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
963that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
964bytes since the lowercase version may be longer than the original character.
965
966The first code point of the lowercased version is returned
21da7284
KW
967(but note, as explained at L<the top of this section|/Character case
968changing>, that there may be more).
1f607577 969
a239b1e2
KW
970The suffix C<_safe> in the function's name indicates that it will not attempt
971to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
972true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
973input character is malformed in some way, the program may croak, or the
974function may return the REPLACEMENT CHARACTER, at the discretion of the
975implementation, and subject to change in future releases.
976
977=for apidoc Am|UV|toLOWER_utf8|U8* p|U8* s|STRLEN* lenp
978This is like C<L</toLOWER_utf8_safe>>, but doesn't have the C<e>
979parameter The function therefore can't check if it is reading
607313a1
KW
980beyond the end of the string. Starting in Perl v5.30, it will take the C<e>
981parameter, becoming a synonym for C<toLOWER_utf8_safe>. At that time every
982program that uses it will have to be changed to successfully compile. In the
983meantime, the first runtime call to C<toLOWER_utf8> from each call point in the
984program will raise a deprecation warning, enabled by default. You can convert
985your program now to use C<toLOWER_utf8_safe>, and avoid the warnings, and get an
986extra measure of protection, or you can wait until v5.30, when you'll be forced
987to add the C<e> parameter.
1f607577 988
25200305
KW
989=for apidoc Am|U8|toTITLE|U8 ch
990Converts the specified character to titlecase. If the input is anything but an
991ASCII lowercase character, that input character itself is returned. Variant
b7d90381 992C<toTITLE_A> is equivalent. (There is no C<toTITLE_L1> for the full Latin1
d0da05db 993range, as the full generality of L</toTITLE_uvchr> is needed there. Titlecase is
b7d90381 994not a concept used in locale handling, so there is no functionality for that.)
25200305 995
d0da05db
KW
996=for apidoc Am|UV|toTITLE_uvchr|UV cp|U8* s|STRLEN* lenp
997Converts the code point C<cp> to its titlecase version, and
998stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
999point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
1000that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
1001bytes since the titlecase version may be longer than the original character.
1002
1003The first code point of the titlecased version is returned
21da7284
KW
1004(but note, as explained at L<the top of this section|/Character case
1005changing>, that there may be more).
1f607577 1006
a239b1e2
KW
1007=for apidoc Am|UV|toTITLE_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
1008Converts the first UTF-8 encoded character in the sequence starting at C<p> and
1009extending no further than S<C<e - 1>> to its titlecase version, and
1f607577
KW
1010stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
1011that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
1012bytes since the titlecase version may be longer than the original character.
1013
1014The first code point of the titlecased version is returned
21da7284
KW
1015(but note, as explained at L<the top of this section|/Character case
1016changing>, that there may be more).
1f607577 1017
a239b1e2
KW
1018The suffix C<_safe> in the function's name indicates that it will not attempt
1019to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
1020true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
1021input character is malformed in some way, the program may croak, or the
1022function may return the REPLACEMENT CHARACTER, at the discretion of the
1023implementation, and subject to change in future releases.
1024
1025=for apidoc Am|UV|toTITLE_utf8|U8* p|U8* s|STRLEN* lenp
1026This is like C<L</toLOWER_utf8_safe>>, but doesn't have the C<e>
1027parameter The function therefore can't check if it is reading
607313a1
KW
1028beyond the end of the string. Starting in Perl v5.30, it will take the C<e>
1029parameter, becoming a synonym for C<toTITLE_utf8_safe>. At that time every
1030program that uses it will have to be changed to successfully compile. In the
1031meantime, the first runtime call to C<toTITLE_utf8> from each call point in the
1032program will raise a deprecation warning, enabled by default. You can convert
1033your program now to use C<toTITLE_utf8_safe>, and avoid the warnings, and get an
1034extra measure of protection, or you can wait until v5.30, when you'll be forced
1035to add the C<e> parameter.
1f607577 1036
954c1994 1037=cut
353c9b6f 1038
d0da05db 1039XXX Still undocumented isVERTWS_uvchr and _utf8; it's unclear what their names
1e222e4f
KW
1040really should be. Also toUPPER_LC and toFOLD_LC, which are subject to change,
1041and aren't general purpose as they don't work on U+DF, and assert against that.
243effed 1042
8a58bdcf 1043Note that these macros are repeated in Devel::PPPort, so should also be
62fa66b6
KW
1044patched there. The file as of this writing is cpan/Devel-PPPort/parts/inc/misc
1045
954c1994
GS
1046*/
1047
7c062697
KW
1048/* Specify the widest unsigned type on the platform. Use U64TYPE because U64
1049 * is known only in the perl core, and this macro can be called from outside
1050 * that */
1051#ifdef HAS_QUAD
1052# define WIDEST_UTYPE U64TYPE
1053#else
1054# define WIDEST_UTYPE U32
1055#endif
1056
3912bc88
KW
1057/* FITS_IN_8_BITS(c) returns true if c doesn't have a bit set other than in
1058 * the lower 8. It is designed to be hopefully bomb-proof, making sure that no
1059 * bits of information are lost even on a 64-bit machine, but to get the
1060 * compiler to optimize it out if possible. This is because Configure makes
1061 * sure that the machine has an 8-bit byte, so if c is stored in a byte, the
1062 * sizeof() guarantees that this evaluates to a constant true at compile time.
7e75d1a1
JH
1063 *
1064 * For Coverity, be always true, because otherwise Coverity thinks
1065 * it finds several expressions that are always true, independent
1066 * of operands. Well, they are, but that is kind of the point.
220c71bf 1067 */
7e75d1a1 1068#ifndef __COVERITY__
6c5b02ac
KW
1069 /* The '| 0' part ensures a compiler error if c is not integer (like e.g., a
1070 * pointer) */
1071#define FITS_IN_8_BITS(c) ( (sizeof(c) == 1) \
ace3ad0f 1072 || !(((WIDEST_UTYPE)((c) | 0)) & ~0xFF))
7e75d1a1
JH
1073#else
1074#define FITS_IN_8_BITS(c) (1)
1075#endif
cf301eb7 1076
41f43cc2 1077#ifdef EBCDIC
b6340bd0 1078# ifndef _ALL_SOURCE
0852beac
KW
1079 /* The native libc isascii() et.al. functions return the wrong results
1080 * on at least z/OS unless this is defined. */
b6340bd0
KW
1081# error _ALL_SOURCE should probably be defined
1082# endif
41f43cc2 1083#else
0852beac
KW
1084 /* There is a simple definition of ASCII for ASCII platforms. But the
1085 * EBCDIC one isn't so simple, so is defined using table look-up like the
9c903d59 1086 * other macros below.
3f3c579d
KW
1087 *
1088 * The cast here is used instead of '(c) >= 0', because some compilers emit
1089 * a warning that that test is always true when the parameter is an
1090 * unsigned type. khw supposes that it could be written as
1091 * && ((c) == '\0' || (c) > 0)
1092 * to avoid the message, but the cast will likely avoid extra branches even
1093 * with stupid compilers.
1094 *
1095 * The '| 0' part ensures a compiler error if c is not integer (like e.g.,
1096 * a pointer) */
9c903d59 1097# define isASCII(c) ((WIDEST_UTYPE)((c) | 0) < 128)
41f43cc2
KW
1098#endif
1099
38694112
KW
1100/* Take the eight possible bit patterns of the lower 3 bits and you get the
1101 * lower 3 bits of the 8 octal digits, in both ASCII and EBCDIC, so those bits
1102 * can be ignored. If the rest match '0', we have an octal */
1103#define isOCTAL_A(c) (((WIDEST_UTYPE)((c) | 0) & ~7) == '0')
c2da0b36 1104
9fb1bf9d 1105#ifdef H_PERL /* If have access to perl.h, lookup in its table */
f4cdb42c 1106
a500dc72
KW
1107/* Character class numbers. For internal core Perl use only. The ones less
1108 * than 32 are used in PL_charclass[] and the ones up through the one that
1109 * corresponds to <_HIGHEST_REGCOMP_DOT_H_SYNC> are used by regcomp.h and
1110 * related files. PL_charclass ones use names used in l1_char_class_tab.h but
1111 * their actual definitions are here. If that file has a name not used here,
1112 * it won't compile.
1709d539
KW
1113 *
1114 * The first group of these is ordered in what I (khw) estimate to be the
31c7f561
KW
1115 * frequency of their use. This gives a slight edge to exiting a loop earlier
1116 * (in reginclass() in regexec.c) */
1709d539
KW
1117# define _CC_WORDCHAR 0 /* \w and [:word:] */
1118# define _CC_DIGIT 1 /* \d and [:digit:] */
1119# define _CC_ALPHA 2 /* [:alpha:] */
1120# define _CC_LOWER 3 /* [:lower:] */
1121# define _CC_UPPER 4 /* [:upper:] */
1122# define _CC_PUNCT 5 /* [:punct:] */
1123# define _CC_PRINT 6 /* [:print:] */
15861f94 1124# define _CC_ALPHANUMERIC 7 /* [:alnum:] */
1709d539 1125# define _CC_GRAPH 8 /* [:graph:] */
359b005e 1126# define _CC_CASED 9 /* [:lower:] or [:upper:] under /i */
1709d539 1127
b0d691b2 1128#define _FIRST_NON_SWASH_CC 10
1709d539
KW
1129/* The character classes above are implemented with swashes. The second group
1130 * (just below) contains the ones implemented without. These are also sorted
1131 * in rough order of the frequency of their use, except that \v should be last,
1132 * as it isn't a real Posix character class, and some (small) inefficiencies in
1133 * regular expression handling would be introduced by putting it in the middle
1134 * of those that are. Also, cntrl and ascii come after the others as it may be
1135 * useful to group these which have no members that match above Latin1, (or
1136 * above ASCII in the latter case) */
1137
779cf272 1138# define _CC_SPACE 10 /* \s, [:space:] */
34aeb2e9
KW
1139# define _CC_PSXSPC _CC_SPACE /* XXX Temporary, can be removed
1140 when the deprecated isFOO_utf8()
1141 functions are removed */
b0d691b2
KW
1142# define _CC_BLANK 11 /* [:blank:] */
1143# define _CC_XDIGIT 12 /* [:xdigit:] */
779cf272
KW
1144# define _CC_CNTRL 13 /* [:cntrl:] */
1145# define _CC_ASCII 14 /* [:ascii:] */
1146# define _CC_VERTSPACE 15 /* \v */
1709d539 1147
a0947d7b
KW
1148# define _HIGHEST_REGCOMP_DOT_H_SYNC _CC_VERTSPACE
1149
1709d539 1150/* The members of the third group below do not need to be coordinated with data
3ffc8c70 1151 * structures in regcomp.[ch] and regexec.c. */
779cf272
KW
1152# define _CC_IDFIRST 16
1153# define _CC_CHARNAME_CONT 17
1154# define _CC_NONLATIN1_FOLD 18
1155# define _CC_NONLATIN1_SIMPLE_FOLD 19
1156# define _CC_QUOTEMETA 20
1157# define _CC_NON_FINAL_FOLD 21
1158# define _CC_IS_IN_SOME_FOLD 22
1159# define _CC_MNEMONIC_CNTRL 23
073c22b3 1160
34aeb2e9
KW
1161# define _CC_IDCONT 24 /* XXX Temporary, can be removed when the deprecated
1162 isFOO_utf8() functions are removed */
1163
073c22b3
KW
1164/* This next group is only used on EBCDIC platforms, so theoretically could be
1165 * shared with something entirely different that's only on ASCII platforms */
5d5376e2
KW
1166# define _CC_UTF8_START_BYTE_IS_FOR_AT_LEAST_SURROGATE 28
1167# define _CC_UTF8_IS_START 29
1168# define _CC_UTF8_IS_DOWNGRADEABLE_START 30
1169# define _CC_UTF8_IS_CONTINUATION 31
1170/* Unused: 24-27
f4cdb42c
KW
1171 * If more bits are needed, one could add a second word for non-64bit
1172 * QUAD_IS_INT systems, using some #ifdefs to distinguish between having a 2nd
37ede926
KW
1173 * word or not. The IS_IN_SOME_FOLD bit is the most easily expendable, as it
1174 * is used only for optimization (as of this writing), and differs in the
1175 * Latin1 range from the ALPHA bit only in two relatively unimportant
a500dc72 1176 * characters: the masculine and feminine ordinal indicators, so removing it
073c22b3
KW
1177 * would just cause /i regexes which match them to run less efficiently.
1178 * Similarly the EBCDIC-only bits are used just for speed, and could be
1179 * replaced by other means */
96ac0975 1180
3a371f2f
KW
1181#if defined(PERL_CORE) || defined(PERL_EXT)
1182/* An enum version of the character class numbers, to help compilers
1183 * optimize */
1184typedef enum {
3a371f2f 1185 _CC_ENUM_ALPHA = _CC_ALPHA,
e8d596e0
KW
1186 _CC_ENUM_ALPHANUMERIC = _CC_ALPHANUMERIC,
1187 _CC_ENUM_ASCII = _CC_ASCII,
1188 _CC_ENUM_BLANK = _CC_BLANK,
b0d691b2 1189 _CC_ENUM_CASED = _CC_CASED,
e8d596e0 1190 _CC_ENUM_CNTRL = _CC_CNTRL,
3a371f2f
KW
1191 _CC_ENUM_DIGIT = _CC_DIGIT,
1192 _CC_ENUM_GRAPH = _CC_GRAPH,
1193 _CC_ENUM_LOWER = _CC_LOWER,
1194 _CC_ENUM_PRINT = _CC_PRINT,
1195 _CC_ENUM_PUNCT = _CC_PUNCT,
e8d596e0 1196 _CC_ENUM_SPACE = _CC_SPACE,
3a371f2f 1197 _CC_ENUM_UPPER = _CC_UPPER,
e8d596e0 1198 _CC_ENUM_VERTSPACE = _CC_VERTSPACE,
3a371f2f 1199 _CC_ENUM_WORDCHAR = _CC_WORDCHAR,
e8d596e0 1200 _CC_ENUM_XDIGIT = _CC_XDIGIT
3a371f2f
KW
1201} _char_class_number;
1202#endif
1203
63c61c3f 1204#define POSIX_SWASH_COUNT _FIRST_NON_SWASH_CC
86f72d56 1205#define POSIX_CC_COUNT (_HIGHEST_REGCOMP_DOT_H_SYNC + 1)
63c61c3f 1206
b7d90381
KW
1207#if defined(PERL_IN_UTF8_C) \
1208 || defined(PERL_IN_REGCOMP_C) \
1209 || defined(PERL_IN_REGEXEC_C)
922e8cb4 1210# if _CC_WORDCHAR != 0 || _CC_DIGIT != 1 || _CC_ALPHA != 2 || _CC_LOWER != 3 \
b0d691b2
KW
1211 || _CC_UPPER != 4 || _CC_PUNCT != 5 || _CC_PRINT != 6 \
1212 || _CC_ALPHANUMERIC != 7 || _CC_GRAPH != 8 || _CC_CASED != 9
922e8cb4
KW
1213 #error Need to adjust order of swash_property_names[]
1214# endif
1215
b44daeeb
KW
1216/* This is declared static in each of the few files that this is #defined for
1217 * to keep them from being publicly accessible. Hence there is a small amount
1218 * of wasted space */
1219
922e8cb4
KW
1220static const char* const swash_property_names[] = {
1221 "XPosixWord",
1222 "XPosixDigit",
1223 "XPosixAlpha",
1224 "XPosixLower",
1225 "XPosixUpper",
1226 "XPosixPunct",
1227 "XPosixPrint",
1228 "XPosixAlnum",
b0d691b2
KW
1229 "XPosixGraph",
1230 "Cased"
922e8cb4
KW
1231};
1232#endif
1233
6635f04f 1234START_EXTERN_C
96ac0975
NC
1235# ifdef DOINIT
1236EXTCONST U32 PL_charclass[] = {
1237# include "l1_char_class_tab.h"
1238};
1239
1240# else /* ! DOINIT */
1241EXTCONST U32 PL_charclass[];
1242# endif
6635f04f 1243END_EXTERN_C
96ac0975 1244
265c1f46 1245 /* The 1U keeps Solaris from griping when shifting sets the uppermost bit */
430b7c70 1246# define _CC_mask(classnum) (1U << (classnum))
4650c663
KW
1247
1248 /* For internal core Perl use only: the base macro for defining macros like
1249 * isALPHA */
ff7ecfc3 1250# define _generic_isCC(c, classnum) cBOOL(FITS_IN_8_BITS(c) \
f4cd282c 1251 && (PL_charclass[(U8) (c)] & _CC_mask(classnum)))
4eeeb416 1252
f4cdb42c
KW
1253 /* The mask for the _A versions of the macros; it just adds in the bit for
1254 * ASCII. */
1255# define _CC_mask_A(classnum) (_CC_mask(classnum) | _CC_mask(_CC_ASCII))
1256
4650c663
KW
1257 /* For internal core Perl use only: the base macro for defining macros like
1258 * isALPHA_A. The foo_A version makes sure that both the desired bit and
1259 * the ASCII bit are present */
b7d90381
KW
1260# define _generic_isCC_A(c, classnum) (FITS_IN_8_BITS(c) \
1261 && ((PL_charclass[(U8) (c)] & _CC_mask_A(classnum)) \
1262 == _CC_mask_A(classnum)))
f4cdb42c 1263
f4cdb42c 1264# define isALPHA_A(c) _generic_isCC_A(c, _CC_ALPHA)
15861f94 1265# define isALPHANUMERIC_A(c) _generic_isCC_A(c, _CC_ALPHANUMERIC)
f4cdb42c
KW
1266# define isBLANK_A(c) _generic_isCC_A(c, _CC_BLANK)
1267# define isCNTRL_A(c) _generic_isCC_A(c, _CC_CNTRL)
96ca48da 1268# define isDIGIT_A(c) _generic_isCC(c, _CC_DIGIT) /* No non-ASCII digits */
f4cdb42c
KW
1269# define isGRAPH_A(c) _generic_isCC_A(c, _CC_GRAPH)
1270# define isLOWER_A(c) _generic_isCC_A(c, _CC_LOWER)
1271# define isPRINT_A(c) _generic_isCC_A(c, _CC_PRINT)
f4cdb42c
KW
1272# define isPUNCT_A(c) _generic_isCC_A(c, _CC_PUNCT)
1273# define isSPACE_A(c) _generic_isCC_A(c, _CC_SPACE)
1274# define isUPPER_A(c) _generic_isCC_A(c, _CC_UPPER)
1275# define isWORDCHAR_A(c) _generic_isCC_A(c, _CC_WORDCHAR)
b7d90381
KW
1276# define isXDIGIT_A(c) _generic_isCC(c, _CC_XDIGIT) /* No non-ASCII xdigits
1277 */
d95f8b6a 1278# define isIDFIRST_A(c) _generic_isCC_A(c, _CC_IDFIRST)
3ded5eb0
KW
1279# define isALPHA_L1(c) _generic_isCC(c, _CC_ALPHA)
1280# define isALPHANUMERIC_L1(c) _generic_isCC(c, _CC_ALPHANUMERIC)
1281# define isBLANK_L1(c) _generic_isCC(c, _CC_BLANK)
1282
1283 /* continuation character for legal NAME in \N{NAME} */
1284# define isCHARNAME_CONT(c) _generic_isCC(c, _CC_CHARNAME_CONT)
1285
1286# define isCNTRL_L1(c) _generic_isCC(c, _CC_CNTRL)
1287# define isGRAPH_L1(c) _generic_isCC(c, _CC_GRAPH)
1288# define isLOWER_L1(c) _generic_isCC(c, _CC_LOWER)
1289# define isPRINT_L1(c) _generic_isCC(c, _CC_PRINT)
b7d90381 1290# define isPSXSPC_L1(c) isSPACE_L1(c)
3ded5eb0
KW
1291# define isPUNCT_L1(c) _generic_isCC(c, _CC_PUNCT)
1292# define isSPACE_L1(c) _generic_isCC(c, _CC_SPACE)
1293# define isUPPER_L1(c) _generic_isCC(c, _CC_UPPER)
1294# define isWORDCHAR_L1(c) _generic_isCC(c, _CC_WORDCHAR)
1295# define isIDFIRST_L1(c) _generic_isCC(c, _CC_IDFIRST)
f4cdb42c 1296
0852beac
KW
1297# ifdef EBCDIC
1298# define isASCII(c) _generic_isCC(c, _CC_ASCII)
1299# endif
1300
f12c0118
KW
1301 /* Participates in a single-character fold with a character above 255 */
1302# 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)))
1303
1304 /* Like the above, but also can be part of a multi-char fold */
f4cd282c 1305# 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 1306
4eeeb416 1307# define _isQUOTEMETA(c) _generic_isCC(c, _CC_QUOTEMETA)
26faadbd 1308# define _IS_NON_FINAL_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
b7d90381 1309 _generic_isCC(c, _CC_NON_FINAL_FOLD)
37ede926 1310# define _IS_IN_SOME_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
b7d90381 1311 _generic_isCC(c, _CC_IS_IN_SOME_FOLD)
8e35b056
KW
1312# define _IS_MNEMONIC_CNTRL_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
1313 _generic_isCC(c, _CC_MNEMONIC_CNTRL)
687c8d01 1314#else /* else we don't have perl.h H_PERL */
3ded5eb0
KW
1315
1316 /* If we don't have perl.h, we are compiling a utility program. Below we
1317 * hard-code various macro definitions that wouldn't otherwise be available
fc273927 1318 * to it. Most are coded based on first principles. These are written to
74665a89 1319 * avoid EBCDIC vs. ASCII #ifdef's as much as possible. */
0852beac
KW
1320# define isDIGIT_A(c) ((c) <= '9' && (c) >= '0')
1321# define isBLANK_A(c) ((c) == ' ' || (c) == '\t')
74665a89
KW
1322# define isSPACE_A(c) (isBLANK_A(c) \
1323 || (c) == '\n' \
1324 || (c) == '\r' \
1325 || (c) == '\v' \
0852beac 1326 || (c) == '\f')
74665a89
KW
1327 /* On EBCDIC, there are gaps between 'i' and 'j'; 'r' and 's'. Same for
1328 * uppercase. The tests for those aren't necessary on ASCII, but hurt only
1329 * performance (if optimization isn't on), and allow the same code to be
1330 * used for both platform types */
1331# define isLOWER_A(c) ((c) >= 'a' && (c) <= 'z' \
1332 && ( (c) <= 'i' \
1333 || ((c) >= 'j' && (c) <= 'r') \
1334 || (c) >= 's'))
1335# define isUPPER_A(c) ((c) >= 'A' && (c) <= 'Z' \
1336 && ( (c) <= 'I' \
1337 || ((c) >= 'J' && (c) <= 'R') \
1338 || (c) >= 'S'))
a4d7a999
KW
1339# define isALPHA_A(c) (isUPPER_A(c) || isLOWER_A(c))
1340# define isALPHANUMERIC_A(c) (isALPHA_A(c) || isDIGIT_A(c))
3ded5eb0 1341# define isWORDCHAR_A(c) (isALPHANUMERIC_A(c) || (c) == '_')
0852beac 1342# define isIDFIRST_A(c) (isALPHA_A(c) || (c) == '_')
74665a89
KW
1343# define isXDIGIT_A(c) (isDIGIT_A(c) \
1344 || ((c) >= 'a' && (c) <= 'f') \
a4d7a999 1345 || ((c) <= 'F' && (c) >= 'A'))
74665a89
KW
1346# define isPUNCT_A(c) ((c) == '-' || (c) == '!' || (c) == '"' \
1347 || (c) == '#' || (c) == '$' || (c) == '%' \
1348 || (c) == '&' || (c) == '\'' || (c) == '(' \
1349 || (c) == ')' || (c) == '*' || (c) == '+' \
1350 || (c) == ',' || (c) == '.' || (c) == '/' \
1351 || (c) == ':' || (c) == ';' || (c) == '<' \
1352 || (c) == '=' || (c) == '>' || (c) == '?' \
1353 || (c) == '@' || (c) == '[' || (c) == '\\' \
1354 || (c) == ']' || (c) == '^' || (c) == '_' \
1355 || (c) == '`' || (c) == '{' || (c) == '|' \
1356 || (c) == '}' || (c) == '~')
1357# define isGRAPH_A(c) (isALPHANUMERIC_A(c) || isPUNCT_A(c))
1358# define isPRINT_A(c) (isGRAPH_A(c) || (c) == ' ')
3ded5eb0 1359
0852beac 1360# ifdef EBCDIC
74665a89
KW
1361 /* The below is accurate for the 3 EBCDIC code pages traditionally
1362 * supported by perl. The only difference between them in the controls
1363 * is the position of \n, and that is represented symbolically below */
1364# define isCNTRL_A(c) ((c) == '\0' || (c) == '\a' || (c) == '\b' \
1365 || (c) == '\f' || (c) == '\n' || (c) == '\r' \
1366 || (c) == '\t' || (c) == '\v' \
1367 || ((c) <= 3 && (c) >= 1) /* SOH, STX, ETX */ \
1368 || (c) == 7 /* U+7F DEL */ \
1369 || ((c) <= 0x13 && (c) >= 0x0E) /* SO, SI */ \
1370 /* DLE, DC[1-3] */ \
1371 || (c) == 0x18 /* U+18 CAN */ \
1372 || (c) == 0x19 /* U+19 EOM */ \
1373 || ((c) <= 0x1F && (c) >= 0x1C) /* [FGRU]S */ \
1374 || (c) == 0x26 /* U+17 ETB */ \
1375 || (c) == 0x27 /* U+1B ESC */ \
1376 || (c) == 0x2D /* U+05 ENQ */ \
1377 || (c) == 0x2E /* U+06 ACK */ \
1378 || (c) == 0x32 /* U+16 SYN */ \
1379 || (c) == 0x37 /* U+04 EOT */ \
1380 || (c) == 0x3C /* U+14 DC4 */ \
1381 || (c) == 0x3D /* U+15 NAK */ \
1382 || (c) == 0x3F)/* U+1A SUB */
0852beac 1383# define isASCII(c) (isCNTRL_A(c) || isPRINT_A(c))
74665a89
KW
1384# else /* isASCII is already defined for ASCII platforms, so can use that to
1385 define isCNTRL */
1386# define isCNTRL_A(c) (isASCII(c) && ! isPRINT_A(c))
0852beac
KW
1387# endif
1388
3ffc8c70 1389 /* The _L1 macros may be unnecessary for the utilities; I (khw) added them
caa94d35
KW
1390 * during debugging, and it seems best to keep them. We may be called
1391 * without NATIVE_TO_LATIN1 being defined. On ASCII platforms, it doesn't
1392 * do anything anyway, so make it not a problem */
1393# if ! defined(EBCDIC) && ! defined(NATIVE_TO_LATIN1)
1394# define NATIVE_TO_LATIN1(ch) (ch)
1395# endif
3ded5eb0
KW
1396# define isALPHA_L1(c) (isUPPER_L1(c) || isLOWER_L1(c))
1397# define isALPHANUMERIC_L1(c) (isALPHA_L1(c) || isDIGIT_A(c))
1398# define isBLANK_L1(c) (isBLANK_A(c) \
1399 || (FITS_IN_8_BITS(c) \
1400 && NATIVE_TO_LATIN1((U8) c) == 0xA0))
1401# define isCNTRL_L1(c) (FITS_IN_8_BITS(c) && (! isPRINT_L1(c)))
1402# define isGRAPH_L1(c) (isPRINT_L1(c) && (! isBLANK_L1(c)))
1403# define isLOWER_L1(c) (isLOWER_A(c) \
1404 || (FITS_IN_8_BITS(c) \
ae683a5f 1405 && (( NATIVE_TO_LATIN1((U8) c) >= 0xDF \
3ded5eb0
KW
1406 && NATIVE_TO_LATIN1((U8) c) != 0xF7) \
1407 || NATIVE_TO_LATIN1((U8) c) == 0xAA \
1408 || NATIVE_TO_LATIN1((U8) c) == 0xBA \
1409 || NATIVE_TO_LATIN1((U8) c) == 0xB5)))
1410# define isPRINT_L1(c) (isPRINT_A(c) \
1411 || (FITS_IN_8_BITS(c) \
1412 && NATIVE_TO_LATIN1((U8) c) >= 0xA0))
3ded5eb0
KW
1413# define isPUNCT_L1(c) (isPUNCT_A(c) \
1414 || (FITS_IN_8_BITS(c) \
ae683a5f 1415 && ( NATIVE_TO_LATIN1((U8) c) == 0xA1 \
3ded5eb0
KW
1416 || NATIVE_TO_LATIN1((U8) c) == 0xA7 \
1417 || NATIVE_TO_LATIN1((U8) c) == 0xAB \
1418 || NATIVE_TO_LATIN1((U8) c) == 0xB6 \
1419 || NATIVE_TO_LATIN1((U8) c) == 0xB7 \
1420 || NATIVE_TO_LATIN1((U8) c) == 0xBB \
1421 || NATIVE_TO_LATIN1((U8) c) == 0xBF)))
1422# define isSPACE_L1(c) (isSPACE_A(c) \
1423 || (FITS_IN_8_BITS(c) \
ae683a5f 1424 && ( NATIVE_TO_LATIN1((U8) c) == 0x85 \
3ded5eb0
KW
1425 || NATIVE_TO_LATIN1((U8) c) == 0xA0)))
1426# define isUPPER_L1(c) (isUPPER_A(c) \
1427 || (FITS_IN_8_BITS(c) \
ae683a5f 1428 && ( NATIVE_TO_LATIN1((U8) c) >= 0xC0 \
3ded5eb0
KW
1429 && NATIVE_TO_LATIN1((U8) c) <= 0xDE \
1430 && NATIVE_TO_LATIN1((U8) c) != 0xD7)))
1431# define isWORDCHAR_L1(c) (isIDFIRST_L1(c) || isDIGIT_A(c))
1432# define isIDFIRST_L1(c) (isALPHA_L1(c) || NATIVE_TO_LATIN1(c) == '_')
1433# define isCHARNAME_CONT(c) (isWORDCHAR_L1(c) \
1434 || isBLANK_L1(c) \
1435 || (c) == '-' \
1436 || (c) == '(' \
1437 || (c) == ')')
1438 /* The following are not fully accurate in the above-ASCII range. I (khw)
1439 * don't think it's necessary to be so for the purposes where this gets
1440 * compiled */
1441# define _isQUOTEMETA(c) (FITS_IN_8_BITS(c) && ! isWORDCHAR_L1(c))
1442# define _IS_IN_SOME_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) isALPHA_L1(c)
1443
1444 /* And these aren't accurate at all. They are useful only for above
1445 * Latin1, which utilities and bootstrapping don't deal with */
1446# define _IS_NON_FINAL_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) 0
6838b41e 1447# define _HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) 0
3ded5eb0
KW
1448# define _HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) 0
1449
1450 /* Many of the macros later in this file are defined in terms of these. By
1451 * implementing them with a function, which converts the class number into
1452 * a call to the desired macro, all of the later ones work. However, that
1453 * function won't be actually defined when building a utility program (no
1454 * perl.h), and so a compiler error will be generated if one is attempted
1455 * to be used. And the above-Latin1 code points require Unicode tables to
1456 * be present, something unlikely to be the case when bootstrapping */
1457# define _generic_isCC(c, classnum) \
1458 (FITS_IN_8_BITS(c) && S_bootstrap_ctype((U8) (c), (classnum), TRUE))
1459# define _generic_isCC_A(c, classnum) \
1460 (FITS_IN_8_BITS(c) && S_bootstrap_ctype((U8) (c), (classnum), FALSE))
687c8d01 1461#endif /* End of no perl.h H_PERL */
8a58bdcf 1462
e66b99e9
KW
1463#define isALPHANUMERIC(c) isALPHANUMERIC_A(c)
1464#define isALPHA(c) isALPHA_A(c)
0852beac
KW
1465#define isASCII_A(c) isASCII(c)
1466#define isASCII_L1(c) isASCII(c)
e66b99e9
KW
1467#define isBLANK(c) isBLANK_A(c)
1468#define isCNTRL(c) isCNTRL_A(c)
1469#define isDIGIT(c) isDIGIT_A(c)
1470#define isGRAPH(c) isGRAPH_A(c)
1471#define isIDFIRST(c) isIDFIRST_A(c)
1472#define isLOWER(c) isLOWER_A(c)
1473#define isPRINT(c) isPRINT_A(c)
779cf272 1474#define isPSXSPC_A(c) isSPACE_A(c)
e66b99e9 1475#define isPSXSPC(c) isPSXSPC_A(c)
779cf272 1476#define isPSXSPC_L1(c) isSPACE_L1(c)
e66b99e9
KW
1477#define isPUNCT(c) isPUNCT_A(c)
1478#define isSPACE(c) isSPACE_A(c)
1479#define isUPPER(c) isUPPER_A(c)
1480#define isWORDCHAR(c) isWORDCHAR_A(c)
1481#define isXDIGIT(c) isXDIGIT_A(c)
1482
1483/* ASCII casing. These could also be written as
1484 #define toLOWER(c) (isASCII(c) ? toLOWER_LATIN1(c) : (c))
1485 #define toUPPER(c) (isASCII(c) ? toUPPER_LATIN1_MOD(c) : (c))
1486 which uses table lookup and mask instead of subtraction. (This would
1487 work because the _MOD does not apply in the ASCII range) */
68067e4e
DM
1488#define toLOWER(c) (isUPPER(c) ? (U8)((c) + ('a' - 'A')) : (c))
1489#define toUPPER(c) (isLOWER(c) ? (U8)((c) - ('a' - 'A')) : (c))
bbce6d69 1490
25200305
KW
1491/* In the ASCII range, these are equivalent to what they're here defined to be.
1492 * But by creating these definitions, other code doesn't have to be aware of
1493 * this detail */
1494#define toFOLD(c) toLOWER(c)
25200305
KW
1495#define toTITLE(c) toUPPER(c)
1496
c753c8d3
KW
1497#define toLOWER_A(c) toLOWER(c)
1498#define toUPPER_A(c) toUPPER(c)
25200305
KW
1499#define toFOLD_A(c) toFOLD(c)
1500#define toTITLE_A(c) toTITLE(c)
1a0901db 1501
4650c663 1502/* Use table lookup for speed; returns the input itself if is out-of-range */
b2bf251f 1503#define toLOWER_LATIN1(c) ((! FITS_IN_8_BITS(c)) \
8e7c6e7d 1504 ? (c) \
f4cd282c 1505 : PL_latin1_lc[ (U8) (c) ])
c753c8d3
KW
1506#define toLOWER_L1(c) toLOWER_LATIN1(c) /* Synonym for consistency */
1507
1a0901db 1508/* Modified uc. Is correct uc except for three non-ascii chars which are
4650c663
KW
1509 * all mapped to one of them, and these need special handling; returns the
1510 * input itself if is out-of-range */
b2bf251f 1511#define toUPPER_LATIN1_MOD(c) ((! FITS_IN_8_BITS(c)) \
8e7c6e7d 1512 ? (c) \
f4cd282c 1513 : PL_mod_latin1_uc[ (U8) (c) ])
31f05a37 1514#define IN_UTF8_CTYPE_LOCALE PL_in_utf8_CTYPE_locale
84061b6a 1515
beab9ebe
KW
1516/* Use foo_LC_uvchr() instead of these for beyond the Latin1 range */
1517
1518/* For internal core Perl use only: the base macro for defining macros like
1519 * isALPHA_LC, which uses the current LC_CTYPE locale. 'c' is the code point
31f05a37
KW
1520 * (0-255) to check. In a UTF-8 locale, the result is the same as calling
1521 * isFOO_L1(); the 'utf8_locale_classnum' parameter is something like
1522 * _CC_UPPER, which gives the class number for doing this. For non-UTF-8
1523 * locales, the code to actually do the test this is passed in 'non_utf8'. If
1524 * 'c' is above 255, 0 is returned. For accessing the full range of possible
1525 * code points under locale rules, use the macros based on _generic_LC_uvchr
1526 * instead of this. */
beab9ebe
KW
1527#define _generic_LC_base(c, utf8_locale_classnum, non_utf8) \
1528 (! FITS_IN_8_BITS(c) \
1529 ? 0 \
31f05a37
KW
1530 : IN_UTF8_CTYPE_LOCALE \
1531 ? cBOOL(PL_charclass[(U8) (c)] & _CC_mask(utf8_locale_classnum)) \
beab9ebe
KW
1532 : cBOOL(non_utf8))
1533
1534/* For internal core Perl use only: a helper macro for defining macros like
1535 * isALPHA_LC. 'c' is the code point (0-255) to check. The function name to
1536 * actually do this test is passed in 'non_utf8_func', which is called on 'c',
1537 * casting 'c' to the macro _LC_CAST, which should not be parenthesized. See
1538 * _generic_LC_base for more info */
1539#define _generic_LC(c, utf8_locale_classnum, non_utf8_func) \
1540 _generic_LC_base(c,utf8_locale_classnum, \
1541 non_utf8_func( (_LC_CAST) (c)))
1542
1543/* For internal core Perl use only: like _generic_LC, but also returns TRUE if
1544 * 'c' is the platform's native underscore character */
1545#define _generic_LC_underscore(c,utf8_locale_classnum,non_utf8_func) \
1546 _generic_LC_base(c, utf8_locale_classnum, \
1547 (non_utf8_func( (_LC_CAST) (c)) \
1548 || (char)(c) == '_'))
1549
1550/* These next three are also for internal core Perl use only: case-change
1551 * helper macros */
1552#define _generic_toLOWER_LC(c, function, cast) (! FITS_IN_8_BITS(c) \
1553 ? (c) \
31f05a37
KW
1554 : (IN_UTF8_CTYPE_LOCALE) \
1555 ? PL_latin1_lc[ (U8) (c) ] \
d2a8cf7f 1556 : (cast)function((cast)(c)))
beab9ebe 1557
31f05a37
KW
1558/* Note that the result can be larger than a byte in a UTF-8 locale. It
1559 * returns a single value, so can't adequately return the upper case of LATIN
1560 * SMALL LETTER SHARP S in a UTF-8 locale (which should be a string of two
1561 * values "SS"); instead it asserts against that under DEBUGGING, and
1562 * otherwise returns its input */
beab9ebe
KW
1563#define _generic_toUPPER_LC(c, function, cast) \
1564 (! FITS_IN_8_BITS(c) \
1565 ? (c) \
31f05a37 1566 : ((! IN_UTF8_CTYPE_LOCALE) \
b7d90381 1567 ? (cast)function((cast)(c)) \
31f05a37
KW
1568 : ((((U8)(c)) == MICRO_SIGN) \
1569 ? GREEK_CAPITAL_LETTER_MU \
1570 : ((((U8)(c)) == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS) \
1571 ? LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS \
1572 : ((((U8)(c)) == LATIN_SMALL_LETTER_SHARP_S) \
1573 ? (__ASSERT_(0) (c)) \
1574 : PL_mod_latin1_uc[ (U8) (c) ])))))
1575
1576/* Note that the result can be larger than a byte in a UTF-8 locale. It
1577 * returns a single value, so can't adequately return the fold case of LATIN
1578 * SMALL LETTER SHARP S in a UTF-8 locale (which should be a string of two
1579 * values "ss"); instead it asserts against that under DEBUGGING, and
1580 * otherwise returns its input */
beab9ebe 1581#define _generic_toFOLD_LC(c, function, cast) \
4d7db1b9 1582 ((UNLIKELY((c) == MICRO_SIGN) && IN_UTF8_CTYPE_LOCALE) \
31f05a37 1583 ? GREEK_SMALL_LETTER_MU \
4d7db1b9
KW
1584 : (__ASSERT_(! IN_UTF8_CTYPE_LOCALE \
1585 || (c) != LATIN_SMALL_LETTER_SHARP_S) \
1586 _generic_toLOWER_LC(c, function, cast)))
beab9ebe 1587
84061b6a 1588/* Use the libc versions for these if available. */
f05550c0 1589#if defined(HAS_ISASCII)
84061b6a
KW
1590# define isASCII_LC(c) (FITS_IN_8_BITS(c) && isascii( (U8) (c)))
1591#else
1592# define isASCII_LC(c) isASCII(c)
1593#endif
1594
f05550c0 1595#if defined(HAS_ISBLANK)
84061b6a 1596# define isBLANK_LC(c) _generic_LC(c, _CC_BLANK, isblank)
95f34b6f 1597#else /* Unlike isASCII, varies if in a UTF-8 locale */
278b2b56 1598# define isBLANK_LC(c) ((IN_UTF8_CTYPE_LOCALE) ? isBLANK_L1(c) : isBLANK(c))
84061b6a
KW
1599#endif
1600
f05550c0 1601#define _LC_CAST U8
bbce6d69 1602
f05550c0 1603#ifdef WIN32
375f5f06
KW
1604 /* The Windows functions don't bother to follow the POSIX standard, which
1605 * for example says that something can't both be a printable and a control.
1606 * But Windows treats the \t control as a printable, and does such things
1607 * as making superscripts into both digits and punctuation. This tames
1608 * these flaws by assuming that the definitions of both controls and space
1609 * are correct, and then making sure that other definitions don't have
1610 * weirdnesses, by making sure that isalnum() isn't also ispunct(), etc.
1611 * Not all possible weirdnesses are checked for, just the ones that were
1612 * detected on actual Microsoft code pages */
1613
b7d90381
KW
1614# define isCNTRL_LC(c) _generic_LC(c, _CC_CNTRL, iscntrl)
1615# define isSPACE_LC(c) _generic_LC(c, _CC_SPACE, isspace)
1616
1617# define isALPHA_LC(c) (_generic_LC(c, _CC_ALPHA, isalpha) \
1618 && isALPHANUMERIC_LC(c))
1619# define isALPHANUMERIC_LC(c) (_generic_LC(c, _CC_ALPHANUMERIC, isalnum) && \
1620 ! isPUNCT_LC(c))
1621# define isDIGIT_LC(c) (_generic_LC(c, _CC_DIGIT, isdigit) && \
1622 isALPHANUMERIC_LC(c))
1623# define isGRAPH_LC(c) (_generic_LC(c, _CC_GRAPH, isgraph) && isPRINT_LC(c))
1624# define isIDFIRST_LC(c) (((c) == '_') \
1625 || (_generic_LC(c, _CC_IDFIRST, isalpha) && ! isPUNCT_LC(c)))
1626# define isLOWER_LC(c) (_generic_LC(c, _CC_LOWER, islower) && isALPHA_LC(c))
1627# define isPRINT_LC(c) (_generic_LC(c, _CC_PRINT, isprint) && ! isCNTRL_LC(c))
1628# define isPUNCT_LC(c) (_generic_LC(c, _CC_PUNCT, ispunct) && ! isCNTRL_LC(c))
1629# define isUPPER_LC(c) (_generic_LC(c, _CC_UPPER, isupper) && isALPHA_LC(c))
f05550c0 1630# define isWORDCHAR_LC(c) (((c) == '_') || isALPHANUMERIC_LC(c))
b7d90381
KW
1631# define isXDIGIT_LC(c) (_generic_LC(c, _CC_XDIGIT, isxdigit) \
1632 && isALPHANUMERIC_LC(c))
f05550c0
BF
1633
1634# define toLOWER_LC(c) _generic_toLOWER_LC((c), tolower, U8)
1635# define toUPPER_LC(c) _generic_toUPPER_LC((c), toupper, U8)
1636# define toFOLD_LC(c) _generic_toFOLD_LC((c), tolower, U8)
1637
1638#elif defined(CTYPE256) || (!defined(isascii) && !defined(HAS_ISASCII))
4650c663 1639 /* For most other platforms */
beab9ebe 1640
f05550c0
BF
1641# define isALPHA_LC(c) _generic_LC(c, _CC_ALPHA, isalpha)
1642# define isALPHANUMERIC_LC(c) _generic_LC(c, _CC_ALPHANUMERIC, isalnum)
1643# define isCNTRL_LC(c) _generic_LC(c, _CC_CNTRL, iscntrl)
1644# define isDIGIT_LC(c) _generic_LC(c, _CC_DIGIT, isdigit)
1645# define isGRAPH_LC(c) _generic_LC(c, _CC_GRAPH, isgraph)
1646# define isIDFIRST_LC(c) _generic_LC_underscore(c, _CC_IDFIRST, isalpha)
1647# define isLOWER_LC(c) _generic_LC(c, _CC_LOWER, islower)
1648# define isPRINT_LC(c) _generic_LC(c, _CC_PRINT, isprint)
1649# define isPUNCT_LC(c) _generic_LC(c, _CC_PUNCT, ispunct)
1650# define isSPACE_LC(c) _generic_LC(c, _CC_SPACE, isspace)
1651# define isUPPER_LC(c) _generic_LC(c, _CC_UPPER, isupper)
1652# define isWORDCHAR_LC(c) _generic_LC_underscore(c, _CC_WORDCHAR, isalnum)
1653# define isXDIGIT_LC(c) _generic_LC(c, _CC_XDIGIT, isxdigit)
1654
1655
1656# define toLOWER_LC(c) _generic_toLOWER_LC((c), tolower, U8)
1657# define toUPPER_LC(c) _generic_toUPPER_LC((c), toupper, U8)
1658# define toFOLD_LC(c) _generic_toFOLD_LC((c), tolower, U8)
1659
1660#else /* The final fallback position */
1661
b7d90381
KW
1662# define isALPHA_LC(c) (isascii(c) && isalpha(c))
1663# define isALPHANUMERIC_LC(c) (isascii(c) && isalnum(c))
1664# define isCNTRL_LC(c) (isascii(c) && iscntrl(c))
1665# define isDIGIT_LC(c) (isascii(c) && isdigit(c))
1666# define isGRAPH_LC(c) (isascii(c) && isgraph(c))
f05550c0 1667# define isIDFIRST_LC(c) (isascii(c) && (isalpha(c) || (c) == '_'))
b7d90381
KW
1668# define isLOWER_LC(c) (isascii(c) && islower(c))
1669# define isPRINT_LC(c) (isascii(c) && isprint(c))
1670# define isPUNCT_LC(c) (isascii(c) && ispunct(c))
1671# define isSPACE_LC(c) (isascii(c) && isspace(c))
1672# define isUPPER_LC(c) (isascii(c) && isupper(c))
f05550c0 1673# define isWORDCHAR_LC(c) (isascii(c) && (isalnum(c) || (c) == '_'))
b7d90381 1674# define isXDIGIT_LC(c) (isascii(c) && isxdigit(c))
f05550c0
BF
1675
1676# define toLOWER_LC(c) (isascii(c) ? tolower(c) : (c))
1677# define toUPPER_LC(c) (isascii(c) ? toupper(c) : (c))
1678# define toFOLD_LC(c) (isascii(c) ? tolower(c) : (c))
bbce6d69 1679
f05550c0 1680#endif
55204971 1681
eba68aa0
KW
1682#define isIDCONT(c) isWORDCHAR(c)
1683#define isIDCONT_A(c) isWORDCHAR_A(c)
1684#define isIDCONT_L1(c) isWORDCHAR_L1(c)
1685#define isIDCONT_LC(c) isWORDCHAR_LC(c)
13380643 1686#define isPSXSPC_LC(c) isSPACE_LC(c)
aaa51d5e 1687
4650c663 1688/* For internal core Perl use only: the base macros for defining macros like
d0da05db
KW
1689 * isALPHA_uvchr. 'c' is the code point to check. 'classnum' is the POSIX class
1690 * number defined earlier in this file. _generic_uvchr() is used for POSIX
4650c663
KW
1691 * classes where there is a macro or function 'above_latin1' that takes the
1692 * single argument 'c' and returns the desired value. These exist for those
1693 * classes which have simple definitions, avoiding the overhead of a hash
d0da05db 1694 * lookup or inversion list binary search. _generic_swash_uvchr() can be used
4650c663 1695 * for classes where that overhead is faster than a direct lookup.
d0da05db 1696 * _generic_uvchr() won't compile if 'c' isn't unsigned, as it won't match the
4650c663
KW
1697 * 'above_latin1' prototype. _generic_isCC() macro does bounds checking, so
1698 * have duplicate checks here, so could create versions of the macros that
1699 * don't, but experiments show that gcc optimizes them out anyway. */
66c17564
KW
1700
1701/* Note that all ignore 'use bytes' */
1e222e4f
KW
1702#define _generic_uvchr(classnum, above_latin1, c) ((c) < 256 \
1703 ? _generic_isCC(c, classnum) \
cd500f2f 1704 : above_latin1(c))
1e222e4f
KW
1705#define _generic_swash_uvchr(classnum, c) ((c) < 256 \
1706 ? _generic_isCC(c, classnum) \
922e8cb4 1707 : _is_uni_FOO(classnum, c))
d0da05db
KW
1708#define isALPHA_uvchr(c) _generic_swash_uvchr(_CC_ALPHA, c)
1709#define isALPHANUMERIC_uvchr(c) _generic_swash_uvchr(_CC_ALPHANUMERIC, c)
1710#define isASCII_uvchr(c) isASCII(c)
1711#define isBLANK_uvchr(c) _generic_uvchr(_CC_BLANK, is_HORIZWS_cp_high, c)
1712#define isCNTRL_uvchr(c) isCNTRL_L1(c) /* All controls are in Latin1 */
1713#define isDIGIT_uvchr(c) _generic_swash_uvchr(_CC_DIGIT, c)
1714#define isGRAPH_uvchr(c) _generic_swash_uvchr(_CC_GRAPH, c)
1e222e4f
KW
1715#define isIDCONT_uvchr(c) \
1716 _generic_uvchr(_CC_WORDCHAR, _is_uni_perl_idcont, c)
1717#define isIDFIRST_uvchr(c) \
1718 _generic_uvchr(_CC_IDFIRST, _is_uni_perl_idstart, c)
d0da05db
KW
1719#define isLOWER_uvchr(c) _generic_swash_uvchr(_CC_LOWER, c)
1720#define isPRINT_uvchr(c) _generic_swash_uvchr(_CC_PRINT, c)
1721
1722#define isPUNCT_uvchr(c) _generic_swash_uvchr(_CC_PUNCT, c)
1723#define isSPACE_uvchr(c) _generic_uvchr(_CC_SPACE, is_XPERLSPACE_cp_high, c)
1724#define isPSXSPC_uvchr(c) isSPACE_uvchr(c)
1725
1726#define isUPPER_uvchr(c) _generic_swash_uvchr(_CC_UPPER, c)
1727#define isVERTWS_uvchr(c) _generic_uvchr(_CC_VERTSPACE, is_VERTWS_cp_high, c)
1728#define isWORDCHAR_uvchr(c) _generic_swash_uvchr(_CC_WORDCHAR, c)
1729#define isXDIGIT_uvchr(c) _generic_uvchr(_CC_XDIGIT, is_XDIGIT_cp_high, c)
1730
1731#define toFOLD_uvchr(c,s,l) to_uni_fold(c,s,l)
1732#define toLOWER_uvchr(c,s,l) to_uni_lower(c,s,l)
1733#define toTITLE_uvchr(c,s,l) to_uni_title(c,s,l)
1734#define toUPPER_uvchr(c,s,l) to_uni_upper(c,s,l)
1735
1736/* For backwards compatibility, even though '_uni' should mean official Unicode
1737 * code points, in Perl it means native for those below 256 */
1738#define isALPHA_uni(c) isALPHA_uvchr(c)
1739#define isALPHANUMERIC_uni(c) isALPHANUMERIC_uvchr(c)
1740#define isASCII_uni(c) isASCII_uvchr(c)
1741#define isBLANK_uni(c) isBLANK_uvchr(c)
1742#define isCNTRL_uni(c) isCNTRL_uvchr(c)
1743#define isDIGIT_uni(c) isDIGIT_uvchr(c)
1744#define isGRAPH_uni(c) isGRAPH_uvchr(c)
1745#define isIDCONT_uni(c) isIDCONT_uvchr(c)
1746#define isIDFIRST_uni(c) isIDFIRST_uvchr(c)
1747#define isLOWER_uni(c) isLOWER_uvchr(c)
1748#define isPRINT_uni(c) isPRINT_uvchr(c)
1749#define isPUNCT_uni(c) isPUNCT_uvchr(c)
1750#define isSPACE_uni(c) isSPACE_uvchr(c)
1751#define isPSXSPC_uni(c) isPSXSPC_uvchr(c)
1752#define isUPPER_uni(c) isUPPER_uvchr(c)
1753#define isVERTWS_uni(c) isVERTWS_uvchr(c)
1754#define isWORDCHAR_uni(c) isWORDCHAR_uvchr(c)
1755#define isXDIGIT_uni(c) isXDIGIT_uvchr(c)
1756#define toFOLD_uni(c,s,l) toFOLD_uvchr(c,s,l)
1757#define toLOWER_uni(c,s,l) toLOWER_uvchr(c,s,l)
1758#define toTITLE_uni(c,s,l) toTITLE_uvchr(c,s,l)
1759#define toUPPER_uni(c,s,l) toUPPER_uvchr(c,s,l)
a0ed51b3 1760
4650c663
KW
1761/* For internal core Perl use only: the base macros for defining macros like
1762 * isALPHA_LC_uvchr. These are like isALPHA_LC, but the input can be any code
d0da05db 1763 * point, not just 0-255. Like _generic_uvchr, there are two versions, one for
4650c663 1764 * simple class definitions; the other for more complex. These are like
d0da05db 1765 * _generic_uvchr, so see it for more info. */
cd500f2f
KW
1766#define _generic_LC_uvchr(latin1, above_latin1, c) \
1767 (c < 256 ? latin1(c) : above_latin1(c))
1768#define _generic_LC_swash_uvchr(latin1, classnum, c) \
1769 (c < 256 ? latin1(c) : _is_uni_FOO(classnum, c))
1770
1771#define isALPHA_LC_uvchr(c) _generic_LC_swash_uvchr(isALPHA_LC, _CC_ALPHA, c)
1772#define isALPHANUMERIC_LC_uvchr(c) _generic_LC_swash_uvchr(isALPHANUMERIC_LC, \
1773 _CC_ALPHANUMERIC, c)
b7d90381
KW
1774#define isASCII_LC_uvchr(c) isASCII_LC(c)
1775#define isBLANK_LC_uvchr(c) _generic_LC_uvchr(isBLANK_LC, \
1776 is_HORIZWS_cp_high, c)
feeab5a9 1777#define isCNTRL_LC_uvchr(c) (c < 256 ? isCNTRL_LC(c) : 0)
cd500f2f
KW
1778#define isDIGIT_LC_uvchr(c) _generic_LC_swash_uvchr(isDIGIT_LC, _CC_DIGIT, c)
1779#define isGRAPH_LC_uvchr(c) _generic_LC_swash_uvchr(isGRAPH_LC, _CC_GRAPH, c)
b7d90381 1780#define isIDCONT_LC_uvchr(c) _generic_LC_uvchr(isIDCONT_LC, \
eba68aa0 1781 _is_uni_perl_idcont, c)
b7d90381 1782#define isIDFIRST_LC_uvchr(c) _generic_LC_uvchr(isIDFIRST_LC, \
cd500f2f
KW
1783 _is_uni_perl_idstart, c)
1784#define isLOWER_LC_uvchr(c) _generic_LC_swash_uvchr(isLOWER_LC, _CC_LOWER, c)
1785#define isPRINT_LC_uvchr(c) _generic_LC_swash_uvchr(isPRINT_LC, _CC_PRINT, c)
b7d90381 1786#define isPSXSPC_LC_uvchr(c) isSPACE_LC_uvchr(c)
cd500f2f 1787#define isPUNCT_LC_uvchr(c) _generic_LC_swash_uvchr(isPUNCT_LC, _CC_PUNCT, c)
b7d90381 1788#define isSPACE_LC_uvchr(c) _generic_LC_uvchr(isSPACE_LC, \
509fb054 1789 is_XPERLSPACE_cp_high, c)
cd500f2f 1790#define isUPPER_LC_uvchr(c) _generic_LC_swash_uvchr(isUPPER_LC, _CC_UPPER, c)
b7d90381 1791#define isWORDCHAR_LC_uvchr(c) _generic_LC_swash_uvchr(isWORDCHAR_LC, \
cd500f2f 1792 _CC_WORDCHAR, c)
b7d90381
KW
1793#define isXDIGIT_LC_uvchr(c) _generic_LC_uvchr(isXDIGIT_LC, \
1794 is_XDIGIT_cp_high, c)
e712593e 1795
b7d90381 1796#define isBLANK_LC_uni(c) isBLANK_LC_uvchr(UNI_TO_NATIVE(c))
aaa51d5e 1797
4650c663
KW
1798/* For internal core Perl use only: the base macros for defining macros like
1799 * isALPHA_utf8. These are like the earlier defined macros, but take an input
1800 * UTF-8 encoded string 'p'. If the input is in the Latin1 range, use
1801 * the Latin1 macro 'classnum' on 'p'. Otherwise use the value given by the
1802 * 'utf8' parameter. This relies on the fact that ASCII characters have the
1803 * same representation whether utf8 or not. Note that it assumes that the utf8
1804 * has been validated, and ignores 'use bytes' */
34aeb2e9
KW
1805#define _base_generic_utf8(enum_name, name, p, use_locale ) \
1806 _is_utf8_FOO(CAT2(_CC_, enum_name), \
1807 (const U8 *) p, \
1808 "is" STRINGIFY(name) "_utf8", \
1809 "is" STRINGIFY(name) "_utf8_safe", \
1810 1, use_locale, __FILE__,__LINE__)
1811
1812#define _generic_utf8(name, p) _base_generic_utf8(name, name, p, 0)
da8c1a98
KW
1813
1814/* The "_safe" macros make sure that we don't attempt to read beyond 'e', but
1815 * they don't otherwise go out of their way to look for malformed UTF-8. If
1816 * they can return accurate results without knowing if the input is otherwise
1817 * malformed, they do so. For example isASCII is accurate in spite of any
1818 * non-length malformations because it looks only at a single byte. Likewise
1819 * isDIGIT looks just at the first byte for code points 0-255, as all UTF-8
1820 * variant ones return FALSE. But, if the input has to be well-formed in order
1821 * for the results to be accurate, the macros will test and if malformed will
1822 * call a routine to die
1823 *
1824 * Except for toke.c, the macros do assume that e > p, asserting that on
1825 * DEBUGGING builds. Much code that calls these depends on this being true,
1826 * for other reasons. toke.c is treated specially as using the regular
1827 * assertion breaks it in many ways. All strings that these operate on there
1828 * are supposed to have an extra NUL character at the end, so that *e = \0. A
1829 * bunch of code in toke.c assumes that this is true, so the assertion allows
1830 * for that */
1831#ifdef PERL_IN_TOKE_C
1832# define _utf8_safe_assert(p,e) ((e) > (p) || ((e) == (p) && *(p) == '\0'))
1833#else
1834# define _utf8_safe_assert(p,e) ((e) > (p))
1835#endif
1836
1837#define _generic_utf8_safe(classnum, p, e, above_latin1) \
1838 (__ASSERT_(_utf8_safe_assert(p, e)) \
1839 (UTF8_IS_INVARIANT(*(p))) \
1840 ? _generic_isCC(*(p), classnum) \
1841 : (UTF8_IS_DOWNGRADEABLE_START(*(p)) \
1842 ? ((LIKELY((e) - (p) > 1 && UTF8_IS_CONTINUATION(*((p)+1)))) \
1843 ? _generic_isCC(EIGHT_BIT_UTF8_TO_NATIVE(*(p), *((p)+1 )), \
1844 classnum) \
1845 : (_force_out_malformed_utf8_message( \
1846 (U8 *) (p), (U8 *) (e), 0, 1), 0)) \
1847 : above_latin1))
b7d90381
KW
1848/* Like the above, but calls 'above_latin1(p)' to get the utf8 value.
1849 * 'above_latin1' can be a macro */
da8c1a98
KW
1850#define _generic_func_utf8_safe(classnum, above_latin1, p, e) \
1851 _generic_utf8_safe(classnum, p, e, above_latin1(p, e))
1852#define _generic_non_swash_utf8_safe(classnum, above_latin1, p, e) \
1853 _generic_utf8_safe(classnum, p, e, \
1854 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
1855 ? (_force_out_malformed_utf8_message( \
1856 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
1857 : above_latin1(p)))
4650c663 1858/* Like the above, but passes classnum to _isFOO_utf8(), instead of having an
cd500f2f 1859 * 'above_latin1' parameter */
da8c1a98
KW
1860#define _generic_swash_utf8_safe(classnum, p, e) \
1861_generic_utf8_safe(classnum, p, e, _is_utf8_FOO_with_len(classnum, p, e))
922e8cb4 1862
cc8ab7c0 1863/* Like the above, but should be used only when it is known that there are no
ff7ecfc3
KW
1864 * characters in the upper-Latin1 range (128-255 on ASCII platforms) which the
1865 * class is TRUE for. Hence it can skip the tests for this range.
1866 * 'above_latin1' should include its arguments */
da8c1a98
KW
1867#define _generic_utf8_safe_no_upper_latin1(classnum, p, e, above_latin1) \
1868 (__ASSERT_(_utf8_safe_assert(p, e)) \
1869 (UTF8_IS_INVARIANT(*(p))) \
1870 ? _generic_isCC(*(p), classnum) \
1871 : (UTF8_IS_DOWNGRADEABLE_START(*(p))) \
1872 ? 0 /* Note that doesn't check validity for latin1 */ \
1873 : above_latin1)
1874
84238efa
KW
1875/* NOTE that some of these macros have very similar ones in regcharclass.h.
1876 * For example, there is (at the time of this writing) an 'is_SPACE_utf8()'
1877 * there, differing in name only by an underscore from the one here
1878 * 'isSPACE_utf8(). The difference is that the ones here are probably more
1879 * efficient and smaller, using an O(1) array lookup for Latin1-range code
1880 * points; the regcharclass.h ones are implemented as a series of
1881 * "if-else-if-else ..." */
1882
34aeb2e9
KW
1883#define isALPHA_utf8(p) _generic_utf8(ALPHA, p)
1884#define isALPHANUMERIC_utf8(p) _generic_utf8(ALPHANUMERIC, p)
1885#define isASCII_utf8(p) _generic_utf8(ASCII, p)
1886#define isBLANK_utf8(p) _generic_utf8(BLANK, p)
1887#define isCNTRL_utf8(p) _generic_utf8(CNTRL, p)
1888#define isDIGIT_utf8(p) _generic_utf8(DIGIT, p)
1889#define isGRAPH_utf8(p) _generic_utf8(GRAPH, p)
1890#define isIDCONT_utf8(p) _generic_utf8(IDCONT, p)
1891#define isIDFIRST_utf8(p) _generic_utf8(IDFIRST, p)
1892#define isLOWER_utf8(p) _generic_utf8(LOWER, p)
1893#define isPRINT_utf8(p) _generic_utf8(PRINT, p)
1894#define isPSXSPC_utf8(p) _generic_utf8(PSXSPC, p)
1895#define isPUNCT_utf8(p) _generic_utf8(PUNCT, p)
1896#define isSPACE_utf8(p) _generic_utf8(SPACE, p)
1897#define isUPPER_utf8(p) _generic_utf8(UPPER, p)
1898#define isVERTWS_utf8(p) _generic_utf8(VERTSPACE, p)
1899#define isWORDCHAR_utf8(p) _generic_utf8(WORDCHAR, p)
1900#define isXDIGIT_utf8(p) _generic_utf8(XDIGIT, p)
e8fa43e2 1901
da8c1a98
KW
1902#define isALPHA_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_ALPHA, p, e)
1903#define isALPHANUMERIC_utf8_safe(p, e) \
1904 _generic_swash_utf8_safe(_CC_ALPHANUMERIC, p, e)
1905#define isASCII_utf8_safe(p, e) \
1906 /* Because ASCII is invariant under utf8, the non-utf8 macro \
1907 * works */ \
1908 (__ASSERT_(_utf8_safe_assert(p, e)) isASCII(*(p)))
1909#define isBLANK_utf8_safe(p, e) \
1910 _generic_non_swash_utf8_safe(_CC_BLANK, is_HORIZWS_high, p, e)
1911
e8fa43e2
KW
1912#ifdef EBCDIC
1913 /* Because all controls are UTF-8 invariants in EBCDIC, we can use this
1914 * more efficient macro instead of the more general one */
da8c1a98 1915# define isCNTRL_utf8_safe(p, e) \
56d02b8c 1916 (__ASSERT_(_utf8_safe_assert(p, e)) isCNTRL_L1(*(p)))
e8fa43e2 1917#else
da8c1a98 1918# define isCNTRL_utf8_safe(p, e) _generic_utf8_safe(_CC_CNTRL, p, e, 0)
e8fa43e2
KW
1919#endif
1920
da8c1a98
KW
1921#define isDIGIT_utf8_safe(p, e) \
1922 _generic_utf8_safe_no_upper_latin1(_CC_DIGIT, p, e, \
1923 _is_utf8_FOO_with_len(_CC_DIGIT, p, e))
1924#define isGRAPH_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_GRAPH, p, e)
1925#define isIDCONT_utf8_safe(p, e) _generic_func_utf8_safe(_CC_WORDCHAR, \
1926 _is_utf8_perl_idcont_with_len, p, e)
e5dcd934 1927
c11ff943
KW
1928/* To prevent S_scan_word in toke.c from hanging, we have to make sure that
1929 * IDFIRST is an alnum. See
c8362b00 1930 * http://rt.perl.org/rt3/Ticket/Display.html?id=74022 for more detail than you
f91dcd13
KW
1931 * ever wanted to know about. (In the ASCII range, there isn't a difference.)
1932 * This used to be not the XID version, but we decided to go with the more
1933 * modern Unicode definition */
da8c1a98
KW
1934#define isIDFIRST_utf8_safe(p, e) \
1935 _generic_func_utf8_safe(_CC_IDFIRST, \
1936 _is_utf8_perl_idstart_with_len, (U8 *) (p), (U8 *) (e))
1937
1938#define isLOWER_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_LOWER, p, e)
1939#define isPRINT_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_PRINT, p, e)
1940#define isPSXSPC_utf8_safe(p, e) isSPACE_utf8_safe(p, e)
1941#define isPUNCT_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_PUNCT, p, e)
1942#define isSPACE_utf8_safe(p, e) \
1943 _generic_non_swash_utf8_safe(_CC_SPACE, is_XPERLSPACE_high, p, e)
1944#define isUPPER_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_UPPER, p, e)
1945#define isVERTWS_utf8_safe(p, e) \
1946 _generic_non_swash_utf8_safe(_CC_VERTSPACE, is_VERTWS_high, p, e)
1947#define isWORDCHAR_utf8_safe(p, e) \
1948 _generic_swash_utf8_safe(_CC_WORDCHAR, p, e)
1949#define isXDIGIT_utf8_safe(p, e) \
1950 _generic_utf8_safe_no_upper_latin1(_CC_XDIGIT, p, e, \
1951 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
1952 ? (_force_out_malformed_utf8_message( \
1953 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
1954 : is_XDIGIT_high(p)))
a0ed51b3 1955
7d2b61dd 1956#define toFOLD_utf8(p,s,l) to_utf8_fold(p,s,l)
509fb054
KW
1957#define toLOWER_utf8(p,s,l) to_utf8_lower(p,s,l)
1958#define toTITLE_utf8(p,s,l) to_utf8_title(p,s,l)
1959#define toUPPER_utf8(p,s,l) to_utf8_upper(p,s,l)
2e8adce6 1960
567b353c 1961/* For internal core use only, subject to change */
607313a1
KW
1962#define _toFOLD_utf8_flags(p,e,s,l,f) _to_utf8_fold_flags (p,e,s,l,f, "", 0)
1963#define _toLOWER_utf8_flags(p,e,s,l,f) _to_utf8_lower_flags(p,e,s,l,f, "", 0)
1964#define _toTITLE_utf8_flags(p,e,s,l,f) _to_utf8_title_flags(p,e,s,l,f, "", 0)
1965#define _toUPPER_utf8_flags(p,e,s,l,f) _to_utf8_upper_flags(p,e,s,l,f, "", 0)
a1a5ec35
KW
1966
1967#define toFOLD_utf8_safe(p,e,s,l) _toFOLD_utf8_flags(p,e,s,l, FOLD_FLAGS_FULL)
1968#define toLOWER_utf8_safe(p,e,s,l) _toLOWER_utf8_flags(p,e,s,l, 0)
1969#define toTITLE_utf8_safe(p,e,s,l) _toTITLE_utf8_flags(p,e,s,l, 0)
1970#define toUPPER_utf8_safe(p,e,s,l) _toUPPER_utf8_flags(p,e,s,l, 0)
567b353c 1971
4650c663
KW
1972/* For internal core Perl use only: the base macros for defining macros like
1973 * isALPHA_LC_utf8. These are like _generic_utf8, but if the first code point
1974 * in 'p' is within the 0-255 range, it uses locale rules from the passed-in
1975 * 'macro' parameter */
34aeb2e9
KW
1976#define _generic_LC_utf8(name, p) _base_generic_utf8(name, name, p, 1)
1977
1978#define isALPHA_LC_utf8(p) _generic_LC_utf8(ALPHA, p)
1979#define isALPHANUMERIC_LC_utf8(p) _generic_LC_utf8(ALPHANUMERIC, p)
1980#define isASCII_LC_utf8(p) _generic_LC_utf8(ASCII, p)
1981#define isBLANK_LC_utf8(p) _generic_LC_utf8(BLANK, p)
1982#define isCNTRL_LC_utf8(p) _generic_LC_utf8(CNTRL, p)
1983#define isDIGIT_LC_utf8(p) _generic_LC_utf8(DIGIT, p)
1984#define isGRAPH_LC_utf8(p) _generic_LC_utf8(GRAPH, p)
1985#define isIDCONT_LC_utf8(p) _generic_LC_utf8(IDCONT, p)
1986#define isIDFIRST_LC_utf8(p) _generic_LC_utf8(IDFIRST, p)
1987#define isLOWER_LC_utf8(p) _generic_LC_utf8(LOWER, p)
1988#define isPRINT_LC_utf8(p) _generic_LC_utf8(PRINT, p)
1989#define isPSXSPC_LC_utf8(p) _generic_LC_utf8(PSXSPC, p)
1990#define isPUNCT_LC_utf8(p) _generic_LC_utf8(PUNCT, p)
1991#define isSPACE_LC_utf8(p) _generic_LC_utf8(SPACE, p)
1992#define isUPPER_LC_utf8(p) _generic_LC_utf8(UPPER, p)
1993#define isWORDCHAR_LC_utf8(p) _generic_LC_utf8(WORDCHAR, p)
1994#define isXDIGIT_LC_utf8(p) _generic_LC_utf8(XDIGIT, p)
1995
da8c1a98
KW
1996/* For internal core Perl use only: the base macros for defining macros like
1997 * isALPHA_LC_utf8_safe. These are like _generic_utf8, but if the first code
1998 * point in 'p' is within the 0-255 range, it uses locale rules from the
1999 * passed-in 'macro' parameter */
2000#define _generic_LC_utf8_safe(macro, p, e, above_latin1) \
2001 (__ASSERT_(_utf8_safe_assert(p, e)) \
2002 (UTF8_IS_INVARIANT(*(p))) \
2003 ? macro(*(p)) \
2004 : (UTF8_IS_DOWNGRADEABLE_START(*(p)) \
2005 ? ((LIKELY((e) - (p) > 1 && UTF8_IS_CONTINUATION(*((p)+1)))) \
2006 ? macro(EIGHT_BIT_UTF8_TO_NATIVE(*(p), *((p)+1))) \
2007 : (_force_out_malformed_utf8_message( \
2008 (U8 *) (p), (U8 *) (e), 0, 1), 0)) \
2009 : above_latin1))
2010
2011#define _generic_LC_swash_utf8_safe(macro, classnum, p, e) \
2012 _generic_LC_utf8_safe(macro, p, e, \
2013 _is_utf8_FOO_with_len(classnum, p, e))
2014
2015#define _generic_LC_func_utf8_safe(macro, above_latin1, p, e) \
2016 _generic_LC_utf8_safe(macro, p, e, above_latin1(p, e))
2017
2018#define _generic_LC_non_swash_utf8_safe(classnum, above_latin1, p, e) \
2019 _generic_LC_utf8_safe(classnum, p, e, \
2020 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
2021 ? (_force_out_malformed_utf8_message( \
2022 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
2023 : above_latin1(p)))
2024
2025#define isALPHANUMERIC_LC_utf8_safe(p, e) \
2026 _generic_LC_swash_utf8_safe(isALPHANUMERIC_LC, \
2027 _CC_ALPHANUMERIC, p, e)
2028#define isALPHA_LC_utf8_safe(p, e) \
2029 _generic_LC_swash_utf8_safe(isALPHA_LC, _CC_ALPHA, p, e)
2030#define isASCII_LC_utf8_safe(p, e) \
2031 (__ASSERT_(_utf8_safe_assert(p, e)) isASCII_LC(*(p)))
2032#define isBLANK_LC_utf8_safe(p, e) \
2033 _generic_LC_non_swash_utf8_safe(isBLANK_LC, is_HORIZWS_high, p, e)
2034#define isCNTRL_LC_utf8_safe(p, e) \
2035 _generic_LC_utf8_safe(isCNTRL_LC, p, e, 0)
2036#define isDIGIT_LC_utf8_safe(p, e) \
2037 _generic_LC_swash_utf8_safe(isDIGIT_LC, _CC_DIGIT, p, e)
2038#define isGRAPH_LC_utf8_safe(p, e) \
2039 _generic_LC_swash_utf8_safe(isGRAPH_LC, _CC_GRAPH, p, e)
2040#define isIDCONT_LC_utf8_safe(p, e) \
2041 _generic_LC_func_utf8_safe(isIDCONT_LC, \
2042 _is_utf8_perl_idcont_with_len, p, e)
2043#define isIDFIRST_LC_utf8_safe(p, e) \
2044 _generic_LC_func_utf8_safe(isIDFIRST_LC, \
2045 _is_utf8_perl_idstart_with_len, p, e)
2046#define isLOWER_LC_utf8_safe(p, e) \
2047 _generic_LC_swash_utf8_safe(isLOWER_LC, _CC_LOWER, p, e)
2048#define isPRINT_LC_utf8_safe(p, e) \
2049 _generic_LC_swash_utf8_safe(isPRINT_LC, _CC_PRINT, p, e)
2050#define isPSXSPC_LC_utf8_safe(p, e) isSPACE_LC_utf8_safe(p, e)
2051#define isPUNCT_LC_utf8_safe(p, e) \
2052 _generic_LC_swash_utf8_safe(isPUNCT_LC, _CC_PUNCT, p, e)
2053#define isSPACE_LC_utf8_safe(p, e) \
2054 _generic_LC_non_swash_utf8_safe(isSPACE_LC, is_XPERLSPACE_high, p, e)
2055#define isUPPER_LC_utf8_safe(p, e) \
2056 _generic_LC_swash_utf8_safe(isUPPER_LC, _CC_UPPER, p, e)
2057#define isWORDCHAR_LC_utf8_safe(p, e) \
2058 _generic_LC_swash_utf8_safe(isWORDCHAR_LC, _CC_WORDCHAR, p, e)
2059#define isXDIGIT_LC_utf8_safe(p, e) \
2060 _generic_LC_non_swash_utf8_safe(isXDIGIT_LC, is_XDIGIT_high, p, e)
aaa51d5e 2061
fbc19f27
KW
2062/* Macros for backwards compatibility and for completeness when the ASCII and
2063 * Latin1 values are identical */
b7d90381
KW
2064#define isALPHAU(c) isALPHA_L1(c)
2065#define isDIGIT_L1(c) isDIGIT_A(c)
2066#define isOCTAL(c) isOCTAL_A(c)
2067#define isOCTAL_L1(c) isOCTAL_A(c)
2068#define isXDIGIT_L1(c) isXDIGIT_A(c)
2069#define isALNUM(c) isWORDCHAR(c)
2070#define isALNUMU(c) isWORDCHAR_L1(c)
2071#define isALNUM_LC(c) isWORDCHAR_LC(c)
2072#define isALNUM_uni(c) isWORDCHAR_uni(c)
2e28f0b9 2073#define isALNUM_LC_uvchr(c) isWORDCHAR_LC_uvchr(c)
b7d90381
KW
2074#define isALNUM_utf8(p) isWORDCHAR_utf8(p)
2075#define isALNUM_LC_utf8(p) isWORDCHAR_LC_utf8(p)
2076#define isALNUMC_A(c) isALPHANUMERIC_A(c) /* Mnemonic: "C's alnum" */
2077#define isALNUMC_L1(c) isALPHANUMERIC_L1(c)
2078#define isALNUMC(c) isALPHANUMERIC(c)
2079#define isALNUMC_LC(c) isALPHANUMERIC_LC(c)
2080#define isALNUMC_uni(c) isALPHANUMERIC_uni(c)
15861f94 2081#define isALNUMC_LC_uvchr(c) isALPHANUMERIC_LC_uvchr(c)
b7d90381 2082#define isALNUMC_utf8(p) isALPHANUMERIC_utf8(p)
15861f94 2083#define isALNUMC_LC_utf8(p) isALPHANUMERIC_LC_utf8(p)
fbc19f27 2084
2bd1cbf6
KW
2085/* On EBCDIC platforms, CTRL-@ is 0, CTRL-A is 1, etc, just like on ASCII,
2086 * except that they don't necessarily mean the same characters, e.g. CTRL-D is
2087 * 4 on both systems, but that is EOT on ASCII; ST on EBCDIC.
2088 * '?' is special-cased on EBCDIC to APC, which is the control there that is
2089 * the outlier from the block that contains the other controls, just like
2090 * toCTRL('?') on ASCII yields DEL, the control that is the outlier from the C0
2091 * block. If it weren't special cased, it would yield a non-control.
88794300
KW
2092 * The conversion works both ways, so toCTRL('D') is 4, and toCTRL(4) is D,
2093 * etc. */
2bd1cbf6 2094#ifndef EBCDIC
75763b3a 2095# define toCTRL(c) (__ASSERT_(FITS_IN_8_BITS(c)) toUPPER(((U8)(c))) ^ 64)
2bd1cbf6 2096#else
75763b3a
KW
2097# define toCTRL(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
2098 ((isPRINT_A(c)) \
2099 ? (UNLIKELY((c) == '?') \
2100 ? QUESTION_MARK_CTRL \
2101 : (NATIVE_TO_LATIN1(toUPPER((U8) (c))) ^ 64)) \
2102 : (UNLIKELY((c) == QUESTION_MARK_CTRL) \
2103 ? '?' \
2104 : (LATIN1_TO_NATIVE(((U8) (c)) ^ 64)))))
2bd1cbf6 2105#endif
bbce6d69 2106
dea28490
JJ
2107/* Line numbers are unsigned, 32 bits. */
2108typedef U32 line_t;
e5dcd934 2109#define NOLINE ((line_t) 4294967295UL) /* = FFFFFFFF */
378cc40b 2110
91152fc1
DG
2111/* Helpful alias for version prescan */
2112#define is_LAX_VERSION(a,b) \
2113 (a != Perl_prescan_version(aTHX_ a, FALSE, b, NULL, NULL, NULL, NULL))
2114
2115#define is_STRICT_VERSION(a,b) \
2116 (a != Perl_prescan_version(aTHX_ a, TRUE, b, NULL, NULL, NULL, NULL))
2117
2118#define BADVERSION(a,b,c) \
2119 if (b) { \
2120 *b = c; \
2121 } \
2122 return a;
8c52afec 2123
cb27eebd
KW
2124/* Converts a character known to represent a hexadecimal digit (0-9, A-F, or
2125 * a-f) to its numeric value. READ_XDIGIT's argument is a string pointer,
2126 * which is advanced. The input is validated only by an assert() in DEBUGGING
2127 * builds. In both ASCII and EBCDIC the last 4 bits of the digits are 0-9; and
2128 * the last 4 bits of A-F and a-f are 1-6, so adding 9 yields 10-15 */
2129#define XDIGIT_VALUE(c) (__ASSERT_(isXDIGIT(c)) (0xf & (isDIGIT(c) \
2130 ? (c) \
2131 : ((c) + 9))))
902008b8
KW
2132#define READ_XDIGIT(s) (__ASSERT_(isXDIGIT(*s)) (0xf & (isDIGIT(*(s)) \
2133 ? (*(s)++) \
2134 : (*(s)++ + 9))))
95a59cab 2135
cb27eebd
KW
2136/* Converts a character known to represent an octal digit (0-7) to its numeric
2137 * value. The input is validated only by an assert() in DEBUGGING builds. In
2138 * both ASCII and EBCDIC the last 3 bits of the octal digits range from 0-7. */
2139#define OCTAL_VALUE(c) (__ASSERT_(isOCTAL(c)) (7 & (c)))
2140
305b8651
KW
2141/* Efficiently returns a boolean as to if two native characters are equivalent
2142 * case-insenstively. At least one of the characters must be one of [A-Za-z];
2143 * the ALPHA in the name is to remind you of that. This is asserted() in
2144 * DEBUGGING builds. Because [A-Za-z] are invariant under UTF-8, this macro
2145 * works (on valid input) for both non- and UTF-8-encoded bytes.
2146 *
2147 * When one of the inputs is a compile-time constant and gets folded by the
2148 * compiler, this reduces to an AND and a TEST. On both EBCDIC and ASCII
2149 * machines, 'A' and 'a' differ by a single bit; the same with the upper and
2150 * lower case of all other ASCII-range alphabetics. On ASCII platforms, they
96ca48da
KW
2151 * are 32 apart; on EBCDIC, they are 64. At compile time, this uses an
2152 * exclusive 'or' to find that bit and then inverts it to form a mask, with
2153 * just a single 0, in the bit position where the upper- and lowercase differ.
2154 * */
305b8651
KW
2155#define isALPHA_FOLD_EQ(c1, c2) \
2156 (__ASSERT_(isALPHA_A(c1) || isALPHA_A(c2)) \
2157 ((c1) & ~('A' ^ 'a')) == ((c2) & ~('A' ^ 'a')))
2158#define isALPHA_FOLD_NE(c1, c2) (! isALPHA_FOLD_EQ((c1), (c2)))
2159
8e84507e 2160/*
ccfc67b7
JH
2161=head1 Memory Management
2162
a02a5408 2163=for apidoc Am|void|Newx|void* ptr|int nitems|type
954c1994
GS
2164The XSUB-writer's interface to the C C<malloc> function.
2165
596f7718 2166Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2167
c5008215
JC
2168In 5.9.3, Newx() and friends replace the older New() API, and drops
2169the first parameter, I<x>, a debug aid which allowed callers to identify
37b8b4c9 2170themselves. This aid has been superseded by a new build option,
d10b4965 2171PERL_MEM_LOG (see L<perlhacktips/PERL_MEM_LOG>). The older API is still
c5008215
JC
2172there for use in XS modules supporting older perls.
2173
a02a5408 2174=for apidoc Am|void|Newxc|void* ptr|int nitems|type|cast
954c1994 2175The XSUB-writer's interface to the C C<malloc> function, with
fbe13c60 2176cast. See also C<L</Newx>>.
954c1994 2177
596f7718 2178Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2179
a02a5408 2180=for apidoc Am|void|Newxz|void* ptr|int nitems|type
954c1994 2181The XSUB-writer's interface to the C C<malloc> function. The allocated
fbe13c60 2182memory is zeroed with C<memzero>. See also C<L</Newx>>.
a02a5408 2183
596f7718 2184Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2185
954c1994
GS
2186=for apidoc Am|void|Renew|void* ptr|int nitems|type
2187The XSUB-writer's interface to the C C<realloc> function.
2188
596f7718 2189Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2190
954c1994
GS
2191=for apidoc Am|void|Renewc|void* ptr|int nitems|type|cast
2192The XSUB-writer's interface to the C C<realloc> function, with
2193cast.
2194
596f7718 2195Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2196
49b8b560 2197=for apidoc Am|void|Safefree|void* ptr
954c1994
GS
2198The XSUB-writer's interface to the C C<free> function.
2199
596f7718 2200This should B<ONLY> be used on memory obtained using L</"Newx"> and friends.
0d7b2759 2201
954c1994
GS
2202=for apidoc Am|void|Move|void* src|void* dest|int nitems|type
2203The XSUB-writer's interface to the C C<memmove> function. The C<src> is the
926bb54c 2204source, C<dest> is the destination, C<nitems> is the number of items, and
fbe13c60 2205C<type> is the type. Can do overlapping moves. See also C<L</Copy>>.
954c1994 2206
e90e2364 2207=for apidoc Am|void *|MoveD|void* src|void* dest|int nitems|type
796b6530 2208Like C<Move> but returns C<dest>. Useful
72d33970 2209for encouraging compilers to tail-call
e90e2364
NC
2210optimise.
2211
954c1994
GS
2212=for apidoc Am|void|Copy|void* src|void* dest|int nitems|type
2213The XSUB-writer's interface to the C C<memcpy> function. The C<src> is the
926bb54c 2214source, C<dest> is the destination, C<nitems> is the number of items, and
fbe13c60 2215C<type> is the type. May fail on overlapping copies. See also C<L</Move>>.
954c1994 2216
e90e2364
NC
2217=for apidoc Am|void *|CopyD|void* src|void* dest|int nitems|type
2218
796b6530 2219Like C<Copy> but returns C<dest>. Useful
72d33970 2220for encouraging compilers to tail-call
e90e2364
NC
2221optimise.
2222
954c1994
GS
2223=for apidoc Am|void|Zero|void* dest|int nitems|type
2224
2225The XSUB-writer's interface to the C C<memzero> function. The C<dest> is the
2226destination, C<nitems> is the number of items, and C<type> is the type.
2227
e90e2364
NC
2228=for apidoc Am|void *|ZeroD|void* dest|int nitems|type
2229
72d33970
FC
2230Like C<Zero> but returns dest. Useful
2231for encouraging compilers to tail-call
e90e2364
NC
2232optimise.
2233
da5d8dbb 2234=for apidoc Am|void|StructCopy|type *src|type *dest|type
4375e838 2235This is an architecture-independent macro to copy one structure to another.
954c1994 2236
7e337ee0
JH
2237=for apidoc Am|void|PoisonWith|void* dest|int nitems|type|U8 byte
2238
2239Fill up memory with a byte pattern (a byte repeated over and over
2240again) that hopefully catches attempts to access uninitialized memory.
2241
2242=for apidoc Am|void|PoisonNew|void* dest|int nitems|type
2243
2244PoisonWith(0xAB) for catching access to allocated but uninitialized memory.
2245
1c12ffb4 2246=for apidoc Am|void|PoisonFree|void* dest|int nitems|type
7e337ee0
JH
2247
2248PoisonWith(0xEF) for catching access to freed memory.
2249
9965345d
JH
2250=for apidoc Am|void|Poison|void* dest|int nitems|type
2251
7e337ee0 2252PoisonWith(0xEF) for catching access to freed memory.
9965345d
JH
2253
2254=cut */
954c1994 2255
561b68a9
SH
2256/* Maintained for backwards-compatibility only. Use newSV() instead. */
2257#ifndef PERL_CORE
ff06c60c 2258#define NEWSV(x,len) newSV(len)
561b68a9 2259#endif
ff06c60c 2260
19a94d75
JH
2261#define MEM_SIZE_MAX ((MEM_SIZE)~0)
2262
e6bdf523 2263
27d5b266 2264#ifdef PERL_MALLOC_WRAP
e6bdf523
DM
2265
2266/* This expression will be constant-folded at compile time. It checks
2267 * whether or not the type of the count n is so small (e.g. U8 or U16, or
2268 * U32 on 64-bit systems) that there's no way a wrap-around could occur.
2269 * As well as avoiding the need for a run-time check in some cases, it's
2270 * designed to avoid compiler warnings like:
2271 * comparison is always false due to limited range of data type
73e8ff00
DM
2272 * It's mathematically equivalent to
2273 * max(n) * sizeof(t) > MEM_SIZE_MAX
e6bdf523
DM
2274 */
2275
2276# define _MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) \
004073ba 2277 (8 * sizeof(n) + sizeof(t) > sizeof(MEM_SIZE))
e6bdf523 2278
88f9f128 2279/* This is written in a slightly odd way to avoid various spurious
d98e5cde
DM
2280 * compiler warnings. We *want* to write the expression as
2281 * _MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) && (n > C)
2282 * (for some compile-time constant C), but even when the LHS
2283 * constant-folds to false at compile-time, g++ insists on emitting
2284 * warnings about the RHS (e.g. "comparison is always false"), so instead
2285 * we write it as
e6bdf523 2286 *
d98e5cde 2287 * (cond ? n : X) > C
88f9f128 2288 *
d98e5cde
DM
2289 * where X is a constant with X > C always false. Choosing a value for X
2290 * is tricky. If 0, some compilers will complain about 0 > C always being
2291 * false; if 1, Coverity complains when n happens to be the constant value
2292 * '1', that cond ? 1 : 1 has the same value on both branches; so use C
2293 * for X and hope that nothing else whines.
e6bdf523
DM
2294 */
2295
2296# define _MEM_WRAP_WILL_WRAP(n,t) \
88f9f128
DM
2297 ((_MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) ? (MEM_SIZE)(n) : \
2298 MEM_SIZE_MAX/sizeof(t)) > MEM_SIZE_MAX/sizeof(t))
e6bdf523
DM
2299
2300# define MEM_WRAP_CHECK(n,t) \
2301 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2302 && (croak_memory_wrap(),0))
2303
2304# define MEM_WRAP_CHECK_1(n,t,a) \
2305 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2306 && (Perl_croak_nocontext("%s",(a)),0))
2307
8b44ba4c 2308#define MEM_WRAP_CHECK_(n,t) MEM_WRAP_CHECK(n,t),
27d5b266 2309
d1decf2b 2310#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
2311#else
2312
410319be
NC
2313#define MEM_WRAP_CHECK(n,t)
2314#define MEM_WRAP_CHECK_1(n,t,a)
2315#define MEM_WRAP_CHECK_2(n,t,a,b)
8b44ba4c
NC
2316#define MEM_WRAP_CHECK_(n,t)
2317
6fff79ce 2318#define PERL_STRLEN_ROUNDUP(n) (((n-1+PERL_STRLEN_ROUNDUP_QUANTUM)&~((MEM_SIZE)PERL_STRLEN_ROUNDUP_QUANTUM-1)))
27d5b266 2319
1936d2a7 2320#endif
8b44ba4c 2321
fe4f188c 2322#ifdef PERL_MEM_LOG
46c6c7e2 2323/*
9f653bb5 2324 * If PERL_MEM_LOG is defined, all Newx()s, Renew()s, and Safefree()s
46c6c7e2
JH
2325 * go through functions, which are handy for debugging breakpoints, but
2326 * which more importantly get the immediate calling environment (file and
e352bcff
JH
2327 * line number, and C function name if available) passed in. This info can
2328 * then be used for logging the calls, for which one gets a sample
73d1d973 2329 * implementation unless -DPERL_MEM_LOG_NOIMPL is also defined.
3609ea0d 2330 *
46c6c7e2 2331 * Known problems:
94e892a6 2332 * - not all memory allocs get logged, only those
46c6c7e2 2333 * that go through Newx() and derivatives (while all
94e892a6 2334 * Safefrees do get logged)
46c6c7e2
JH
2335 * - __FILE__ and __LINE__ do not work everywhere
2336 * - __func__ or __FUNCTION__ even less so
2337 * - I think more goes on after the perlio frees but
2338 * the thing is that STDERR gets closed (as do all
2339 * the file descriptors)
2340 * - no deeper calling stack than the caller of the Newx()
2341 * or the kind, but do I look like a C reflection/introspection
2342 * utility to you?
2343 * - the function prototypes for the logging functions
2344 * probably should maybe be somewhere else than handy.h
2345 * - one could consider inlining (macrofying) the logging
2346 * for speed, but I am too lazy
2347 * - one could imagine recording the allocations in a hash,
2348 * (keyed by the allocation address?), and maintain that
2349 * through reallocs and frees, but how to do that without
2350 * any News() happening...?
73d1d973 2351 * - lots of -Ddefines to get useful/controllable output
b953482e 2352 * - lots of ENV reads
46c6c7e2
JH
2353 */
2354
0b0ab801 2355# ifdef PERL_CORE
73d1d973 2356# ifndef PERL_MEM_LOG_NOIMPL
0b0ab801
MHM
2357enum mem_log_type {
2358 MLT_ALLOC,
2359 MLT_REALLOC,
d7a2c63c
MHM
2360 MLT_FREE,
2361 MLT_NEW_SV,
2362 MLT_DEL_SV
0b0ab801
MHM
2363};
2364# endif
12754f92 2365# if defined(PERL_IN_SV_C) /* those are only used in sv.c */
d7a2c63c
MHM
2366void Perl_mem_log_new_sv(const SV *sv, const char *filename, const int linenumber, const char *funcname);
2367void Perl_mem_log_del_sv(const SV *sv, const char *filename, const int linenumber, const char *funcname);
12754f92 2368# endif
0b0ab801
MHM
2369# endif
2370
fe4f188c
JH
2371#endif
2372
2373#ifdef PERL_MEM_LOG
d1401ee9
MHM
2374#define MEM_LOG_ALLOC(n,t,a) Perl_mem_log_alloc(n,sizeof(t),STRINGIFY(t),a,__FILE__,__LINE__,FUNCTION__)
2375#define MEM_LOG_REALLOC(n,t,v,a) Perl_mem_log_realloc(n,sizeof(t),STRINGIFY(t),v,a,__FILE__,__LINE__,FUNCTION__)
46c6c7e2 2376#define MEM_LOG_FREE(a) Perl_mem_log_free(a,__FILE__,__LINE__,FUNCTION__)
fe4f188c
JH
2377#endif
2378
2379#ifndef MEM_LOG_ALLOC
2380#define MEM_LOG_ALLOC(n,t,a) (a)
2381#endif
2382#ifndef MEM_LOG_REALLOC
2383#define MEM_LOG_REALLOC(n,t,v,a) (a)
2384#endif
2385#ifndef MEM_LOG_FREE
2386#define MEM_LOG_FREE(a) (a)
2387#endif
2388
d1401ee9
MHM
2389#define Newx(v,n,t) (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_ALLOC(n,t,safemalloc((MEM_SIZE)((n)*sizeof(t))))))
2390#define Newxc(v,n,t,c) (v = (MEM_WRAP_CHECK_(n,t) (c*)MEM_LOG_ALLOC(n,t,safemalloc((MEM_SIZE)((n)*sizeof(t))))))
2391#define Newxz(v,n,t) (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_ALLOC(n,t,safecalloc((n),sizeof(t)))))
a6f6820f
NC
2392
2393#ifndef PERL_CORE
a02a5408
JC
2394/* pre 5.9.x compatibility */
2395#define New(x,v,n,t) Newx(v,n,t)
2396#define Newc(x,v,n,t,c) Newxc(v,n,t,c)
4541904d 2397#define Newz(x,v,n,t) Newxz(v,n,t)
a6f6820f 2398#endif
a02a5408 2399
ff68c719 2400#define Renew(v,n,t) \
d1401ee9 2401 (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_REALLOC(n,t,v,saferealloc((Malloc_t)(v),(MEM_SIZE)((n)*sizeof(t))))))
ff68c719 2402#define Renewc(v,n,t,c) \
d1401ee9 2403 (v = (MEM_WRAP_CHECK_(n,t) (c*)MEM_LOG_REALLOC(n,t,v,saferealloc((Malloc_t)(v),(MEM_SIZE)((n)*sizeof(t))))))
94010e71
NC
2404
2405#ifdef PERL_POISON
2406#define Safefree(d) \
06c0cc96 2407 ((d) ? (void)(safefree(MEM_LOG_FREE((Malloc_t)(d))), Poison(&(d), 1, Malloc_t)) : (void) 0)
94010e71 2408#else
fe4f188c 2409#define Safefree(d) safefree(MEM_LOG_FREE((Malloc_t)(d)))
94010e71 2410#endif
55497cff 2411
f14cf363
TC
2412#define Move(s,d,n,t) (MEM_WRAP_CHECK_(n,t) assert(d), assert(s), (void)memmove((char*)(d),(const char*)(s), (n) * sizeof(t)))
2413#define Copy(s,d,n,t) (MEM_WRAP_CHECK_(n,t) assert(d), assert(s), (void)memcpy((char*)(d),(const char*)(s), (n) * sizeof(t)))
2414#define Zero(d,n,t) (MEM_WRAP_CHECK_(n,t) assert(d), (void)memzero((char*)(d), (n) * sizeof(t)))
55497cff 2415
f14cf363
TC
2416#define MoveD(s,d,n,t) (MEM_WRAP_CHECK_(n,t) assert(d), assert(s), memmove((char*)(d),(const char*)(s), (n) * sizeof(t)))
2417#define CopyD(s,d,n,t) (MEM_WRAP_CHECK_(n,t) assert(d), assert(s), memcpy((char*)(d),(const char*)(s), (n) * sizeof(t)))
e90e2364 2418#ifdef HAS_MEMSET
f14cf363 2419#define ZeroD(d,n,t) (MEM_WRAP_CHECK_(n,t) assert(d), memzero((char*)(d), (n) * sizeof(t)))
e90e2364 2420#else
8b44ba4c 2421/* Using bzero(), which returns void. */
f14cf363 2422#define ZeroD(d,n,t) (MEM_WRAP_CHECK_(n,t) assert(d), memzero((char*)(d), (n) * sizeof(t)),d)
e90e2364
NC
2423#endif
2424
7e337ee0
JH
2425#define PoisonWith(d,n,t,b) (MEM_WRAP_CHECK_(n,t) (void)memset((char*)(d), (U8)(b), (n) * sizeof(t)))
2426#define PoisonNew(d,n,t) PoisonWith(d,n,t,0xAB)
2427#define PoisonFree(d,n,t) PoisonWith(d,n,t,0xEF)
2428#define Poison(d,n,t) PoisonFree(d,n,t)
27d5b266 2429
caa674f3
DD
2430#ifdef PERL_POISON
2431# define PERL_POISON_EXPR(x) x
2432#else
2433# define PERL_POISON_EXPR(x)
2434#endif
2435
2304df62 2436#ifdef USE_STRUCT_COPY
ff68c719 2437#define StructCopy(s,d,t) (*((t*)(d)) = *((t*)(s)))
bee1dbe2
LW
2438#else
2439#define StructCopy(s,d,t) Copy(s,d,1,t)
2440#endif
2cc61e15 2441
c3caa5c3
JH
2442/* C_ARRAY_LENGTH is the number of elements in the C array (so you
2443 * want your zero-based indices to be less than but not equal to).
2444 *
2445 * C_ARRAY_END is one past the last: half-open/half-closed range,
2446 * not last-inclusive range. */
622913ab 2447#define C_ARRAY_LENGTH(a) (sizeof(a)/sizeof((a)[0]))
c3caa5c3 2448#define C_ARRAY_END(a) ((a) + C_ARRAY_LENGTH(a))
622913ab 2449
2cc61e15
DD
2450#ifdef NEED_VA_COPY
2451# ifdef va_copy
2452# define Perl_va_copy(s, d) va_copy(d, s)
2cc61e15 2453# else
a1866d1b
JH
2454# if defined(__va_copy)
2455# define Perl_va_copy(s, d) __va_copy(d, s)
2456# else
2457# define Perl_va_copy(s, d) Copy(s, d, 1, va_list)
2458# endif
2cc61e15
DD
2459# endif
2460#endif
2461
472d47bc
SB
2462/* convenience debug macros */
2463#ifdef USE_ITHREADS
2464#define pTHX_FORMAT "Perl interpreter: 0x%p"
2465#define pTHX__FORMAT ", Perl interpreter: 0x%p"
f54cb97a
AL
2466#define pTHX_VALUE_ (void *)my_perl,
2467#define pTHX_VALUE (void *)my_perl
2468#define pTHX__VALUE_ ,(void *)my_perl,
2469#define pTHX__VALUE ,(void *)my_perl
472d47bc 2470#else
3609ea0d 2471#define pTHX_FORMAT
472d47bc 2472#define pTHX__FORMAT
3609ea0d 2473#define pTHX_VALUE_
472d47bc 2474#define pTHX_VALUE
3609ea0d 2475#define pTHX__VALUE_
472d47bc
SB
2476#define pTHX__VALUE
2477#endif /* USE_ITHREADS */
3609ea0d 2478
2acdbac1
NC
2479/* Perl_deprecate was not part of the public API, and did not have a deprecate()
2480 shortcut macro defined without -DPERL_CORE. Neither codesearch.google.com nor
2481 CPAN::Unpack show any users outside the core. */
2482#ifdef PERL_CORE
dc6e8de0
A
2483# define deprecate(s) Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2484 "Use of " s " is deprecated")
c9680906
A
2485# define deprecate_disappears_in(when,message) \
2486 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2487 message ", and will disappear in Perl " when)
ac641426
A
2488# define deprecate_fatal_in(when,message) \
2489 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2490 message ". Its use will be fatal in Perl " when)
2acdbac1
NC
2491#endif
2492
dfff4baf
BF
2493/* Internal macros to deal with gids and uids */
2494#ifdef PERL_CORE
2495
2496# if Uid_t_size > IVSIZE
2497# define sv_setuid(sv, uid) sv_setnv((sv), (NV)(uid))
2498# define SvUID(sv) SvNV(sv)
2499# else
2500# if Uid_t_sign <= 0
2501# define sv_setuid(sv, uid) sv_setiv((sv), (IV)(uid))
2502# define SvUID(sv) SvIV(sv)
2503# else
2504# define sv_setuid(sv, uid) sv_setuv((sv), (UV)(uid))
2505# define SvUID(sv) SvUV(sv)
2506# endif
2507# endif /* Uid_t_size */
2508
2509# if Gid_t_size > IVSIZE
2510# define sv_setgid(sv, gid) sv_setnv((sv), (NV)(gid))
2511# define SvGID(sv) SvNV(sv)
2512# else
2513# if Gid_t_sign <= 0
2514# define sv_setgid(sv, gid) sv_setiv((sv), (IV)(gid))
2515# define SvGID(sv) SvIV(sv)
2516# else
2517# define sv_setgid(sv, gid) sv_setuv((sv), (UV)(gid))
2518# define SvGID(sv) SvUV(sv)
2519# endif
2520# endif /* Gid_t_size */
2521
2522#endif
2523
6a5bc5ac 2524#endif /* PERL_HANDY_H_ */
9d745869 2525
e9a8c099 2526/*
14d04a33 2527 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 2528 */