This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
perlapi: Document cBOOL
[perl5.git] / handy.h
CommitLineData
a0d0e21e 1/* handy.h
a687059c 2 *
1129b882 3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1999, 2000,
da5d8dbb 4 * 2001, 2002, 2004, 2005, 2006, 2007, 2008, 2012 by Larry Wall and others
a687059c 5 *
6e21c824
LW
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
8d063cd8 8 *
8d063cd8
LW
9 */
10
4650c663
KW
11/* IMPORTANT NOTE: Everything whose name begins with an underscore is for
12 * internal core Perl use only. */
13
6a5bc5ac
KW
14#ifndef PERL_HANDY_H_ /* Guard against nested #inclusion */
15#define PERL_HANDY_H_
9d745869 16
24792b8d
NC
17#ifndef PERL_CORE
18# define Null(type) ((type)NULL)
954c1994
GS
19
20/*
ccfc67b7 21=head1 Handy Values
954c1994 22
78342678 23=for apidoc AmnU||Nullch
72d33970
FC
24Null character pointer. (No longer available when C<PERL_CORE> is
25defined.)
2307c6d0 26
78342678 27=for apidoc AmnU||Nullsv
72d33970 28Null SV pointer. (No longer available when C<PERL_CORE> is defined.)
954c1994
GS
29
30=cut
31*/
32
24792b8d
NC
33# define Nullch Null(char*)
34# define Nullfp Null(PerlIO*)
35# define Nullsv Null(SV*)
36#endif
8d063cd8 37
641d3f0b
PP
38#ifdef TRUE
39#undef TRUE
40#endif
41#ifdef FALSE
42#undef FALSE
43#endif
44#define TRUE (1)
45#define FALSE (0)
46
cf3f0ffb
DM
47/* The MUTABLE_*() macros cast pointers to the types shown, in such a way
48 * (compiler permitting) that casting away const-ness will give a warning;
49 * e.g.:
50 *
51 * const SV *sv = ...;
52 * AV *av1 = (AV*)sv; <== BAD: the const has been silently cast away
53 * AV *av2 = MUTABLE_AV(sv); <== GOOD: it may warn
54 */
55
b1bc3f34 56#if defined(__GNUC__) && !defined(PERL_GCC_BRACE_GROUPS_FORBIDDEN)
f065df04 57# define MUTABLE_PTR(p) ({ void *_p = (p); _p; })
b1bc3f34
NC
58#else
59# define MUTABLE_PTR(p) ((void *) (p))
60#endif
61
a062e10d 62#define MUTABLE_AV(p) ((AV *)MUTABLE_PTR(p))
ea726b52 63#define MUTABLE_CV(p) ((CV *)MUTABLE_PTR(p))
159b6efe 64#define MUTABLE_GV(p) ((GV *)MUTABLE_PTR(p))
dbebbdb4 65#define MUTABLE_HV(p) ((HV *)MUTABLE_PTR(p))
a45c7426 66#define MUTABLE_IO(p) ((IO *)MUTABLE_PTR(p))
b1bc3f34 67#define MUTABLE_SV(p) ((SV *)MUTABLE_PTR(p))
27d4fb96 68
f789f6a4 69#if defined(I_STDBOOL) && !defined(PERL_BOOL_AS_CHAR)
bd31be4b
NC
70# include <stdbool.h>
71# ifndef HAS_BOOL
72# define HAS_BOOL 1
73# endif
74#endif
75
8e84507e 76/* bool is built-in for g++-2.6.3 and later, which might be used
c1d22f6b
GS
77 for extensions. <_G_config.h> defines _G_HAVE_BOOL, but we can't
78 be sure _G_config.h will be included before this file. _G_config.h
8e84507e 79 also defines _G_HAVE_BOOL for both gcc and g++, but only g++
c1d22f6b
GS
80 actually has bool. Hence, _G_HAVE_BOOL is pretty useless for us.
81 g++ can be identified by __GNUG__.
82 Andy Dougherty February 2000
5d94fbed 83*/
3609ea0d 84#ifdef __GNUG__ /* GNU g++ has bool built-in */
f789f6a4 85# ifndef PERL_BOOL_AS_CHAR
5d94fbed 86# ifndef HAS_BOOL
c1d22f6b 87# define HAS_BOOL 1
5d94fbed 88# endif
f789f6a4 89# endif
5d94fbed
AD
90#endif
91
92#ifndef HAS_BOOL
f789f6a4
FC
93# ifdef bool
94# undef bool
95# endif
70d5cb32 96# define bool char
c1d22f6b 97# define HAS_BOOL 1
a687059c 98#endif
0d3e774c 99
25ba28ce
KW
100/*
101=for apidoc Am|bool|cBOOL|bool expr
102
103Cast-to-bool. A simple S<C<(bool) I<expr>>> cast may not do the right thing:
104if C<bool> is defined as C<char>, for example, then the cast from C<int> is
105implementation-defined.
106
107C<(bool)!!(cbool)> in a ternary triggers a bug in xlc on AIX
108
109=cut
110*/
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__
07798b17
AC
119#elif (defined(USING_MSVC6)) || /* MSVC6 has neither __func__ nor __FUNCTION and no good workarounds, either. */ \
120 (defined(__DECC_VER)) /* Tru64 or VMS, and strict C89 being used, but not modern enough cc (in Tur64, -c99 not known, only -std1). */
121# define FUNCTION__ ""
46c6c7e2 122#else
07798b17 123# define FUNCTION__ __FUNCTION__ /* Common extension. */
46c6c7e2
JH
124#endif
125
27d4fb96
PP
126/* XXX A note on the perl source internal type system. The
127 original intent was that I32 be *exactly* 32 bits.
128
129 Currently, we only guarantee that I32 is *at least* 32 bits.
130 Specifically, if int is 64 bits, then so is I32. (This is the case
131 for the Cray.) This has the advantage of meshing nicely with
132 standard library calls (where we pass an I32 and the library is
133 expecting an int), but the disadvantage that an I32 is not 32 bits.
134 Andy Dougherty August 1996
24fef2a7 135
dc45a647
MB
136 There is no guarantee that there is *any* integral type with
137 exactly 32 bits. It is perfectly legal for a system to have
138 sizeof(short) == sizeof(int) == sizeof(long) == 8.
693762b4 139
dc45a647
MB
140 Similarly, there is no guarantee that I16 and U16 have exactly 16
141 bits.
693762b4 142
8e84507e
NIS
143 For dealing with issues that may arise from various 32/64-bit
144 systems, we will ask Configure to check out
8175356b 145
3609ea0d
JH
146 SHORTSIZE == sizeof(short)
147 INTSIZE == sizeof(int)
148 LONGSIZE == sizeof(long)
dc45a647 149 LONGLONGSIZE == sizeof(long long) (if HAS_LONG_LONG)
3609ea0d 150 PTRSIZE == sizeof(void *)
dc45a647
MB
151 DOUBLESIZE == sizeof(double)
152 LONG_DOUBLESIZE == sizeof(long double) (if HAS_LONG_DOUBLE).
8175356b 153
27d4fb96
PP
154*/
155
69512466
JH
156#ifdef I_INTTYPES /* e.g. Linux has int64_t without <inttypes.h> */
157# include <inttypes.h>
dd0eed91
JH
158# ifdef INT32_MIN_BROKEN
159# undef INT32_MIN
160# define INT32_MIN (-2147483647-1)
161# endif
162# ifdef INT64_MIN_BROKEN
163# undef INT64_MIN
164# define INT64_MIN (-9223372036854775807LL-1)
165# endif
69512466
JH
166#endif
167
8175356b
JH
168typedef I8TYPE I8;
169typedef U8TYPE U8;
170typedef I16TYPE I16;
171typedef U16TYPE U16;
172typedef I32TYPE I32;
173typedef U32TYPE U32;
16d89be8 174
74b807c7 175#ifdef QUADKIND
8175356b
JH
176typedef I64TYPE I64;
177typedef U64TYPE U64;
16d89be8 178#endif
8175356b 179
d8668976 180#if defined(UINT8_MAX) && defined(INT16_MAX) && defined(INT32_MAX)
5ff3f7a4 181
5ff3f7a4
GS
182/* I8_MAX and I8_MIN constants are not defined, as I8 is an ambiguous type.
183 Please search CHAR_MAX in perl.h for further details. */
184#define U8_MAX UINT8_MAX
185#define U8_MIN UINT8_MIN
186
5ff3f7a4
GS
187#define I16_MAX INT16_MAX
188#define I16_MIN INT16_MIN
189#define U16_MAX UINT16_MAX
190#define U16_MIN UINT16_MIN
191
5ff3f7a4
GS
192#define I32_MAX INT32_MAX
193#define I32_MIN INT32_MIN
0e983133
GS
194#ifndef UINT32_MAX_BROKEN /* e.g. HP-UX with gcc messes this up */
195# define U32_MAX UINT32_MAX
196#else
197# define U32_MAX 4294967295U
198#endif
5ff3f7a4
GS
199#define U32_MIN UINT32_MIN
200
201#else
202
5c9fa16e
KA
203/* I8_MAX and I8_MIN constants are not defined, as I8 is an ambiguous type.
204 Please search CHAR_MAX in perl.h for further details. */
27d4fb96
PP
205#define U8_MAX PERL_UCHAR_MAX
206#define U8_MIN PERL_UCHAR_MIN
79072805 207
27d4fb96
PP
208#define I16_MAX PERL_SHORT_MAX
209#define I16_MIN PERL_SHORT_MIN
210#define U16_MAX PERL_USHORT_MAX
211#define U16_MIN PERL_USHORT_MIN
79072805 212
c4f23d77 213#if LONGSIZE > 4
27d4fb96
PP
214# define I32_MAX PERL_INT_MAX
215# define I32_MIN PERL_INT_MIN
216# define U32_MAX PERL_UINT_MAX
217# define U32_MIN PERL_UINT_MIN
79072805 218#else
27d4fb96
PP
219# define I32_MAX PERL_LONG_MAX
220# define I32_MIN PERL_LONG_MIN
221# define U32_MAX PERL_ULONG_MAX
222# define U32_MIN PERL_ULONG_MIN
79072805
LW
223#endif
224
5ff3f7a4
GS
225#endif
226
247cee9f
KW
227/* These C99 typedefs are useful sometimes for, say, loop variables whose
228 * maximum values are small, but for which speed trumps size. If we have a C99
229 * compiler, use that. Otherwise, a plain 'int' should be good enough.
230 *
231 * Restrict these to core for now until we are more certain this is a good
232 * idea. */
233#if defined(PERL_CORE) || defined(PERL_EXT)
234# ifdef I_STDINT
235 typedef int_fast8_t PERL_INT_FAST8_T;
236 typedef uint_fast8_t PERL_UINT_FAST8_T;
237 typedef int_fast16_t PERL_INT_FAST16_T;
238 typedef uint_fast16_t PERL_UINT_FAST16_T;
239# else
240 typedef int PERL_INT_FAST8_T;
241 typedef unsigned int PERL_UINT_FAST8_T;
242 typedef int PERL_INT_FAST16_T;
243 typedef unsigned int PERL_UINT_FAST16_T;
244# endif
245#endif
246
464decb6
KW
247/* log(2) (i.e., log base 10 of 2) is pretty close to 0.30103, just in case
248 * anyone is grepping for it */
249#define BIT_DIGITS(N) (((N)*146)/485 + 1) /* log10(2) =~ 146/485 */
fc36a67e
PP
250#define TYPE_DIGITS(T) BIT_DIGITS(sizeof(T) * 8)
251#define TYPE_CHARS(T) (TYPE_DIGITS(T) + 2) /* sign, NUL */
252
88794300 253/* Unused by core; should be deprecated */
ff68c719 254#define Ctl(ch) ((ch) & 037)
8d063cd8 255
98fce2a4
KW
256#if defined(PERL_CORE) || defined(PERL_EXT)
257# ifndef MIN
258# define MIN(a,b) ((a) < (b) ? (a) : (b))
259# endif
260# ifndef MAX
261# define MAX(a,b) ((a) > (b) ? (a) : (b))
262# endif
263#endif
264
84ff4fa9
KW
265/* Returns a boolean as to whether the input unsigned number is a power of 2
266 * (2**0, 2**1, etc). In other words if it has just a single bit set.
267 * If not, subtracting 1 would leave the uppermost bit set, so the & would
268 * yield non-zero */
269#if defined(PERL_CORE) || defined(PERL_EXT)
011b1419 270# define isPOWER_OF_2(n) ((n) && ((n) & ((n)-1)) == 0)
84ff4fa9
KW
271#endif
272
4650c663
KW
273/* This is a helper macro to avoid preprocessor issues, replaced by nothing
274 * unless under DEBUGGING, where it expands to an assert of its argument,
275 * followed by a comma (hence the comma operator). If we just used a straight
3e94db23
JH
276 * assert(), we would get a comma with nothing before it when not DEBUGGING.
277 *
278 * We also use empty definition under Coverity since the __ASSERT__
279 * checks often check for things that Really Cannot Happen, and Coverity
280 * detects that and gets all excited. */
281
282#if defined(DEBUGGING) && !defined(__COVERITY__)
0f092d08
KW
283# define __ASSERT_(statement) assert(statement),
284#else
285# define __ASSERT_(statement)
286#endif
287
3fe05580 288/*
a4ee4fb5 289=head1 SV Manipulation Functions
3fe05580 290
1568d13a
Z
291=for apidoc Ama|SV*|newSVpvs|"literal string" s
292Like C<newSVpvn>, but takes a literal string instead of a
30a15352 293string/length pair.
3fe05580 294
1568d13a
Z
295=for apidoc Ama|SV*|newSVpvs_flags|"literal string" s|U32 flags
296Like C<newSVpvn_flags>, but takes a literal string instead of
30a15352 297a string/length pair.
84bafc02 298
1568d13a
Z
299=for apidoc Ama|SV*|newSVpvs_share|"literal string" s
300Like C<newSVpvn_share>, but takes a literal string instead of
30a15352 301a string/length pair and omits the hash parameter.
3fe05580 302
1568d13a
Z
303=for apidoc Am|void|sv_catpvs_flags|SV* sv|"literal string" s|I32 flags
304Like C<sv_catpvn_flags>, but takes a literal string instead
30a15352 305of a string/length pair.
9dcc53ea 306
1568d13a
Z
307=for apidoc Am|void|sv_catpvs_nomg|SV* sv|"literal string" s
308Like C<sv_catpvn_nomg>, but takes a literal string instead of
0c395ea5 309a string/length pair.
9dcc53ea 310
1568d13a
Z
311=for apidoc Am|void|sv_catpvs|SV* sv|"literal string" s
312Like C<sv_catpvn>, but takes a literal string instead of a
0c395ea5 313string/length pair.
3fe05580 314
1568d13a
Z
315=for apidoc Am|void|sv_catpvs_mg|SV* sv|"literal string" s
316Like C<sv_catpvn_mg>, but takes a literal string instead of a
9dcc53ea
Z
317string/length pair.
318
1568d13a
Z
319=for apidoc Am|void|sv_setpvs|SV* sv|"literal string" s
320Like C<sv_setpvn>, but takes a literal string instead of a
0c395ea5 321string/length pair.
3fe05580 322
1568d13a
Z
323=for apidoc Am|void|sv_setpvs_mg|SV* sv|"literal string" s
324Like C<sv_setpvn_mg>, but takes a literal string instead of a
9dcc53ea
Z
325string/length pair.
326
78342678 327=for apidoc Am|SV *|sv_setref_pvs|SV *const rv|const char *const classname|"literal string" s
1568d13a 328Like C<sv_setref_pvn>, but takes a literal string instead of
0c395ea5 329a string/length pair.
9dcc53ea 330
3fe05580
MHM
331=head1 Memory Management
332
1568d13a
Z
333=for apidoc Ama|char*|savepvs|"literal string" s
334Like C<savepvn>, but takes a literal string instead of a
30a15352 335string/length pair.
3fe05580 336
1568d13a 337=for apidoc Ama|char*|savesharedpvs|"literal string" s
9dcc53ea
Z
338A version of C<savepvs()> which allocates the duplicate string in memory
339which is shared between threads.
340
3fe05580
MHM
341=head1 GV Functions
342
1568d13a
Z
343=for apidoc Am|HV*|gv_stashpvs|"literal string" name|I32 create
344Like C<gv_stashpvn>, but takes a literal string instead of a
0c395ea5 345string/length pair.
3fe05580
MHM
346
347=head1 Hash Manipulation Functions
348
1568d13a
Z
349=for apidoc Am|SV**|hv_fetchs|HV* tb|"literal string" key|I32 lval
350Like C<hv_fetch>, but takes a literal string instead of a
0c395ea5 351string/length pair.
3fe05580 352
d106cb46 353=for apidoc Am|SV**|hv_stores|HV* tb|"literal string" key|SV* val
1568d13a 354Like C<hv_store>, but takes a literal string instead of a
0c395ea5 355string/length pair
3fe05580
MHM
356and omits the hash parameter.
357
510966aa
Z
358=head1 Lexer interface
359
1568d13a 360=for apidoc Amx|void|lex_stuff_pvs|"literal string" pv|U32 flags
510966aa 361
1568d13a 362Like L</lex_stuff_pvn>, but takes a literal string instead of
0c395ea5 363a string/length pair.
510966aa 364
3fe05580
MHM
365=cut
366*/
367
926bb54c
KW
368/* concatenating with "" ensures that only literal strings are accepted as
369 * argument */
26bfd924 370#define STR_WITH_LEN(s) ("" s ""), (sizeof(s)-1)
2efa8cc7 371
926bb54c
KW
372/* note that STR_WITH_LEN() can't be used as argument to macros or functions
373 * that under some configurations might be macros, which means that it requires
374 * the full Perl_xxx(aTHX_ ...) form for any API calls where it's used.
ba3a79e7
GA
375 */
376
377/* STR_WITH_LEN() shortcuts */
378#define newSVpvs(str) Perl_newSVpvn(aTHX_ STR_WITH_LEN(str))
84bafc02
NC
379#define newSVpvs_flags(str,flags) \
380 Perl_newSVpvn_flags(aTHX_ STR_WITH_LEN(str), flags)
ba3a79e7 381#define newSVpvs_share(str) Perl_newSVpvn_share(aTHX_ STR_WITH_LEN(str), 0)
9dcc53ea
Z
382#define sv_catpvs_flags(sv, str, flags) \
383 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), flags)
384#define sv_catpvs_nomg(sv, str) \
385 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), 0)
386#define sv_catpvs(sv, str) \
387 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), SV_GMAGIC)
388#define sv_catpvs_mg(sv, str) \
389 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), SV_GMAGIC|SV_SMAGIC)
3fe05580 390#define sv_setpvs(sv, str) Perl_sv_setpvn(aTHX_ sv, STR_WITH_LEN(str))
9dcc53ea
Z
391#define sv_setpvs_mg(sv, str) Perl_sv_setpvn_mg(aTHX_ sv, STR_WITH_LEN(str))
392#define sv_setref_pvs(rv, classname, str) \
393 Perl_sv_setref_pvn(aTHX_ rv, classname, STR_WITH_LEN(str))
ba3a79e7 394#define savepvs(str) Perl_savepvn(aTHX_ STR_WITH_LEN(str))
9dcc53ea
Z
395#define savesharedpvs(str) Perl_savesharedpvn(aTHX_ STR_WITH_LEN(str))
396#define gv_stashpvs(str, create) \
397 Perl_gv_stashpvn(aTHX_ STR_WITH_LEN(str), create)
398#define gv_fetchpvs(namebeg, add, sv_type) \
399 Perl_gv_fetchpvn_flags(aTHX_ STR_WITH_LEN(namebeg), add, sv_type)
400#define gv_fetchpvn(namebeg, len, add, sv_type) \
401 Perl_gv_fetchpvn_flags(aTHX_ namebeg, len, add, sv_type)
402#define sv_catxmlpvs(dsv, str, utf8) \
403 Perl_sv_catxmlpvn(aTHX_ dsv, STR_WITH_LEN(str), utf8)
4ac46235 404
ba3a79e7 405
510966aa
Z
406#define lex_stuff_pvs(pv,flags) Perl_lex_stuff_pvn(aTHX_ STR_WITH_LEN(pv), flags)
407
b96d8cd9
NC
408#define get_cvs(str, flags) \
409 Perl_get_cvn_flags(aTHX_ STR_WITH_LEN(str), (flags))
5c1737d1 410
954c1994 411/*
ccfc67b7
JH
412=head1 Miscellaneous Functions
413
954c1994 414=for apidoc Am|bool|strNE|char* s1|char* s2
dc6b0978
KW
415Test two C<NUL>-terminated strings to see if they are different. Returns true
416or false.
954c1994
GS
417
418=for apidoc Am|bool|strEQ|char* s1|char* s2
dc6b0978
KW
419Test two C<NUL>-terminated strings to see if they are equal. Returns true or
420false.
954c1994
GS
421
422=for apidoc Am|bool|strLT|char* s1|char* s2
dc6b0978
KW
423Test two C<NUL>-terminated strings to see if the first, C<s1>, is less than the
424second, C<s2>. Returns true or false.
954c1994
GS
425
426=for apidoc Am|bool|strLE|char* s1|char* s2
dc6b0978
KW
427Test two C<NUL>-terminated strings to see if the first, C<s1>, is less than or
428equal to the second, C<s2>. Returns true or false.
954c1994
GS
429
430=for apidoc Am|bool|strGT|char* s1|char* s2
dc6b0978
KW
431Test two C<NUL>-terminated strings to see if the first, C<s1>, is greater than
432the second, C<s2>. Returns true or false.
954c1994
GS
433
434=for apidoc Am|bool|strGE|char* s1|char* s2
dc6b0978
KW
435Test two C<NUL>-terminated strings to see if the first, C<s1>, is greater than
436or equal to the second, C<s2>. Returns true or false.
954c1994
GS
437
438=for apidoc Am|bool|strnNE|char* s1|char* s2|STRLEN len
dc6b0978
KW
439Test two C<NUL>-terminated strings to see if they are different. The C<len>
440parameter indicates the number of bytes to compare. Returns true or false. (A
954c1994
GS
441wrapper for C<strncmp>).
442
443=for apidoc Am|bool|strnEQ|char* s1|char* s2|STRLEN len
dc6b0978
KW
444Test two C<NUL>-terminated strings to see if they are equal. The C<len>
445parameter indicates the number of bytes to compare. Returns true or false. (A
446wrapper for C<strncmp>).
954c1994 447
bd18bd40
KW
448=for apidoc Am|bool|memEQ|char* s1|char* s2|STRLEN len
449Test two buffers (which may contain embedded C<NUL> characters, to see if they
450are equal. The C<len> parameter indicates the number of bytes to compare.
451Returns zero if equal, or non-zero if non-equal.
452
453=for apidoc Am|bool|memNE|char* s1|char* s2|STRLEN len
454Test two buffers (which may contain embedded C<NUL> characters, to see if they
455are not equal. The C<len> parameter indicates the number of bytes to compare.
456Returns zero if non-equal, or non-zero if equal.
457
954c1994 458=cut
fc169e00
KW
459
460New macros should use the following conventions for their names (which are
461based on the underlying C library functions):
462
463 (mem | str n? ) (EQ | NE | LT | GT | GE | (( BEGIN | END ) P? )) l? s?
464
465 Each has two main parameters, string-like operands that are compared
466 against each other, as specified by the macro name. Some macros may
467 additionally have one or potentially even two length parameters. If a length
468 parameter applies to both string parameters, it will be positioned third;
469 otherwise any length parameter immediately follows the string parameter it
470 applies to.
471
472 If the prefix to the name is 'str', the string parameter is a pointer to a C
473 language string. Such a string does not contain embedded NUL bytes; its
474 length may be unknown, but can be calculated by C<strlen()>, since it is
475 terminated by a NUL, which isn't included in its length.
476
477 The optional 'n' following 'str' means that that there is a third parameter,
478 giving the maximum number of bytes to look at in each string. Even if both
479 strings are longer than the length parameter, those extra bytes will be
480 unexamined.
481
482 The 's' suffix means that the 2nd byte string parameter is a literal C
483 double-quoted string. Its length will automatically be calculated by the
484 macro, so no length parameter will ever be needed for it.
485
486 If the prefix is 'mem', the string parameters don't have to be C strings;
487 they may contain embedded NUL bytes, do not necessarily have a terminating
488 NUL, and their lengths can be known only through other means, which in
489 practice are additional parameter(s) passed to the function. All 'mem'
490 functions have at least one length parameter. Barring any 'l' or 's' suffix,
491 there is a single length parameter, in position 3, which applies to both
492 string parameters. The 's' suffix means, as described above, that the 2nd
493 string is a literal double-quoted C string (hence its length is calculated by
494 the macro, and the length parameter to the function applies just to the first
495 string parameter, and hence is positioned just after it). An 'l' suffix
496 means that the 2nd string parameter has its own length parameter, and the
497 signature will look like memFOOl(s1, l1, s2, l2).
498
499 BEGIN (and END) are for testing if the 2nd string is an initial (or final)
500 substring of the 1st string. 'P' if present indicates that the substring
501 must be a "proper" one in tha mathematical sense that the first one must be
502 strictly larger than the 2nd.
503
954c1994
GS
504*/
505
62946e08 506
75400963
KW
507#define strNE(s1,s2) (strcmp(s1,s2) != 0)
508#define strEQ(s1,s2) (strcmp(s1,s2) == 0)
8d063cd8
LW
509#define strLT(s1,s2) (strcmp(s1,s2) < 0)
510#define strLE(s1,s2) (strcmp(s1,s2) <= 0)
511#define strGT(s1,s2) (strcmp(s1,s2) > 0)
512#define strGE(s1,s2) (strcmp(s1,s2) >= 0)
62946e08 513
75400963
KW
514#define strnNE(s1,s2,l) (strncmp(s1,s2,l) != 0)
515#define strnEQ(s1,s2,l) (strncmp(s1,s2,l) == 0)
378cc40b 516
9d3980bc
KW
517#define memEQ(s1,s2,l) (memcmp(((const void *) (s1)), ((const void *) (s2)), l) == 0)
518#define memNE(s1,s2,l) (! memEQ(s1,s2,l))
36477c24 519
085b7534 520/* memEQ and memNE where second comparand is a string constant */
568a785a 521#define memEQs(s1, l, s2) \
777fa2cb 522 (((sizeof(s2)-1) == (l)) && memEQ((s1), ("" s2 ""), (sizeof(s2)-1)))
5f50c6c9 523#define memNEs(s1, l, s2) (! memEQs(s1, l, s2))
568a785a 524
fdbb9a7c
KW
525/* Keep these private until we decide it was a good idea */
526#if defined(PERL_CORE) || defined(PERL_EXT) || defined(PERL_EXT_POSIX)
527
528#define strBEGINs(s1,s2) (strncmp(s1,"" s2 "", sizeof(s2)-1) == 0)
529
bdb7e3f0 530#define memBEGINs(s1, l, s2) \
30a6480c 531 ( (Ptrdiff_t) (l) >= (Ptrdiff_t) sizeof(s2) - 1 \
bdb7e3f0 532 && memEQ(s1, "" s2 "", sizeof(s2)-1))
de627158 533#define memBEGINPs(s1, l, s2) \
30a6480c 534 ( (Ptrdiff_t) (l) > (Ptrdiff_t) sizeof(s2) - 1 \
de627158 535 && memEQ(s1, "" s2 "", sizeof(s2)-1))
bdb7e3f0 536#define memENDs(s1, l, s2) \
30a6480c 537 ( (Ptrdiff_t) (l) >= (Ptrdiff_t) sizeof(s2) - 1 \
bdb7e3f0 538 && memEQ(s1 + (l) - (sizeof(s2) - 1), "" s2 "", sizeof(s2)-1))
b80f8424 539#define memENDPs(s1, l, s2) \
30a6480c 540 ( (Ptrdiff_t) (l) > (Ptrdiff_t) sizeof(s2) \
b80f8424 541 && memEQ(s1 + (l) - (sizeof(s2) - 1), "" s2 "", sizeof(s2)-1))
fdbb9a7c 542#endif /* End of making macros private */
bdb7e3f0 543
062b6850
KW
544#define memLT(s1,s2,l) (memcmp(s1,s2,l) < 0)
545#define memLE(s1,s2,l) (memcmp(s1,s2,l) <= 0)
546#define memGT(s1,s2,l) (memcmp(s1,s2,l) > 0)
547#define memGE(s1,s2,l) (memcmp(s1,s2,l) >= 0)
548
bbce6d69
PP
549/*
550 * Character classes.
551 *
552 * Unfortunately, the introduction of locales means that we
553 * can't trust isupper(), etc. to tell the truth. And when
554 * it comes to /\w+/ with tainting enabled, we *must* be able
555 * to trust our character classes.
556 *
557 * Therefore, the default tests in the text of Perl will be
558 * independent of locale. Any code that wants to depend on
559 * the current locale will use the tests that begin with "lc".
560 */
561
2304df62
AD
562#ifdef HAS_SETLOCALE /* XXX Is there a better test for this? */
563# ifndef CTYPE256
564# define CTYPE256
565# endif
566#endif
567
954c1994 568/*
ccfc67b7 569
dcccc8ff 570=head1 Character classification
243effed
KW
571This section is about functions (really macros) that classify characters
572into types, such as punctuation versus alphabetic, etc. Most of these are
573analogous to regular expression character classes. (See
574L<perlrecharclass/POSIX Character Classes>.) There are several variants for
575each class. (Not all macros have all variants; each item below lists the
576ones valid for it.) None are affected by C<use bytes>, and only the ones
577with C<LC> in the name are affected by the current locale.
578
d713f9d9
KW
579The base function, e.g., C<isALPHA()>, takes any signed or unsigned value,
580treating it as a code point, and returns a boolean as to whether or not the
581character represented by it is (or on non-ASCII platforms, corresponds to) an
6aff1f14
KW
582ASCII character in the named class based on platform, Unicode, and Perl rules.
583If the input is a number that doesn't fit in an octet, FALSE is returned.
243effed 584
c98722a4 585Variant C<isI<FOO>_A> (e.g., C<isALPHA_A()>) is identical to the base function
550da823
KW
586with no suffix C<"_A">. This variant is used to emphasize by its name that
587only ASCII-range characters can return TRUE.
4b9734bf 588
d60679e1 589Variant C<isI<FOO>_L1> imposes the Latin-1 (or EBCDIC equivalent) character set
4b9734bf
KW
590onto the platform. That is, the code points that are ASCII are unaffected,
591since ASCII is a subset of Latin-1. But the non-ASCII code points are treated
592as if they are Latin-1 characters. For example, C<isWORDCHAR_L1()> will return
593true when called with the code point 0xDF, which is a word character in both
4650c663 594ASCII and EBCDIC (though it represents different characters in each).
d713f9d9
KW
595If the input is a number that doesn't fit in an octet, FALSE is returned.
596(Perl's documentation uses a colloquial definition of Latin-1, to include all
597code points below 256.)
243effed 598
d713f9d9
KW
599Variant C<isI<FOO>_uvchr> is exactly like the C<isI<FOO>_L1> variant, for
600inputs below 256, but if the code point is larger than 255, Unicode rules are
601used to determine if it is in the character class. For example,
d0da05db 602C<isWORDCHAR_uvchr(0x100)> returns TRUE, since 0x100 is LATIN CAPITAL LETTER A
6aff1f14 603WITH MACRON in Unicode, and is a word character.
243effed 604
c98722a4 605Variant C<isI<FOO>_utf8_safe> is like C<isI<FOO>_uvchr>, but is used for UTF-8
d713f9d9
KW
606encoded strings. Each call classifies the first character of the string. This
607variant takes two parameters. The first, C<p>, is a
da8c1a98
KW
608pointer to the first byte of the character to be classified. (Recall that it
609may take more than one byte to represent a character in UTF-8 strings.) The
610second parameter, C<e>, points to anywhere in the string beyond the first
611character, up to one byte past the end of the entire string. The suffix
612C<_safe> in the function's name indicates that it will not attempt to read
613beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is true (this
614is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the input
615character is malformed in some way, the program may croak, or the function may
616return FALSE, at the discretion of the implementation, and subject to change in
617future releases.
618
c98722a4 619Variant C<isI<FOO>_utf8> is like C<isI<FOO>_utf8_safe>, but takes just a single
da8c1a98 620parameter, C<p>, which has the same meaning as the corresponding parameter does
c98722a4 621in C<isI<FOO>_utf8_safe>. The function therefore can't check if it is reading
8b4b30c5 622beyond the end of the string. Starting in Perl v5.32, it will take a second
c98722a4 623parameter, becoming a synonym for C<isI<FOO>_utf8_safe>. At that time every
34aeb2e9 624program that uses it will have to be changed to successfully compile. In the
c98722a4 625meantime, the first runtime call to C<isI<FOO>_utf8> from each call point in the
34aeb2e9 626program will raise a deprecation warning, enabled by default. You can convert
c98722a4 627your program now to use C<isI<FOO>_utf8_safe>, and avoid the warnings, and get an
8b4b30c5 628extra measure of protection, or you can wait until v5.32, when you'll be forced
34aeb2e9 629to add the C<e> parameter.
243effed 630
d713f9d9
KW
631Variant C<isI<FOO>_LC> is like the C<isI<FOO>_A> and C<isI<FOO>_L1> variants,
632but the result is based on the current locale, which is what C<LC> in the name
633stands for. If Perl can determine that the current locale is a UTF-8 locale,
634it uses the published Unicode rules; otherwise, it uses the C library function
635that gives the named classification. For example, C<isDIGIT_LC()> when not in
636a UTF-8 locale returns the result of calling C<isdigit()>. FALSE is always
1a83413c
KW
637returned if the input won't fit into an octet. On some platforms where the C
638library function is known to be defective, Perl changes its result to follow
639the POSIX standard's rules.
243effed 640
d713f9d9
KW
641Variant C<isI<FOO>_LC_uvchr> acts exactly like C<isI<FOO>_LC> for inputs less
642than 256, but for larger ones it returns the Unicode classification of the code
643point.
243effed 644
c98722a4 645Variant C<isI<FOO>_LC_utf8_safe> is like C<isI<FOO>_LC_uvchr>, but is used for UTF-8
d713f9d9
KW
646encoded strings. Each call classifies the first character of the string. This
647variant takes two parameters. The first, C<p>, is a pointer to the first byte
648of the character to be classified. (Recall that it may take more than one byte
649to represent a character in UTF-8 strings.) The second parameter, C<e>,
650points to anywhere in the string beyond the first character, up to one byte
651past the end of the entire string. The suffix C<_safe> in the function's name
652indicates that it will not attempt to read beyond S<C<e - 1>>, provided that
653the constraint S<C<s E<lt> e>> is true (this is asserted for in C<-DDEBUGGING>
654builds). If the UTF-8 for the input character is malformed in some way, the
655program may croak, or the function may return FALSE, at the discretion of the
656implementation, and subject to change in future releases.
da8c1a98 657
c98722a4 658Variant C<isI<FOO>_LC_utf8> is like C<isI<FOO>_LC_utf8_safe>, but takes just a single
da8c1a98 659parameter, C<p>, which has the same meaning as the corresponding parameter does
c98722a4 660in C<isI<FOO>_LC_utf8_safe>. The function therefore can't check if it is reading
8b4b30c5 661beyond the end of the string. Starting in Perl v5.32, it will take a second
c98722a4 662parameter, becoming a synonym for C<isI<FOO>_LC_utf8_safe>. At that time every
34aeb2e9 663program that uses it will have to be changed to successfully compile. In the
c98722a4 664meantime, the first runtime call to C<isI<FOO>_LC_utf8> from each call point in
34aeb2e9 665the program will raise a deprecation warning, enabled by default. You can
c98722a4 666convert your program now to use C<isI<FOO>_LC_utf8_safe>, and avoid the warnings,
8b4b30c5 667and get an extra measure of protection, or you can wait until v5.32, when
34aeb2e9 668you'll be forced to add the C<e> parameter.
ccfc67b7 669
d713f9d9
KW
670=for apidoc Am|bool|isALPHA|int ch
671Returns a boolean indicating whether the specified input is one of C<[A-Za-z]>,
672analogous to C<m/[[:alpha:]]/>.
dcccc8ff
KW
673See the L<top of this section|/Character classification> for an explanation of
674variants
da8c1a98
KW
675C<isALPHA_A>, C<isALPHA_L1>, C<isALPHA_uvchr>, C<isALPHA_utf8_safe>,
676C<isALPHA_LC>, C<isALPHA_LC_uvchr>, and C<isALPHA_LC_utf8_safe>.
8a58bdcf 677
d713f9d9
KW
678=for apidoc Am|bool|isALPHANUMERIC|int ch
679Returns a boolean indicating whether the specified character is one of
680C<[A-Za-z0-9]>, analogous to C<m/[[:alnum:]]/>.
dcccc8ff
KW
681See the L<top of this section|/Character classification> for an explanation of
682variants
d0da05db 683C<isALPHANUMERIC_A>, C<isALPHANUMERIC_L1>, C<isALPHANUMERIC_uvchr>,
da8c1a98
KW
684C<isALPHANUMERIC_utf8_safe>, C<isALPHANUMERIC_LC>, C<isALPHANUMERIC_LC_uvchr>,
685and C<isALPHANUMERIC_LC_utf8_safe>.
15861f94 686
255b632a
KW
687A (discouraged from use) synonym is C<isALNUMC> (where the C<C> suffix means
688this corresponds to the C language alphanumeric definition). Also
689there are the variants
690C<isALNUMC_A>, C<isALNUMC_L1>
691C<isALNUMC_LC>, and C<isALNUMC_LC_uvchr>.
692
d713f9d9 693=for apidoc Am|bool|isASCII|int ch
8a58bdcf 694Returns a boolean indicating whether the specified character is one of the 128
243effed 695characters in the ASCII character set, analogous to C<m/[[:ascii:]]/>.
e5ad6aba 696On non-ASCII platforms, it returns TRUE iff this
8a58bdcf
KW
697character corresponds to an ASCII character. Variants C<isASCII_A()> and
698C<isASCII_L1()> are identical to C<isASCII()>.
dcccc8ff
KW
699See the L<top of this section|/Character classification> for an explanation of
700variants
da8c1a98
KW
701C<isASCII_uvchr>, C<isASCII_utf8_safe>, C<isASCII_LC>, C<isASCII_LC_uvchr>, and
702C<isASCII_LC_utf8_safe>. Note, however, that some platforms do not have the C
243effed
KW
703library routine C<isascii()>. In these cases, the variants whose names contain
704C<LC> are the same as the corresponding ones without.
705
d98532ea
KW
706Also note, that because all ASCII characters are UTF-8 invariant (meaning they
707have the exact same representation (always a single byte) whether encoded in
708UTF-8 or not), C<isASCII> will give the correct results when called with any
da8c1a98
KW
709byte in any string encoded or not in UTF-8. And similarly C<isASCII_utf8_safe>
710will work properly on any string encoded or not in UTF-8.
d98532ea 711
243effed
KW
712=for apidoc Am|bool|isBLANK|char ch
713Returns a boolean indicating whether the specified character is a
6aff1f14 714character considered to be a blank, analogous to C<m/[[:blank:]]/>.
dcccc8ff
KW
715See the L<top of this section|/Character classification> for an explanation of
716variants
da8c1a98
KW
717C<isBLANK_A>, C<isBLANK_L1>, C<isBLANK_uvchr>, C<isBLANK_utf8_safe>,
718C<isBLANK_LC>, C<isBLANK_LC_uvchr>, and C<isBLANK_LC_utf8_safe>. Note,
719however, that some platforms do not have the C library routine
720C<isblank()>. In these cases, the variants whose names contain C<LC> are
721the same as the corresponding ones without.
243effed
KW
722
723=for apidoc Am|bool|isCNTRL|char ch
724Returns a boolean indicating whether the specified character is a
6aff1f14 725control character, analogous to C<m/[[:cntrl:]]/>.
dcccc8ff
KW
726See the L<top of this section|/Character classification> for an explanation of
727variants
da8c1a98
KW
728C<isCNTRL_A>, C<isCNTRL_L1>, C<isCNTRL_uvchr>, C<isCNTRL_utf8_safe>,
729C<isCNTRL_LC>, C<isCNTRL_LC_uvchr>, and C<isCNTRL_LC_utf8_safe> On EBCDIC
730platforms, you almost always want to use the C<isCNTRL_L1> variant.
954c1994 731
954c1994 732=for apidoc Am|bool|isDIGIT|char ch
2787a470 733Returns a boolean indicating whether the specified character is a
6aff1f14 734digit, analogous to C<m/[[:digit:]]/>.
8a58bdcf 735Variants C<isDIGIT_A> and C<isDIGIT_L1> are identical to C<isDIGIT>.
dcccc8ff
KW
736See the L<top of this section|/Character classification> for an explanation of
737variants
da8c1a98
KW
738C<isDIGIT_uvchr>, C<isDIGIT_utf8_safe>, C<isDIGIT_LC>, C<isDIGIT_LC_uvchr>, and
739C<isDIGIT_LC_utf8_safe>.
243effed
KW
740
741=for apidoc Am|bool|isGRAPH|char ch
742Returns a boolean indicating whether the specified character is a
6aff1f14 743graphic character, analogous to C<m/[[:graph:]]/>.
dcccc8ff 744See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
745variants C<isGRAPH_A>, C<isGRAPH_L1>, C<isGRAPH_uvchr>, C<isGRAPH_utf8_safe>,
746C<isGRAPH_LC>, C<isGRAPH_LC_uvchr>, and C<isGRAPH_LC_utf8_safe>.
954c1994 747
0c82b6df 748=for apidoc Am|bool|isLOWER|char ch
2787a470 749Returns a boolean indicating whether the specified character is a
6aff1f14 750lowercase character, analogous to C<m/[[:lower:]]/>.
dcccc8ff
KW
751See the L<top of this section|/Character classification> for an explanation of
752variants
da8c1a98
KW
753C<isLOWER_A>, C<isLOWER_L1>, C<isLOWER_uvchr>, C<isLOWER_utf8_safe>,
754C<isLOWER_LC>, C<isLOWER_LC_uvchr>, and C<isLOWER_LC_utf8_safe>.
0c82b6df 755
c99e91e9 756=for apidoc Am|bool|isOCTAL|char ch
2787a470 757Returns a boolean indicating whether the specified character is an
6aff1f14 758octal digit, [0-7].
243effed
KW
759The only two variants are C<isOCTAL_A> and C<isOCTAL_L1>; each is identical to
760C<isOCTAL>.
761
762=for apidoc Am|bool|isPUNCT|char ch
763Returns a boolean indicating whether the specified character is a
6aff1f14
KW
764punctuation character, analogous to C<m/[[:punct:]]/>.
765Note that the definition of what is punctuation isn't as
243effed
KW
766straightforward as one might desire. See L<perlrecharclass/POSIX Character
767Classes> for details.
dcccc8ff 768See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
769variants C<isPUNCT_A>, C<isPUNCT_L1>, C<isPUNCT_uvchr>, C<isPUNCT_utf8_safe>,
770C<isPUNCT_LC>, C<isPUNCT_LC_uvchr>, and C<isPUNCT_LC_utf8_safe>.
c99e91e9 771
0c82b6df 772=for apidoc Am|bool|isSPACE|char ch
2787a470 773Returns a boolean indicating whether the specified character is a
6aff1f14 774whitespace character. This is analogous
398d098a 775to what C<m/\s/> matches in a regular expression. Starting in Perl 5.18
779cf272 776this also matches what C<m/[[:space:]]/> does. Prior to 5.18, only the
398d098a
KW
777locale forms of this macro (the ones with C<LC> in their names) matched
778precisely what C<m/[[:space:]]/> does. In those releases, the only difference,
779in the non-locale variants, was that C<isSPACE()> did not match a vertical tab.
780(See L</isPSXSPC> for a macro that matches a vertical tab in all releases.)
dcccc8ff
KW
781See the L<top of this section|/Character classification> for an explanation of
782variants
da8c1a98
KW
783C<isSPACE_A>, C<isSPACE_L1>, C<isSPACE_uvchr>, C<isSPACE_utf8_safe>,
784C<isSPACE_LC>, C<isSPACE_LC_uvchr>, and C<isSPACE_LC_utf8_safe>.
0c82b6df 785
398d098a
KW
786=for apidoc Am|bool|isPSXSPC|char ch
787(short for Posix Space)
779cf272
KW
788Starting in 5.18, this is identical in all its forms to the
789corresponding C<isSPACE()> macros.
398d098a
KW
790The locale forms of this macro are identical to their corresponding
791C<isSPACE()> forms in all Perl releases. In releases prior to 5.18, the
792non-locale forms differ from their C<isSPACE()> forms only in that the
793C<isSPACE()> forms don't match a Vertical Tab, and the C<isPSXSPC()> forms do.
794Otherwise they are identical. Thus this macro is analogous to what
795C<m/[[:space:]]/> matches in a regular expression.
dcccc8ff 796See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
797variants C<isPSXSPC_A>, C<isPSXSPC_L1>, C<isPSXSPC_uvchr>, C<isPSXSPC_utf8_safe>,
798C<isPSXSPC_LC>, C<isPSXSPC_LC_uvchr>, and C<isPSXSPC_LC_utf8_safe>.
398d098a 799
954c1994 800=for apidoc Am|bool|isUPPER|char ch
2787a470 801Returns a boolean indicating whether the specified character is an
6aff1f14 802uppercase character, analogous to C<m/[[:upper:]]/>.
dcccc8ff 803See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
804variants C<isUPPER_A>, C<isUPPER_L1>, C<isUPPER_uvchr>, C<isUPPER_utf8_safe>,
805C<isUPPER_LC>, C<isUPPER_LC_uvchr>, and C<isUPPER_LC_utf8_safe>.
954c1994 806
243effed 807=for apidoc Am|bool|isPRINT|char ch
8eea39dd 808Returns a boolean indicating whether the specified character is a
6aff1f14 809printable character, analogous to C<m/[[:print:]]/>.
dcccc8ff
KW
810See the L<top of this section|/Character classification> for an explanation of
811variants
da8c1a98
KW
812C<isPRINT_A>, C<isPRINT_L1>, C<isPRINT_uvchr>, C<isPRINT_utf8_safe>,
813C<isPRINT_LC>, C<isPRINT_LC_uvchr>, and C<isPRINT_LC_utf8_safe>.
243effed
KW
814
815=for apidoc Am|bool|isWORDCHAR|char ch
816Returns a boolean indicating whether the specified character is a character
817that is a word character, analogous to what C<m/\w/> and C<m/[[:word:]]/> match
818in a regular expression. A word character is an alphabetic character, a
819decimal digit, a connecting punctuation character (such as an underscore), or
820a "mark" character that attaches to one of those (like some sort of accent).
821C<isALNUM()> is a synonym provided for backward compatibility, even though a
822word character includes more than the standard C language meaning of
823alphanumeric.
dcccc8ff 824See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
825variants C<isWORDCHAR_A>, C<isWORDCHAR_L1>, C<isWORDCHAR_uvchr>, and
826C<isWORDCHAR_utf8_safe>. C<isWORDCHAR_LC>, C<isWORDCHAR_LC_uvchr>, and
827C<isWORDCHAR_LC_utf8_safe> are also as described there, but additionally
828include the platform's native underscore.
8a58bdcf
KW
829
830=for apidoc Am|bool|isXDIGIT|char ch
831Returns a boolean indicating whether the specified character is a hexadecimal
243effed
KW
832digit. In the ASCII range these are C<[0-9A-Fa-f]>. Variants C<isXDIGIT_A()>
833and C<isXDIGIT_L1()> are identical to C<isXDIGIT()>.
dcccc8ff
KW
834See the L<top of this section|/Character classification> for an explanation of
835variants
da8c1a98
KW
836C<isXDIGIT_uvchr>, C<isXDIGIT_utf8_safe>, C<isXDIGIT_LC>, C<isXDIGIT_LC_uvchr>,
837and C<isXDIGIT_LC_utf8_safe>.
243effed 838
3c3ecf18
KW
839=for apidoc Am|bool|isIDFIRST|char ch
840Returns a boolean indicating whether the specified character can be the first
841character of an identifier. This is very close to, but not quite the same as
842the official Unicode property C<XID_Start>. The difference is that this
843returns true only if the input character also matches L</isWORDCHAR>.
dcccc8ff
KW
844See the L<top of this section|/Character classification> for an explanation of
845variants
da8c1a98
KW
846C<isIDFIRST_A>, C<isIDFIRST_L1>, C<isIDFIRST_uvchr>, C<isIDFIRST_utf8_safe>,
847C<isIDFIRST_LC>, C<isIDFIRST_LC_uvchr>, and C<isIDFIRST_LC_utf8_safe>.
3c3ecf18
KW
848
849=for apidoc Am|bool|isIDCONT|char ch
850Returns a boolean indicating whether the specified character can be the
851second or succeeding character of an identifier. This is very close to, but
852not quite the same as the official Unicode property C<XID_Continue>. The
853difference is that this returns true only if the input character also matches
dcccc8ff
KW
854L</isWORDCHAR>. See the L<top of this section|/Character classification> for
855an
d0da05db 856explanation of variants C<isIDCONT_A>, C<isIDCONT_L1>, C<isIDCONT_uvchr>,
da8c1a98
KW
857C<isIDCONT_utf8_safe>, C<isIDCONT_LC>, C<isIDCONT_LC_uvchr>, and
858C<isIDCONT_LC_utf8_safe>.
3c3ecf18 859
243effed 860=head1 Miscellaneous Functions
8eea39dd 861
95a59cab 862=for apidoc Am|U8|READ_XDIGIT|char str*
243effed 863Returns the value of an ASCII-range hex digit and advances the string pointer.
95a59cab
YO
864Behaviour is only well defined when isXDIGIT(*str) is true.
865
e7c1e6c1 866=head1 Character case changing
21da7284
KW
867Perl uses "full" Unicode case mappings. This means that converting a single
868character to another case may result in a sequence of more than one character.
869For example, the uppercase of C<E<223>> (LATIN SMALL LETTER SHARP S) is the two
870character sequence C<SS>. This presents some complications The lowercase of
871all characters in the range 0..255 is a single character, and thus
872C<L</toLOWER_L1>> is furnished. But, C<toUPPER_L1> can't exist, as it couldn't
873return a valid result for all legal inputs. Instead C<L</toUPPER_uvchr>> has
874an API that does allow every possible legal result to be returned.) Likewise
875no other function that is crippled by not being able to give the correct
876results for the full range of possible inputs has been implemented here.
e7c1e6c1 877
d713f9d9 878=for apidoc Am|U8|toUPPER|int ch
1f607577
KW
879Converts the specified character to uppercase. If the input is anything but an
880ASCII lowercase character, that input character itself is returned. Variant
c753c8d3 881C<toUPPER_A> is equivalent.
954c1994 882
d0da05db
KW
883=for apidoc Am|UV|toUPPER_uvchr|UV cp|U8* s|STRLEN* lenp
884Converts the code point C<cp> to its uppercase version, and
885stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
886point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
887that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
888bytes since the uppercase version may be longer than the original character.
889
890The first code point of the uppercased version is returned
21da7284
KW
891(but note, as explained at L<the top of this section|/Character case
892changing>, that there may be more.)
1f607577 893
a239b1e2
KW
894=for apidoc Am|UV|toUPPER_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
895Converts the first UTF-8 encoded character in the sequence starting at C<p> and
896extending no further than S<C<e - 1>> to its uppercase version, and
1f607577
KW
897stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
898that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
899bytes since the uppercase version may be longer than the original character.
900
901The first code point of the uppercased version is returned
21da7284
KW
902(but note, as explained at L<the top of this section|/Character case
903changing>, that there may be more).
1f607577 904
a239b1e2
KW
905The suffix C<_safe> in the function's name indicates that it will not attempt
906to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
907true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
908input character is malformed in some way, the program may croak, or the
909function may return the REPLACEMENT CHARACTER, at the discretion of the
910implementation, and subject to change in future releases.
911
912=for apidoc Am|UV|toUPPER_utf8|U8* p|U8* s|STRLEN* lenp
913This is like C<L</toUPPER_utf8_safe>>, but doesn't have the C<e>
914parameter The function therefore can't check if it is reading
8b4b30c5 915beyond the end of the string. Starting in Perl v5.32, it will take the C<e>
607313a1
KW
916parameter, becoming a synonym for C<toUPPER_utf8_safe>. At that time every
917program that uses it will have to be changed to successfully compile. In the
918meantime, the first runtime call to C<toUPPER_utf8> from each call point in the
919program will raise a deprecation warning, enabled by default. You can convert
920your program now to use C<toUPPER_utf8_safe>, and avoid the warnings, and get an
8b4b30c5 921extra measure of protection, or you can wait until v5.32, when you'll be forced
607313a1 922to add the C<e> parameter.
1f607577 923
25200305
KW
924=for apidoc Am|U8|toFOLD|U8 ch
925Converts the specified character to foldcase. If the input is anything but an
926ASCII uppercase character, that input character itself is returned. Variant
927C<toFOLD_A> is equivalent. (There is no equivalent C<to_FOLD_L1> for the full
d0da05db 928Latin1 range, as the full generality of L</toFOLD_uvchr> is needed there.)
25200305 929
d0da05db
KW
930=for apidoc Am|UV|toFOLD_uvchr|UV cp|U8* s|STRLEN* lenp
931Converts the code point C<cp> to its foldcase version, and
932stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
933point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
934that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
935bytes since the foldcase version may be longer than the original character.
936
937The first code point of the foldcased version is returned
21da7284
KW
938(but note, as explained at L<the top of this section|/Character case
939changing>, that there may be more).
1f607577 940
a239b1e2
KW
941=for apidoc Am|UV|toFOLD_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
942Converts the first UTF-8 encoded character in the sequence starting at C<p> and
943extending no further than S<C<e - 1>> to its foldcase version, and
1f607577
KW
944stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
945that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
946bytes since the foldcase version may be longer than the original character.
947
948The first code point of the foldcased version is returned
21da7284
KW
949(but note, as explained at L<the top of this section|/Character case
950changing>, that there may be more).
1f607577 951
a239b1e2
KW
952The suffix C<_safe> in the function's name indicates that it will not attempt
953to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
954true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
955input character is malformed in some way, the program may croak, or the
956function may return the REPLACEMENT CHARACTER, at the discretion of the
957implementation, and subject to change in future releases.
958
959=for apidoc Am|UV|toFOLD_utf8|U8* p|U8* s|STRLEN* lenp
960This is like C<L</toFOLD_utf8_safe>>, but doesn't have the C<e>
961parameter The function therefore can't check if it is reading
8b4b30c5 962beyond the end of the string. Starting in Perl v5.32, it will take the C<e>
607313a1
KW
963parameter, becoming a synonym for C<toFOLD_utf8_safe>. At that time every
964program that uses it will have to be changed to successfully compile. In the
965meantime, the first runtime call to C<toFOLD_utf8> from each call point in the
966program will raise a deprecation warning, enabled by default. You can convert
967your program now to use C<toFOLD_utf8_safe>, and avoid the warnings, and get an
8b4b30c5 968extra measure of protection, or you can wait until v5.32, when you'll be forced
607313a1 969to add the C<e> parameter.
1f607577
KW
970
971=for apidoc Am|U8|toLOWER|U8 ch
972Converts the specified character to lowercase. If the input is anything but an
973ASCII uppercase character, that input character itself is returned. Variant
c753c8d3 974C<toLOWER_A> is equivalent.
954c1994 975
1f607577 976=for apidoc Am|U8|toLOWER_L1|U8 ch
b7d90381
KW
977Converts the specified Latin1 character to lowercase. The results are
978undefined if the input doesn't fit in a byte.
1f607577
KW
979
980=for apidoc Am|U8|toLOWER_LC|U8 ch
981Converts the specified character to lowercase using the current locale's rules,
982if possible; otherwise returns the input character itself.
983
d0da05db
KW
984=for apidoc Am|UV|toLOWER_uvchr|UV cp|U8* s|STRLEN* lenp
985Converts the code point C<cp> to its lowercase version, and
986stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
987point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
988that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
989bytes since the lowercase version may be longer than the original character.
990
991The first code point of the lowercased version is returned
21da7284
KW
992(but note, as explained at L<the top of this section|/Character case
993changing>, that there may be more).
1f607577 994
a239b1e2
KW
995
996=for apidoc Am|UV|toLOWER_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
997Converts the first UTF-8 encoded character in the sequence starting at C<p> and
998extending no further than S<C<e - 1>> to its lowercase version, and
1f607577
KW
999stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
1000that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
1001bytes since the lowercase version may be longer than the original character.
1002
1003The first code point of the lowercased 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
1007The suffix C<_safe> in the function's name indicates that it will not attempt
1008to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
1009true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
1010input character is malformed in some way, the program may croak, or the
1011function may return the REPLACEMENT CHARACTER, at the discretion of the
1012implementation, and subject to change in future releases.
1013
1014=for apidoc Am|UV|toLOWER_utf8|U8* p|U8* s|STRLEN* lenp
1015This is like C<L</toLOWER_utf8_safe>>, but doesn't have the C<e>
1016parameter The function therefore can't check if it is reading
8b4b30c5 1017beyond the end of the string. Starting in Perl v5.32, it will take the C<e>
607313a1
KW
1018parameter, becoming a synonym for C<toLOWER_utf8_safe>. At that time every
1019program that uses it will have to be changed to successfully compile. In the
1020meantime, the first runtime call to C<toLOWER_utf8> from each call point in the
1021program will raise a deprecation warning, enabled by default. You can convert
1022your program now to use C<toLOWER_utf8_safe>, and avoid the warnings, and get an
8b4b30c5 1023extra measure of protection, or you can wait until v5.32, when you'll be forced
607313a1 1024to add the C<e> parameter.
1f607577 1025
25200305
KW
1026=for apidoc Am|U8|toTITLE|U8 ch
1027Converts the specified character to titlecase. If the input is anything but an
1028ASCII lowercase character, that input character itself is returned. Variant
b7d90381 1029C<toTITLE_A> is equivalent. (There is no C<toTITLE_L1> for the full Latin1
d0da05db 1030range, as the full generality of L</toTITLE_uvchr> is needed there. Titlecase is
b7d90381 1031not a concept used in locale handling, so there is no functionality for that.)
25200305 1032
d0da05db
KW
1033=for apidoc Am|UV|toTITLE_uvchr|UV cp|U8* s|STRLEN* lenp
1034Converts the code point C<cp> to its titlecase version, and
1035stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
1036point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
1037that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
1038bytes since the titlecase version may be longer than the original character.
1039
1040The first code point of the titlecased version is returned
21da7284
KW
1041(but note, as explained at L<the top of this section|/Character case
1042changing>, that there may be more).
1f607577 1043
a239b1e2
KW
1044=for apidoc Am|UV|toTITLE_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
1045Converts the first UTF-8 encoded character in the sequence starting at C<p> and
1046extending no further than S<C<e - 1>> to its titlecase version, and
1f607577
KW
1047stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
1048that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
1049bytes since the titlecase version may be longer than the original character.
1050
1051The first code point of the titlecased version is returned
21da7284
KW
1052(but note, as explained at L<the top of this section|/Character case
1053changing>, that there may be more).
1f607577 1054
a239b1e2
KW
1055The suffix C<_safe> in the function's name indicates that it will not attempt
1056to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
1057true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
1058input character is malformed in some way, the program may croak, or the
1059function may return the REPLACEMENT CHARACTER, at the discretion of the
1060implementation, and subject to change in future releases.
1061
1062=for apidoc Am|UV|toTITLE_utf8|U8* p|U8* s|STRLEN* lenp
1063This is like C<L</toLOWER_utf8_safe>>, but doesn't have the C<e>
1064parameter The function therefore can't check if it is reading
8b4b30c5 1065beyond the end of the string. Starting in Perl v5.32, it will take the C<e>
607313a1
KW
1066parameter, becoming a synonym for C<toTITLE_utf8_safe>. At that time every
1067program that uses it will have to be changed to successfully compile. In the
1068meantime, the first runtime call to C<toTITLE_utf8> from each call point in the
1069program will raise a deprecation warning, enabled by default. You can convert
1070your program now to use C<toTITLE_utf8_safe>, and avoid the warnings, and get an
8b4b30c5 1071extra measure of protection, or you can wait until v5.32, when you'll be forced
607313a1 1072to add the C<e> parameter.
1f607577 1073
954c1994 1074=cut
353c9b6f 1075
d0da05db 1076XXX Still undocumented isVERTWS_uvchr and _utf8; it's unclear what their names
1e222e4f
KW
1077really should be. Also toUPPER_LC and toFOLD_LC, which are subject to change,
1078and aren't general purpose as they don't work on U+DF, and assert against that.
243effed 1079
8a58bdcf 1080Note that these macros are repeated in Devel::PPPort, so should also be
62fa66b6
KW
1081patched there. The file as of this writing is cpan/Devel-PPPort/parts/inc/misc
1082
954c1994
GS
1083*/
1084
de9e2639
KW
1085/* Specify the widest unsigned type on the platform. */
1086#ifdef QUADKIND
1087# define WIDEST_UTYPE U64
7c062697
KW
1088#else
1089# define WIDEST_UTYPE U32
1090#endif
1091
3912bc88
KW
1092/* FITS_IN_8_BITS(c) returns true if c doesn't have a bit set other than in
1093 * the lower 8. It is designed to be hopefully bomb-proof, making sure that no
1094 * bits of information are lost even on a 64-bit machine, but to get the
1095 * compiler to optimize it out if possible. This is because Configure makes
1096 * sure that the machine has an 8-bit byte, so if c is stored in a byte, the
1097 * sizeof() guarantees that this evaluates to a constant true at compile time.
7e75d1a1
JH
1098 *
1099 * For Coverity, be always true, because otherwise Coverity thinks
1100 * it finds several expressions that are always true, independent
1101 * of operands. Well, they are, but that is kind of the point.
220c71bf 1102 */
7e75d1a1 1103#ifndef __COVERITY__
6c5b02ac
KW
1104 /* The '| 0' part ensures a compiler error if c is not integer (like e.g., a
1105 * pointer) */
1106#define FITS_IN_8_BITS(c) ( (sizeof(c) == 1) \
ace3ad0f 1107 || !(((WIDEST_UTYPE)((c) | 0)) & ~0xFF))
7e75d1a1
JH
1108#else
1109#define FITS_IN_8_BITS(c) (1)
1110#endif
cf301eb7 1111
94250c4f
KW
1112/* Returns true if c is in the range l..u, where 'l' is non-negative
1113 * Written this way so that after optimization, only one conditional test is
1114 * needed.
1115 *
1116 * This isn't fully general, except for the special cased 'signed char' (which
1117 * should be resolved at compile time): It won't work if 'c' is negative, and
1118 * 'l' is larger than the max for that signed type. Thus if 'c' is a negative
1119 * int, and 'l' is larger than INT_MAX, it will fail. To protect agains this
1120 * happening, there is an assert that will generate a warning if c is larger
1121 * than e.g. INT_MAX if it is an 'unsigned int'. This could be a false
1122 * positive, but khw couldn't figure out a way to make it better. It's good
1123 * enough so far */
1124#define inRANGE(c, l, u) (__ASSERT_((l) >= 0) __ASSERT_((u) >= (l)) \
1125 ((sizeof(c) == 1) \
1126 ? (((WIDEST_UTYPE) ((((U8) (c))|0) - (l))) <= ((WIDEST_UTYPE) ((u) - (l)))) \
1127 : (__ASSERT_( (((WIDEST_UTYPE) 1) << (CHARBITS * sizeof(c) - 1) & (c)) \
1128 /* sign bit of c is 0 */ == 0 \
1129 || (((~ ((WIDEST_UTYPE) 1) << ((CHARBITS * sizeof(c) - 1) - 1))\
1130 /* l not larger than largest value in c's signed type */ \
1131 & ~ ((WIDEST_UTYPE) 0)) & (l)) == 0) \
1132 ((WIDEST_UTYPE) (((c) - (l)) | 0) <= ((WIDEST_UTYPE) ((u) - (l)))))))
305fe86e 1133
41f43cc2 1134#ifdef EBCDIC
b6340bd0 1135# ifndef _ALL_SOURCE
0852beac
KW
1136 /* The native libc isascii() et.al. functions return the wrong results
1137 * on at least z/OS unless this is defined. */
b6340bd0
KW
1138# error _ALL_SOURCE should probably be defined
1139# endif
41f43cc2 1140#else
0852beac
KW
1141 /* There is a simple definition of ASCII for ASCII platforms. But the
1142 * EBCDIC one isn't so simple, so is defined using table look-up like the
9c903d59 1143 * other macros below.
3f3c579d
KW
1144 *
1145 * The cast here is used instead of '(c) >= 0', because some compilers emit
1146 * a warning that that test is always true when the parameter is an
1147 * unsigned type. khw supposes that it could be written as
1148 * && ((c) == '\0' || (c) > 0)
1149 * to avoid the message, but the cast will likely avoid extra branches even
1150 * with stupid compilers.
1151 *
1152 * The '| 0' part ensures a compiler error if c is not integer (like e.g.,
1153 * a pointer) */
9c903d59 1154# define isASCII(c) ((WIDEST_UTYPE)((c) | 0) < 128)
41f43cc2
KW
1155#endif
1156
38694112
KW
1157/* Take the eight possible bit patterns of the lower 3 bits and you get the
1158 * lower 3 bits of the 8 octal digits, in both ASCII and EBCDIC, so those bits
1159 * can be ignored. If the rest match '0', we have an octal */
1160#define isOCTAL_A(c) (((WIDEST_UTYPE)((c) | 0) & ~7) == '0')
c2da0b36 1161
9fb1bf9d 1162#ifdef H_PERL /* If have access to perl.h, lookup in its table */
f4cdb42c 1163
a500dc72
KW
1164/* Character class numbers. For internal core Perl use only. The ones less
1165 * than 32 are used in PL_charclass[] and the ones up through the one that
1166 * corresponds to <_HIGHEST_REGCOMP_DOT_H_SYNC> are used by regcomp.h and
1167 * related files. PL_charclass ones use names used in l1_char_class_tab.h but
1168 * their actual definitions are here. If that file has a name not used here,
1169 * it won't compile.
1709d539
KW
1170 *
1171 * The first group of these is ordered in what I (khw) estimate to be the
31c7f561 1172 * frequency of their use. This gives a slight edge to exiting a loop earlier
58a3ba2c
KW
1173 * (in reginclass() in regexec.c). Except \v should be last, as it isn't a
1174 * real Posix character class, and some (small) inefficiencies in regular
1175 * expression handling would be introduced by putting it in the middle of those
1176 * that are. Also, cntrl and ascii come after the others as it may be useful
1177 * to group these which have no members that match above Latin1, (or above
1178 * ASCII in the latter case) */
1179
1709d539
KW
1180# define _CC_WORDCHAR 0 /* \w and [:word:] */
1181# define _CC_DIGIT 1 /* \d and [:digit:] */
1182# define _CC_ALPHA 2 /* [:alpha:] */
1183# define _CC_LOWER 3 /* [:lower:] */
1184# define _CC_UPPER 4 /* [:upper:] */
1185# define _CC_PUNCT 5 /* [:punct:] */
1186# define _CC_PRINT 6 /* [:print:] */
15861f94 1187# define _CC_ALPHANUMERIC 7 /* [:alnum:] */
1709d539 1188# define _CC_GRAPH 8 /* [:graph:] */
359b005e 1189# define _CC_CASED 9 /* [:lower:] or [:upper:] under /i */
779cf272 1190# define _CC_SPACE 10 /* \s, [:space:] */
34aeb2e9
KW
1191# define _CC_PSXSPC _CC_SPACE /* XXX Temporary, can be removed
1192 when the deprecated isFOO_utf8()
1193 functions are removed */
b0d691b2
KW
1194# define _CC_BLANK 11 /* [:blank:] */
1195# define _CC_XDIGIT 12 /* [:xdigit:] */
779cf272
KW
1196# define _CC_CNTRL 13 /* [:cntrl:] */
1197# define _CC_ASCII 14 /* [:ascii:] */
1198# define _CC_VERTSPACE 15 /* \v */
1709d539 1199
a0947d7b
KW
1200# define _HIGHEST_REGCOMP_DOT_H_SYNC _CC_VERTSPACE
1201
1709d539 1202/* The members of the third group below do not need to be coordinated with data
3ffc8c70 1203 * structures in regcomp.[ch] and regexec.c. */
779cf272
KW
1204# define _CC_IDFIRST 16
1205# define _CC_CHARNAME_CONT 17
1206# define _CC_NONLATIN1_FOLD 18
1207# define _CC_NONLATIN1_SIMPLE_FOLD 19
1208# define _CC_QUOTEMETA 20
1209# define _CC_NON_FINAL_FOLD 21
1210# define _CC_IS_IN_SOME_FOLD 22
1211# define _CC_MNEMONIC_CNTRL 23
073c22b3 1212
34aeb2e9
KW
1213# define _CC_IDCONT 24 /* XXX Temporary, can be removed when the deprecated
1214 isFOO_utf8() functions are removed */
1215
073c22b3
KW
1216/* This next group is only used on EBCDIC platforms, so theoretically could be
1217 * shared with something entirely different that's only on ASCII platforms */
5d5376e2
KW
1218# define _CC_UTF8_START_BYTE_IS_FOR_AT_LEAST_SURROGATE 28
1219# define _CC_UTF8_IS_START 29
1220# define _CC_UTF8_IS_DOWNGRADEABLE_START 30
1221# define _CC_UTF8_IS_CONTINUATION 31
1222/* Unused: 24-27
f4cdb42c
KW
1223 * If more bits are needed, one could add a second word for non-64bit
1224 * QUAD_IS_INT systems, using some #ifdefs to distinguish between having a 2nd
37ede926
KW
1225 * word or not. The IS_IN_SOME_FOLD bit is the most easily expendable, as it
1226 * is used only for optimization (as of this writing), and differs in the
1227 * Latin1 range from the ALPHA bit only in two relatively unimportant
a500dc72 1228 * characters: the masculine and feminine ordinal indicators, so removing it
073c22b3
KW
1229 * would just cause /i regexes which match them to run less efficiently.
1230 * Similarly the EBCDIC-only bits are used just for speed, and could be
1231 * replaced by other means */
96ac0975 1232
3a371f2f
KW
1233#if defined(PERL_CORE) || defined(PERL_EXT)
1234/* An enum version of the character class numbers, to help compilers
1235 * optimize */
1236typedef enum {
3a371f2f 1237 _CC_ENUM_ALPHA = _CC_ALPHA,
e8d596e0
KW
1238 _CC_ENUM_ALPHANUMERIC = _CC_ALPHANUMERIC,
1239 _CC_ENUM_ASCII = _CC_ASCII,
1240 _CC_ENUM_BLANK = _CC_BLANK,
b0d691b2 1241 _CC_ENUM_CASED = _CC_CASED,
e8d596e0 1242 _CC_ENUM_CNTRL = _CC_CNTRL,
3a371f2f
KW
1243 _CC_ENUM_DIGIT = _CC_DIGIT,
1244 _CC_ENUM_GRAPH = _CC_GRAPH,
1245 _CC_ENUM_LOWER = _CC_LOWER,
1246 _CC_ENUM_PRINT = _CC_PRINT,
1247 _CC_ENUM_PUNCT = _CC_PUNCT,
e8d596e0 1248 _CC_ENUM_SPACE = _CC_SPACE,
3a371f2f 1249 _CC_ENUM_UPPER = _CC_UPPER,
e8d596e0 1250 _CC_ENUM_VERTSPACE = _CC_VERTSPACE,
3a371f2f 1251 _CC_ENUM_WORDCHAR = _CC_WORDCHAR,
e8d596e0 1252 _CC_ENUM_XDIGIT = _CC_XDIGIT
3a371f2f
KW
1253} _char_class_number;
1254#endif
1255
86f72d56 1256#define POSIX_CC_COUNT (_HIGHEST_REGCOMP_DOT_H_SYNC + 1)
63c61c3f 1257
6635f04f 1258START_EXTERN_C
96ac0975
NC
1259# ifdef DOINIT
1260EXTCONST U32 PL_charclass[] = {
1261# include "l1_char_class_tab.h"
1262};
1263
1264# else /* ! DOINIT */
1265EXTCONST U32 PL_charclass[];
1266# endif
6635f04f 1267END_EXTERN_C
96ac0975 1268
265c1f46 1269 /* The 1U keeps Solaris from griping when shifting sets the uppermost bit */
430b7c70 1270# define _CC_mask(classnum) (1U << (classnum))
4650c663
KW
1271
1272 /* For internal core Perl use only: the base macro for defining macros like
1273 * isALPHA */
ff7ecfc3 1274# define _generic_isCC(c, classnum) cBOOL(FITS_IN_8_BITS(c) \
f4cd282c 1275 && (PL_charclass[(U8) (c)] & _CC_mask(classnum)))
4eeeb416 1276
f4cdb42c
KW
1277 /* The mask for the _A versions of the macros; it just adds in the bit for
1278 * ASCII. */
1279# define _CC_mask_A(classnum) (_CC_mask(classnum) | _CC_mask(_CC_ASCII))
1280
4650c663
KW
1281 /* For internal core Perl use only: the base macro for defining macros like
1282 * isALPHA_A. The foo_A version makes sure that both the desired bit and
1283 * the ASCII bit are present */
b7d90381
KW
1284# define _generic_isCC_A(c, classnum) (FITS_IN_8_BITS(c) \
1285 && ((PL_charclass[(U8) (c)] & _CC_mask_A(classnum)) \
1286 == _CC_mask_A(classnum)))
f4cdb42c 1287
26c1d9d8
KW
1288/* On ASCII platforms certain classes form a single range. It's faster to
1289 * special case these. isDIGIT is a single range on all platforms */
b877c1ff
KW
1290# ifdef EBCDIC
1291# define isALPHA_A(c) _generic_isCC_A(c, _CC_ALPHA)
1292# define isGRAPH_A(c) _generic_isCC_A(c, _CC_GRAPH)
1293# define isLOWER_A(c) _generic_isCC_A(c, _CC_LOWER)
1294# define isPRINT_A(c) _generic_isCC_A(c, _CC_PRINT)
1295# define isUPPER_A(c) _generic_isCC_A(c, _CC_UPPER)
1296# else
26c1d9d8 1297 /* By folding the upper and lowercase, we can use a single range */
b877c1ff 1298# define isALPHA_A(c) inRANGE((~('A' ^ 'a') & (c)), 'A', 'Z')
26c1d9d8 1299# define isGRAPH_A(c) inRANGE(c, ' ' + 1, 0x7e)
b877c1ff
KW
1300# define isLOWER_A(c) inRANGE(c, 'a', 'z')
1301# define isPRINT_A(c) inRANGE(c, ' ', 0x7e)
1302# define isUPPER_A(c) inRANGE(c, 'A', 'Z')
1303# endif
15861f94 1304# define isALPHANUMERIC_A(c) _generic_isCC_A(c, _CC_ALPHANUMERIC)
f4cdb42c
KW
1305# define isBLANK_A(c) _generic_isCC_A(c, _CC_BLANK)
1306# define isCNTRL_A(c) _generic_isCC_A(c, _CC_CNTRL)
b877c1ff 1307# define isDIGIT_A(c) inRANGE(c, '0', '9')
f4cdb42c
KW
1308# define isPUNCT_A(c) _generic_isCC_A(c, _CC_PUNCT)
1309# define isSPACE_A(c) _generic_isCC_A(c, _CC_SPACE)
f4cdb42c 1310# define isWORDCHAR_A(c) _generic_isCC_A(c, _CC_WORDCHAR)
b7d90381
KW
1311# define isXDIGIT_A(c) _generic_isCC(c, _CC_XDIGIT) /* No non-ASCII xdigits
1312 */
d95f8b6a 1313# define isIDFIRST_A(c) _generic_isCC_A(c, _CC_IDFIRST)
3ded5eb0
KW
1314# define isALPHA_L1(c) _generic_isCC(c, _CC_ALPHA)
1315# define isALPHANUMERIC_L1(c) _generic_isCC(c, _CC_ALPHANUMERIC)
1316# define isBLANK_L1(c) _generic_isCC(c, _CC_BLANK)
1317
1318 /* continuation character for legal NAME in \N{NAME} */
1319# define isCHARNAME_CONT(c) _generic_isCC(c, _CC_CHARNAME_CONT)
1320
1321# define isCNTRL_L1(c) _generic_isCC(c, _CC_CNTRL)
1322# define isGRAPH_L1(c) _generic_isCC(c, _CC_GRAPH)
1323# define isLOWER_L1(c) _generic_isCC(c, _CC_LOWER)
1324# define isPRINT_L1(c) _generic_isCC(c, _CC_PRINT)
b7d90381 1325# define isPSXSPC_L1(c) isSPACE_L1(c)
3ded5eb0
KW
1326# define isPUNCT_L1(c) _generic_isCC(c, _CC_PUNCT)
1327# define isSPACE_L1(c) _generic_isCC(c, _CC_SPACE)
1328# define isUPPER_L1(c) _generic_isCC(c, _CC_UPPER)
1329# define isWORDCHAR_L1(c) _generic_isCC(c, _CC_WORDCHAR)
1330# define isIDFIRST_L1(c) _generic_isCC(c, _CC_IDFIRST)
f4cdb42c 1331
0852beac
KW
1332# ifdef EBCDIC
1333# define isASCII(c) _generic_isCC(c, _CC_ASCII)
1334# endif
1335
f12c0118
KW
1336 /* Participates in a single-character fold with a character above 255 */
1337# 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)))
1338
1339 /* Like the above, but also can be part of a multi-char fold */
f4cd282c 1340# 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 1341
4eeeb416 1342# define _isQUOTEMETA(c) _generic_isCC(c, _CC_QUOTEMETA)
26faadbd 1343# define _IS_NON_FINAL_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
b7d90381 1344 _generic_isCC(c, _CC_NON_FINAL_FOLD)
37ede926 1345# define _IS_IN_SOME_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
b7d90381 1346 _generic_isCC(c, _CC_IS_IN_SOME_FOLD)
8e35b056
KW
1347# define _IS_MNEMONIC_CNTRL_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
1348 _generic_isCC(c, _CC_MNEMONIC_CNTRL)
687c8d01 1349#else /* else we don't have perl.h H_PERL */
3ded5eb0
KW
1350
1351 /* If we don't have perl.h, we are compiling a utility program. Below we
1352 * hard-code various macro definitions that wouldn't otherwise be available
fc273927 1353 * to it. Most are coded based on first principles. These are written to
74665a89 1354 * avoid EBCDIC vs. ASCII #ifdef's as much as possible. */
182c4ace 1355# define isDIGIT_A(c) inRANGE(c, '0', '9')
0852beac 1356# define isBLANK_A(c) ((c) == ' ' || (c) == '\t')
74665a89
KW
1357# define isSPACE_A(c) (isBLANK_A(c) \
1358 || (c) == '\n' \
1359 || (c) == '\r' \
1360 || (c) == '\v' \
0852beac 1361 || (c) == '\f')
74665a89
KW
1362 /* On EBCDIC, there are gaps between 'i' and 'j'; 'r' and 's'. Same for
1363 * uppercase. The tests for those aren't necessary on ASCII, but hurt only
1364 * performance (if optimization isn't on), and allow the same code to be
1365 * used for both platform types */
182c4ace
KW
1366# define isLOWER_A(c) inRANGE((c), 'a', 'i') \
1367 || inRANGE((c), 'j', 'r') \
1368 || inRANGE((c), 's', 'z')
1369# define isUPPER_A(c) inRANGE((c), 'A', 'I') \
1370 || inRANGE((c), 'J', 'R') \
1371 || inRANGE((c), 'S', 'Z')
a4d7a999
KW
1372# define isALPHA_A(c) (isUPPER_A(c) || isLOWER_A(c))
1373# define isALPHANUMERIC_A(c) (isALPHA_A(c) || isDIGIT_A(c))
3ded5eb0 1374# define isWORDCHAR_A(c) (isALPHANUMERIC_A(c) || (c) == '_')
0852beac 1375# define isIDFIRST_A(c) (isALPHA_A(c) || (c) == '_')
182c4ace
KW
1376# define isXDIGIT_A(c) ( isDIGIT_A(c) \
1377 || inRANGE((c), 'a', 'f') \
1378 || inRANGE((c), 'A', 'F')
74665a89
KW
1379# define isPUNCT_A(c) ((c) == '-' || (c) == '!' || (c) == '"' \
1380 || (c) == '#' || (c) == '$' || (c) == '%' \
1381 || (c) == '&' || (c) == '\'' || (c) == '(' \
1382 || (c) == ')' || (c) == '*' || (c) == '+' \
1383 || (c) == ',' || (c) == '.' || (c) == '/' \
1384 || (c) == ':' || (c) == ';' || (c) == '<' \
1385 || (c) == '=' || (c) == '>' || (c) == '?' \
1386 || (c) == '@' || (c) == '[' || (c) == '\\' \
1387 || (c) == ']' || (c) == '^' || (c) == '_' \
1388 || (c) == '`' || (c) == '{' || (c) == '|' \
1389 || (c) == '}' || (c) == '~')
1390# define isGRAPH_A(c) (isALPHANUMERIC_A(c) || isPUNCT_A(c))
1391# define isPRINT_A(c) (isGRAPH_A(c) || (c) == ' ')
3ded5eb0 1392
0852beac 1393# ifdef EBCDIC
74665a89
KW
1394 /* The below is accurate for the 3 EBCDIC code pages traditionally
1395 * supported by perl. The only difference between them in the controls
1396 * is the position of \n, and that is represented symbolically below */
1397# define isCNTRL_A(c) ((c) == '\0' || (c) == '\a' || (c) == '\b' \
1398 || (c) == '\f' || (c) == '\n' || (c) == '\r' \
1399 || (c) == '\t' || (c) == '\v' \
182c4ace 1400 || inRANGE((c), 1, 3) /* SOH, STX, ETX */ \
74665a89 1401 || (c) == 7 /* U+7F DEL */ \
182c4ace
KW
1402 || inRANGE((c), 0x0E, 0x13) /* SO SI DLE \
1403 DC[1-3] */ \
74665a89
KW
1404 || (c) == 0x18 /* U+18 CAN */ \
1405 || (c) == 0x19 /* U+19 EOM */ \
182c4ace 1406 || inRANGE((c), 0x1C, 0x1F) /* [FGRU]S */ \
74665a89
KW
1407 || (c) == 0x26 /* U+17 ETB */ \
1408 || (c) == 0x27 /* U+1B ESC */ \
1409 || (c) == 0x2D /* U+05 ENQ */ \
1410 || (c) == 0x2E /* U+06 ACK */ \
1411 || (c) == 0x32 /* U+16 SYN */ \
1412 || (c) == 0x37 /* U+04 EOT */ \
1413 || (c) == 0x3C /* U+14 DC4 */ \
1414 || (c) == 0x3D /* U+15 NAK */ \
1415 || (c) == 0x3F)/* U+1A SUB */
0852beac 1416# define isASCII(c) (isCNTRL_A(c) || isPRINT_A(c))
74665a89
KW
1417# else /* isASCII is already defined for ASCII platforms, so can use that to
1418 define isCNTRL */
1419# define isCNTRL_A(c) (isASCII(c) && ! isPRINT_A(c))
0852beac
KW
1420# endif
1421
3ffc8c70 1422 /* The _L1 macros may be unnecessary for the utilities; I (khw) added them
caa94d35
KW
1423 * during debugging, and it seems best to keep them. We may be called
1424 * without NATIVE_TO_LATIN1 being defined. On ASCII platforms, it doesn't
1425 * do anything anyway, so make it not a problem */
1426# if ! defined(EBCDIC) && ! defined(NATIVE_TO_LATIN1)
1427# define NATIVE_TO_LATIN1(ch) (ch)
1428# endif
3ded5eb0
KW
1429# define isALPHA_L1(c) (isUPPER_L1(c) || isLOWER_L1(c))
1430# define isALPHANUMERIC_L1(c) (isALPHA_L1(c) || isDIGIT_A(c))
1431# define isBLANK_L1(c) (isBLANK_A(c) \
1432 || (FITS_IN_8_BITS(c) \
1433 && NATIVE_TO_LATIN1((U8) c) == 0xA0))
1434# define isCNTRL_L1(c) (FITS_IN_8_BITS(c) && (! isPRINT_L1(c)))
1435# define isGRAPH_L1(c) (isPRINT_L1(c) && (! isBLANK_L1(c)))
1436# define isLOWER_L1(c) (isLOWER_A(c) \
1437 || (FITS_IN_8_BITS(c) \
ae683a5f 1438 && (( NATIVE_TO_LATIN1((U8) c) >= 0xDF \
3ded5eb0
KW
1439 && NATIVE_TO_LATIN1((U8) c) != 0xF7) \
1440 || NATIVE_TO_LATIN1((U8) c) == 0xAA \
1441 || NATIVE_TO_LATIN1((U8) c) == 0xBA \
1442 || NATIVE_TO_LATIN1((U8) c) == 0xB5)))
1443# define isPRINT_L1(c) (isPRINT_A(c) \
1444 || (FITS_IN_8_BITS(c) \
1445 && NATIVE_TO_LATIN1((U8) c) >= 0xA0))
3ded5eb0
KW
1446# define isPUNCT_L1(c) (isPUNCT_A(c) \
1447 || (FITS_IN_8_BITS(c) \
ae683a5f 1448 && ( NATIVE_TO_LATIN1((U8) c) == 0xA1 \
3ded5eb0
KW
1449 || NATIVE_TO_LATIN1((U8) c) == 0xA7 \
1450 || NATIVE_TO_LATIN1((U8) c) == 0xAB \
1451 || NATIVE_TO_LATIN1((U8) c) == 0xB6 \
1452 || NATIVE_TO_LATIN1((U8) c) == 0xB7 \
1453 || NATIVE_TO_LATIN1((U8) c) == 0xBB \
1454 || NATIVE_TO_LATIN1((U8) c) == 0xBF)))
1455# define isSPACE_L1(c) (isSPACE_A(c) \
1456 || (FITS_IN_8_BITS(c) \
ae683a5f 1457 && ( NATIVE_TO_LATIN1((U8) c) == 0x85 \
3ded5eb0
KW
1458 || NATIVE_TO_LATIN1((U8) c) == 0xA0)))
1459# define isUPPER_L1(c) (isUPPER_A(c) \
1460 || (FITS_IN_8_BITS(c) \
182c4ace
KW
1461 && ( IN_RANGE(NATIVE_TO_LATIN1((U8) c), \
1462 0xC0, 0xDE) \
3ded5eb0
KW
1463 && NATIVE_TO_LATIN1((U8) c) != 0xD7)))
1464# define isWORDCHAR_L1(c) (isIDFIRST_L1(c) || isDIGIT_A(c))
1465# define isIDFIRST_L1(c) (isALPHA_L1(c) || NATIVE_TO_LATIN1(c) == '_')
1466# define isCHARNAME_CONT(c) (isWORDCHAR_L1(c) \
1467 || isBLANK_L1(c) \
1468 || (c) == '-' \
1469 || (c) == '(' \
1470 || (c) == ')')
1471 /* The following are not fully accurate in the above-ASCII range. I (khw)
1472 * don't think it's necessary to be so for the purposes where this gets
1473 * compiled */
1474# define _isQUOTEMETA(c) (FITS_IN_8_BITS(c) && ! isWORDCHAR_L1(c))
1475# define _IS_IN_SOME_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) isALPHA_L1(c)
1476
1477 /* And these aren't accurate at all. They are useful only for above
1478 * Latin1, which utilities and bootstrapping don't deal with */
1479# define _IS_NON_FINAL_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) 0
6838b41e 1480# define _HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) 0
3ded5eb0
KW
1481# define _HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) 0
1482
1483 /* Many of the macros later in this file are defined in terms of these. By
1484 * implementing them with a function, which converts the class number into
1485 * a call to the desired macro, all of the later ones work. However, that
1486 * function won't be actually defined when building a utility program (no
1487 * perl.h), and so a compiler error will be generated if one is attempted
1488 * to be used. And the above-Latin1 code points require Unicode tables to
1489 * be present, something unlikely to be the case when bootstrapping */
1490# define _generic_isCC(c, classnum) \
1491 (FITS_IN_8_BITS(c) && S_bootstrap_ctype((U8) (c), (classnum), TRUE))
1492# define _generic_isCC_A(c, classnum) \
1493 (FITS_IN_8_BITS(c) && S_bootstrap_ctype((U8) (c), (classnum), FALSE))
687c8d01 1494#endif /* End of no perl.h H_PERL */
8a58bdcf 1495
e66b99e9
KW
1496#define isALPHANUMERIC(c) isALPHANUMERIC_A(c)
1497#define isALPHA(c) isALPHA_A(c)
0852beac
KW
1498#define isASCII_A(c) isASCII(c)
1499#define isASCII_L1(c) isASCII(c)
e66b99e9
KW
1500#define isBLANK(c) isBLANK_A(c)
1501#define isCNTRL(c) isCNTRL_A(c)
1502#define isDIGIT(c) isDIGIT_A(c)
1503#define isGRAPH(c) isGRAPH_A(c)
1504#define isIDFIRST(c) isIDFIRST_A(c)
1505#define isLOWER(c) isLOWER_A(c)
1506#define isPRINT(c) isPRINT_A(c)
779cf272 1507#define isPSXSPC_A(c) isSPACE_A(c)
e66b99e9 1508#define isPSXSPC(c) isPSXSPC_A(c)
779cf272 1509#define isPSXSPC_L1(c) isSPACE_L1(c)
e66b99e9
KW
1510#define isPUNCT(c) isPUNCT_A(c)
1511#define isSPACE(c) isSPACE_A(c)
1512#define isUPPER(c) isUPPER_A(c)
1513#define isWORDCHAR(c) isWORDCHAR_A(c)
1514#define isXDIGIT(c) isXDIGIT_A(c)
1515
1516/* ASCII casing. These could also be written as
1517 #define toLOWER(c) (isASCII(c) ? toLOWER_LATIN1(c) : (c))
1518 #define toUPPER(c) (isASCII(c) ? toUPPER_LATIN1_MOD(c) : (c))
1519 which uses table lookup and mask instead of subtraction. (This would
c5e9991e
KW
1520 work because the _MOD does not apply in the ASCII range).
1521
1522 These actually are UTF-8 invariant casing, not just ASCII, as any non-ASCII
1523 UTF-8 invariants are neither upper nor lower. (Only on EBCDIC platforms are
1524 there non-ASCII invariants, and all of them are controls.) */
68067e4e
DM
1525#define toLOWER(c) (isUPPER(c) ? (U8)((c) + ('a' - 'A')) : (c))
1526#define toUPPER(c) (isLOWER(c) ? (U8)((c) - ('a' - 'A')) : (c))
bbce6d69 1527
25200305
KW
1528/* In the ASCII range, these are equivalent to what they're here defined to be.
1529 * But by creating these definitions, other code doesn't have to be aware of
c5e9991e
KW
1530 * this detail. Actually this works for all UTF-8 invariants, not just the
1531 * ASCII range. (EBCDIC platforms can have non-ASCII invariants.) */
25200305 1532#define toFOLD(c) toLOWER(c)
25200305
KW
1533#define toTITLE(c) toUPPER(c)
1534
c753c8d3
KW
1535#define toLOWER_A(c) toLOWER(c)
1536#define toUPPER_A(c) toUPPER(c)
25200305
KW
1537#define toFOLD_A(c) toFOLD(c)
1538#define toTITLE_A(c) toTITLE(c)
1a0901db 1539
4650c663 1540/* Use table lookup for speed; returns the input itself if is out-of-range */
b2bf251f 1541#define toLOWER_LATIN1(c) ((! FITS_IN_8_BITS(c)) \
8e7c6e7d 1542 ? (c) \
f4cd282c 1543 : PL_latin1_lc[ (U8) (c) ])
c753c8d3
KW
1544#define toLOWER_L1(c) toLOWER_LATIN1(c) /* Synonym for consistency */
1545
1a0901db 1546/* Modified uc. Is correct uc except for three non-ascii chars which are
4650c663
KW
1547 * all mapped to one of them, and these need special handling; returns the
1548 * input itself if is out-of-range */
b2bf251f 1549#define toUPPER_LATIN1_MOD(c) ((! FITS_IN_8_BITS(c)) \
8e7c6e7d 1550 ? (c) \
f4cd282c 1551 : PL_mod_latin1_uc[ (U8) (c) ])
31f05a37 1552#define IN_UTF8_CTYPE_LOCALE PL_in_utf8_CTYPE_locale
84061b6a 1553
beab9ebe
KW
1554/* Use foo_LC_uvchr() instead of these for beyond the Latin1 range */
1555
1556/* For internal core Perl use only: the base macro for defining macros like
1557 * isALPHA_LC, which uses the current LC_CTYPE locale. 'c' is the code point
31f05a37
KW
1558 * (0-255) to check. In a UTF-8 locale, the result is the same as calling
1559 * isFOO_L1(); the 'utf8_locale_classnum' parameter is something like
1560 * _CC_UPPER, which gives the class number for doing this. For non-UTF-8
1561 * locales, the code to actually do the test this is passed in 'non_utf8'. If
1562 * 'c' is above 255, 0 is returned. For accessing the full range of possible
1563 * code points under locale rules, use the macros based on _generic_LC_uvchr
1564 * instead of this. */
beab9ebe
KW
1565#define _generic_LC_base(c, utf8_locale_classnum, non_utf8) \
1566 (! FITS_IN_8_BITS(c) \
1567 ? 0 \
31f05a37
KW
1568 : IN_UTF8_CTYPE_LOCALE \
1569 ? cBOOL(PL_charclass[(U8) (c)] & _CC_mask(utf8_locale_classnum)) \
beab9ebe
KW
1570 : cBOOL(non_utf8))
1571
1572/* For internal core Perl use only: a helper macro for defining macros like
1573 * isALPHA_LC. 'c' is the code point (0-255) to check. The function name to
1574 * actually do this test is passed in 'non_utf8_func', which is called on 'c',
1575 * casting 'c' to the macro _LC_CAST, which should not be parenthesized. See
1576 * _generic_LC_base for more info */
1577#define _generic_LC(c, utf8_locale_classnum, non_utf8_func) \
1578 _generic_LC_base(c,utf8_locale_classnum, \
1579 non_utf8_func( (_LC_CAST) (c)))
1580
1581/* For internal core Perl use only: like _generic_LC, but also returns TRUE if
1582 * 'c' is the platform's native underscore character */
1583#define _generic_LC_underscore(c,utf8_locale_classnum,non_utf8_func) \
1584 _generic_LC_base(c, utf8_locale_classnum, \
1585 (non_utf8_func( (_LC_CAST) (c)) \
1586 || (char)(c) == '_'))
1587
1588/* These next three are also for internal core Perl use only: case-change
247985d4
KW
1589 * helper macros. The reason for using the PL_latin arrays is in case the
1590 * system function is defective; it ensures uniform results that conform to the
b257a28c 1591 * Unicod standard. It does not handle the anomalies in UTF-8 Turkic locales */
beab9ebe
KW
1592#define _generic_toLOWER_LC(c, function, cast) (! FITS_IN_8_BITS(c) \
1593 ? (c) \
31f05a37
KW
1594 : (IN_UTF8_CTYPE_LOCALE) \
1595 ? PL_latin1_lc[ (U8) (c) ] \
5a10328c 1596 : (cast)function((cast)(c)))
beab9ebe 1597
31f05a37
KW
1598/* Note that the result can be larger than a byte in a UTF-8 locale. It
1599 * returns a single value, so can't adequately return the upper case of LATIN
1600 * SMALL LETTER SHARP S in a UTF-8 locale (which should be a string of two
1601 * values "SS"); instead it asserts against that under DEBUGGING, and
b257a28c
KW
1602 * otherwise returns its input. It does not handle the anomalies in UTF-8
1603 * Turkic locales. */
beab9ebe
KW
1604#define _generic_toUPPER_LC(c, function, cast) \
1605 (! FITS_IN_8_BITS(c) \
1606 ? (c) \
31f05a37 1607 : ((! IN_UTF8_CTYPE_LOCALE) \
b7d90381 1608 ? (cast)function((cast)(c)) \
31f05a37
KW
1609 : ((((U8)(c)) == MICRO_SIGN) \
1610 ? GREEK_CAPITAL_LETTER_MU \
1611 : ((((U8)(c)) == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS) \
1612 ? LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS \
1613 : ((((U8)(c)) == LATIN_SMALL_LETTER_SHARP_S) \
1614 ? (__ASSERT_(0) (c)) \
1615 : PL_mod_latin1_uc[ (U8) (c) ])))))
1616
1617/* Note that the result can be larger than a byte in a UTF-8 locale. It
1618 * returns a single value, so can't adequately return the fold case of LATIN
1619 * SMALL LETTER SHARP S in a UTF-8 locale (which should be a string of two
1620 * values "ss"); instead it asserts against that under DEBUGGING, and
b257a28c
KW
1621 * otherwise returns its input. It does not handle the anomalies in UTF-8
1622 * Turkic locales */
beab9ebe 1623#define _generic_toFOLD_LC(c, function, cast) \
4d7db1b9 1624 ((UNLIKELY((c) == MICRO_SIGN) && IN_UTF8_CTYPE_LOCALE) \
31f05a37 1625 ? GREEK_SMALL_LETTER_MU \
4d7db1b9
KW
1626 : (__ASSERT_(! IN_UTF8_CTYPE_LOCALE \
1627 || (c) != LATIN_SMALL_LETTER_SHARP_S) \
1628 _generic_toLOWER_LC(c, function, cast)))
beab9ebe 1629
84061b6a 1630/* Use the libc versions for these if available. */
f05550c0 1631#if defined(HAS_ISASCII)
84061b6a
KW
1632# define isASCII_LC(c) (FITS_IN_8_BITS(c) && isascii( (U8) (c)))
1633#else
1634# define isASCII_LC(c) isASCII(c)
1635#endif
1636
f05550c0 1637#if defined(HAS_ISBLANK)
84061b6a 1638# define isBLANK_LC(c) _generic_LC(c, _CC_BLANK, isblank)
95f34b6f 1639#else /* Unlike isASCII, varies if in a UTF-8 locale */
278b2b56 1640# define isBLANK_LC(c) ((IN_UTF8_CTYPE_LOCALE) ? isBLANK_L1(c) : isBLANK(c))
84061b6a
KW
1641#endif
1642
f05550c0 1643#define _LC_CAST U8
bbce6d69 1644
f05550c0 1645#ifdef WIN32
375f5f06
KW
1646 /* The Windows functions don't bother to follow the POSIX standard, which
1647 * for example says that something can't both be a printable and a control.
1648 * But Windows treats the \t control as a printable, and does such things
1649 * as making superscripts into both digits and punctuation. This tames
1650 * these flaws by assuming that the definitions of both controls and space
1651 * are correct, and then making sure that other definitions don't have
1652 * weirdnesses, by making sure that isalnum() isn't also ispunct(), etc.
1653 * Not all possible weirdnesses are checked for, just the ones that were
1654 * detected on actual Microsoft code pages */
1655
b7d90381
KW
1656# define isCNTRL_LC(c) _generic_LC(c, _CC_CNTRL, iscntrl)
1657# define isSPACE_LC(c) _generic_LC(c, _CC_SPACE, isspace)
1658
1659# define isALPHA_LC(c) (_generic_LC(c, _CC_ALPHA, isalpha) \
1660 && isALPHANUMERIC_LC(c))
1661# define isALPHANUMERIC_LC(c) (_generic_LC(c, _CC_ALPHANUMERIC, isalnum) && \
1662 ! isPUNCT_LC(c))
1663# define isDIGIT_LC(c) (_generic_LC(c, _CC_DIGIT, isdigit) && \
1664 isALPHANUMERIC_LC(c))
1665# define isGRAPH_LC(c) (_generic_LC(c, _CC_GRAPH, isgraph) && isPRINT_LC(c))
1666# define isIDFIRST_LC(c) (((c) == '_') \
1667 || (_generic_LC(c, _CC_IDFIRST, isalpha) && ! isPUNCT_LC(c)))
1668# define isLOWER_LC(c) (_generic_LC(c, _CC_LOWER, islower) && isALPHA_LC(c))
1669# define isPRINT_LC(c) (_generic_LC(c, _CC_PRINT, isprint) && ! isCNTRL_LC(c))
1670# define isPUNCT_LC(c) (_generic_LC(c, _CC_PUNCT, ispunct) && ! isCNTRL_LC(c))
1671# define isUPPER_LC(c) (_generic_LC(c, _CC_UPPER, isupper) && isALPHA_LC(c))
f05550c0 1672# define isWORDCHAR_LC(c) (((c) == '_') || isALPHANUMERIC_LC(c))
b7d90381
KW
1673# define isXDIGIT_LC(c) (_generic_LC(c, _CC_XDIGIT, isxdigit) \
1674 && isALPHANUMERIC_LC(c))
f05550c0
BF
1675
1676# define toLOWER_LC(c) _generic_toLOWER_LC((c), tolower, U8)
1677# define toUPPER_LC(c) _generic_toUPPER_LC((c), toupper, U8)
1678# define toFOLD_LC(c) _generic_toFOLD_LC((c), tolower, U8)
1679
1680#elif defined(CTYPE256) || (!defined(isascii) && !defined(HAS_ISASCII))
4650c663 1681 /* For most other platforms */
beab9ebe 1682
f05550c0
BF
1683# define isALPHA_LC(c) _generic_LC(c, _CC_ALPHA, isalpha)
1684# define isALPHANUMERIC_LC(c) _generic_LC(c, _CC_ALPHANUMERIC, isalnum)
1685# define isCNTRL_LC(c) _generic_LC(c, _CC_CNTRL, iscntrl)
1686# define isDIGIT_LC(c) _generic_LC(c, _CC_DIGIT, isdigit)
1687# define isGRAPH_LC(c) _generic_LC(c, _CC_GRAPH, isgraph)
1688# define isIDFIRST_LC(c) _generic_LC_underscore(c, _CC_IDFIRST, isalpha)
1689# define isLOWER_LC(c) _generic_LC(c, _CC_LOWER, islower)
1690# define isPRINT_LC(c) _generic_LC(c, _CC_PRINT, isprint)
1691# define isPUNCT_LC(c) _generic_LC(c, _CC_PUNCT, ispunct)
1692# define isSPACE_LC(c) _generic_LC(c, _CC_SPACE, isspace)
1693# define isUPPER_LC(c) _generic_LC(c, _CC_UPPER, isupper)
1694# define isWORDCHAR_LC(c) _generic_LC_underscore(c, _CC_WORDCHAR, isalnum)
1695# define isXDIGIT_LC(c) _generic_LC(c, _CC_XDIGIT, isxdigit)
1696
1697
1698# define toLOWER_LC(c) _generic_toLOWER_LC((c), tolower, U8)
1699# define toUPPER_LC(c) _generic_toUPPER_LC((c), toupper, U8)
1700# define toFOLD_LC(c) _generic_toFOLD_LC((c), tolower, U8)
1701
1702#else /* The final fallback position */
1703
b7d90381
KW
1704# define isALPHA_LC(c) (isascii(c) && isalpha(c))
1705# define isALPHANUMERIC_LC(c) (isascii(c) && isalnum(c))
1706# define isCNTRL_LC(c) (isascii(c) && iscntrl(c))
1707# define isDIGIT_LC(c) (isascii(c) && isdigit(c))
1708# define isGRAPH_LC(c) (isascii(c) && isgraph(c))
f05550c0 1709# define isIDFIRST_LC(c) (isascii(c) && (isalpha(c) || (c) == '_'))
b7d90381
KW
1710# define isLOWER_LC(c) (isascii(c) && islower(c))
1711# define isPRINT_LC(c) (isascii(c) && isprint(c))
1712# define isPUNCT_LC(c) (isascii(c) && ispunct(c))
1713# define isSPACE_LC(c) (isascii(c) && isspace(c))
1714# define isUPPER_LC(c) (isascii(c) && isupper(c))
f05550c0 1715# define isWORDCHAR_LC(c) (isascii(c) && (isalnum(c) || (c) == '_'))
b7d90381 1716# define isXDIGIT_LC(c) (isascii(c) && isxdigit(c))
f05550c0
BF
1717
1718# define toLOWER_LC(c) (isascii(c) ? tolower(c) : (c))
1719# define toUPPER_LC(c) (isascii(c) ? toupper(c) : (c))
1720# define toFOLD_LC(c) (isascii(c) ? tolower(c) : (c))
bbce6d69 1721
f05550c0 1722#endif
55204971 1723
eba68aa0
KW
1724#define isIDCONT(c) isWORDCHAR(c)
1725#define isIDCONT_A(c) isWORDCHAR_A(c)
1726#define isIDCONT_L1(c) isWORDCHAR_L1(c)
1727#define isIDCONT_LC(c) isWORDCHAR_LC(c)
13380643 1728#define isPSXSPC_LC(c) isSPACE_LC(c)
aaa51d5e 1729
4650c663 1730/* For internal core Perl use only: the base macros for defining macros like
d0da05db
KW
1731 * isALPHA_uvchr. 'c' is the code point to check. 'classnum' is the POSIX class
1732 * number defined earlier in this file. _generic_uvchr() is used for POSIX
4650c663
KW
1733 * classes where there is a macro or function 'above_latin1' that takes the
1734 * single argument 'c' and returns the desired value. These exist for those
1735 * classes which have simple definitions, avoiding the overhead of a hash
d0da05db 1736 * lookup or inversion list binary search. _generic_swash_uvchr() can be used
4650c663 1737 * for classes where that overhead is faster than a direct lookup.
d0da05db 1738 * _generic_uvchr() won't compile if 'c' isn't unsigned, as it won't match the
4650c663
KW
1739 * 'above_latin1' prototype. _generic_isCC() macro does bounds checking, so
1740 * have duplicate checks here, so could create versions of the macros that
1741 * don't, but experiments show that gcc optimizes them out anyway. */
66c17564
KW
1742
1743/* Note that all ignore 'use bytes' */
1e222e4f
KW
1744#define _generic_uvchr(classnum, above_latin1, c) ((c) < 256 \
1745 ? _generic_isCC(c, classnum) \
cd500f2f 1746 : above_latin1(c))
1e222e4f
KW
1747#define _generic_swash_uvchr(classnum, c) ((c) < 256 \
1748 ? _generic_isCC(c, classnum) \
922e8cb4 1749 : _is_uni_FOO(classnum, c))
d0da05db
KW
1750#define isALPHA_uvchr(c) _generic_swash_uvchr(_CC_ALPHA, c)
1751#define isALPHANUMERIC_uvchr(c) _generic_swash_uvchr(_CC_ALPHANUMERIC, c)
1752#define isASCII_uvchr(c) isASCII(c)
1753#define isBLANK_uvchr(c) _generic_uvchr(_CC_BLANK, is_HORIZWS_cp_high, c)
1754#define isCNTRL_uvchr(c) isCNTRL_L1(c) /* All controls are in Latin1 */
1755#define isDIGIT_uvchr(c) _generic_swash_uvchr(_CC_DIGIT, c)
1756#define isGRAPH_uvchr(c) _generic_swash_uvchr(_CC_GRAPH, c)
1e222e4f
KW
1757#define isIDCONT_uvchr(c) \
1758 _generic_uvchr(_CC_WORDCHAR, _is_uni_perl_idcont, c)
1759#define isIDFIRST_uvchr(c) \
1760 _generic_uvchr(_CC_IDFIRST, _is_uni_perl_idstart, c)
d0da05db
KW
1761#define isLOWER_uvchr(c) _generic_swash_uvchr(_CC_LOWER, c)
1762#define isPRINT_uvchr(c) _generic_swash_uvchr(_CC_PRINT, c)
1763
1764#define isPUNCT_uvchr(c) _generic_swash_uvchr(_CC_PUNCT, c)
1765#define isSPACE_uvchr(c) _generic_uvchr(_CC_SPACE, is_XPERLSPACE_cp_high, c)
1766#define isPSXSPC_uvchr(c) isSPACE_uvchr(c)
1767
1768#define isUPPER_uvchr(c) _generic_swash_uvchr(_CC_UPPER, c)
1769#define isVERTWS_uvchr(c) _generic_uvchr(_CC_VERTSPACE, is_VERTWS_cp_high, c)
1770#define isWORDCHAR_uvchr(c) _generic_swash_uvchr(_CC_WORDCHAR, c)
1771#define isXDIGIT_uvchr(c) _generic_uvchr(_CC_XDIGIT, is_XDIGIT_cp_high, c)
1772
1773#define toFOLD_uvchr(c,s,l) to_uni_fold(c,s,l)
1774#define toLOWER_uvchr(c,s,l) to_uni_lower(c,s,l)
1775#define toTITLE_uvchr(c,s,l) to_uni_title(c,s,l)
1776#define toUPPER_uvchr(c,s,l) to_uni_upper(c,s,l)
1777
1778/* For backwards compatibility, even though '_uni' should mean official Unicode
1779 * code points, in Perl it means native for those below 256 */
1780#define isALPHA_uni(c) isALPHA_uvchr(c)
1781#define isALPHANUMERIC_uni(c) isALPHANUMERIC_uvchr(c)
1782#define isASCII_uni(c) isASCII_uvchr(c)
1783#define isBLANK_uni(c) isBLANK_uvchr(c)
1784#define isCNTRL_uni(c) isCNTRL_uvchr(c)
1785#define isDIGIT_uni(c) isDIGIT_uvchr(c)
1786#define isGRAPH_uni(c) isGRAPH_uvchr(c)
1787#define isIDCONT_uni(c) isIDCONT_uvchr(c)
1788#define isIDFIRST_uni(c) isIDFIRST_uvchr(c)
1789#define isLOWER_uni(c) isLOWER_uvchr(c)
1790#define isPRINT_uni(c) isPRINT_uvchr(c)
1791#define isPUNCT_uni(c) isPUNCT_uvchr(c)
1792#define isSPACE_uni(c) isSPACE_uvchr(c)
1793#define isPSXSPC_uni(c) isPSXSPC_uvchr(c)
1794#define isUPPER_uni(c) isUPPER_uvchr(c)
1795#define isVERTWS_uni(c) isVERTWS_uvchr(c)
1796#define isWORDCHAR_uni(c) isWORDCHAR_uvchr(c)
1797#define isXDIGIT_uni(c) isXDIGIT_uvchr(c)
1798#define toFOLD_uni(c,s,l) toFOLD_uvchr(c,s,l)
1799#define toLOWER_uni(c,s,l) toLOWER_uvchr(c,s,l)
1800#define toTITLE_uni(c,s,l) toTITLE_uvchr(c,s,l)
1801#define toUPPER_uni(c,s,l) toUPPER_uvchr(c,s,l)
a0ed51b3 1802
4650c663
KW
1803/* For internal core Perl use only: the base macros for defining macros like
1804 * isALPHA_LC_uvchr. These are like isALPHA_LC, but the input can be any code
d0da05db 1805 * point, not just 0-255. Like _generic_uvchr, there are two versions, one for
4650c663 1806 * simple class definitions; the other for more complex. These are like
d0da05db 1807 * _generic_uvchr, so see it for more info. */
cd500f2f
KW
1808#define _generic_LC_uvchr(latin1, above_latin1, c) \
1809 (c < 256 ? latin1(c) : above_latin1(c))
1810#define _generic_LC_swash_uvchr(latin1, classnum, c) \
1811 (c < 256 ? latin1(c) : _is_uni_FOO(classnum, c))
1812
1813#define isALPHA_LC_uvchr(c) _generic_LC_swash_uvchr(isALPHA_LC, _CC_ALPHA, c)
1814#define isALPHANUMERIC_LC_uvchr(c) _generic_LC_swash_uvchr(isALPHANUMERIC_LC, \
1815 _CC_ALPHANUMERIC, c)
b7d90381
KW
1816#define isASCII_LC_uvchr(c) isASCII_LC(c)
1817#define isBLANK_LC_uvchr(c) _generic_LC_uvchr(isBLANK_LC, \
1818 is_HORIZWS_cp_high, c)
feeab5a9 1819#define isCNTRL_LC_uvchr(c) (c < 256 ? isCNTRL_LC(c) : 0)
cd500f2f
KW
1820#define isDIGIT_LC_uvchr(c) _generic_LC_swash_uvchr(isDIGIT_LC, _CC_DIGIT, c)
1821#define isGRAPH_LC_uvchr(c) _generic_LC_swash_uvchr(isGRAPH_LC, _CC_GRAPH, c)
b7d90381 1822#define isIDCONT_LC_uvchr(c) _generic_LC_uvchr(isIDCONT_LC, \
eba68aa0 1823 _is_uni_perl_idcont, c)
b7d90381 1824#define isIDFIRST_LC_uvchr(c) _generic_LC_uvchr(isIDFIRST_LC, \
cd500f2f
KW
1825 _is_uni_perl_idstart, c)
1826#define isLOWER_LC_uvchr(c) _generic_LC_swash_uvchr(isLOWER_LC, _CC_LOWER, c)
1827#define isPRINT_LC_uvchr(c) _generic_LC_swash_uvchr(isPRINT_LC, _CC_PRINT, c)
b7d90381 1828#define isPSXSPC_LC_uvchr(c) isSPACE_LC_uvchr(c)
cd500f2f 1829#define isPUNCT_LC_uvchr(c) _generic_LC_swash_uvchr(isPUNCT_LC, _CC_PUNCT, c)
b7d90381 1830#define isSPACE_LC_uvchr(c) _generic_LC_uvchr(isSPACE_LC, \
509fb054 1831 is_XPERLSPACE_cp_high, c)
cd500f2f 1832#define isUPPER_LC_uvchr(c) _generic_LC_swash_uvchr(isUPPER_LC, _CC_UPPER, c)
b7d90381 1833#define isWORDCHAR_LC_uvchr(c) _generic_LC_swash_uvchr(isWORDCHAR_LC, \
cd500f2f 1834 _CC_WORDCHAR, c)
b7d90381
KW
1835#define isXDIGIT_LC_uvchr(c) _generic_LC_uvchr(isXDIGIT_LC, \
1836 is_XDIGIT_cp_high, c)
e712593e 1837
b7d90381 1838#define isBLANK_LC_uni(c) isBLANK_LC_uvchr(UNI_TO_NATIVE(c))
aaa51d5e 1839
4650c663
KW
1840/* For internal core Perl use only: the base macros for defining macros like
1841 * isALPHA_utf8. These are like the earlier defined macros, but take an input
1842 * UTF-8 encoded string 'p'. If the input is in the Latin1 range, use
1843 * the Latin1 macro 'classnum' on 'p'. Otherwise use the value given by the
1844 * 'utf8' parameter. This relies on the fact that ASCII characters have the
1845 * same representation whether utf8 or not. Note that it assumes that the utf8
1846 * has been validated, and ignores 'use bytes' */
34aeb2e9
KW
1847#define _base_generic_utf8(enum_name, name, p, use_locale ) \
1848 _is_utf8_FOO(CAT2(_CC_, enum_name), \
1849 (const U8 *) p, \
1850 "is" STRINGIFY(name) "_utf8", \
1851 "is" STRINGIFY(name) "_utf8_safe", \
1852 1, use_locale, __FILE__,__LINE__)
1853
1854#define _generic_utf8(name, p) _base_generic_utf8(name, name, p, 0)
da8c1a98
KW
1855
1856/* The "_safe" macros make sure that we don't attempt to read beyond 'e', but
1857 * they don't otherwise go out of their way to look for malformed UTF-8. If
1858 * they can return accurate results without knowing if the input is otherwise
1859 * malformed, they do so. For example isASCII is accurate in spite of any
1860 * non-length malformations because it looks only at a single byte. Likewise
1861 * isDIGIT looks just at the first byte for code points 0-255, as all UTF-8
1862 * variant ones return FALSE. But, if the input has to be well-formed in order
1863 * for the results to be accurate, the macros will test and if malformed will
1864 * call a routine to die
1865 *
1866 * Except for toke.c, the macros do assume that e > p, asserting that on
1867 * DEBUGGING builds. Much code that calls these depends on this being true,
1868 * for other reasons. toke.c is treated specially as using the regular
1869 * assertion breaks it in many ways. All strings that these operate on there
1870 * are supposed to have an extra NUL character at the end, so that *e = \0. A
1871 * bunch of code in toke.c assumes that this is true, so the assertion allows
1872 * for that */
1873#ifdef PERL_IN_TOKE_C
1874# define _utf8_safe_assert(p,e) ((e) > (p) || ((e) == (p) && *(p) == '\0'))
1875#else
1876# define _utf8_safe_assert(p,e) ((e) > (p))
1877#endif
1878
1879#define _generic_utf8_safe(classnum, p, e, above_latin1) \
1880 (__ASSERT_(_utf8_safe_assert(p, e)) \
1881 (UTF8_IS_INVARIANT(*(p))) \
1882 ? _generic_isCC(*(p), classnum) \
1883 : (UTF8_IS_DOWNGRADEABLE_START(*(p)) \
1884 ? ((LIKELY((e) - (p) > 1 && UTF8_IS_CONTINUATION(*((p)+1)))) \
1885 ? _generic_isCC(EIGHT_BIT_UTF8_TO_NATIVE(*(p), *((p)+1 )), \
1886 classnum) \
1887 : (_force_out_malformed_utf8_message( \
1888 (U8 *) (p), (U8 *) (e), 0, 1), 0)) \
1889 : above_latin1))
b7d90381
KW
1890/* Like the above, but calls 'above_latin1(p)' to get the utf8 value.
1891 * 'above_latin1' can be a macro */
da8c1a98
KW
1892#define _generic_func_utf8_safe(classnum, above_latin1, p, e) \
1893 _generic_utf8_safe(classnum, p, e, above_latin1(p, e))
1894#define _generic_non_swash_utf8_safe(classnum, above_latin1, p, e) \
1895 _generic_utf8_safe(classnum, p, e, \
1896 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
1897 ? (_force_out_malformed_utf8_message( \
1898 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
1899 : above_latin1(p)))
4650c663 1900/* Like the above, but passes classnum to _isFOO_utf8(), instead of having an
cd500f2f 1901 * 'above_latin1' parameter */
da8c1a98
KW
1902#define _generic_swash_utf8_safe(classnum, p, e) \
1903_generic_utf8_safe(classnum, p, e, _is_utf8_FOO_with_len(classnum, p, e))
922e8cb4 1904
cc8ab7c0 1905/* Like the above, but should be used only when it is known that there are no
ff7ecfc3
KW
1906 * characters in the upper-Latin1 range (128-255 on ASCII platforms) which the
1907 * class is TRUE for. Hence it can skip the tests for this range.
1908 * 'above_latin1' should include its arguments */
da8c1a98
KW
1909#define _generic_utf8_safe_no_upper_latin1(classnum, p, e, above_latin1) \
1910 (__ASSERT_(_utf8_safe_assert(p, e)) \
1911 (UTF8_IS_INVARIANT(*(p))) \
1912 ? _generic_isCC(*(p), classnum) \
1913 : (UTF8_IS_DOWNGRADEABLE_START(*(p))) \
1914 ? 0 /* Note that doesn't check validity for latin1 */ \
1915 : above_latin1)
1916
84238efa 1917
34aeb2e9
KW
1918#define isALPHA_utf8(p) _generic_utf8(ALPHA, p)
1919#define isALPHANUMERIC_utf8(p) _generic_utf8(ALPHANUMERIC, p)
1920#define isASCII_utf8(p) _generic_utf8(ASCII, p)
1921#define isBLANK_utf8(p) _generic_utf8(BLANK, p)
1922#define isCNTRL_utf8(p) _generic_utf8(CNTRL, p)
1923#define isDIGIT_utf8(p) _generic_utf8(DIGIT, p)
1924#define isGRAPH_utf8(p) _generic_utf8(GRAPH, p)
1925#define isIDCONT_utf8(p) _generic_utf8(IDCONT, p)
1926#define isIDFIRST_utf8(p) _generic_utf8(IDFIRST, p)
1927#define isLOWER_utf8(p) _generic_utf8(LOWER, p)
1928#define isPRINT_utf8(p) _generic_utf8(PRINT, p)
1929#define isPSXSPC_utf8(p) _generic_utf8(PSXSPC, p)
1930#define isPUNCT_utf8(p) _generic_utf8(PUNCT, p)
1931#define isSPACE_utf8(p) _generic_utf8(SPACE, p)
1932#define isUPPER_utf8(p) _generic_utf8(UPPER, p)
1933#define isVERTWS_utf8(p) _generic_utf8(VERTSPACE, p)
1934#define isWORDCHAR_utf8(p) _generic_utf8(WORDCHAR, p)
1935#define isXDIGIT_utf8(p) _generic_utf8(XDIGIT, p)
e8fa43e2 1936
da8c1a98
KW
1937#define isALPHA_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_ALPHA, p, e)
1938#define isALPHANUMERIC_utf8_safe(p, e) \
1939 _generic_swash_utf8_safe(_CC_ALPHANUMERIC, p, e)
1940#define isASCII_utf8_safe(p, e) \
1941 /* Because ASCII is invariant under utf8, the non-utf8 macro \
1942 * works */ \
1943 (__ASSERT_(_utf8_safe_assert(p, e)) isASCII(*(p)))
1944#define isBLANK_utf8_safe(p, e) \
1945 _generic_non_swash_utf8_safe(_CC_BLANK, is_HORIZWS_high, p, e)
1946
e8fa43e2
KW
1947#ifdef EBCDIC
1948 /* Because all controls are UTF-8 invariants in EBCDIC, we can use this
1949 * more efficient macro instead of the more general one */
da8c1a98 1950# define isCNTRL_utf8_safe(p, e) \
56d02b8c 1951 (__ASSERT_(_utf8_safe_assert(p, e)) isCNTRL_L1(*(p)))
e8fa43e2 1952#else
da8c1a98 1953# define isCNTRL_utf8_safe(p, e) _generic_utf8_safe(_CC_CNTRL, p, e, 0)
e8fa43e2
KW
1954#endif
1955
da8c1a98
KW
1956#define isDIGIT_utf8_safe(p, e) \
1957 _generic_utf8_safe_no_upper_latin1(_CC_DIGIT, p, e, \
1958 _is_utf8_FOO_with_len(_CC_DIGIT, p, e))
1959#define isGRAPH_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_GRAPH, p, e)
1960#define isIDCONT_utf8_safe(p, e) _generic_func_utf8_safe(_CC_WORDCHAR, \
1961 _is_utf8_perl_idcont_with_len, p, e)
e5dcd934 1962
c11ff943
KW
1963/* To prevent S_scan_word in toke.c from hanging, we have to make sure that
1964 * IDFIRST is an alnum. See
c8362b00 1965 * http://rt.perl.org/rt3/Ticket/Display.html?id=74022 for more detail than you
f91dcd13
KW
1966 * ever wanted to know about. (In the ASCII range, there isn't a difference.)
1967 * This used to be not the XID version, but we decided to go with the more
1968 * modern Unicode definition */
da8c1a98
KW
1969#define isIDFIRST_utf8_safe(p, e) \
1970 _generic_func_utf8_safe(_CC_IDFIRST, \
1971 _is_utf8_perl_idstart_with_len, (U8 *) (p), (U8 *) (e))
1972
1973#define isLOWER_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_LOWER, p, e)
1974#define isPRINT_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_PRINT, p, e)
1975#define isPSXSPC_utf8_safe(p, e) isSPACE_utf8_safe(p, e)
1976#define isPUNCT_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_PUNCT, p, e)
1977#define isSPACE_utf8_safe(p, e) \
1978 _generic_non_swash_utf8_safe(_CC_SPACE, is_XPERLSPACE_high, p, e)
1979#define isUPPER_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_UPPER, p, e)
1980#define isVERTWS_utf8_safe(p, e) \
1981 _generic_non_swash_utf8_safe(_CC_VERTSPACE, is_VERTWS_high, p, e)
1982#define isWORDCHAR_utf8_safe(p, e) \
1983 _generic_swash_utf8_safe(_CC_WORDCHAR, p, e)
1984#define isXDIGIT_utf8_safe(p, e) \
1985 _generic_utf8_safe_no_upper_latin1(_CC_XDIGIT, p, e, \
1986 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
1987 ? (_force_out_malformed_utf8_message( \
1988 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
1989 : is_XDIGIT_high(p)))
a0ed51b3 1990
7d2b61dd 1991#define toFOLD_utf8(p,s,l) to_utf8_fold(p,s,l)
509fb054
KW
1992#define toLOWER_utf8(p,s,l) to_utf8_lower(p,s,l)
1993#define toTITLE_utf8(p,s,l) to_utf8_title(p,s,l)
1994#define toUPPER_utf8(p,s,l) to_utf8_upper(p,s,l)
2e8adce6 1995
567b353c 1996/* For internal core use only, subject to change */
607313a1
KW
1997#define _toFOLD_utf8_flags(p,e,s,l,f) _to_utf8_fold_flags (p,e,s,l,f, "", 0)
1998#define _toLOWER_utf8_flags(p,e,s,l,f) _to_utf8_lower_flags(p,e,s,l,f, "", 0)
1999#define _toTITLE_utf8_flags(p,e,s,l,f) _to_utf8_title_flags(p,e,s,l,f, "", 0)
2000#define _toUPPER_utf8_flags(p,e,s,l,f) _to_utf8_upper_flags(p,e,s,l,f, "", 0)
a1a5ec35
KW
2001
2002#define toFOLD_utf8_safe(p,e,s,l) _toFOLD_utf8_flags(p,e,s,l, FOLD_FLAGS_FULL)
2003#define toLOWER_utf8_safe(p,e,s,l) _toLOWER_utf8_flags(p,e,s,l, 0)
2004#define toTITLE_utf8_safe(p,e,s,l) _toTITLE_utf8_flags(p,e,s,l, 0)
2005#define toUPPER_utf8_safe(p,e,s,l) _toUPPER_utf8_flags(p,e,s,l, 0)
567b353c 2006
4650c663
KW
2007/* For internal core Perl use only: the base macros for defining macros like
2008 * isALPHA_LC_utf8. These are like _generic_utf8, but if the first code point
2009 * in 'p' is within the 0-255 range, it uses locale rules from the passed-in
2010 * 'macro' parameter */
34aeb2e9
KW
2011#define _generic_LC_utf8(name, p) _base_generic_utf8(name, name, p, 1)
2012
2013#define isALPHA_LC_utf8(p) _generic_LC_utf8(ALPHA, p)
2014#define isALPHANUMERIC_LC_utf8(p) _generic_LC_utf8(ALPHANUMERIC, p)
2015#define isASCII_LC_utf8(p) _generic_LC_utf8(ASCII, p)
2016#define isBLANK_LC_utf8(p) _generic_LC_utf8(BLANK, p)
2017#define isCNTRL_LC_utf8(p) _generic_LC_utf8(CNTRL, p)
2018#define isDIGIT_LC_utf8(p) _generic_LC_utf8(DIGIT, p)
2019#define isGRAPH_LC_utf8(p) _generic_LC_utf8(GRAPH, p)
2020#define isIDCONT_LC_utf8(p) _generic_LC_utf8(IDCONT, p)
2021#define isIDFIRST_LC_utf8(p) _generic_LC_utf8(IDFIRST, p)
2022#define isLOWER_LC_utf8(p) _generic_LC_utf8(LOWER, p)
2023#define isPRINT_LC_utf8(p) _generic_LC_utf8(PRINT, p)
2024#define isPSXSPC_LC_utf8(p) _generic_LC_utf8(PSXSPC, p)
2025#define isPUNCT_LC_utf8(p) _generic_LC_utf8(PUNCT, p)
2026#define isSPACE_LC_utf8(p) _generic_LC_utf8(SPACE, p)
2027#define isUPPER_LC_utf8(p) _generic_LC_utf8(UPPER, p)
2028#define isWORDCHAR_LC_utf8(p) _generic_LC_utf8(WORDCHAR, p)
2029#define isXDIGIT_LC_utf8(p) _generic_LC_utf8(XDIGIT, p)
2030
da8c1a98
KW
2031/* For internal core Perl use only: the base macros for defining macros like
2032 * isALPHA_LC_utf8_safe. These are like _generic_utf8, but if the first code
2033 * point in 'p' is within the 0-255 range, it uses locale rules from the
2034 * passed-in 'macro' parameter */
2035#define _generic_LC_utf8_safe(macro, p, e, above_latin1) \
2036 (__ASSERT_(_utf8_safe_assert(p, e)) \
2037 (UTF8_IS_INVARIANT(*(p))) \
2038 ? macro(*(p)) \
2039 : (UTF8_IS_DOWNGRADEABLE_START(*(p)) \
2040 ? ((LIKELY((e) - (p) > 1 && UTF8_IS_CONTINUATION(*((p)+1)))) \
2041 ? macro(EIGHT_BIT_UTF8_TO_NATIVE(*(p), *((p)+1))) \
2042 : (_force_out_malformed_utf8_message( \
2043 (U8 *) (p), (U8 *) (e), 0, 1), 0)) \
2044 : above_latin1))
2045
2046#define _generic_LC_swash_utf8_safe(macro, classnum, p, e) \
2047 _generic_LC_utf8_safe(macro, p, e, \
2048 _is_utf8_FOO_with_len(classnum, p, e))
2049
2050#define _generic_LC_func_utf8_safe(macro, above_latin1, p, e) \
2051 _generic_LC_utf8_safe(macro, p, e, above_latin1(p, e))
2052
2053#define _generic_LC_non_swash_utf8_safe(classnum, above_latin1, p, e) \
2054 _generic_LC_utf8_safe(classnum, p, e, \
2055 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
2056 ? (_force_out_malformed_utf8_message( \
2057 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
2058 : above_latin1(p)))
2059
2060#define isALPHANUMERIC_LC_utf8_safe(p, e) \
2061 _generic_LC_swash_utf8_safe(isALPHANUMERIC_LC, \
2062 _CC_ALPHANUMERIC, p, e)
2063#define isALPHA_LC_utf8_safe(p, e) \
2064 _generic_LC_swash_utf8_safe(isALPHA_LC, _CC_ALPHA, p, e)
2065#define isASCII_LC_utf8_safe(p, e) \
2066 (__ASSERT_(_utf8_safe_assert(p, e)) isASCII_LC(*(p)))
2067#define isBLANK_LC_utf8_safe(p, e) \
2068 _generic_LC_non_swash_utf8_safe(isBLANK_LC, is_HORIZWS_high, p, e)
2069#define isCNTRL_LC_utf8_safe(p, e) \
2070 _generic_LC_utf8_safe(isCNTRL_LC, p, e, 0)
2071#define isDIGIT_LC_utf8_safe(p, e) \
2072 _generic_LC_swash_utf8_safe(isDIGIT_LC, _CC_DIGIT, p, e)
2073#define isGRAPH_LC_utf8_safe(p, e) \
2074 _generic_LC_swash_utf8_safe(isGRAPH_LC, _CC_GRAPH, p, e)
2075#define isIDCONT_LC_utf8_safe(p, e) \
2076 _generic_LC_func_utf8_safe(isIDCONT_LC, \
2077 _is_utf8_perl_idcont_with_len, p, e)
2078#define isIDFIRST_LC_utf8_safe(p, e) \
2079 _generic_LC_func_utf8_safe(isIDFIRST_LC, \
2080 _is_utf8_perl_idstart_with_len, p, e)
2081#define isLOWER_LC_utf8_safe(p, e) \
2082 _generic_LC_swash_utf8_safe(isLOWER_LC, _CC_LOWER, p, e)
2083#define isPRINT_LC_utf8_safe(p, e) \
2084 _generic_LC_swash_utf8_safe(isPRINT_LC, _CC_PRINT, p, e)
2085#define isPSXSPC_LC_utf8_safe(p, e) isSPACE_LC_utf8_safe(p, e)
2086#define isPUNCT_LC_utf8_safe(p, e) \
2087 _generic_LC_swash_utf8_safe(isPUNCT_LC, _CC_PUNCT, p, e)
2088#define isSPACE_LC_utf8_safe(p, e) \
2089 _generic_LC_non_swash_utf8_safe(isSPACE_LC, is_XPERLSPACE_high, p, e)
2090#define isUPPER_LC_utf8_safe(p, e) \
2091 _generic_LC_swash_utf8_safe(isUPPER_LC, _CC_UPPER, p, e)
2092#define isWORDCHAR_LC_utf8_safe(p, e) \
2093 _generic_LC_swash_utf8_safe(isWORDCHAR_LC, _CC_WORDCHAR, p, e)
2094#define isXDIGIT_LC_utf8_safe(p, e) \
2095 _generic_LC_non_swash_utf8_safe(isXDIGIT_LC, is_XDIGIT_high, p, e)
aaa51d5e 2096
fbc19f27
KW
2097/* Macros for backwards compatibility and for completeness when the ASCII and
2098 * Latin1 values are identical */
b7d90381
KW
2099#define isALPHAU(c) isALPHA_L1(c)
2100#define isDIGIT_L1(c) isDIGIT_A(c)
2101#define isOCTAL(c) isOCTAL_A(c)
2102#define isOCTAL_L1(c) isOCTAL_A(c)
2103#define isXDIGIT_L1(c) isXDIGIT_A(c)
2104#define isALNUM(c) isWORDCHAR(c)
a377c856 2105#define isALNUM_A(c) isALNUM(c)
b7d90381
KW
2106#define isALNUMU(c) isWORDCHAR_L1(c)
2107#define isALNUM_LC(c) isWORDCHAR_LC(c)
2108#define isALNUM_uni(c) isWORDCHAR_uni(c)
2e28f0b9 2109#define isALNUM_LC_uvchr(c) isWORDCHAR_LC_uvchr(c)
b7d90381
KW
2110#define isALNUM_utf8(p) isWORDCHAR_utf8(p)
2111#define isALNUM_LC_utf8(p) isWORDCHAR_LC_utf8(p)
2112#define isALNUMC_A(c) isALPHANUMERIC_A(c) /* Mnemonic: "C's alnum" */
2113#define isALNUMC_L1(c) isALPHANUMERIC_L1(c)
2114#define isALNUMC(c) isALPHANUMERIC(c)
2115#define isALNUMC_LC(c) isALPHANUMERIC_LC(c)
2116#define isALNUMC_uni(c) isALPHANUMERIC_uni(c)
15861f94 2117#define isALNUMC_LC_uvchr(c) isALPHANUMERIC_LC_uvchr(c)
b7d90381 2118#define isALNUMC_utf8(p) isALPHANUMERIC_utf8(p)
15861f94 2119#define isALNUMC_LC_utf8(p) isALPHANUMERIC_LC_utf8(p)
fbc19f27 2120
2bd1cbf6
KW
2121/* On EBCDIC platforms, CTRL-@ is 0, CTRL-A is 1, etc, just like on ASCII,
2122 * except that they don't necessarily mean the same characters, e.g. CTRL-D is
2123 * 4 on both systems, but that is EOT on ASCII; ST on EBCDIC.
2124 * '?' is special-cased on EBCDIC to APC, which is the control there that is
2125 * the outlier from the block that contains the other controls, just like
2126 * toCTRL('?') on ASCII yields DEL, the control that is the outlier from the C0
2127 * block. If it weren't special cased, it would yield a non-control.
88794300
KW
2128 * The conversion works both ways, so toCTRL('D') is 4, and toCTRL(4) is D,
2129 * etc. */
2bd1cbf6 2130#ifndef EBCDIC
75763b3a 2131# define toCTRL(c) (__ASSERT_(FITS_IN_8_BITS(c)) toUPPER(((U8)(c))) ^ 64)
2bd1cbf6 2132#else
75763b3a
KW
2133# define toCTRL(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
2134 ((isPRINT_A(c)) \
2135 ? (UNLIKELY((c) == '?') \
2136 ? QUESTION_MARK_CTRL \
2137 : (NATIVE_TO_LATIN1(toUPPER((U8) (c))) ^ 64)) \
2138 : (UNLIKELY((c) == QUESTION_MARK_CTRL) \
2139 ? '?' \
2140 : (LATIN1_TO_NATIVE(((U8) (c)) ^ 64)))))
2bd1cbf6 2141#endif
bbce6d69 2142
dea28490
JJ
2143/* Line numbers are unsigned, 32 bits. */
2144typedef U32 line_t;
e5dcd934 2145#define NOLINE ((line_t) 4294967295UL) /* = FFFFFFFF */
378cc40b 2146
91152fc1
DG
2147/* Helpful alias for version prescan */
2148#define is_LAX_VERSION(a,b) \
2149 (a != Perl_prescan_version(aTHX_ a, FALSE, b, NULL, NULL, NULL, NULL))
2150
2151#define is_STRICT_VERSION(a,b) \
2152 (a != Perl_prescan_version(aTHX_ a, TRUE, b, NULL, NULL, NULL, NULL))
2153
2154#define BADVERSION(a,b,c) \
2155 if (b) { \
2156 *b = c; \
2157 } \
2158 return a;
8c52afec 2159
cb27eebd
KW
2160/* Converts a character known to represent a hexadecimal digit (0-9, A-F, or
2161 * a-f) to its numeric value. READ_XDIGIT's argument is a string pointer,
2162 * which is advanced. The input is validated only by an assert() in DEBUGGING
2163 * builds. In both ASCII and EBCDIC the last 4 bits of the digits are 0-9; and
2164 * the last 4 bits of A-F and a-f are 1-6, so adding 9 yields 10-15 */
2165#define XDIGIT_VALUE(c) (__ASSERT_(isXDIGIT(c)) (0xf & (isDIGIT(c) \
2166 ? (c) \
2167 : ((c) + 9))))
902008b8
KW
2168#define READ_XDIGIT(s) (__ASSERT_(isXDIGIT(*s)) (0xf & (isDIGIT(*(s)) \
2169 ? (*(s)++) \
2170 : (*(s)++ + 9))))
95a59cab 2171
cb27eebd
KW
2172/* Converts a character known to represent an octal digit (0-7) to its numeric
2173 * value. The input is validated only by an assert() in DEBUGGING builds. In
2174 * both ASCII and EBCDIC the last 3 bits of the octal digits range from 0-7. */
2175#define OCTAL_VALUE(c) (__ASSERT_(isOCTAL(c)) (7 & (c)))
2176
305b8651
KW
2177/* Efficiently returns a boolean as to if two native characters are equivalent
2178 * case-insenstively. At least one of the characters must be one of [A-Za-z];
2179 * the ALPHA in the name is to remind you of that. This is asserted() in
2180 * DEBUGGING builds. Because [A-Za-z] are invariant under UTF-8, this macro
2181 * works (on valid input) for both non- and UTF-8-encoded bytes.
2182 *
2183 * When one of the inputs is a compile-time constant and gets folded by the
2184 * compiler, this reduces to an AND and a TEST. On both EBCDIC and ASCII
2185 * machines, 'A' and 'a' differ by a single bit; the same with the upper and
2186 * lower case of all other ASCII-range alphabetics. On ASCII platforms, they
96ca48da
KW
2187 * are 32 apart; on EBCDIC, they are 64. At compile time, this uses an
2188 * exclusive 'or' to find that bit and then inverts it to form a mask, with
2189 * just a single 0, in the bit position where the upper- and lowercase differ.
2190 * */
305b8651
KW
2191#define isALPHA_FOLD_EQ(c1, c2) \
2192 (__ASSERT_(isALPHA_A(c1) || isALPHA_A(c2)) \
2193 ((c1) & ~('A' ^ 'a')) == ((c2) & ~('A' ^ 'a')))
2194#define isALPHA_FOLD_NE(c1, c2) (! isALPHA_FOLD_EQ((c1), (c2)))
2195
8e84507e 2196/*
ccfc67b7
JH
2197=head1 Memory Management
2198
a02a5408 2199=for apidoc Am|void|Newx|void* ptr|int nitems|type
954c1994
GS
2200The XSUB-writer's interface to the C C<malloc> function.
2201
596f7718 2202Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2203
c5008215
JC
2204In 5.9.3, Newx() and friends replace the older New() API, and drops
2205the first parameter, I<x>, a debug aid which allowed callers to identify
37b8b4c9 2206themselves. This aid has been superseded by a new build option,
d10b4965 2207PERL_MEM_LOG (see L<perlhacktips/PERL_MEM_LOG>). The older API is still
c5008215
JC
2208there for use in XS modules supporting older perls.
2209
a02a5408 2210=for apidoc Am|void|Newxc|void* ptr|int nitems|type|cast
954c1994 2211The XSUB-writer's interface to the C C<malloc> function, with
fbe13c60 2212cast. See also C<L</Newx>>.
954c1994 2213
596f7718 2214Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2215
a02a5408 2216=for apidoc Am|void|Newxz|void* ptr|int nitems|type
954c1994 2217The XSUB-writer's interface to the C C<malloc> function. The allocated
fbe13c60 2218memory is zeroed with C<memzero>. See also C<L</Newx>>.
a02a5408 2219
596f7718 2220Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2221
954c1994
GS
2222=for apidoc Am|void|Renew|void* ptr|int nitems|type
2223The XSUB-writer's interface to the C C<realloc> function.
2224
596f7718 2225Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2226
954c1994
GS
2227=for apidoc Am|void|Renewc|void* ptr|int nitems|type|cast
2228The XSUB-writer's interface to the C C<realloc> function, with
2229cast.
2230
596f7718 2231Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2232
49b8b560 2233=for apidoc Am|void|Safefree|void* ptr
954c1994
GS
2234The XSUB-writer's interface to the C C<free> function.
2235
596f7718 2236This should B<ONLY> be used on memory obtained using L</"Newx"> and friends.
0d7b2759 2237
954c1994
GS
2238=for apidoc Am|void|Move|void* src|void* dest|int nitems|type
2239The XSUB-writer's interface to the C C<memmove> function. The C<src> is the
926bb54c 2240source, C<dest> is the destination, C<nitems> is the number of items, and
fbe13c60 2241C<type> is the type. Can do overlapping moves. See also C<L</Copy>>.
954c1994 2242
e90e2364 2243=for apidoc Am|void *|MoveD|void* src|void* dest|int nitems|type
796b6530 2244Like C<Move> but returns C<dest>. Useful
72d33970 2245for encouraging compilers to tail-call
e90e2364
NC
2246optimise.
2247
954c1994
GS
2248=for apidoc Am|void|Copy|void* src|void* dest|int nitems|type
2249The XSUB-writer's interface to the C C<memcpy> function. The C<src> is the
926bb54c 2250source, C<dest> is the destination, C<nitems> is the number of items, and
fbe13c60 2251C<type> is the type. May fail on overlapping copies. See also C<L</Move>>.
954c1994 2252
e90e2364
NC
2253=for apidoc Am|void *|CopyD|void* src|void* dest|int nitems|type
2254
796b6530 2255Like C<Copy> but returns C<dest>. Useful
72d33970 2256for encouraging compilers to tail-call
e90e2364
NC
2257optimise.
2258
954c1994
GS
2259=for apidoc Am|void|Zero|void* dest|int nitems|type
2260
2261The XSUB-writer's interface to the C C<memzero> function. The C<dest> is the
2262destination, C<nitems> is the number of items, and C<type> is the type.
2263
e90e2364
NC
2264=for apidoc Am|void *|ZeroD|void* dest|int nitems|type
2265
72d33970
FC
2266Like C<Zero> but returns dest. Useful
2267for encouraging compilers to tail-call
e90e2364
NC
2268optimise.
2269
da5d8dbb 2270=for apidoc Am|void|StructCopy|type *src|type *dest|type
4375e838 2271This is an architecture-independent macro to copy one structure to another.
954c1994 2272
7e337ee0
JH
2273=for apidoc Am|void|PoisonWith|void* dest|int nitems|type|U8 byte
2274
2275Fill up memory with a byte pattern (a byte repeated over and over
2276again) that hopefully catches attempts to access uninitialized memory.
2277
2278=for apidoc Am|void|PoisonNew|void* dest|int nitems|type
2279
2280PoisonWith(0xAB) for catching access to allocated but uninitialized memory.
2281
1c12ffb4 2282=for apidoc Am|void|PoisonFree|void* dest|int nitems|type
7e337ee0
JH
2283
2284PoisonWith(0xEF) for catching access to freed memory.
2285
9965345d
JH
2286=for apidoc Am|void|Poison|void* dest|int nitems|type
2287
7e337ee0 2288PoisonWith(0xEF) for catching access to freed memory.
9965345d
JH
2289
2290=cut */
954c1994 2291
561b68a9
SH
2292/* Maintained for backwards-compatibility only. Use newSV() instead. */
2293#ifndef PERL_CORE
ff06c60c 2294#define NEWSV(x,len) newSV(len)
561b68a9 2295#endif
ff06c60c 2296
b7112dce 2297#define MEM_SIZE_MAX ((MEM_SIZE)-1)
19a94d75 2298
a500027b 2299#define _PERL_STRLEN_ROUNDUP_UNCHECKED(n) (((n) - 1 + PERL_STRLEN_ROUNDUP_QUANTUM) & ~((MEM_SIZE)PERL_STRLEN_ROUNDUP_QUANTUM - 1))
e6bdf523 2300
27d5b266 2301#ifdef PERL_MALLOC_WRAP
e6bdf523
DM
2302
2303/* This expression will be constant-folded at compile time. It checks
2304 * whether or not the type of the count n is so small (e.g. U8 or U16, or
2305 * U32 on 64-bit systems) that there's no way a wrap-around could occur.
2306 * As well as avoiding the need for a run-time check in some cases, it's
2307 * designed to avoid compiler warnings like:
2308 * comparison is always false due to limited range of data type
73e8ff00
DM
2309 * It's mathematically equivalent to
2310 * max(n) * sizeof(t) > MEM_SIZE_MAX
e6bdf523
DM
2311 */
2312
2313# define _MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) \
445198b9
LM
2314 ( sizeof(MEM_SIZE) < sizeof(n) \
2315 || sizeof(t) > ((MEM_SIZE)1 << 8*(sizeof(MEM_SIZE) - sizeof(n))))
e6bdf523 2316
88f9f128 2317/* This is written in a slightly odd way to avoid various spurious
d98e5cde
DM
2318 * compiler warnings. We *want* to write the expression as
2319 * _MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) && (n > C)
2320 * (for some compile-time constant C), but even when the LHS
2321 * constant-folds to false at compile-time, g++ insists on emitting
2322 * warnings about the RHS (e.g. "comparison is always false"), so instead
2323 * we write it as
e6bdf523 2324 *
d98e5cde 2325 * (cond ? n : X) > C
88f9f128 2326 *
d98e5cde
DM
2327 * where X is a constant with X > C always false. Choosing a value for X
2328 * is tricky. If 0, some compilers will complain about 0 > C always being
2329 * false; if 1, Coverity complains when n happens to be the constant value
2330 * '1', that cond ? 1 : 1 has the same value on both branches; so use C
2331 * for X and hope that nothing else whines.
e6bdf523
DM
2332 */
2333
2334# define _MEM_WRAP_WILL_WRAP(n,t) \
88f9f128
DM
2335 ((_MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) ? (MEM_SIZE)(n) : \
2336 MEM_SIZE_MAX/sizeof(t)) > MEM_SIZE_MAX/sizeof(t))
e6bdf523
DM
2337
2338# define MEM_WRAP_CHECK(n,t) \
2339 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2340 && (croak_memory_wrap(),0))
2341
2342# define MEM_WRAP_CHECK_1(n,t,a) \
2343 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2344 && (Perl_croak_nocontext("%s",(a)),0))
2345
814eedc8
DD
2346/* "a" arg must be a string literal */
2347# define MEM_WRAP_CHECK_s(n,t,a) \
2348 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2349 && (Perl_croak_nocontext("" a ""),0))
2350
8b44ba4c 2351#define MEM_WRAP_CHECK_(n,t) MEM_WRAP_CHECK(n,t),
27d5b266 2352
a500027b 2353#define PERL_STRLEN_ROUNDUP(n) ((void)(((n) > MEM_SIZE_MAX - 2 * PERL_STRLEN_ROUNDUP_QUANTUM) ? (croak_memory_wrap(),0) : 0), _PERL_STRLEN_ROUNDUP_UNCHECKED(n))
27d5b266
JH
2354#else
2355
410319be
NC
2356#define MEM_WRAP_CHECK(n,t)
2357#define MEM_WRAP_CHECK_1(n,t,a)
814eedc8 2358#define MEM_WRAP_CHECK_s(n,t,a)
8b44ba4c
NC
2359#define MEM_WRAP_CHECK_(n,t)
2360
a500027b 2361#define PERL_STRLEN_ROUNDUP(n) _PERL_STRLEN_ROUNDUP_UNCHECKED(n)
27d5b266 2362
1936d2a7 2363#endif
8b44ba4c 2364
fe4f188c 2365#ifdef PERL_MEM_LOG
46c6c7e2 2366/*
9f653bb5 2367 * If PERL_MEM_LOG is defined, all Newx()s, Renew()s, and Safefree()s
46c6c7e2
JH
2368 * go through functions, which are handy for debugging breakpoints, but
2369 * which more importantly get the immediate calling environment (file and
e352bcff
JH
2370 * line number, and C function name if available) passed in. This info can
2371 * then be used for logging the calls, for which one gets a sample
73d1d973 2372 * implementation unless -DPERL_MEM_LOG_NOIMPL is also defined.
3609ea0d 2373 *
46c6c7e2 2374 * Known problems:
94e892a6 2375 * - not all memory allocs get logged, only those
46c6c7e2 2376 * that go through Newx() and derivatives (while all
94e892a6 2377 * Safefrees do get logged)
46c6c7e2
JH
2378 * - __FILE__ and __LINE__ do not work everywhere
2379 * - __func__ or __FUNCTION__ even less so
2380 * - I think more goes on after the perlio frees but
2381 * the thing is that STDERR gets closed (as do all
2382 * the file descriptors)
2383 * - no deeper calling stack than the caller of the Newx()
2384 * or the kind, but do I look like a C reflection/introspection
2385 * utility to you?
2386 * - the function prototypes for the logging functions
2387 * probably should maybe be somewhere else than handy.h
2388 * - one could consider inlining (macrofying) the logging
2389 * for speed, but I am too lazy
2390 * - one could imagine recording the allocations in a hash,
2391 * (keyed by the allocation address?), and maintain that
2392 * through reallocs and frees, but how to do that without
2393 * any News() happening...?
73d1d973 2394 * - lots of -Ddefines to get useful/controllable output
b953482e 2395 * - lots of ENV reads
46c6c7e2
JH
2396 */
2397
0b0ab801 2398# ifdef PERL_CORE
73d1d973 2399# ifndef PERL_MEM_LOG_NOIMPL
0b0ab801
MHM
2400enum mem_log_type {
2401 MLT_ALLOC,
2402 MLT_REALLOC,
d7a2c63c
MHM
2403 MLT_FREE,
2404 MLT_NEW_SV,
2405 MLT_DEL_SV
0b0ab801
MHM
2406};
2407# endif
12754f92 2408# if defined(PERL_IN_SV_C) /* those are only used in sv.c */
d7a2c63c
MHM
2409void Perl_mem_log_new_sv(const SV *sv, const char *filename, const int linenumber, const char *funcname);
2410void Perl_mem_log_del_sv(const SV *sv, const char *filename, const int linenumber, const char *funcname);
12754f92 2411# endif
0b0ab801
MHM
2412# endif
2413
fe4f188c
JH
2414#endif
2415
2416#ifdef PERL_MEM_LOG
d1401ee9
MHM
2417#define MEM_LOG_ALLOC(n,t,a) Perl_mem_log_alloc(n,sizeof(t),STRINGIFY(t),a,__FILE__,__LINE__,FUNCTION__)
2418#define MEM_LOG_REALLOC(n,t,v,a) Perl_mem_log_realloc(n,sizeof(t),STRINGIFY(t),v,a,__FILE__,__LINE__,FUNCTION__)
46c6c7e2 2419#define MEM_LOG_FREE(a) Perl_mem_log_free(a,__FILE__,__LINE__,FUNCTION__)
fe4f188c
JH
2420#endif
2421
2422#ifndef MEM_LOG_ALLOC
2423#define MEM_LOG_ALLOC(n,t,a) (a)
2424#endif
2425#ifndef MEM_LOG_REALLOC
2426#define MEM_LOG_REALLOC(n,t,v,a) (a)
2427#endif
2428#ifndef MEM_LOG_FREE
2429#define MEM_LOG_FREE(a) (a)
2430#endif
2431
d1401ee9
MHM
2432#define Newx(v,n,t) (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_ALLOC(n,t,safemalloc((MEM_SIZE)((n)*sizeof(t))))))
2433#define Newxc(v,n,t,c) (v = (MEM_WRAP_CHECK_(n,t) (c*)MEM_LOG_ALLOC(n,t,safemalloc((MEM_SIZE)((n)*sizeof(t))))))
2434#define Newxz(v,n,t) (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_ALLOC(n,t,safecalloc((n),sizeof(t)))))
a6f6820f
NC
2435
2436#ifndef PERL_CORE
a02a5408
JC
2437/* pre 5.9.x compatibility */
2438#define New(x,v,n,t) Newx(v,n,t)
2439#define Newc(x,v,n,t,c) Newxc(v,n,t,c)
4541904d 2440#define Newz(x,v,n,t) Newxz(v,n,t)
a6f6820f 2441#endif
a02a5408 2442
ff68c719 2443#define Renew(v,n,t) \
d1401ee9 2444 (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_REALLOC(n,t,v,saferealloc((Malloc_t)(v),(MEM_SIZE)((n)*sizeof(t))))))
ff68c719 2445#define Renewc(v,n,t,c) \
d1401ee9 2446 (v = (MEM_WRAP_CHECK_(n,t) (c*)MEM_LOG_REALLOC(n,t,v,saferealloc((Malloc_t)(v),(MEM_SIZE)((n)*sizeof(t))))))
94010e71
NC
2447
2448#ifdef PERL_POISON
2449#define Safefree(d) \
06c0cc96 2450 ((d) ? (void)(safefree(MEM_LOG_FREE((Malloc_t)(d))), Poison(&(d), 1, Malloc_t)) : (void) 0)
94010e71 2451#else
fe4f188c 2452#define Safefree(d) safefree(MEM_LOG_FREE((Malloc_t)(d)))
94010e71 2453#endif
55497cff 2454
dbb57106
YO
2455/* assert that a valid ptr has been supplied - use this instead of assert(ptr) *
2456 * as it handles cases like constant string arguments without throwing warnings *
2457 * the cast is required, as is the inequality check, to avoid warnings */
45908e4d 2458#define perl_assert_ptr(p) assert( ((void*)(p)) != 0 )
55497cff 2459
45908e4d
YO
2460
2461#define Move(s,d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), perl_assert_ptr(s), (void)memmove((char*)(d),(const char*)(s), (n) * sizeof(t)))
2462#define Copy(s,d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), perl_assert_ptr(s), (void)memcpy((char*)(d),(const char*)(s), (n) * sizeof(t)))
2463#define Zero(d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), (void)memzero((char*)(d), (n) * sizeof(t)))
2464
bdd1531d 2465/* Like above, but returns a pointer to 'd' */
45908e4d
YO
2466#define MoveD(s,d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), perl_assert_ptr(s), memmove((char*)(d),(const char*)(s), (n) * sizeof(t)))
2467#define CopyD(s,d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), perl_assert_ptr(s), memcpy((char*)(d),(const char*)(s), (n) * sizeof(t)))
45908e4d 2468#define ZeroD(d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), memzero((char*)(d), (n) * sizeof(t)))
e90e2364 2469
7e337ee0
JH
2470#define PoisonWith(d,n,t,b) (MEM_WRAP_CHECK_(n,t) (void)memset((char*)(d), (U8)(b), (n) * sizeof(t)))
2471#define PoisonNew(d,n,t) PoisonWith(d,n,t,0xAB)
2472#define PoisonFree(d,n,t) PoisonWith(d,n,t,0xEF)
2473#define Poison(d,n,t) PoisonFree(d,n,t)
27d5b266 2474
caa674f3
DD
2475#ifdef PERL_POISON
2476# define PERL_POISON_EXPR(x) x
2477#else
2478# define PERL_POISON_EXPR(x)
2479#endif
2480
ff68c719 2481#define StructCopy(s,d,t) (*((t*)(d)) = *((t*)(s)))
2cc61e15 2482
c3caa5c3
JH
2483/* C_ARRAY_LENGTH is the number of elements in the C array (so you
2484 * want your zero-based indices to be less than but not equal to).
2485 *
2486 * C_ARRAY_END is one past the last: half-open/half-closed range,
2487 * not last-inclusive range. */
622913ab 2488#define C_ARRAY_LENGTH(a) (sizeof(a)/sizeof((a)[0]))
c3caa5c3 2489#define C_ARRAY_END(a) ((a) + C_ARRAY_LENGTH(a))
622913ab 2490
2cc61e15
DD
2491#ifdef NEED_VA_COPY
2492# ifdef va_copy
2493# define Perl_va_copy(s, d) va_copy(d, s)
07798b17
AC
2494# elif defined(__va_copy)
2495# define Perl_va_copy(s, d) __va_copy(d, s)
2cc61e15 2496# else
07798b17 2497# define Perl_va_copy(s, d) Copy(s, d, 1, va_list)
2cc61e15
DD
2498# endif
2499#endif
2500
472d47bc
SB
2501/* convenience debug macros */
2502#ifdef USE_ITHREADS
2503#define pTHX_FORMAT "Perl interpreter: 0x%p"
2504#define pTHX__FORMAT ", Perl interpreter: 0x%p"
f54cb97a
AL
2505#define pTHX_VALUE_ (void *)my_perl,
2506#define pTHX_VALUE (void *)my_perl
2507#define pTHX__VALUE_ ,(void *)my_perl,
2508#define pTHX__VALUE ,(void *)my_perl
472d47bc 2509#else
3609ea0d 2510#define pTHX_FORMAT
472d47bc 2511#define pTHX__FORMAT
3609ea0d 2512#define pTHX_VALUE_
472d47bc 2513#define pTHX_VALUE
3609ea0d 2514#define pTHX__VALUE_
472d47bc
SB
2515#define pTHX__VALUE
2516#endif /* USE_ITHREADS */
3609ea0d 2517
2acdbac1
NC
2518/* Perl_deprecate was not part of the public API, and did not have a deprecate()
2519 shortcut macro defined without -DPERL_CORE. Neither codesearch.google.com nor
2520 CPAN::Unpack show any users outside the core. */
2521#ifdef PERL_CORE
dc6e8de0
A
2522# define deprecate(s) Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2523 "Use of " s " is deprecated")
c9680906
A
2524# define deprecate_disappears_in(when,message) \
2525 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2526 message ", and will disappear in Perl " when)
ac641426
A
2527# define deprecate_fatal_in(when,message) \
2528 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2529 message ". Its use will be fatal in Perl " when)
2acdbac1
NC
2530#endif
2531
dfff4baf
BF
2532/* Internal macros to deal with gids and uids */
2533#ifdef PERL_CORE
2534
2535# if Uid_t_size > IVSIZE
2536# define sv_setuid(sv, uid) sv_setnv((sv), (NV)(uid))
2537# define SvUID(sv) SvNV(sv)
07798b17
AC
2538# elif Uid_t_sign <= 0
2539# define sv_setuid(sv, uid) sv_setiv((sv), (IV)(uid))
2540# define SvUID(sv) SvIV(sv)
dfff4baf 2541# else
07798b17
AC
2542# define sv_setuid(sv, uid) sv_setuv((sv), (UV)(uid))
2543# define SvUID(sv) SvUV(sv)
dfff4baf
BF
2544# endif /* Uid_t_size */
2545
2546# if Gid_t_size > IVSIZE
2547# define sv_setgid(sv, gid) sv_setnv((sv), (NV)(gid))
2548# define SvGID(sv) SvNV(sv)
07798b17
AC
2549# elif Gid_t_sign <= 0
2550# define sv_setgid(sv, gid) sv_setiv((sv), (IV)(gid))
2551# define SvGID(sv) SvIV(sv)
dfff4baf 2552# else
07798b17
AC
2553# define sv_setgid(sv, gid) sv_setuv((sv), (UV)(gid))
2554# define SvGID(sv) SvUV(sv)
dfff4baf
BF
2555# endif /* Gid_t_size */
2556
2557#endif
2558
6a5bc5ac 2559#endif /* PERL_HANDY_H_ */
9d745869 2560
e9a8c099 2561/*
14d04a33 2562 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 2563 */