This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
File::Glob bsd_glob.c dont use "loop" as a goto label
[perl5.git] / handy.h
CommitLineData
a0d0e21e 1/* handy.h
a687059c 2 *
1129b882 3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1999, 2000,
da5d8dbb 4 * 2001, 2002, 2004, 2005, 2006, 2007, 2008, 2012 by Larry Wall and others
a687059c 5 *
6e21c824
LW
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
8d063cd8 8 *
8d063cd8
LW
9 */
10
4650c663
KW
11/* IMPORTANT NOTE: Everything whose name begins with an underscore is for
12 * internal core Perl use only. */
13
6a5bc5ac
KW
14#ifndef PERL_HANDY_H_ /* Guard against nested #inclusion */
15#define PERL_HANDY_H_
9d745869 16
24792b8d
NC
17#ifndef PERL_CORE
18# define Null(type) ((type)NULL)
954c1994
GS
19
20/*
ccfc67b7 21=head1 Handy Values
954c1994 22
3609ea0d 23=for apidoc AmU||Nullch
72d33970
FC
24Null character pointer. (No longer available when C<PERL_CORE> is
25defined.)
2307c6d0 26
954c1994 27=for apidoc AmU||Nullsv
72d33970 28Null SV pointer. (No longer available when C<PERL_CORE> is defined.)
954c1994
GS
29
30=cut
31*/
32
24792b8d
NC
33# define Nullch Null(char*)
34# define Nullfp Null(PerlIO*)
35# define Nullsv Null(SV*)
36#endif
8d063cd8 37
641d3f0b
PP
38#ifdef TRUE
39#undef TRUE
40#endif
41#ifdef FALSE
42#undef FALSE
43#endif
44#define TRUE (1)
45#define FALSE (0)
46
cf3f0ffb
DM
47/* The MUTABLE_*() macros cast pointers to the types shown, in such a way
48 * (compiler permitting) that casting away const-ness will give a warning;
49 * e.g.:
50 *
51 * const SV *sv = ...;
52 * AV *av1 = (AV*)sv; <== BAD: the const has been silently cast away
53 * AV *av2 = MUTABLE_AV(sv); <== GOOD: it may warn
54 */
55
b1bc3f34 56#if defined(__GNUC__) && !defined(PERL_GCC_BRACE_GROUPS_FORBIDDEN)
f065df04 57# define MUTABLE_PTR(p) ({ void *_p = (p); _p; })
b1bc3f34
NC
58#else
59# define MUTABLE_PTR(p) ((void *) (p))
60#endif
61
a062e10d 62#define MUTABLE_AV(p) ((AV *)MUTABLE_PTR(p))
ea726b52 63#define MUTABLE_CV(p) ((CV *)MUTABLE_PTR(p))
159b6efe 64#define MUTABLE_GV(p) ((GV *)MUTABLE_PTR(p))
dbebbdb4 65#define MUTABLE_HV(p) ((HV *)MUTABLE_PTR(p))
a45c7426 66#define MUTABLE_IO(p) ((IO *)MUTABLE_PTR(p))
b1bc3f34 67#define MUTABLE_SV(p) ((SV *)MUTABLE_PTR(p))
27d4fb96 68
f789f6a4 69#if defined(I_STDBOOL) && !defined(PERL_BOOL_AS_CHAR)
bd31be4b
NC
70# include <stdbool.h>
71# ifndef HAS_BOOL
72# define HAS_BOOL 1
73# endif
74#endif
75
8e84507e 76/* bool is built-in for g++-2.6.3 and later, which might be used
c1d22f6b
GS
77 for extensions. <_G_config.h> defines _G_HAVE_BOOL, but we can't
78 be sure _G_config.h will be included before this file. _G_config.h
8e84507e 79 also defines _G_HAVE_BOOL for both gcc and g++, but only g++
c1d22f6b
GS
80 actually has bool. Hence, _G_HAVE_BOOL is pretty useless for us.
81 g++ can be identified by __GNUG__.
82 Andy Dougherty February 2000
5d94fbed 83*/
3609ea0d 84#ifdef __GNUG__ /* GNU g++ has bool built-in */
f789f6a4 85# ifndef PERL_BOOL_AS_CHAR
5d94fbed 86# ifndef HAS_BOOL
c1d22f6b 87# define HAS_BOOL 1
5d94fbed 88# endif
f789f6a4 89# endif
5d94fbed
AD
90#endif
91
92#ifndef HAS_BOOL
f789f6a4
FC
93# ifdef bool
94# undef bool
95# endif
70d5cb32 96# define bool char
c1d22f6b 97# define HAS_BOOL 1
a687059c 98#endif
0d3e774c 99
236415e2
KW
100/* cast-to-bool. A simple (bool) cast may not do the right thing: if bool is
101 * defined as char for example, then the cast from int is
102 * implementation-defined (bool)!!(cbool) in a ternary triggers a bug in xlc on
103 * AIX */
18f5643b 104#define cBOOL(cbool) ((cbool) ? (bool)1 : (bool)0)
f2338a2e 105
46c6c7e2 106/* Try to figure out __func__ or __FUNCTION__ equivalent, if any.
e352bcff
JH
107 * XXX Should really be a Configure probe, with HAS__FUNCTION__
108 * and FUNCTION__ as results.
109 * XXX Similarly, a Configure probe for __FILE__ and __LINE__ is needed. */
46c6c7e2
JH
110#if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L) || (defined(__SUNPRO_C)) /* C99 or close enough. */
111# define FUNCTION__ __func__
07798b17
AC
112#elif (defined(USING_MSVC6)) || /* MSVC6 has neither __func__ nor __FUNCTION and no good workarounds, either. */ \
113 (defined(__DECC_VER)) /* Tru64 or VMS, and strict C89 being used, but not modern enough cc (in Tur64, -c99 not known, only -std1). */
114# define FUNCTION__ ""
46c6c7e2 115#else
07798b17 116# define FUNCTION__ __FUNCTION__ /* Common extension. */
46c6c7e2
JH
117#endif
118
27d4fb96
PP
119/* XXX A note on the perl source internal type system. The
120 original intent was that I32 be *exactly* 32 bits.
121
122 Currently, we only guarantee that I32 is *at least* 32 bits.
123 Specifically, if int is 64 bits, then so is I32. (This is the case
124 for the Cray.) This has the advantage of meshing nicely with
125 standard library calls (where we pass an I32 and the library is
126 expecting an int), but the disadvantage that an I32 is not 32 bits.
127 Andy Dougherty August 1996
24fef2a7 128
dc45a647
MB
129 There is no guarantee that there is *any* integral type with
130 exactly 32 bits. It is perfectly legal for a system to have
131 sizeof(short) == sizeof(int) == sizeof(long) == 8.
693762b4 132
dc45a647
MB
133 Similarly, there is no guarantee that I16 and U16 have exactly 16
134 bits.
693762b4 135
8e84507e
NIS
136 For dealing with issues that may arise from various 32/64-bit
137 systems, we will ask Configure to check out
8175356b 138
3609ea0d
JH
139 SHORTSIZE == sizeof(short)
140 INTSIZE == sizeof(int)
141 LONGSIZE == sizeof(long)
dc45a647 142 LONGLONGSIZE == sizeof(long long) (if HAS_LONG_LONG)
3609ea0d 143 PTRSIZE == sizeof(void *)
dc45a647
MB
144 DOUBLESIZE == sizeof(double)
145 LONG_DOUBLESIZE == sizeof(long double) (if HAS_LONG_DOUBLE).
8175356b 146
27d4fb96
PP
147*/
148
69512466
JH
149#ifdef I_INTTYPES /* e.g. Linux has int64_t without <inttypes.h> */
150# include <inttypes.h>
dd0eed91
JH
151# ifdef INT32_MIN_BROKEN
152# undef INT32_MIN
153# define INT32_MIN (-2147483647-1)
154# endif
155# ifdef INT64_MIN_BROKEN
156# undef INT64_MIN
157# define INT64_MIN (-9223372036854775807LL-1)
158# endif
69512466
JH
159#endif
160
8175356b
JH
161typedef I8TYPE I8;
162typedef U8TYPE U8;
163typedef I16TYPE I16;
164typedef U16TYPE U16;
165typedef I32TYPE I32;
166typedef U32TYPE U32;
16d89be8 167
74b807c7 168#ifdef QUADKIND
8175356b
JH
169typedef I64TYPE I64;
170typedef U64TYPE U64;
16d89be8 171#endif
8175356b 172
d8668976 173#if defined(UINT8_MAX) && defined(INT16_MAX) && defined(INT32_MAX)
5ff3f7a4 174
5ff3f7a4
GS
175/* I8_MAX and I8_MIN constants are not defined, as I8 is an ambiguous type.
176 Please search CHAR_MAX in perl.h for further details. */
177#define U8_MAX UINT8_MAX
178#define U8_MIN UINT8_MIN
179
5ff3f7a4
GS
180#define I16_MAX INT16_MAX
181#define I16_MIN INT16_MIN
182#define U16_MAX UINT16_MAX
183#define U16_MIN UINT16_MIN
184
5ff3f7a4
GS
185#define I32_MAX INT32_MAX
186#define I32_MIN INT32_MIN
0e983133
GS
187#ifndef UINT32_MAX_BROKEN /* e.g. HP-UX with gcc messes this up */
188# define U32_MAX UINT32_MAX
189#else
190# define U32_MAX 4294967295U
191#endif
5ff3f7a4
GS
192#define U32_MIN UINT32_MIN
193
194#else
195
5c9fa16e
KA
196/* I8_MAX and I8_MIN constants are not defined, as I8 is an ambiguous type.
197 Please search CHAR_MAX in perl.h for further details. */
27d4fb96
PP
198#define U8_MAX PERL_UCHAR_MAX
199#define U8_MIN PERL_UCHAR_MIN
79072805 200
27d4fb96
PP
201#define I16_MAX PERL_SHORT_MAX
202#define I16_MIN PERL_SHORT_MIN
203#define U16_MAX PERL_USHORT_MAX
204#define U16_MIN PERL_USHORT_MIN
79072805 205
c4f23d77 206#if LONGSIZE > 4
27d4fb96
PP
207# define I32_MAX PERL_INT_MAX
208# define I32_MIN PERL_INT_MIN
209# define U32_MAX PERL_UINT_MAX
210# define U32_MIN PERL_UINT_MIN
79072805 211#else
27d4fb96
PP
212# define I32_MAX PERL_LONG_MAX
213# define I32_MIN PERL_LONG_MIN
214# define U32_MAX PERL_ULONG_MAX
215# define U32_MIN PERL_ULONG_MIN
79072805
LW
216#endif
217
5ff3f7a4
GS
218#endif
219
247cee9f
KW
220/* These C99 typedefs are useful sometimes for, say, loop variables whose
221 * maximum values are small, but for which speed trumps size. If we have a C99
222 * compiler, use that. Otherwise, a plain 'int' should be good enough.
223 *
224 * Restrict these to core for now until we are more certain this is a good
225 * idea. */
226#if defined(PERL_CORE) || defined(PERL_EXT)
227# ifdef I_STDINT
228 typedef int_fast8_t PERL_INT_FAST8_T;
229 typedef uint_fast8_t PERL_UINT_FAST8_T;
230 typedef int_fast16_t PERL_INT_FAST16_T;
231 typedef uint_fast16_t PERL_UINT_FAST16_T;
232# else
233 typedef int PERL_INT_FAST8_T;
234 typedef unsigned int PERL_UINT_FAST8_T;
235 typedef int PERL_INT_FAST16_T;
236 typedef unsigned int PERL_UINT_FAST16_T;
237# endif
238#endif
239
464decb6
KW
240/* log(2) (i.e., log base 10 of 2) is pretty close to 0.30103, just in case
241 * anyone is grepping for it */
242#define BIT_DIGITS(N) (((N)*146)/485 + 1) /* log10(2) =~ 146/485 */
fc36a67e
PP
243#define TYPE_DIGITS(T) BIT_DIGITS(sizeof(T) * 8)
244#define TYPE_CHARS(T) (TYPE_DIGITS(T) + 2) /* sign, NUL */
245
88794300 246/* Unused by core; should be deprecated */
ff68c719 247#define Ctl(ch) ((ch) & 037)
8d063cd8 248
98fce2a4
KW
249#if defined(PERL_CORE) || defined(PERL_EXT)
250# ifndef MIN
251# define MIN(a,b) ((a) < (b) ? (a) : (b))
252# endif
253# ifndef MAX
254# define MAX(a,b) ((a) > (b) ? (a) : (b))
255# endif
256#endif
257
4650c663
KW
258/* This is a helper macro to avoid preprocessor issues, replaced by nothing
259 * unless under DEBUGGING, where it expands to an assert of its argument,
260 * followed by a comma (hence the comma operator). If we just used a straight
3e94db23
JH
261 * assert(), we would get a comma with nothing before it when not DEBUGGING.
262 *
263 * We also use empty definition under Coverity since the __ASSERT__
264 * checks often check for things that Really Cannot Happen, and Coverity
265 * detects that and gets all excited. */
266
267#if defined(DEBUGGING) && !defined(__COVERITY__)
0f092d08
KW
268# define __ASSERT_(statement) assert(statement),
269#else
270# define __ASSERT_(statement)
271#endif
272
3fe05580
MHM
273/*
274=head1 SV-Body Allocation
275
1568d13a
Z
276=for apidoc Ama|SV*|newSVpvs|"literal string" s
277Like C<newSVpvn>, but takes a literal string instead of a
30a15352 278string/length pair.
3fe05580 279
1568d13a
Z
280=for apidoc Ama|SV*|newSVpvs_flags|"literal string" s|U32 flags
281Like C<newSVpvn_flags>, but takes a literal string instead of
30a15352 282a string/length pair.
84bafc02 283
1568d13a
Z
284=for apidoc Ama|SV*|newSVpvs_share|"literal string" s
285Like C<newSVpvn_share>, but takes a literal string instead of
30a15352 286a string/length pair and omits the hash parameter.
3fe05580 287
1568d13a
Z
288=for apidoc Am|void|sv_catpvs_flags|SV* sv|"literal string" s|I32 flags
289Like C<sv_catpvn_flags>, but takes a literal string instead
30a15352 290of a string/length pair.
9dcc53ea 291
1568d13a
Z
292=for apidoc Am|void|sv_catpvs_nomg|SV* sv|"literal string" s
293Like C<sv_catpvn_nomg>, but takes a literal string instead of
0c395ea5 294a string/length pair.
9dcc53ea 295
1568d13a
Z
296=for apidoc Am|void|sv_catpvs|SV* sv|"literal string" s
297Like C<sv_catpvn>, but takes a literal string instead of a
0c395ea5 298string/length pair.
3fe05580 299
1568d13a
Z
300=for apidoc Am|void|sv_catpvs_mg|SV* sv|"literal string" s
301Like C<sv_catpvn_mg>, but takes a literal string instead of a
9dcc53ea
Z
302string/length pair.
303
1568d13a
Z
304=for apidoc Am|void|sv_setpvs|SV* sv|"literal string" s
305Like C<sv_setpvn>, but takes a literal string instead of a
0c395ea5 306string/length pair.
3fe05580 307
1568d13a
Z
308=for apidoc Am|void|sv_setpvs_mg|SV* sv|"literal string" s
309Like C<sv_setpvn_mg>, but takes a literal string instead of a
9dcc53ea
Z
310string/length pair.
311
1568d13a
Z
312=for apidoc Am|SV *|sv_setref_pvs|"literal string" s
313Like C<sv_setref_pvn>, but takes a literal string instead of
0c395ea5 314a string/length pair.
9dcc53ea 315
3fe05580
MHM
316=head1 Memory Management
317
1568d13a
Z
318=for apidoc Ama|char*|savepvs|"literal string" s
319Like C<savepvn>, but takes a literal string instead of a
30a15352 320string/length pair.
3fe05580 321
1568d13a 322=for apidoc Ama|char*|savesharedpvs|"literal string" s
9dcc53ea
Z
323A version of C<savepvs()> which allocates the duplicate string in memory
324which is shared between threads.
325
3fe05580
MHM
326=head1 GV Functions
327
1568d13a
Z
328=for apidoc Am|HV*|gv_stashpvs|"literal string" name|I32 create
329Like C<gv_stashpvn>, but takes a literal string instead of a
0c395ea5 330string/length pair.
3fe05580
MHM
331
332=head1 Hash Manipulation Functions
333
1568d13a
Z
334=for apidoc Am|SV**|hv_fetchs|HV* tb|"literal string" key|I32 lval
335Like C<hv_fetch>, but takes a literal string instead of a
0c395ea5 336string/length pair.
3fe05580 337
d106cb46 338=for apidoc Am|SV**|hv_stores|HV* tb|"literal string" key|SV* val
1568d13a 339Like C<hv_store>, but takes a literal string instead of a
0c395ea5 340string/length pair
3fe05580
MHM
341and omits the hash parameter.
342
510966aa
Z
343=head1 Lexer interface
344
1568d13a 345=for apidoc Amx|void|lex_stuff_pvs|"literal string" pv|U32 flags
510966aa 346
1568d13a 347Like L</lex_stuff_pvn>, but takes a literal string instead of
0c395ea5 348a string/length pair.
510966aa 349
3fe05580
MHM
350=cut
351*/
352
926bb54c
KW
353/* concatenating with "" ensures that only literal strings are accepted as
354 * argument */
26bfd924 355#define STR_WITH_LEN(s) ("" s ""), (sizeof(s)-1)
2efa8cc7 356
926bb54c
KW
357/* note that STR_WITH_LEN() can't be used as argument to macros or functions
358 * that under some configurations might be macros, which means that it requires
359 * the full Perl_xxx(aTHX_ ...) form for any API calls where it's used.
ba3a79e7
GA
360 */
361
362/* STR_WITH_LEN() shortcuts */
363#define newSVpvs(str) Perl_newSVpvn(aTHX_ STR_WITH_LEN(str))
84bafc02
NC
364#define newSVpvs_flags(str,flags) \
365 Perl_newSVpvn_flags(aTHX_ STR_WITH_LEN(str), flags)
ba3a79e7 366#define newSVpvs_share(str) Perl_newSVpvn_share(aTHX_ STR_WITH_LEN(str), 0)
9dcc53ea
Z
367#define sv_catpvs_flags(sv, str, flags) \
368 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), flags)
369#define sv_catpvs_nomg(sv, str) \
370 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), 0)
371#define sv_catpvs(sv, str) \
372 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), SV_GMAGIC)
373#define sv_catpvs_mg(sv, str) \
374 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), SV_GMAGIC|SV_SMAGIC)
3fe05580 375#define sv_setpvs(sv, str) Perl_sv_setpvn(aTHX_ sv, STR_WITH_LEN(str))
9dcc53ea
Z
376#define sv_setpvs_mg(sv, str) Perl_sv_setpvn_mg(aTHX_ sv, STR_WITH_LEN(str))
377#define sv_setref_pvs(rv, classname, str) \
378 Perl_sv_setref_pvn(aTHX_ rv, classname, STR_WITH_LEN(str))
ba3a79e7 379#define savepvs(str) Perl_savepvn(aTHX_ STR_WITH_LEN(str))
9dcc53ea
Z
380#define savesharedpvs(str) Perl_savesharedpvn(aTHX_ STR_WITH_LEN(str))
381#define gv_stashpvs(str, create) \
382 Perl_gv_stashpvn(aTHX_ STR_WITH_LEN(str), create)
383#define gv_fetchpvs(namebeg, add, sv_type) \
384 Perl_gv_fetchpvn_flags(aTHX_ STR_WITH_LEN(namebeg), add, sv_type)
385#define gv_fetchpvn(namebeg, len, add, sv_type) \
386 Perl_gv_fetchpvn_flags(aTHX_ namebeg, len, add, sv_type)
387#define sv_catxmlpvs(dsv, str, utf8) \
388 Perl_sv_catxmlpvn(aTHX_ dsv, STR_WITH_LEN(str), utf8)
4ac46235 389
ba3a79e7 390
510966aa
Z
391#define lex_stuff_pvs(pv,flags) Perl_lex_stuff_pvn(aTHX_ STR_WITH_LEN(pv), flags)
392
b96d8cd9
NC
393#define get_cvs(str, flags) \
394 Perl_get_cvn_flags(aTHX_ STR_WITH_LEN(str), (flags))
5c1737d1 395
954c1994 396/*
ccfc67b7
JH
397=head1 Miscellaneous Functions
398
954c1994 399=for apidoc Am|bool|strNE|char* s1|char* s2
dc6b0978
KW
400Test two C<NUL>-terminated strings to see if they are different. Returns true
401or false.
954c1994
GS
402
403=for apidoc Am|bool|strEQ|char* s1|char* s2
dc6b0978
KW
404Test two C<NUL>-terminated strings to see if they are equal. Returns true or
405false.
954c1994
GS
406
407=for apidoc Am|bool|strLT|char* s1|char* s2
dc6b0978
KW
408Test two C<NUL>-terminated strings to see if the first, C<s1>, is less than the
409second, C<s2>. Returns true or false.
954c1994
GS
410
411=for apidoc Am|bool|strLE|char* s1|char* s2
dc6b0978
KW
412Test two C<NUL>-terminated strings to see if the first, C<s1>, is less than or
413equal to the second, C<s2>. Returns true or false.
954c1994
GS
414
415=for apidoc Am|bool|strGT|char* s1|char* s2
dc6b0978
KW
416Test two C<NUL>-terminated strings to see if the first, C<s1>, is greater than
417the second, C<s2>. Returns true or false.
954c1994
GS
418
419=for apidoc Am|bool|strGE|char* s1|char* s2
dc6b0978
KW
420Test two C<NUL>-terminated strings to see if the first, C<s1>, is greater than
421or equal to the second, C<s2>. Returns true or false.
954c1994
GS
422
423=for apidoc Am|bool|strnNE|char* s1|char* s2|STRLEN len
dc6b0978
KW
424Test two C<NUL>-terminated strings to see if they are different. The C<len>
425parameter indicates the number of bytes to compare. Returns true or false. (A
954c1994
GS
426wrapper for C<strncmp>).
427
428=for apidoc Am|bool|strnEQ|char* s1|char* s2|STRLEN len
dc6b0978
KW
429Test two C<NUL>-terminated strings to see if they are equal. The C<len>
430parameter indicates the number of bytes to compare. Returns true or false. (A
431wrapper for C<strncmp>).
954c1994 432
bd18bd40
KW
433=for apidoc Am|bool|memEQ|char* s1|char* s2|STRLEN len
434Test two buffers (which may contain embedded C<NUL> characters, to see if they
435are equal. The C<len> parameter indicates the number of bytes to compare.
436Returns zero if equal, or non-zero if non-equal.
437
438=for apidoc Am|bool|memNE|char* s1|char* s2|STRLEN len
439Test two buffers (which may contain embedded C<NUL> characters, to see if they
440are not equal. The C<len> parameter indicates the number of bytes to compare.
441Returns zero if non-equal, or non-zero if equal.
442
954c1994 443=cut
fc169e00
KW
444
445New macros should use the following conventions for their names (which are
446based on the underlying C library functions):
447
448 (mem | str n? ) (EQ | NE | LT | GT | GE | (( BEGIN | END ) P? )) l? s?
449
450 Each has two main parameters, string-like operands that are compared
451 against each other, as specified by the macro name. Some macros may
452 additionally have one or potentially even two length parameters. If a length
453 parameter applies to both string parameters, it will be positioned third;
454 otherwise any length parameter immediately follows the string parameter it
455 applies to.
456
457 If the prefix to the name is 'str', the string parameter is a pointer to a C
458 language string. Such a string does not contain embedded NUL bytes; its
459 length may be unknown, but can be calculated by C<strlen()>, since it is
460 terminated by a NUL, which isn't included in its length.
461
462 The optional 'n' following 'str' means that that there is a third parameter,
463 giving the maximum number of bytes to look at in each string. Even if both
464 strings are longer than the length parameter, those extra bytes will be
465 unexamined.
466
467 The 's' suffix means that the 2nd byte string parameter is a literal C
468 double-quoted string. Its length will automatically be calculated by the
469 macro, so no length parameter will ever be needed for it.
470
471 If the prefix is 'mem', the string parameters don't have to be C strings;
472 they may contain embedded NUL bytes, do not necessarily have a terminating
473 NUL, and their lengths can be known only through other means, which in
474 practice are additional parameter(s) passed to the function. All 'mem'
475 functions have at least one length parameter. Barring any 'l' or 's' suffix,
476 there is a single length parameter, in position 3, which applies to both
477 string parameters. The 's' suffix means, as described above, that the 2nd
478 string is a literal double-quoted C string (hence its length is calculated by
479 the macro, and the length parameter to the function applies just to the first
480 string parameter, and hence is positioned just after it). An 'l' suffix
481 means that the 2nd string parameter has its own length parameter, and the
482 signature will look like memFOOl(s1, l1, s2, l2).
483
484 BEGIN (and END) are for testing if the 2nd string is an initial (or final)
485 substring of the 1st string. 'P' if present indicates that the substring
486 must be a "proper" one in tha mathematical sense that the first one must be
487 strictly larger than the 2nd.
488
954c1994
GS
489*/
490
62946e08 491
75400963
KW
492#define strNE(s1,s2) (strcmp(s1,s2) != 0)
493#define strEQ(s1,s2) (strcmp(s1,s2) == 0)
8d063cd8
LW
494#define strLT(s1,s2) (strcmp(s1,s2) < 0)
495#define strLE(s1,s2) (strcmp(s1,s2) <= 0)
496#define strGT(s1,s2) (strcmp(s1,s2) > 0)
497#define strGE(s1,s2) (strcmp(s1,s2) >= 0)
62946e08 498
75400963
KW
499#define strnNE(s1,s2,l) (strncmp(s1,s2,l) != 0)
500#define strnEQ(s1,s2,l) (strncmp(s1,s2,l) == 0)
378cc40b 501
75400963
KW
502#define memNE(s1,s2,l) (memcmp(s1,s2,l) != 0)
503#define memEQ(s1,s2,l) (memcmp(s1,s2,l) == 0)
36477c24 504
085b7534 505/* memEQ and memNE where second comparand is a string constant */
568a785a 506#define memEQs(s1, l, s2) \
777fa2cb 507 (((sizeof(s2)-1) == (l)) && memEQ((s1), ("" s2 ""), (sizeof(s2)-1)))
5f50c6c9 508#define memNEs(s1, l, s2) (! memEQs(s1, l, s2))
568a785a 509
fdbb9a7c
KW
510/* Keep these private until we decide it was a good idea */
511#if defined(PERL_CORE) || defined(PERL_EXT) || defined(PERL_EXT_POSIX)
512
513#define strBEGINs(s1,s2) (strncmp(s1,"" s2 "", sizeof(s2)-1) == 0)
514
bdb7e3f0
KW
515#define memBEGINs(s1, l, s2) \
516 ( (l) >= sizeof(s2) - 1 \
517 && memEQ(s1, "" s2 "", sizeof(s2)-1))
de627158
KW
518#define memBEGINPs(s1, l, s2) \
519 ( (l) > sizeof(s2) - 1 \
520 && memEQ(s1, "" s2 "", sizeof(s2)-1))
bdb7e3f0
KW
521#define memENDs(s1, l, s2) \
522 ( (l) >= sizeof(s2) - 1 \
523 && memEQ(s1 + (l) - (sizeof(s2) - 1), "" s2 "", sizeof(s2)-1))
b80f8424
KW
524#define memENDPs(s1, l, s2) \
525 ( (l) > sizeof(s2) \
526 && memEQ(s1 + (l) - (sizeof(s2) - 1), "" s2 "", sizeof(s2)-1))
fdbb9a7c 527#endif /* End of making macros private */
bdb7e3f0 528
062b6850
KW
529#define memLT(s1,s2,l) (memcmp(s1,s2,l) < 0)
530#define memLE(s1,s2,l) (memcmp(s1,s2,l) <= 0)
531#define memGT(s1,s2,l) (memcmp(s1,s2,l) > 0)
532#define memGE(s1,s2,l) (memcmp(s1,s2,l) >= 0)
533
bbce6d69
PP
534/*
535 * Character classes.
536 *
537 * Unfortunately, the introduction of locales means that we
538 * can't trust isupper(), etc. to tell the truth. And when
539 * it comes to /\w+/ with tainting enabled, we *must* be able
540 * to trust our character classes.
541 *
542 * Therefore, the default tests in the text of Perl will be
543 * independent of locale. Any code that wants to depend on
544 * the current locale will use the tests that begin with "lc".
545 */
546
2304df62
AD
547#ifdef HAS_SETLOCALE /* XXX Is there a better test for this? */
548# ifndef CTYPE256
549# define CTYPE256
550# endif
551#endif
552
954c1994 553/*
ccfc67b7 554
dcccc8ff 555=head1 Character classification
243effed
KW
556This section is about functions (really macros) that classify characters
557into types, such as punctuation versus alphabetic, etc. Most of these are
558analogous to regular expression character classes. (See
559L<perlrecharclass/POSIX Character Classes>.) There are several variants for
560each class. (Not all macros have all variants; each item below lists the
561ones valid for it.) None are affected by C<use bytes>, and only the ones
562with C<LC> in the name are affected by the current locale.
563
564The base function, e.g., C<isALPHA()>, takes an octet (either a C<char> or a
565C<U8>) as input and returns a boolean as to whether or not the character
6aff1f14
KW
566represented by that octet is (or on non-ASCII platforms, corresponds to) an
567ASCII character in the named class based on platform, Unicode, and Perl rules.
568If the input is a number that doesn't fit in an octet, FALSE is returned.
243effed 569
c98722a4 570Variant C<isI<FOO>_A> (e.g., C<isALPHA_A()>) is identical to the base function
550da823
KW
571with no suffix C<"_A">. This variant is used to emphasize by its name that
572only ASCII-range characters can return TRUE.
4b9734bf 573
d60679e1 574Variant C<isI<FOO>_L1> imposes the Latin-1 (or EBCDIC equivalent) character set
4b9734bf
KW
575onto the platform. That is, the code points that are ASCII are unaffected,
576since ASCII is a subset of Latin-1. But the non-ASCII code points are treated
577as if they are Latin-1 characters. For example, C<isWORDCHAR_L1()> will return
578true when called with the code point 0xDF, which is a word character in both
4650c663 579ASCII and EBCDIC (though it represents different characters in each).
243effed 580
c98722a4 581Variant C<isI<FOO>_uvchr> is like the C<isI<FOO>_L1> variant, but accepts any UV code
243effed
KW
582point as input. If the code point is larger than 255, Unicode rules are used
583to determine if it is in the character class. For example,
d0da05db 584C<isWORDCHAR_uvchr(0x100)> returns TRUE, since 0x100 is LATIN CAPITAL LETTER A
6aff1f14 585WITH MACRON in Unicode, and is a word character.
243effed 586
c98722a4 587Variant C<isI<FOO>_utf8_safe> is like C<isI<FOO>_uvchr>, but is used for UTF-8
da8c1a98
KW
588encoded strings. Each call classifies one character, even if the string
589contains many. This variant takes two parameters. The first, C<p>, is a
590pointer to the first byte of the character to be classified. (Recall that it
591may take more than one byte to represent a character in UTF-8 strings.) The
592second parameter, C<e>, points to anywhere in the string beyond the first
593character, up to one byte past the end of the entire string. The suffix
594C<_safe> in the function's name indicates that it will not attempt to read
595beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is true (this
596is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the input
597character is malformed in some way, the program may croak, or the function may
598return FALSE, at the discretion of the implementation, and subject to change in
599future releases.
600
c98722a4 601Variant C<isI<FOO>_utf8> is like C<isI<FOO>_utf8_safe>, but takes just a single
da8c1a98 602parameter, C<p>, which has the same meaning as the corresponding parameter does
c98722a4 603in C<isI<FOO>_utf8_safe>. The function therefore can't check if it is reading
34aeb2e9 604beyond the end of the string. Starting in Perl v5.30, it will take a second
c98722a4 605parameter, becoming a synonym for C<isI<FOO>_utf8_safe>. At that time every
34aeb2e9 606program that uses it will have to be changed to successfully compile. In the
c98722a4 607meantime, the first runtime call to C<isI<FOO>_utf8> from each call point in the
34aeb2e9 608program will raise a deprecation warning, enabled by default. You can convert
c98722a4 609your program now to use C<isI<FOO>_utf8_safe>, and avoid the warnings, and get an
34aeb2e9
KW
610extra measure of protection, or you can wait until v5.30, when you'll be forced
611to add the C<e> parameter.
243effed 612
c98722a4 613Variant C<isI<FOO>_LC> is like the C<isI<FOO>_A> and C<isI<FOO>_L1> variants, but the
31f05a37
KW
614result is based on the current locale, which is what C<LC> in the name stands
615for. If Perl can determine that the current locale is a UTF-8 locale, it uses
616the published Unicode rules; otherwise, it uses the C library function that
617gives the named classification. For example, C<isDIGIT_LC()> when not in a
618UTF-8 locale returns the result of calling C<isdigit()>. FALSE is always
1a83413c
KW
619returned if the input won't fit into an octet. On some platforms where the C
620library function is known to be defective, Perl changes its result to follow
621the POSIX standard's rules.
243effed 622
c98722a4
KW
623Variant C<isI<FOO>_LC_uvchr> is like C<isI<FOO>_LC>, but is defined on any UV. It
624returns the same as C<isI<FOO>_LC> for input code points less than 256, and
243effed
KW
625returns the hard-coded, not-affected-by-locale, Unicode results for larger ones.
626
c98722a4 627Variant C<isI<FOO>_LC_utf8_safe> is like C<isI<FOO>_LC_uvchr>, but is used for UTF-8
da8c1a98
KW
628encoded strings. Each call classifies one character, even if the string
629contains many. This variant takes two parameters. The first, C<p>, is a
630pointer to the first byte of the character to be classified. (Recall that it
631may take more than one byte to represent a character in UTF-8 strings.) The
632second parameter, C<e>, points to anywhere in the string beyond the first
633character, up to one byte past the end of the entire string. The suffix
634C<_safe> in the function's name indicates that it will not attempt to read
635beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is true (this
636is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the input
637character is malformed in some way, the program may croak, or the function may
638return FALSE, at the discretion of the implementation, and subject to change in
639future releases.
640
c98722a4 641Variant C<isI<FOO>_LC_utf8> is like C<isI<FOO>_LC_utf8_safe>, but takes just a single
da8c1a98 642parameter, C<p>, which has the same meaning as the corresponding parameter does
c98722a4 643in C<isI<FOO>_LC_utf8_safe>. The function therefore can't check if it is reading
34aeb2e9 644beyond the end of the string. Starting in Perl v5.30, it will take a second
c98722a4 645parameter, becoming a synonym for C<isI<FOO>_LC_utf8_safe>. At that time every
34aeb2e9 646program that uses it will have to be changed to successfully compile. In the
c98722a4 647meantime, the first runtime call to C<isI<FOO>_LC_utf8> from each call point in
34aeb2e9 648the program will raise a deprecation warning, enabled by default. You can
c98722a4 649convert your program now to use C<isI<FOO>_LC_utf8_safe>, and avoid the warnings,
34aeb2e9
KW
650and get an extra measure of protection, or you can wait until v5.30, when
651you'll be forced to add the C<e> parameter.
ccfc67b7 652
954c1994 653=for apidoc Am|bool|isALPHA|char ch
2787a470 654Returns a boolean indicating whether the specified character is an
6aff1f14 655alphabetic character, analogous to C<m/[[:alpha:]]/>.
dcccc8ff
KW
656See the L<top of this section|/Character classification> for an explanation of
657variants
da8c1a98
KW
658C<isALPHA_A>, C<isALPHA_L1>, C<isALPHA_uvchr>, C<isALPHA_utf8_safe>,
659C<isALPHA_LC>, C<isALPHA_LC_uvchr>, and C<isALPHA_LC_utf8_safe>.
8a58bdcf 660
15861f94
KW
661=for apidoc Am|bool|isALPHANUMERIC|char ch
662Returns a boolean indicating whether the specified character is a either an
6aff1f14 663alphabetic character or decimal digit, analogous to C<m/[[:alnum:]]/>.
dcccc8ff
KW
664See the L<top of this section|/Character classification> for an explanation of
665variants
d0da05db 666C<isALPHANUMERIC_A>, C<isALPHANUMERIC_L1>, C<isALPHANUMERIC_uvchr>,
da8c1a98
KW
667C<isALPHANUMERIC_utf8_safe>, C<isALPHANUMERIC_LC>, C<isALPHANUMERIC_LC_uvchr>,
668and C<isALPHANUMERIC_LC_utf8_safe>.
15861f94 669
8a58bdcf
KW
670=for apidoc Am|bool|isASCII|char ch
671Returns a boolean indicating whether the specified character is one of the 128
243effed 672characters in the ASCII character set, analogous to C<m/[[:ascii:]]/>.
e5ad6aba 673On non-ASCII platforms, it returns TRUE iff this
8a58bdcf
KW
674character corresponds to an ASCII character. Variants C<isASCII_A()> and
675C<isASCII_L1()> are identical to C<isASCII()>.
dcccc8ff
KW
676See the L<top of this section|/Character classification> for an explanation of
677variants
da8c1a98
KW
678C<isASCII_uvchr>, C<isASCII_utf8_safe>, C<isASCII_LC>, C<isASCII_LC_uvchr>, and
679C<isASCII_LC_utf8_safe>. Note, however, that some platforms do not have the C
243effed
KW
680library routine C<isascii()>. In these cases, the variants whose names contain
681C<LC> are the same as the corresponding ones without.
682
d98532ea
KW
683Also note, that because all ASCII characters are UTF-8 invariant (meaning they
684have the exact same representation (always a single byte) whether encoded in
685UTF-8 or not), C<isASCII> will give the correct results when called with any
da8c1a98
KW
686byte in any string encoded or not in UTF-8. And similarly C<isASCII_utf8_safe>
687will work properly on any string encoded or not in UTF-8.
d98532ea 688
243effed
KW
689=for apidoc Am|bool|isBLANK|char ch
690Returns a boolean indicating whether the specified character is a
6aff1f14 691character considered to be a blank, analogous to C<m/[[:blank:]]/>.
dcccc8ff
KW
692See the L<top of this section|/Character classification> for an explanation of
693variants
da8c1a98
KW
694C<isBLANK_A>, C<isBLANK_L1>, C<isBLANK_uvchr>, C<isBLANK_utf8_safe>,
695C<isBLANK_LC>, C<isBLANK_LC_uvchr>, and C<isBLANK_LC_utf8_safe>. Note,
696however, that some platforms do not have the C library routine
697C<isblank()>. In these cases, the variants whose names contain C<LC> are
698the same as the corresponding ones without.
243effed
KW
699
700=for apidoc Am|bool|isCNTRL|char ch
701Returns a boolean indicating whether the specified character is a
6aff1f14 702control character, analogous to C<m/[[:cntrl:]]/>.
dcccc8ff
KW
703See the L<top of this section|/Character classification> for an explanation of
704variants
da8c1a98
KW
705C<isCNTRL_A>, C<isCNTRL_L1>, C<isCNTRL_uvchr>, C<isCNTRL_utf8_safe>,
706C<isCNTRL_LC>, C<isCNTRL_LC_uvchr>, and C<isCNTRL_LC_utf8_safe> On EBCDIC
707platforms, you almost always want to use the C<isCNTRL_L1> variant.
954c1994 708
954c1994 709=for apidoc Am|bool|isDIGIT|char ch
2787a470 710Returns a boolean indicating whether the specified character is a
6aff1f14 711digit, analogous to C<m/[[:digit:]]/>.
8a58bdcf 712Variants C<isDIGIT_A> and C<isDIGIT_L1> are identical to C<isDIGIT>.
dcccc8ff
KW
713See the L<top of this section|/Character classification> for an explanation of
714variants
da8c1a98
KW
715C<isDIGIT_uvchr>, C<isDIGIT_utf8_safe>, C<isDIGIT_LC>, C<isDIGIT_LC_uvchr>, and
716C<isDIGIT_LC_utf8_safe>.
243effed
KW
717
718=for apidoc Am|bool|isGRAPH|char ch
719Returns a boolean indicating whether the specified character is a
6aff1f14 720graphic character, analogous to C<m/[[:graph:]]/>.
dcccc8ff 721See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
722variants C<isGRAPH_A>, C<isGRAPH_L1>, C<isGRAPH_uvchr>, C<isGRAPH_utf8_safe>,
723C<isGRAPH_LC>, C<isGRAPH_LC_uvchr>, and C<isGRAPH_LC_utf8_safe>.
954c1994 724
0c82b6df 725=for apidoc Am|bool|isLOWER|char ch
2787a470 726Returns a boolean indicating whether the specified character is a
6aff1f14 727lowercase character, analogous to C<m/[[:lower:]]/>.
dcccc8ff
KW
728See the L<top of this section|/Character classification> for an explanation of
729variants
da8c1a98
KW
730C<isLOWER_A>, C<isLOWER_L1>, C<isLOWER_uvchr>, C<isLOWER_utf8_safe>,
731C<isLOWER_LC>, C<isLOWER_LC_uvchr>, and C<isLOWER_LC_utf8_safe>.
0c82b6df 732
c99e91e9 733=for apidoc Am|bool|isOCTAL|char ch
2787a470 734Returns a boolean indicating whether the specified character is an
6aff1f14 735octal digit, [0-7].
243effed
KW
736The only two variants are C<isOCTAL_A> and C<isOCTAL_L1>; each is identical to
737C<isOCTAL>.
738
739=for apidoc Am|bool|isPUNCT|char ch
740Returns a boolean indicating whether the specified character is a
6aff1f14
KW
741punctuation character, analogous to C<m/[[:punct:]]/>.
742Note that the definition of what is punctuation isn't as
243effed
KW
743straightforward as one might desire. See L<perlrecharclass/POSIX Character
744Classes> for details.
dcccc8ff 745See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
746variants C<isPUNCT_A>, C<isPUNCT_L1>, C<isPUNCT_uvchr>, C<isPUNCT_utf8_safe>,
747C<isPUNCT_LC>, C<isPUNCT_LC_uvchr>, and C<isPUNCT_LC_utf8_safe>.
c99e91e9 748
0c82b6df 749=for apidoc Am|bool|isSPACE|char ch
2787a470 750Returns a boolean indicating whether the specified character is a
6aff1f14 751whitespace character. This is analogous
398d098a 752to what C<m/\s/> matches in a regular expression. Starting in Perl 5.18
779cf272 753this also matches what C<m/[[:space:]]/> does. Prior to 5.18, only the
398d098a
KW
754locale forms of this macro (the ones with C<LC> in their names) matched
755precisely what C<m/[[:space:]]/> does. In those releases, the only difference,
756in the non-locale variants, was that C<isSPACE()> did not match a vertical tab.
757(See L</isPSXSPC> for a macro that matches a vertical tab in all releases.)
dcccc8ff
KW
758See the L<top of this section|/Character classification> for an explanation of
759variants
da8c1a98
KW
760C<isSPACE_A>, C<isSPACE_L1>, C<isSPACE_uvchr>, C<isSPACE_utf8_safe>,
761C<isSPACE_LC>, C<isSPACE_LC_uvchr>, and C<isSPACE_LC_utf8_safe>.
0c82b6df 762
398d098a
KW
763=for apidoc Am|bool|isPSXSPC|char ch
764(short for Posix Space)
779cf272
KW
765Starting in 5.18, this is identical in all its forms to the
766corresponding C<isSPACE()> macros.
398d098a
KW
767The locale forms of this macro are identical to their corresponding
768C<isSPACE()> forms in all Perl releases. In releases prior to 5.18, the
769non-locale forms differ from their C<isSPACE()> forms only in that the
770C<isSPACE()> forms don't match a Vertical Tab, and the C<isPSXSPC()> forms do.
771Otherwise they are identical. Thus this macro is analogous to what
772C<m/[[:space:]]/> matches in a regular expression.
dcccc8ff 773See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
774variants C<isPSXSPC_A>, C<isPSXSPC_L1>, C<isPSXSPC_uvchr>, C<isPSXSPC_utf8_safe>,
775C<isPSXSPC_LC>, C<isPSXSPC_LC_uvchr>, and C<isPSXSPC_LC_utf8_safe>.
398d098a 776
954c1994 777=for apidoc Am|bool|isUPPER|char ch
2787a470 778Returns a boolean indicating whether the specified character is an
6aff1f14 779uppercase character, analogous to C<m/[[:upper:]]/>.
dcccc8ff 780See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
781variants C<isUPPER_A>, C<isUPPER_L1>, C<isUPPER_uvchr>, C<isUPPER_utf8_safe>,
782C<isUPPER_LC>, C<isUPPER_LC_uvchr>, and C<isUPPER_LC_utf8_safe>.
954c1994 783
243effed 784=for apidoc Am|bool|isPRINT|char ch
8eea39dd 785Returns a boolean indicating whether the specified character is a
6aff1f14 786printable character, analogous to C<m/[[:print:]]/>.
dcccc8ff
KW
787See the L<top of this section|/Character classification> for an explanation of
788variants
da8c1a98
KW
789C<isPRINT_A>, C<isPRINT_L1>, C<isPRINT_uvchr>, C<isPRINT_utf8_safe>,
790C<isPRINT_LC>, C<isPRINT_LC_uvchr>, and C<isPRINT_LC_utf8_safe>.
243effed
KW
791
792=for apidoc Am|bool|isWORDCHAR|char ch
793Returns a boolean indicating whether the specified character is a character
794that is a word character, analogous to what C<m/\w/> and C<m/[[:word:]]/> match
795in a regular expression. A word character is an alphabetic character, a
796decimal digit, a connecting punctuation character (such as an underscore), or
797a "mark" character that attaches to one of those (like some sort of accent).
798C<isALNUM()> is a synonym provided for backward compatibility, even though a
799word character includes more than the standard C language meaning of
800alphanumeric.
dcccc8ff 801See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
802variants C<isWORDCHAR_A>, C<isWORDCHAR_L1>, C<isWORDCHAR_uvchr>, and
803C<isWORDCHAR_utf8_safe>. C<isWORDCHAR_LC>, C<isWORDCHAR_LC_uvchr>, and
804C<isWORDCHAR_LC_utf8_safe> are also as described there, but additionally
805include the platform's native underscore.
8a58bdcf
KW
806
807=for apidoc Am|bool|isXDIGIT|char ch
808Returns a boolean indicating whether the specified character is a hexadecimal
243effed
KW
809digit. In the ASCII range these are C<[0-9A-Fa-f]>. Variants C<isXDIGIT_A()>
810and C<isXDIGIT_L1()> are identical to C<isXDIGIT()>.
dcccc8ff
KW
811See the L<top of this section|/Character classification> for an explanation of
812variants
da8c1a98
KW
813C<isXDIGIT_uvchr>, C<isXDIGIT_utf8_safe>, C<isXDIGIT_LC>, C<isXDIGIT_LC_uvchr>,
814and C<isXDIGIT_LC_utf8_safe>.
243effed 815
3c3ecf18
KW
816=for apidoc Am|bool|isIDFIRST|char ch
817Returns a boolean indicating whether the specified character can be the first
818character of an identifier. This is very close to, but not quite the same as
819the official Unicode property C<XID_Start>. The difference is that this
820returns true only if the input character also matches L</isWORDCHAR>.
dcccc8ff
KW
821See the L<top of this section|/Character classification> for an explanation of
822variants
da8c1a98
KW
823C<isIDFIRST_A>, C<isIDFIRST_L1>, C<isIDFIRST_uvchr>, C<isIDFIRST_utf8_safe>,
824C<isIDFIRST_LC>, C<isIDFIRST_LC_uvchr>, and C<isIDFIRST_LC_utf8_safe>.
3c3ecf18
KW
825
826=for apidoc Am|bool|isIDCONT|char ch
827Returns a boolean indicating whether the specified character can be the
828second or succeeding character of an identifier. This is very close to, but
829not quite the same as the official Unicode property C<XID_Continue>. The
830difference is that this returns true only if the input character also matches
dcccc8ff
KW
831L</isWORDCHAR>. See the L<top of this section|/Character classification> for
832an
d0da05db 833explanation of variants C<isIDCONT_A>, C<isIDCONT_L1>, C<isIDCONT_uvchr>,
da8c1a98
KW
834C<isIDCONT_utf8_safe>, C<isIDCONT_LC>, C<isIDCONT_LC_uvchr>, and
835C<isIDCONT_LC_utf8_safe>.
3c3ecf18 836
243effed 837=head1 Miscellaneous Functions
8eea39dd 838
95a59cab 839=for apidoc Am|U8|READ_XDIGIT|char str*
243effed 840Returns the value of an ASCII-range hex digit and advances the string pointer.
95a59cab
YO
841Behaviour is only well defined when isXDIGIT(*str) is true.
842
e7c1e6c1 843=head1 Character case changing
21da7284
KW
844Perl uses "full" Unicode case mappings. This means that converting a single
845character to another case may result in a sequence of more than one character.
846For example, the uppercase of C<E<223>> (LATIN SMALL LETTER SHARP S) is the two
847character sequence C<SS>. This presents some complications The lowercase of
848all characters in the range 0..255 is a single character, and thus
849C<L</toLOWER_L1>> is furnished. But, C<toUPPER_L1> can't exist, as it couldn't
850return a valid result for all legal inputs. Instead C<L</toUPPER_uvchr>> has
851an API that does allow every possible legal result to be returned.) Likewise
852no other function that is crippled by not being able to give the correct
853results for the full range of possible inputs has been implemented here.
e7c1e6c1 854
1f607577
KW
855=for apidoc Am|U8|toUPPER|U8 ch
856Converts the specified character to uppercase. If the input is anything but an
857ASCII lowercase character, that input character itself is returned. Variant
c753c8d3 858C<toUPPER_A> is equivalent.
954c1994 859
d0da05db
KW
860=for apidoc Am|UV|toUPPER_uvchr|UV cp|U8* s|STRLEN* lenp
861Converts the code point C<cp> to its uppercase version, and
862stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
863point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
864that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
865bytes since the uppercase version may be longer than the original character.
866
867The first code point of the uppercased version is returned
21da7284
KW
868(but note, as explained at L<the top of this section|/Character case
869changing>, that there may be more.)
1f607577 870
a239b1e2
KW
871=for apidoc Am|UV|toUPPER_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
872Converts the first UTF-8 encoded character in the sequence starting at C<p> and
873extending no further than S<C<e - 1>> to its uppercase version, and
1f607577
KW
874stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
875that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
876bytes since the uppercase version may be longer than the original character.
877
878The first code point of the uppercased version is returned
21da7284
KW
879(but note, as explained at L<the top of this section|/Character case
880changing>, that there may be more).
1f607577 881
a239b1e2
KW
882The suffix C<_safe> in the function's name indicates that it will not attempt
883to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
884true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
885input character is malformed in some way, the program may croak, or the
886function may return the REPLACEMENT CHARACTER, at the discretion of the
887implementation, and subject to change in future releases.
888
889=for apidoc Am|UV|toUPPER_utf8|U8* p|U8* s|STRLEN* lenp
890This is like C<L</toUPPER_utf8_safe>>, but doesn't have the C<e>
891parameter The function therefore can't check if it is reading
607313a1
KW
892beyond the end of the string. Starting in Perl v5.30, it will take the C<e>
893parameter, becoming a synonym for C<toUPPER_utf8_safe>. At that time every
894program that uses it will have to be changed to successfully compile. In the
895meantime, the first runtime call to C<toUPPER_utf8> from each call point in the
896program will raise a deprecation warning, enabled by default. You can convert
897your program now to use C<toUPPER_utf8_safe>, and avoid the warnings, and get an
898extra measure of protection, or you can wait until v5.30, when you'll be forced
899to add the C<e> parameter.
1f607577 900
25200305
KW
901=for apidoc Am|U8|toFOLD|U8 ch
902Converts the specified character to foldcase. If the input is anything but an
903ASCII uppercase character, that input character itself is returned. Variant
904C<toFOLD_A> is equivalent. (There is no equivalent C<to_FOLD_L1> for the full
d0da05db 905Latin1 range, as the full generality of L</toFOLD_uvchr> is needed there.)
25200305 906
d0da05db
KW
907=for apidoc Am|UV|toFOLD_uvchr|UV cp|U8* s|STRLEN* lenp
908Converts the code point C<cp> to its foldcase version, and
909stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
910point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
911that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
912bytes since the foldcase version may be longer than the original character.
913
914The first code point of the foldcased version is returned
21da7284
KW
915(but note, as explained at L<the top of this section|/Character case
916changing>, that there may be more).
1f607577 917
a239b1e2
KW
918=for apidoc Am|UV|toFOLD_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
919Converts the first UTF-8 encoded character in the sequence starting at C<p> and
920extending no further than S<C<e - 1>> to its foldcase version, and
1f607577
KW
921stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
922that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
923bytes since the foldcase version may be longer than the original character.
924
925The first code point of the foldcased version is returned
21da7284
KW
926(but note, as explained at L<the top of this section|/Character case
927changing>, that there may be more).
1f607577 928
a239b1e2
KW
929The suffix C<_safe> in the function's name indicates that it will not attempt
930to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
931true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
932input character is malformed in some way, the program may croak, or the
933function may return the REPLACEMENT CHARACTER, at the discretion of the
934implementation, and subject to change in future releases.
935
936=for apidoc Am|UV|toFOLD_utf8|U8* p|U8* s|STRLEN* lenp
937This is like C<L</toFOLD_utf8_safe>>, but doesn't have the C<e>
938parameter The function therefore can't check if it is reading
607313a1
KW
939beyond the end of the string. Starting in Perl v5.30, it will take the C<e>
940parameter, becoming a synonym for C<toFOLD_utf8_safe>. At that time every
941program that uses it will have to be changed to successfully compile. In the
942meantime, the first runtime call to C<toFOLD_utf8> from each call point in the
943program will raise a deprecation warning, enabled by default. You can convert
944your program now to use C<toFOLD_utf8_safe>, and avoid the warnings, and get an
945extra measure of protection, or you can wait until v5.30, when you'll be forced
946to add the C<e> parameter.
1f607577
KW
947
948=for apidoc Am|U8|toLOWER|U8 ch
949Converts the specified character to lowercase. If the input is anything but an
950ASCII uppercase character, that input character itself is returned. Variant
c753c8d3 951C<toLOWER_A> is equivalent.
954c1994 952
1f607577 953=for apidoc Am|U8|toLOWER_L1|U8 ch
b7d90381
KW
954Converts the specified Latin1 character to lowercase. The results are
955undefined if the input doesn't fit in a byte.
1f607577
KW
956
957=for apidoc Am|U8|toLOWER_LC|U8 ch
958Converts the specified character to lowercase using the current locale's rules,
959if possible; otherwise returns the input character itself.
960
d0da05db
KW
961=for apidoc Am|UV|toLOWER_uvchr|UV cp|U8* s|STRLEN* lenp
962Converts the code point C<cp> to its lowercase version, and
963stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
964point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
965that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
966bytes since the lowercase version may be longer than the original character.
967
968The first code point of the lowercased version is returned
21da7284
KW
969(but note, as explained at L<the top of this section|/Character case
970changing>, that there may be more).
1f607577 971
a239b1e2
KW
972
973=for apidoc Am|UV|toLOWER_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
974Converts the first UTF-8 encoded character in the sequence starting at C<p> and
975extending no further than S<C<e - 1>> to its lowercase version, and
1f607577
KW
976stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
977that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
978bytes since the lowercase version may be longer than the original character.
979
980The first code point of the lowercased version is returned
21da7284
KW
981(but note, as explained at L<the top of this section|/Character case
982changing>, that there may be more).
1f607577 983
a239b1e2
KW
984The suffix C<_safe> in the function's name indicates that it will not attempt
985to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
986true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
987input character is malformed in some way, the program may croak, or the
988function may return the REPLACEMENT CHARACTER, at the discretion of the
989implementation, and subject to change in future releases.
990
991=for apidoc Am|UV|toLOWER_utf8|U8* p|U8* s|STRLEN* lenp
992This is like C<L</toLOWER_utf8_safe>>, but doesn't have the C<e>
993parameter The function therefore can't check if it is reading
607313a1
KW
994beyond the end of the string. Starting in Perl v5.30, it will take the C<e>
995parameter, becoming a synonym for C<toLOWER_utf8_safe>. At that time every
996program that uses it will have to be changed to successfully compile. In the
997meantime, the first runtime call to C<toLOWER_utf8> from each call point in the
998program will raise a deprecation warning, enabled by default. You can convert
999your program now to use C<toLOWER_utf8_safe>, and avoid the warnings, and get an
1000extra measure of protection, or you can wait until v5.30, when you'll be forced
1001to add the C<e> parameter.
1f607577 1002
25200305
KW
1003=for apidoc Am|U8|toTITLE|U8 ch
1004Converts the specified character to titlecase. If the input is anything but an
1005ASCII lowercase character, that input character itself is returned. Variant
b7d90381 1006C<toTITLE_A> is equivalent. (There is no C<toTITLE_L1> for the full Latin1
d0da05db 1007range, as the full generality of L</toTITLE_uvchr> is needed there. Titlecase is
b7d90381 1008not a concept used in locale handling, so there is no functionality for that.)
25200305 1009
d0da05db
KW
1010=for apidoc Am|UV|toTITLE_uvchr|UV cp|U8* s|STRLEN* lenp
1011Converts the code point C<cp> to its titlecase version, and
1012stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
1013point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
1014that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
1015bytes since the titlecase version may be longer than the original character.
1016
1017The first code point of the titlecased version is returned
21da7284
KW
1018(but note, as explained at L<the top of this section|/Character case
1019changing>, that there may be more).
1f607577 1020
a239b1e2
KW
1021=for apidoc Am|UV|toTITLE_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
1022Converts the first UTF-8 encoded character in the sequence starting at C<p> and
1023extending no further than S<C<e - 1>> to its titlecase version, and
1f607577
KW
1024stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
1025that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
1026bytes since the titlecase version may be longer than the original character.
1027
1028The first code point of the titlecased version is returned
21da7284
KW
1029(but note, as explained at L<the top of this section|/Character case
1030changing>, that there may be more).
1f607577 1031
a239b1e2
KW
1032The suffix C<_safe> in the function's name indicates that it will not attempt
1033to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
1034true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
1035input character is malformed in some way, the program may croak, or the
1036function may return the REPLACEMENT CHARACTER, at the discretion of the
1037implementation, and subject to change in future releases.
1038
1039=for apidoc Am|UV|toTITLE_utf8|U8* p|U8* s|STRLEN* lenp
1040This is like C<L</toLOWER_utf8_safe>>, but doesn't have the C<e>
1041parameter The function therefore can't check if it is reading
607313a1
KW
1042beyond the end of the string. Starting in Perl v5.30, it will take the C<e>
1043parameter, becoming a synonym for C<toTITLE_utf8_safe>. At that time every
1044program that uses it will have to be changed to successfully compile. In the
1045meantime, the first runtime call to C<toTITLE_utf8> from each call point in the
1046program will raise a deprecation warning, enabled by default. You can convert
1047your program now to use C<toTITLE_utf8_safe>, and avoid the warnings, and get an
1048extra measure of protection, or you can wait until v5.30, when you'll be forced
1049to add the C<e> parameter.
1f607577 1050
954c1994 1051=cut
353c9b6f 1052
d0da05db 1053XXX Still undocumented isVERTWS_uvchr and _utf8; it's unclear what their names
1e222e4f
KW
1054really should be. Also toUPPER_LC and toFOLD_LC, which are subject to change,
1055and aren't general purpose as they don't work on U+DF, and assert against that.
243effed 1056
8a58bdcf 1057Note that these macros are repeated in Devel::PPPort, so should also be
62fa66b6
KW
1058patched there. The file as of this writing is cpan/Devel-PPPort/parts/inc/misc
1059
954c1994
GS
1060*/
1061
de9e2639
KW
1062/* Specify the widest unsigned type on the platform. */
1063#ifdef QUADKIND
1064# define WIDEST_UTYPE U64
7c062697
KW
1065#else
1066# define WIDEST_UTYPE U32
1067#endif
1068
3912bc88
KW
1069/* FITS_IN_8_BITS(c) returns true if c doesn't have a bit set other than in
1070 * the lower 8. It is designed to be hopefully bomb-proof, making sure that no
1071 * bits of information are lost even on a 64-bit machine, but to get the
1072 * compiler to optimize it out if possible. This is because Configure makes
1073 * sure that the machine has an 8-bit byte, so if c is stored in a byte, the
1074 * sizeof() guarantees that this evaluates to a constant true at compile time.
7e75d1a1
JH
1075 *
1076 * For Coverity, be always true, because otherwise Coverity thinks
1077 * it finds several expressions that are always true, independent
1078 * of operands. Well, they are, but that is kind of the point.
220c71bf 1079 */
7e75d1a1 1080#ifndef __COVERITY__
6c5b02ac
KW
1081 /* The '| 0' part ensures a compiler error if c is not integer (like e.g., a
1082 * pointer) */
1083#define FITS_IN_8_BITS(c) ( (sizeof(c) == 1) \
ace3ad0f 1084 || !(((WIDEST_UTYPE)((c) | 0)) & ~0xFF))
7e75d1a1
JH
1085#else
1086#define FITS_IN_8_BITS(c) (1)
1087#endif
cf301eb7 1088
41f43cc2 1089#ifdef EBCDIC
b6340bd0 1090# ifndef _ALL_SOURCE
0852beac
KW
1091 /* The native libc isascii() et.al. functions return the wrong results
1092 * on at least z/OS unless this is defined. */
b6340bd0
KW
1093# error _ALL_SOURCE should probably be defined
1094# endif
41f43cc2 1095#else
0852beac
KW
1096 /* There is a simple definition of ASCII for ASCII platforms. But the
1097 * EBCDIC one isn't so simple, so is defined using table look-up like the
9c903d59 1098 * other macros below.
3f3c579d
KW
1099 *
1100 * The cast here is used instead of '(c) >= 0', because some compilers emit
1101 * a warning that that test is always true when the parameter is an
1102 * unsigned type. khw supposes that it could be written as
1103 * && ((c) == '\0' || (c) > 0)
1104 * to avoid the message, but the cast will likely avoid extra branches even
1105 * with stupid compilers.
1106 *
1107 * The '| 0' part ensures a compiler error if c is not integer (like e.g.,
1108 * a pointer) */
9c903d59 1109# define isASCII(c) ((WIDEST_UTYPE)((c) | 0) < 128)
41f43cc2
KW
1110#endif
1111
38694112
KW
1112/* Take the eight possible bit patterns of the lower 3 bits and you get the
1113 * lower 3 bits of the 8 octal digits, in both ASCII and EBCDIC, so those bits
1114 * can be ignored. If the rest match '0', we have an octal */
1115#define isOCTAL_A(c) (((WIDEST_UTYPE)((c) | 0) & ~7) == '0')
c2da0b36 1116
9fb1bf9d 1117#ifdef H_PERL /* If have access to perl.h, lookup in its table */
f4cdb42c 1118
a500dc72
KW
1119/* Character class numbers. For internal core Perl use only. The ones less
1120 * than 32 are used in PL_charclass[] and the ones up through the one that
1121 * corresponds to <_HIGHEST_REGCOMP_DOT_H_SYNC> are used by regcomp.h and
1122 * related files. PL_charclass ones use names used in l1_char_class_tab.h but
1123 * their actual definitions are here. If that file has a name not used here,
1124 * it won't compile.
1709d539
KW
1125 *
1126 * The first group of these is ordered in what I (khw) estimate to be the
31c7f561
KW
1127 * frequency of their use. This gives a slight edge to exiting a loop earlier
1128 * (in reginclass() in regexec.c) */
1709d539
KW
1129# define _CC_WORDCHAR 0 /* \w and [:word:] */
1130# define _CC_DIGIT 1 /* \d and [:digit:] */
1131# define _CC_ALPHA 2 /* [:alpha:] */
1132# define _CC_LOWER 3 /* [:lower:] */
1133# define _CC_UPPER 4 /* [:upper:] */
1134# define _CC_PUNCT 5 /* [:punct:] */
1135# define _CC_PRINT 6 /* [:print:] */
15861f94 1136# define _CC_ALPHANUMERIC 7 /* [:alnum:] */
1709d539 1137# define _CC_GRAPH 8 /* [:graph:] */
359b005e 1138# define _CC_CASED 9 /* [:lower:] or [:upper:] under /i */
1709d539 1139
b0d691b2 1140#define _FIRST_NON_SWASH_CC 10
1709d539
KW
1141/* The character classes above are implemented with swashes. The second group
1142 * (just below) contains the ones implemented without. These are also sorted
1143 * in rough order of the frequency of their use, except that \v should be last,
1144 * as it isn't a real Posix character class, and some (small) inefficiencies in
1145 * regular expression handling would be introduced by putting it in the middle
1146 * of those that are. Also, cntrl and ascii come after the others as it may be
1147 * useful to group these which have no members that match above Latin1, (or
1148 * above ASCII in the latter case) */
1149
779cf272 1150# define _CC_SPACE 10 /* \s, [:space:] */
34aeb2e9
KW
1151# define _CC_PSXSPC _CC_SPACE /* XXX Temporary, can be removed
1152 when the deprecated isFOO_utf8()
1153 functions are removed */
b0d691b2
KW
1154# define _CC_BLANK 11 /* [:blank:] */
1155# define _CC_XDIGIT 12 /* [:xdigit:] */
779cf272
KW
1156# define _CC_CNTRL 13 /* [:cntrl:] */
1157# define _CC_ASCII 14 /* [:ascii:] */
1158# define _CC_VERTSPACE 15 /* \v */
1709d539 1159
a0947d7b
KW
1160# define _HIGHEST_REGCOMP_DOT_H_SYNC _CC_VERTSPACE
1161
1709d539 1162/* The members of the third group below do not need to be coordinated with data
3ffc8c70 1163 * structures in regcomp.[ch] and regexec.c. */
779cf272
KW
1164# define _CC_IDFIRST 16
1165# define _CC_CHARNAME_CONT 17
1166# define _CC_NONLATIN1_FOLD 18
1167# define _CC_NONLATIN1_SIMPLE_FOLD 19
1168# define _CC_QUOTEMETA 20
1169# define _CC_NON_FINAL_FOLD 21
1170# define _CC_IS_IN_SOME_FOLD 22
1171# define _CC_MNEMONIC_CNTRL 23
073c22b3 1172
34aeb2e9
KW
1173# define _CC_IDCONT 24 /* XXX Temporary, can be removed when the deprecated
1174 isFOO_utf8() functions are removed */
1175
073c22b3
KW
1176/* This next group is only used on EBCDIC platforms, so theoretically could be
1177 * shared with something entirely different that's only on ASCII platforms */
5d5376e2
KW
1178# define _CC_UTF8_START_BYTE_IS_FOR_AT_LEAST_SURROGATE 28
1179# define _CC_UTF8_IS_START 29
1180# define _CC_UTF8_IS_DOWNGRADEABLE_START 30
1181# define _CC_UTF8_IS_CONTINUATION 31
1182/* Unused: 24-27
f4cdb42c
KW
1183 * If more bits are needed, one could add a second word for non-64bit
1184 * QUAD_IS_INT systems, using some #ifdefs to distinguish between having a 2nd
37ede926
KW
1185 * word or not. The IS_IN_SOME_FOLD bit is the most easily expendable, as it
1186 * is used only for optimization (as of this writing), and differs in the
1187 * Latin1 range from the ALPHA bit only in two relatively unimportant
a500dc72 1188 * characters: the masculine and feminine ordinal indicators, so removing it
073c22b3
KW
1189 * would just cause /i regexes which match them to run less efficiently.
1190 * Similarly the EBCDIC-only bits are used just for speed, and could be
1191 * replaced by other means */
96ac0975 1192
3a371f2f
KW
1193#if defined(PERL_CORE) || defined(PERL_EXT)
1194/* An enum version of the character class numbers, to help compilers
1195 * optimize */
1196typedef enum {
3a371f2f 1197 _CC_ENUM_ALPHA = _CC_ALPHA,
e8d596e0
KW
1198 _CC_ENUM_ALPHANUMERIC = _CC_ALPHANUMERIC,
1199 _CC_ENUM_ASCII = _CC_ASCII,
1200 _CC_ENUM_BLANK = _CC_BLANK,
b0d691b2 1201 _CC_ENUM_CASED = _CC_CASED,
e8d596e0 1202 _CC_ENUM_CNTRL = _CC_CNTRL,
3a371f2f
KW
1203 _CC_ENUM_DIGIT = _CC_DIGIT,
1204 _CC_ENUM_GRAPH = _CC_GRAPH,
1205 _CC_ENUM_LOWER = _CC_LOWER,
1206 _CC_ENUM_PRINT = _CC_PRINT,
1207 _CC_ENUM_PUNCT = _CC_PUNCT,
e8d596e0 1208 _CC_ENUM_SPACE = _CC_SPACE,
3a371f2f 1209 _CC_ENUM_UPPER = _CC_UPPER,
e8d596e0 1210 _CC_ENUM_VERTSPACE = _CC_VERTSPACE,
3a371f2f 1211 _CC_ENUM_WORDCHAR = _CC_WORDCHAR,
e8d596e0 1212 _CC_ENUM_XDIGIT = _CC_XDIGIT
3a371f2f
KW
1213} _char_class_number;
1214#endif
1215
63c61c3f 1216#define POSIX_SWASH_COUNT _FIRST_NON_SWASH_CC
86f72d56 1217#define POSIX_CC_COUNT (_HIGHEST_REGCOMP_DOT_H_SYNC + 1)
63c61c3f 1218
f1bfbdda 1219#if defined(PERL_IN_UTF8_C)
922e8cb4 1220# if _CC_WORDCHAR != 0 || _CC_DIGIT != 1 || _CC_ALPHA != 2 || _CC_LOWER != 3 \
b0d691b2
KW
1221 || _CC_UPPER != 4 || _CC_PUNCT != 5 || _CC_PRINT != 6 \
1222 || _CC_ALPHANUMERIC != 7 || _CC_GRAPH != 8 || _CC_CASED != 9
922e8cb4
KW
1223 #error Need to adjust order of swash_property_names[]
1224# endif
1225
b44daeeb
KW
1226/* This is declared static in each of the few files that this is #defined for
1227 * to keep them from being publicly accessible. Hence there is a small amount
1228 * of wasted space */
1229
922e8cb4
KW
1230static const char* const swash_property_names[] = {
1231 "XPosixWord",
1232 "XPosixDigit",
1233 "XPosixAlpha",
1234 "XPosixLower",
1235 "XPosixUpper",
1236 "XPosixPunct",
1237 "XPosixPrint",
1238 "XPosixAlnum",
b0d691b2
KW
1239 "XPosixGraph",
1240 "Cased"
922e8cb4
KW
1241};
1242#endif
1243
6635f04f 1244START_EXTERN_C
96ac0975
NC
1245# ifdef DOINIT
1246EXTCONST U32 PL_charclass[] = {
1247# include "l1_char_class_tab.h"
1248};
1249
1250# else /* ! DOINIT */
1251EXTCONST U32 PL_charclass[];
1252# endif
6635f04f 1253END_EXTERN_C
96ac0975 1254
265c1f46 1255 /* The 1U keeps Solaris from griping when shifting sets the uppermost bit */
430b7c70 1256# define _CC_mask(classnum) (1U << (classnum))
4650c663
KW
1257
1258 /* For internal core Perl use only: the base macro for defining macros like
1259 * isALPHA */
ff7ecfc3 1260# define _generic_isCC(c, classnum) cBOOL(FITS_IN_8_BITS(c) \
f4cd282c 1261 && (PL_charclass[(U8) (c)] & _CC_mask(classnum)))
4eeeb416 1262
f4cdb42c
KW
1263 /* The mask for the _A versions of the macros; it just adds in the bit for
1264 * ASCII. */
1265# define _CC_mask_A(classnum) (_CC_mask(classnum) | _CC_mask(_CC_ASCII))
1266
4650c663
KW
1267 /* For internal core Perl use only: the base macro for defining macros like
1268 * isALPHA_A. The foo_A version makes sure that both the desired bit and
1269 * the ASCII bit are present */
b7d90381
KW
1270# define _generic_isCC_A(c, classnum) (FITS_IN_8_BITS(c) \
1271 && ((PL_charclass[(U8) (c)] & _CC_mask_A(classnum)) \
1272 == _CC_mask_A(classnum)))
f4cdb42c 1273
f4cdb42c 1274# define isALPHA_A(c) _generic_isCC_A(c, _CC_ALPHA)
15861f94 1275# define isALPHANUMERIC_A(c) _generic_isCC_A(c, _CC_ALPHANUMERIC)
f4cdb42c
KW
1276# define isBLANK_A(c) _generic_isCC_A(c, _CC_BLANK)
1277# define isCNTRL_A(c) _generic_isCC_A(c, _CC_CNTRL)
96ca48da 1278# define isDIGIT_A(c) _generic_isCC(c, _CC_DIGIT) /* No non-ASCII digits */
f4cdb42c
KW
1279# define isGRAPH_A(c) _generic_isCC_A(c, _CC_GRAPH)
1280# define isLOWER_A(c) _generic_isCC_A(c, _CC_LOWER)
1281# define isPRINT_A(c) _generic_isCC_A(c, _CC_PRINT)
f4cdb42c
KW
1282# define isPUNCT_A(c) _generic_isCC_A(c, _CC_PUNCT)
1283# define isSPACE_A(c) _generic_isCC_A(c, _CC_SPACE)
1284# define isUPPER_A(c) _generic_isCC_A(c, _CC_UPPER)
1285# define isWORDCHAR_A(c) _generic_isCC_A(c, _CC_WORDCHAR)
b7d90381
KW
1286# define isXDIGIT_A(c) _generic_isCC(c, _CC_XDIGIT) /* No non-ASCII xdigits
1287 */
d95f8b6a 1288# define isIDFIRST_A(c) _generic_isCC_A(c, _CC_IDFIRST)
3ded5eb0
KW
1289# define isALPHA_L1(c) _generic_isCC(c, _CC_ALPHA)
1290# define isALPHANUMERIC_L1(c) _generic_isCC(c, _CC_ALPHANUMERIC)
1291# define isBLANK_L1(c) _generic_isCC(c, _CC_BLANK)
1292
1293 /* continuation character for legal NAME in \N{NAME} */
1294# define isCHARNAME_CONT(c) _generic_isCC(c, _CC_CHARNAME_CONT)
1295
1296# define isCNTRL_L1(c) _generic_isCC(c, _CC_CNTRL)
1297# define isGRAPH_L1(c) _generic_isCC(c, _CC_GRAPH)
1298# define isLOWER_L1(c) _generic_isCC(c, _CC_LOWER)
1299# define isPRINT_L1(c) _generic_isCC(c, _CC_PRINT)
b7d90381 1300# define isPSXSPC_L1(c) isSPACE_L1(c)
3ded5eb0
KW
1301# define isPUNCT_L1(c) _generic_isCC(c, _CC_PUNCT)
1302# define isSPACE_L1(c) _generic_isCC(c, _CC_SPACE)
1303# define isUPPER_L1(c) _generic_isCC(c, _CC_UPPER)
1304# define isWORDCHAR_L1(c) _generic_isCC(c, _CC_WORDCHAR)
1305# define isIDFIRST_L1(c) _generic_isCC(c, _CC_IDFIRST)
f4cdb42c 1306
0852beac
KW
1307# ifdef EBCDIC
1308# define isASCII(c) _generic_isCC(c, _CC_ASCII)
1309# endif
1310
f12c0118
KW
1311 /* Participates in a single-character fold with a character above 255 */
1312# 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)))
1313
1314 /* Like the above, but also can be part of a multi-char fold */
f4cd282c 1315# 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 1316
4eeeb416 1317# define _isQUOTEMETA(c) _generic_isCC(c, _CC_QUOTEMETA)
26faadbd 1318# define _IS_NON_FINAL_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
b7d90381 1319 _generic_isCC(c, _CC_NON_FINAL_FOLD)
37ede926 1320# define _IS_IN_SOME_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
b7d90381 1321 _generic_isCC(c, _CC_IS_IN_SOME_FOLD)
8e35b056
KW
1322# define _IS_MNEMONIC_CNTRL_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
1323 _generic_isCC(c, _CC_MNEMONIC_CNTRL)
687c8d01 1324#else /* else we don't have perl.h H_PERL */
3ded5eb0
KW
1325
1326 /* If we don't have perl.h, we are compiling a utility program. Below we
1327 * hard-code various macro definitions that wouldn't otherwise be available
fc273927 1328 * to it. Most are coded based on first principles. These are written to
74665a89 1329 * avoid EBCDIC vs. ASCII #ifdef's as much as possible. */
0852beac
KW
1330# define isDIGIT_A(c) ((c) <= '9' && (c) >= '0')
1331# define isBLANK_A(c) ((c) == ' ' || (c) == '\t')
74665a89
KW
1332# define isSPACE_A(c) (isBLANK_A(c) \
1333 || (c) == '\n' \
1334 || (c) == '\r' \
1335 || (c) == '\v' \
0852beac 1336 || (c) == '\f')
74665a89
KW
1337 /* On EBCDIC, there are gaps between 'i' and 'j'; 'r' and 's'. Same for
1338 * uppercase. The tests for those aren't necessary on ASCII, but hurt only
1339 * performance (if optimization isn't on), and allow the same code to be
1340 * used for both platform types */
1341# define isLOWER_A(c) ((c) >= 'a' && (c) <= 'z' \
1342 && ( (c) <= 'i' \
1343 || ((c) >= 'j' && (c) <= 'r') \
1344 || (c) >= 's'))
1345# define isUPPER_A(c) ((c) >= 'A' && (c) <= 'Z' \
1346 && ( (c) <= 'I' \
1347 || ((c) >= 'J' && (c) <= 'R') \
1348 || (c) >= 'S'))
a4d7a999
KW
1349# define isALPHA_A(c) (isUPPER_A(c) || isLOWER_A(c))
1350# define isALPHANUMERIC_A(c) (isALPHA_A(c) || isDIGIT_A(c))
3ded5eb0 1351# define isWORDCHAR_A(c) (isALPHANUMERIC_A(c) || (c) == '_')
0852beac 1352# define isIDFIRST_A(c) (isALPHA_A(c) || (c) == '_')
74665a89
KW
1353# define isXDIGIT_A(c) (isDIGIT_A(c) \
1354 || ((c) >= 'a' && (c) <= 'f') \
a4d7a999 1355 || ((c) <= 'F' && (c) >= 'A'))
74665a89
KW
1356# define isPUNCT_A(c) ((c) == '-' || (c) == '!' || (c) == '"' \
1357 || (c) == '#' || (c) == '$' || (c) == '%' \
1358 || (c) == '&' || (c) == '\'' || (c) == '(' \
1359 || (c) == ')' || (c) == '*' || (c) == '+' \
1360 || (c) == ',' || (c) == '.' || (c) == '/' \
1361 || (c) == ':' || (c) == ';' || (c) == '<' \
1362 || (c) == '=' || (c) == '>' || (c) == '?' \
1363 || (c) == '@' || (c) == '[' || (c) == '\\' \
1364 || (c) == ']' || (c) == '^' || (c) == '_' \
1365 || (c) == '`' || (c) == '{' || (c) == '|' \
1366 || (c) == '}' || (c) == '~')
1367# define isGRAPH_A(c) (isALPHANUMERIC_A(c) || isPUNCT_A(c))
1368# define isPRINT_A(c) (isGRAPH_A(c) || (c) == ' ')
3ded5eb0 1369
0852beac 1370# ifdef EBCDIC
74665a89
KW
1371 /* The below is accurate for the 3 EBCDIC code pages traditionally
1372 * supported by perl. The only difference between them in the controls
1373 * is the position of \n, and that is represented symbolically below */
1374# define isCNTRL_A(c) ((c) == '\0' || (c) == '\a' || (c) == '\b' \
1375 || (c) == '\f' || (c) == '\n' || (c) == '\r' \
1376 || (c) == '\t' || (c) == '\v' \
1377 || ((c) <= 3 && (c) >= 1) /* SOH, STX, ETX */ \
1378 || (c) == 7 /* U+7F DEL */ \
1379 || ((c) <= 0x13 && (c) >= 0x0E) /* SO, SI */ \
1380 /* DLE, DC[1-3] */ \
1381 || (c) == 0x18 /* U+18 CAN */ \
1382 || (c) == 0x19 /* U+19 EOM */ \
1383 || ((c) <= 0x1F && (c) >= 0x1C) /* [FGRU]S */ \
1384 || (c) == 0x26 /* U+17 ETB */ \
1385 || (c) == 0x27 /* U+1B ESC */ \
1386 || (c) == 0x2D /* U+05 ENQ */ \
1387 || (c) == 0x2E /* U+06 ACK */ \
1388 || (c) == 0x32 /* U+16 SYN */ \
1389 || (c) == 0x37 /* U+04 EOT */ \
1390 || (c) == 0x3C /* U+14 DC4 */ \
1391 || (c) == 0x3D /* U+15 NAK */ \
1392 || (c) == 0x3F)/* U+1A SUB */
0852beac 1393# define isASCII(c) (isCNTRL_A(c) || isPRINT_A(c))
74665a89
KW
1394# else /* isASCII is already defined for ASCII platforms, so can use that to
1395 define isCNTRL */
1396# define isCNTRL_A(c) (isASCII(c) && ! isPRINT_A(c))
0852beac
KW
1397# endif
1398
3ffc8c70 1399 /* The _L1 macros may be unnecessary for the utilities; I (khw) added them
caa94d35
KW
1400 * during debugging, and it seems best to keep them. We may be called
1401 * without NATIVE_TO_LATIN1 being defined. On ASCII platforms, it doesn't
1402 * do anything anyway, so make it not a problem */
1403# if ! defined(EBCDIC) && ! defined(NATIVE_TO_LATIN1)
1404# define NATIVE_TO_LATIN1(ch) (ch)
1405# endif
3ded5eb0
KW
1406# define isALPHA_L1(c) (isUPPER_L1(c) || isLOWER_L1(c))
1407# define isALPHANUMERIC_L1(c) (isALPHA_L1(c) || isDIGIT_A(c))
1408# define isBLANK_L1(c) (isBLANK_A(c) \
1409 || (FITS_IN_8_BITS(c) \
1410 && NATIVE_TO_LATIN1((U8) c) == 0xA0))
1411# define isCNTRL_L1(c) (FITS_IN_8_BITS(c) && (! isPRINT_L1(c)))
1412# define isGRAPH_L1(c) (isPRINT_L1(c) && (! isBLANK_L1(c)))
1413# define isLOWER_L1(c) (isLOWER_A(c) \
1414 || (FITS_IN_8_BITS(c) \
ae683a5f 1415 && (( NATIVE_TO_LATIN1((U8) c) >= 0xDF \
3ded5eb0
KW
1416 && NATIVE_TO_LATIN1((U8) c) != 0xF7) \
1417 || NATIVE_TO_LATIN1((U8) c) == 0xAA \
1418 || NATIVE_TO_LATIN1((U8) c) == 0xBA \
1419 || NATIVE_TO_LATIN1((U8) c) == 0xB5)))
1420# define isPRINT_L1(c) (isPRINT_A(c) \
1421 || (FITS_IN_8_BITS(c) \
1422 && NATIVE_TO_LATIN1((U8) c) >= 0xA0))
3ded5eb0
KW
1423# define isPUNCT_L1(c) (isPUNCT_A(c) \
1424 || (FITS_IN_8_BITS(c) \
ae683a5f 1425 && ( NATIVE_TO_LATIN1((U8) c) == 0xA1 \
3ded5eb0
KW
1426 || NATIVE_TO_LATIN1((U8) c) == 0xA7 \
1427 || NATIVE_TO_LATIN1((U8) c) == 0xAB \
1428 || NATIVE_TO_LATIN1((U8) c) == 0xB6 \
1429 || NATIVE_TO_LATIN1((U8) c) == 0xB7 \
1430 || NATIVE_TO_LATIN1((U8) c) == 0xBB \
1431 || NATIVE_TO_LATIN1((U8) c) == 0xBF)))
1432# define isSPACE_L1(c) (isSPACE_A(c) \
1433 || (FITS_IN_8_BITS(c) \
ae683a5f 1434 && ( NATIVE_TO_LATIN1((U8) c) == 0x85 \
3ded5eb0
KW
1435 || NATIVE_TO_LATIN1((U8) c) == 0xA0)))
1436# define isUPPER_L1(c) (isUPPER_A(c) \
1437 || (FITS_IN_8_BITS(c) \
ae683a5f 1438 && ( NATIVE_TO_LATIN1((U8) c) >= 0xC0 \
3ded5eb0
KW
1439 && NATIVE_TO_LATIN1((U8) c) <= 0xDE \
1440 && NATIVE_TO_LATIN1((U8) c) != 0xD7)))
1441# define isWORDCHAR_L1(c) (isIDFIRST_L1(c) || isDIGIT_A(c))
1442# define isIDFIRST_L1(c) (isALPHA_L1(c) || NATIVE_TO_LATIN1(c) == '_')
1443# define isCHARNAME_CONT(c) (isWORDCHAR_L1(c) \
1444 || isBLANK_L1(c) \
1445 || (c) == '-' \
1446 || (c) == '(' \
1447 || (c) == ')')
1448 /* The following are not fully accurate in the above-ASCII range. I (khw)
1449 * don't think it's necessary to be so for the purposes where this gets
1450 * compiled */
1451# define _isQUOTEMETA(c) (FITS_IN_8_BITS(c) && ! isWORDCHAR_L1(c))
1452# define _IS_IN_SOME_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) isALPHA_L1(c)
1453
1454 /* And these aren't accurate at all. They are useful only for above
1455 * Latin1, which utilities and bootstrapping don't deal with */
1456# define _IS_NON_FINAL_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) 0
6838b41e 1457# define _HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) 0
3ded5eb0
KW
1458# define _HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) 0
1459
1460 /* Many of the macros later in this file are defined in terms of these. By
1461 * implementing them with a function, which converts the class number into
1462 * a call to the desired macro, all of the later ones work. However, that
1463 * function won't be actually defined when building a utility program (no
1464 * perl.h), and so a compiler error will be generated if one is attempted
1465 * to be used. And the above-Latin1 code points require Unicode tables to
1466 * be present, something unlikely to be the case when bootstrapping */
1467# define _generic_isCC(c, classnum) \
1468 (FITS_IN_8_BITS(c) && S_bootstrap_ctype((U8) (c), (classnum), TRUE))
1469# define _generic_isCC_A(c, classnum) \
1470 (FITS_IN_8_BITS(c) && S_bootstrap_ctype((U8) (c), (classnum), FALSE))
687c8d01 1471#endif /* End of no perl.h H_PERL */
8a58bdcf 1472
e66b99e9
KW
1473#define isALPHANUMERIC(c) isALPHANUMERIC_A(c)
1474#define isALPHA(c) isALPHA_A(c)
0852beac
KW
1475#define isASCII_A(c) isASCII(c)
1476#define isASCII_L1(c) isASCII(c)
e66b99e9
KW
1477#define isBLANK(c) isBLANK_A(c)
1478#define isCNTRL(c) isCNTRL_A(c)
1479#define isDIGIT(c) isDIGIT_A(c)
1480#define isGRAPH(c) isGRAPH_A(c)
1481#define isIDFIRST(c) isIDFIRST_A(c)
1482#define isLOWER(c) isLOWER_A(c)
1483#define isPRINT(c) isPRINT_A(c)
779cf272 1484#define isPSXSPC_A(c) isSPACE_A(c)
e66b99e9 1485#define isPSXSPC(c) isPSXSPC_A(c)
779cf272 1486#define isPSXSPC_L1(c) isSPACE_L1(c)
e66b99e9
KW
1487#define isPUNCT(c) isPUNCT_A(c)
1488#define isSPACE(c) isSPACE_A(c)
1489#define isUPPER(c) isUPPER_A(c)
1490#define isWORDCHAR(c) isWORDCHAR_A(c)
1491#define isXDIGIT(c) isXDIGIT_A(c)
1492
1493/* ASCII casing. These could also be written as
1494 #define toLOWER(c) (isASCII(c) ? toLOWER_LATIN1(c) : (c))
1495 #define toUPPER(c) (isASCII(c) ? toUPPER_LATIN1_MOD(c) : (c))
1496 which uses table lookup and mask instead of subtraction. (This would
1497 work because the _MOD does not apply in the ASCII range) */
68067e4e
DM
1498#define toLOWER(c) (isUPPER(c) ? (U8)((c) + ('a' - 'A')) : (c))
1499#define toUPPER(c) (isLOWER(c) ? (U8)((c) - ('a' - 'A')) : (c))
bbce6d69 1500
25200305
KW
1501/* In the ASCII range, these are equivalent to what they're here defined to be.
1502 * But by creating these definitions, other code doesn't have to be aware of
1503 * this detail */
1504#define toFOLD(c) toLOWER(c)
25200305
KW
1505#define toTITLE(c) toUPPER(c)
1506
c753c8d3
KW
1507#define toLOWER_A(c) toLOWER(c)
1508#define toUPPER_A(c) toUPPER(c)
25200305
KW
1509#define toFOLD_A(c) toFOLD(c)
1510#define toTITLE_A(c) toTITLE(c)
1a0901db 1511
4650c663 1512/* Use table lookup for speed; returns the input itself if is out-of-range */
b2bf251f 1513#define toLOWER_LATIN1(c) ((! FITS_IN_8_BITS(c)) \
8e7c6e7d 1514 ? (c) \
f4cd282c 1515 : PL_latin1_lc[ (U8) (c) ])
c753c8d3
KW
1516#define toLOWER_L1(c) toLOWER_LATIN1(c) /* Synonym for consistency */
1517
1a0901db 1518/* Modified uc. Is correct uc except for three non-ascii chars which are
4650c663
KW
1519 * all mapped to one of them, and these need special handling; returns the
1520 * input itself if is out-of-range */
b2bf251f 1521#define toUPPER_LATIN1_MOD(c) ((! FITS_IN_8_BITS(c)) \
8e7c6e7d 1522 ? (c) \
f4cd282c 1523 : PL_mod_latin1_uc[ (U8) (c) ])
31f05a37 1524#define IN_UTF8_CTYPE_LOCALE PL_in_utf8_CTYPE_locale
84061b6a 1525
beab9ebe
KW
1526/* Use foo_LC_uvchr() instead of these for beyond the Latin1 range */
1527
1528/* For internal core Perl use only: the base macro for defining macros like
1529 * isALPHA_LC, which uses the current LC_CTYPE locale. 'c' is the code point
31f05a37
KW
1530 * (0-255) to check. In a UTF-8 locale, the result is the same as calling
1531 * isFOO_L1(); the 'utf8_locale_classnum' parameter is something like
1532 * _CC_UPPER, which gives the class number for doing this. For non-UTF-8
1533 * locales, the code to actually do the test this is passed in 'non_utf8'. If
1534 * 'c' is above 255, 0 is returned. For accessing the full range of possible
1535 * code points under locale rules, use the macros based on _generic_LC_uvchr
1536 * instead of this. */
beab9ebe
KW
1537#define _generic_LC_base(c, utf8_locale_classnum, non_utf8) \
1538 (! FITS_IN_8_BITS(c) \
1539 ? 0 \
31f05a37
KW
1540 : IN_UTF8_CTYPE_LOCALE \
1541 ? cBOOL(PL_charclass[(U8) (c)] & _CC_mask(utf8_locale_classnum)) \
beab9ebe
KW
1542 : cBOOL(non_utf8))
1543
1544/* For internal core Perl use only: a helper macro for defining macros like
1545 * isALPHA_LC. 'c' is the code point (0-255) to check. The function name to
1546 * actually do this test is passed in 'non_utf8_func', which is called on 'c',
1547 * casting 'c' to the macro _LC_CAST, which should not be parenthesized. See
1548 * _generic_LC_base for more info */
1549#define _generic_LC(c, utf8_locale_classnum, non_utf8_func) \
1550 _generic_LC_base(c,utf8_locale_classnum, \
1551 non_utf8_func( (_LC_CAST) (c)))
1552
1553/* For internal core Perl use only: like _generic_LC, but also returns TRUE if
1554 * 'c' is the platform's native underscore character */
1555#define _generic_LC_underscore(c,utf8_locale_classnum,non_utf8_func) \
1556 _generic_LC_base(c, utf8_locale_classnum, \
1557 (non_utf8_func( (_LC_CAST) (c)) \
1558 || (char)(c) == '_'))
1559
1560/* These next three are also for internal core Perl use only: case-change
1561 * helper macros */
1562#define _generic_toLOWER_LC(c, function, cast) (! FITS_IN_8_BITS(c) \
1563 ? (c) \
31f05a37
KW
1564 : (IN_UTF8_CTYPE_LOCALE) \
1565 ? PL_latin1_lc[ (U8) (c) ] \
d2a8cf7f 1566 : (cast)function((cast)(c)))
beab9ebe 1567
31f05a37
KW
1568/* Note that the result can be larger than a byte in a UTF-8 locale. It
1569 * returns a single value, so can't adequately return the upper case of LATIN
1570 * SMALL LETTER SHARP S in a UTF-8 locale (which should be a string of two
1571 * values "SS"); instead it asserts against that under DEBUGGING, and
1572 * otherwise returns its input */
beab9ebe
KW
1573#define _generic_toUPPER_LC(c, function, cast) \
1574 (! FITS_IN_8_BITS(c) \
1575 ? (c) \
31f05a37 1576 : ((! IN_UTF8_CTYPE_LOCALE) \
b7d90381 1577 ? (cast)function((cast)(c)) \
31f05a37
KW
1578 : ((((U8)(c)) == MICRO_SIGN) \
1579 ? GREEK_CAPITAL_LETTER_MU \
1580 : ((((U8)(c)) == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS) \
1581 ? LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS \
1582 : ((((U8)(c)) == LATIN_SMALL_LETTER_SHARP_S) \
1583 ? (__ASSERT_(0) (c)) \
1584 : PL_mod_latin1_uc[ (U8) (c) ])))))
1585
1586/* Note that the result can be larger than a byte in a UTF-8 locale. It
1587 * returns a single value, so can't adequately return the fold case of LATIN
1588 * SMALL LETTER SHARP S in a UTF-8 locale (which should be a string of two
1589 * values "ss"); instead it asserts against that under DEBUGGING, and
1590 * otherwise returns its input */
beab9ebe 1591#define _generic_toFOLD_LC(c, function, cast) \
4d7db1b9 1592 ((UNLIKELY((c) == MICRO_SIGN) && IN_UTF8_CTYPE_LOCALE) \
31f05a37 1593 ? GREEK_SMALL_LETTER_MU \
4d7db1b9
KW
1594 : (__ASSERT_(! IN_UTF8_CTYPE_LOCALE \
1595 || (c) != LATIN_SMALL_LETTER_SHARP_S) \
1596 _generic_toLOWER_LC(c, function, cast)))
beab9ebe 1597
84061b6a 1598/* Use the libc versions for these if available. */
f05550c0 1599#if defined(HAS_ISASCII)
84061b6a
KW
1600# define isASCII_LC(c) (FITS_IN_8_BITS(c) && isascii( (U8) (c)))
1601#else
1602# define isASCII_LC(c) isASCII(c)
1603#endif
1604
f05550c0 1605#if defined(HAS_ISBLANK)
84061b6a 1606# define isBLANK_LC(c) _generic_LC(c, _CC_BLANK, isblank)
95f34b6f 1607#else /* Unlike isASCII, varies if in a UTF-8 locale */
278b2b56 1608# define isBLANK_LC(c) ((IN_UTF8_CTYPE_LOCALE) ? isBLANK_L1(c) : isBLANK(c))
84061b6a
KW
1609#endif
1610
f05550c0 1611#define _LC_CAST U8
bbce6d69 1612
f05550c0 1613#ifdef WIN32
375f5f06
KW
1614 /* The Windows functions don't bother to follow the POSIX standard, which
1615 * for example says that something can't both be a printable and a control.
1616 * But Windows treats the \t control as a printable, and does such things
1617 * as making superscripts into both digits and punctuation. This tames
1618 * these flaws by assuming that the definitions of both controls and space
1619 * are correct, and then making sure that other definitions don't have
1620 * weirdnesses, by making sure that isalnum() isn't also ispunct(), etc.
1621 * Not all possible weirdnesses are checked for, just the ones that were
1622 * detected on actual Microsoft code pages */
1623
b7d90381
KW
1624# define isCNTRL_LC(c) _generic_LC(c, _CC_CNTRL, iscntrl)
1625# define isSPACE_LC(c) _generic_LC(c, _CC_SPACE, isspace)
1626
1627# define isALPHA_LC(c) (_generic_LC(c, _CC_ALPHA, isalpha) \
1628 && isALPHANUMERIC_LC(c))
1629# define isALPHANUMERIC_LC(c) (_generic_LC(c, _CC_ALPHANUMERIC, isalnum) && \
1630 ! isPUNCT_LC(c))
1631# define isDIGIT_LC(c) (_generic_LC(c, _CC_DIGIT, isdigit) && \
1632 isALPHANUMERIC_LC(c))
1633# define isGRAPH_LC(c) (_generic_LC(c, _CC_GRAPH, isgraph) && isPRINT_LC(c))
1634# define isIDFIRST_LC(c) (((c) == '_') \
1635 || (_generic_LC(c, _CC_IDFIRST, isalpha) && ! isPUNCT_LC(c)))
1636# define isLOWER_LC(c) (_generic_LC(c, _CC_LOWER, islower) && isALPHA_LC(c))
1637# define isPRINT_LC(c) (_generic_LC(c, _CC_PRINT, isprint) && ! isCNTRL_LC(c))
1638# define isPUNCT_LC(c) (_generic_LC(c, _CC_PUNCT, ispunct) && ! isCNTRL_LC(c))
1639# define isUPPER_LC(c) (_generic_LC(c, _CC_UPPER, isupper) && isALPHA_LC(c))
f05550c0 1640# define isWORDCHAR_LC(c) (((c) == '_') || isALPHANUMERIC_LC(c))
b7d90381
KW
1641# define isXDIGIT_LC(c) (_generic_LC(c, _CC_XDIGIT, isxdigit) \
1642 && isALPHANUMERIC_LC(c))
f05550c0
BF
1643
1644# define toLOWER_LC(c) _generic_toLOWER_LC((c), tolower, U8)
1645# define toUPPER_LC(c) _generic_toUPPER_LC((c), toupper, U8)
1646# define toFOLD_LC(c) _generic_toFOLD_LC((c), tolower, U8)
1647
1648#elif defined(CTYPE256) || (!defined(isascii) && !defined(HAS_ISASCII))
4650c663 1649 /* For most other platforms */
beab9ebe 1650
f05550c0
BF
1651# define isALPHA_LC(c) _generic_LC(c, _CC_ALPHA, isalpha)
1652# define isALPHANUMERIC_LC(c) _generic_LC(c, _CC_ALPHANUMERIC, isalnum)
1653# define isCNTRL_LC(c) _generic_LC(c, _CC_CNTRL, iscntrl)
1654# define isDIGIT_LC(c) _generic_LC(c, _CC_DIGIT, isdigit)
1655# define isGRAPH_LC(c) _generic_LC(c, _CC_GRAPH, isgraph)
1656# define isIDFIRST_LC(c) _generic_LC_underscore(c, _CC_IDFIRST, isalpha)
1657# define isLOWER_LC(c) _generic_LC(c, _CC_LOWER, islower)
1658# define isPRINT_LC(c) _generic_LC(c, _CC_PRINT, isprint)
1659# define isPUNCT_LC(c) _generic_LC(c, _CC_PUNCT, ispunct)
1660# define isSPACE_LC(c) _generic_LC(c, _CC_SPACE, isspace)
1661# define isUPPER_LC(c) _generic_LC(c, _CC_UPPER, isupper)
1662# define isWORDCHAR_LC(c) _generic_LC_underscore(c, _CC_WORDCHAR, isalnum)
1663# define isXDIGIT_LC(c) _generic_LC(c, _CC_XDIGIT, isxdigit)
1664
1665
1666# define toLOWER_LC(c) _generic_toLOWER_LC((c), tolower, U8)
1667# define toUPPER_LC(c) _generic_toUPPER_LC((c), toupper, U8)
1668# define toFOLD_LC(c) _generic_toFOLD_LC((c), tolower, U8)
1669
1670#else /* The final fallback position */
1671
b7d90381
KW
1672# define isALPHA_LC(c) (isascii(c) && isalpha(c))
1673# define isALPHANUMERIC_LC(c) (isascii(c) && isalnum(c))
1674# define isCNTRL_LC(c) (isascii(c) && iscntrl(c))
1675# define isDIGIT_LC(c) (isascii(c) && isdigit(c))
1676# define isGRAPH_LC(c) (isascii(c) && isgraph(c))
f05550c0 1677# define isIDFIRST_LC(c) (isascii(c) && (isalpha(c) || (c) == '_'))
b7d90381
KW
1678# define isLOWER_LC(c) (isascii(c) && islower(c))
1679# define isPRINT_LC(c) (isascii(c) && isprint(c))
1680# define isPUNCT_LC(c) (isascii(c) && ispunct(c))
1681# define isSPACE_LC(c) (isascii(c) && isspace(c))
1682# define isUPPER_LC(c) (isascii(c) && isupper(c))
f05550c0 1683# define isWORDCHAR_LC(c) (isascii(c) && (isalnum(c) || (c) == '_'))
b7d90381 1684# define isXDIGIT_LC(c) (isascii(c) && isxdigit(c))
f05550c0
BF
1685
1686# define toLOWER_LC(c) (isascii(c) ? tolower(c) : (c))
1687# define toUPPER_LC(c) (isascii(c) ? toupper(c) : (c))
1688# define toFOLD_LC(c) (isascii(c) ? tolower(c) : (c))
bbce6d69 1689
f05550c0 1690#endif
55204971 1691
eba68aa0
KW
1692#define isIDCONT(c) isWORDCHAR(c)
1693#define isIDCONT_A(c) isWORDCHAR_A(c)
1694#define isIDCONT_L1(c) isWORDCHAR_L1(c)
1695#define isIDCONT_LC(c) isWORDCHAR_LC(c)
13380643 1696#define isPSXSPC_LC(c) isSPACE_LC(c)
aaa51d5e 1697
4650c663 1698/* For internal core Perl use only: the base macros for defining macros like
d0da05db
KW
1699 * isALPHA_uvchr. 'c' is the code point to check. 'classnum' is the POSIX class
1700 * number defined earlier in this file. _generic_uvchr() is used for POSIX
4650c663
KW
1701 * classes where there is a macro or function 'above_latin1' that takes the
1702 * single argument 'c' and returns the desired value. These exist for those
1703 * classes which have simple definitions, avoiding the overhead of a hash
d0da05db 1704 * lookup or inversion list binary search. _generic_swash_uvchr() can be used
4650c663 1705 * for classes where that overhead is faster than a direct lookup.
d0da05db 1706 * _generic_uvchr() won't compile if 'c' isn't unsigned, as it won't match the
4650c663
KW
1707 * 'above_latin1' prototype. _generic_isCC() macro does bounds checking, so
1708 * have duplicate checks here, so could create versions of the macros that
1709 * don't, but experiments show that gcc optimizes them out anyway. */
66c17564
KW
1710
1711/* Note that all ignore 'use bytes' */
1e222e4f
KW
1712#define _generic_uvchr(classnum, above_latin1, c) ((c) < 256 \
1713 ? _generic_isCC(c, classnum) \
cd500f2f 1714 : above_latin1(c))
1e222e4f
KW
1715#define _generic_swash_uvchr(classnum, c) ((c) < 256 \
1716 ? _generic_isCC(c, classnum) \
922e8cb4 1717 : _is_uni_FOO(classnum, c))
d0da05db
KW
1718#define isALPHA_uvchr(c) _generic_swash_uvchr(_CC_ALPHA, c)
1719#define isALPHANUMERIC_uvchr(c) _generic_swash_uvchr(_CC_ALPHANUMERIC, c)
1720#define isASCII_uvchr(c) isASCII(c)
1721#define isBLANK_uvchr(c) _generic_uvchr(_CC_BLANK, is_HORIZWS_cp_high, c)
1722#define isCNTRL_uvchr(c) isCNTRL_L1(c) /* All controls are in Latin1 */
1723#define isDIGIT_uvchr(c) _generic_swash_uvchr(_CC_DIGIT, c)
1724#define isGRAPH_uvchr(c) _generic_swash_uvchr(_CC_GRAPH, c)
1e222e4f
KW
1725#define isIDCONT_uvchr(c) \
1726 _generic_uvchr(_CC_WORDCHAR, _is_uni_perl_idcont, c)
1727#define isIDFIRST_uvchr(c) \
1728 _generic_uvchr(_CC_IDFIRST, _is_uni_perl_idstart, c)
d0da05db
KW
1729#define isLOWER_uvchr(c) _generic_swash_uvchr(_CC_LOWER, c)
1730#define isPRINT_uvchr(c) _generic_swash_uvchr(_CC_PRINT, c)
1731
1732#define isPUNCT_uvchr(c) _generic_swash_uvchr(_CC_PUNCT, c)
1733#define isSPACE_uvchr(c) _generic_uvchr(_CC_SPACE, is_XPERLSPACE_cp_high, c)
1734#define isPSXSPC_uvchr(c) isSPACE_uvchr(c)
1735
1736#define isUPPER_uvchr(c) _generic_swash_uvchr(_CC_UPPER, c)
1737#define isVERTWS_uvchr(c) _generic_uvchr(_CC_VERTSPACE, is_VERTWS_cp_high, c)
1738#define isWORDCHAR_uvchr(c) _generic_swash_uvchr(_CC_WORDCHAR, c)
1739#define isXDIGIT_uvchr(c) _generic_uvchr(_CC_XDIGIT, is_XDIGIT_cp_high, c)
1740
1741#define toFOLD_uvchr(c,s,l) to_uni_fold(c,s,l)
1742#define toLOWER_uvchr(c,s,l) to_uni_lower(c,s,l)
1743#define toTITLE_uvchr(c,s,l) to_uni_title(c,s,l)
1744#define toUPPER_uvchr(c,s,l) to_uni_upper(c,s,l)
1745
1746/* For backwards compatibility, even though '_uni' should mean official Unicode
1747 * code points, in Perl it means native for those below 256 */
1748#define isALPHA_uni(c) isALPHA_uvchr(c)
1749#define isALPHANUMERIC_uni(c) isALPHANUMERIC_uvchr(c)
1750#define isASCII_uni(c) isASCII_uvchr(c)
1751#define isBLANK_uni(c) isBLANK_uvchr(c)
1752#define isCNTRL_uni(c) isCNTRL_uvchr(c)
1753#define isDIGIT_uni(c) isDIGIT_uvchr(c)
1754#define isGRAPH_uni(c) isGRAPH_uvchr(c)
1755#define isIDCONT_uni(c) isIDCONT_uvchr(c)
1756#define isIDFIRST_uni(c) isIDFIRST_uvchr(c)
1757#define isLOWER_uni(c) isLOWER_uvchr(c)
1758#define isPRINT_uni(c) isPRINT_uvchr(c)
1759#define isPUNCT_uni(c) isPUNCT_uvchr(c)
1760#define isSPACE_uni(c) isSPACE_uvchr(c)
1761#define isPSXSPC_uni(c) isPSXSPC_uvchr(c)
1762#define isUPPER_uni(c) isUPPER_uvchr(c)
1763#define isVERTWS_uni(c) isVERTWS_uvchr(c)
1764#define isWORDCHAR_uni(c) isWORDCHAR_uvchr(c)
1765#define isXDIGIT_uni(c) isXDIGIT_uvchr(c)
1766#define toFOLD_uni(c,s,l) toFOLD_uvchr(c,s,l)
1767#define toLOWER_uni(c,s,l) toLOWER_uvchr(c,s,l)
1768#define toTITLE_uni(c,s,l) toTITLE_uvchr(c,s,l)
1769#define toUPPER_uni(c,s,l) toUPPER_uvchr(c,s,l)
a0ed51b3 1770
4650c663
KW
1771/* For internal core Perl use only: the base macros for defining macros like
1772 * isALPHA_LC_uvchr. These are like isALPHA_LC, but the input can be any code
d0da05db 1773 * point, not just 0-255. Like _generic_uvchr, there are two versions, one for
4650c663 1774 * simple class definitions; the other for more complex. These are like
d0da05db 1775 * _generic_uvchr, so see it for more info. */
cd500f2f
KW
1776#define _generic_LC_uvchr(latin1, above_latin1, c) \
1777 (c < 256 ? latin1(c) : above_latin1(c))
1778#define _generic_LC_swash_uvchr(latin1, classnum, c) \
1779 (c < 256 ? latin1(c) : _is_uni_FOO(classnum, c))
1780
1781#define isALPHA_LC_uvchr(c) _generic_LC_swash_uvchr(isALPHA_LC, _CC_ALPHA, c)
1782#define isALPHANUMERIC_LC_uvchr(c) _generic_LC_swash_uvchr(isALPHANUMERIC_LC, \
1783 _CC_ALPHANUMERIC, c)
b7d90381
KW
1784#define isASCII_LC_uvchr(c) isASCII_LC(c)
1785#define isBLANK_LC_uvchr(c) _generic_LC_uvchr(isBLANK_LC, \
1786 is_HORIZWS_cp_high, c)
feeab5a9 1787#define isCNTRL_LC_uvchr(c) (c < 256 ? isCNTRL_LC(c) : 0)
cd500f2f
KW
1788#define isDIGIT_LC_uvchr(c) _generic_LC_swash_uvchr(isDIGIT_LC, _CC_DIGIT, c)
1789#define isGRAPH_LC_uvchr(c) _generic_LC_swash_uvchr(isGRAPH_LC, _CC_GRAPH, c)
b7d90381 1790#define isIDCONT_LC_uvchr(c) _generic_LC_uvchr(isIDCONT_LC, \
eba68aa0 1791 _is_uni_perl_idcont, c)
b7d90381 1792#define isIDFIRST_LC_uvchr(c) _generic_LC_uvchr(isIDFIRST_LC, \
cd500f2f
KW
1793 _is_uni_perl_idstart, c)
1794#define isLOWER_LC_uvchr(c) _generic_LC_swash_uvchr(isLOWER_LC, _CC_LOWER, c)
1795#define isPRINT_LC_uvchr(c) _generic_LC_swash_uvchr(isPRINT_LC, _CC_PRINT, c)
b7d90381 1796#define isPSXSPC_LC_uvchr(c) isSPACE_LC_uvchr(c)
cd500f2f 1797#define isPUNCT_LC_uvchr(c) _generic_LC_swash_uvchr(isPUNCT_LC, _CC_PUNCT, c)
b7d90381 1798#define isSPACE_LC_uvchr(c) _generic_LC_uvchr(isSPACE_LC, \
509fb054 1799 is_XPERLSPACE_cp_high, c)
cd500f2f 1800#define isUPPER_LC_uvchr(c) _generic_LC_swash_uvchr(isUPPER_LC, _CC_UPPER, c)
b7d90381 1801#define isWORDCHAR_LC_uvchr(c) _generic_LC_swash_uvchr(isWORDCHAR_LC, \
cd500f2f 1802 _CC_WORDCHAR, c)
b7d90381
KW
1803#define isXDIGIT_LC_uvchr(c) _generic_LC_uvchr(isXDIGIT_LC, \
1804 is_XDIGIT_cp_high, c)
e712593e 1805
b7d90381 1806#define isBLANK_LC_uni(c) isBLANK_LC_uvchr(UNI_TO_NATIVE(c))
aaa51d5e 1807
4650c663
KW
1808/* For internal core Perl use only: the base macros for defining macros like
1809 * isALPHA_utf8. These are like the earlier defined macros, but take an input
1810 * UTF-8 encoded string 'p'. If the input is in the Latin1 range, use
1811 * the Latin1 macro 'classnum' on 'p'. Otherwise use the value given by the
1812 * 'utf8' parameter. This relies on the fact that ASCII characters have the
1813 * same representation whether utf8 or not. Note that it assumes that the utf8
1814 * has been validated, and ignores 'use bytes' */
34aeb2e9
KW
1815#define _base_generic_utf8(enum_name, name, p, use_locale ) \
1816 _is_utf8_FOO(CAT2(_CC_, enum_name), \
1817 (const U8 *) p, \
1818 "is" STRINGIFY(name) "_utf8", \
1819 "is" STRINGIFY(name) "_utf8_safe", \
1820 1, use_locale, __FILE__,__LINE__)
1821
1822#define _generic_utf8(name, p) _base_generic_utf8(name, name, p, 0)
da8c1a98
KW
1823
1824/* The "_safe" macros make sure that we don't attempt to read beyond 'e', but
1825 * they don't otherwise go out of their way to look for malformed UTF-8. If
1826 * they can return accurate results without knowing if the input is otherwise
1827 * malformed, they do so. For example isASCII is accurate in spite of any
1828 * non-length malformations because it looks only at a single byte. Likewise
1829 * isDIGIT looks just at the first byte for code points 0-255, as all UTF-8
1830 * variant ones return FALSE. But, if the input has to be well-formed in order
1831 * for the results to be accurate, the macros will test and if malformed will
1832 * call a routine to die
1833 *
1834 * Except for toke.c, the macros do assume that e > p, asserting that on
1835 * DEBUGGING builds. Much code that calls these depends on this being true,
1836 * for other reasons. toke.c is treated specially as using the regular
1837 * assertion breaks it in many ways. All strings that these operate on there
1838 * are supposed to have an extra NUL character at the end, so that *e = \0. A
1839 * bunch of code in toke.c assumes that this is true, so the assertion allows
1840 * for that */
1841#ifdef PERL_IN_TOKE_C
1842# define _utf8_safe_assert(p,e) ((e) > (p) || ((e) == (p) && *(p) == '\0'))
1843#else
1844# define _utf8_safe_assert(p,e) ((e) > (p))
1845#endif
1846
1847#define _generic_utf8_safe(classnum, p, e, above_latin1) \
1848 (__ASSERT_(_utf8_safe_assert(p, e)) \
1849 (UTF8_IS_INVARIANT(*(p))) \
1850 ? _generic_isCC(*(p), classnum) \
1851 : (UTF8_IS_DOWNGRADEABLE_START(*(p)) \
1852 ? ((LIKELY((e) - (p) > 1 && UTF8_IS_CONTINUATION(*((p)+1)))) \
1853 ? _generic_isCC(EIGHT_BIT_UTF8_TO_NATIVE(*(p), *((p)+1 )), \
1854 classnum) \
1855 : (_force_out_malformed_utf8_message( \
1856 (U8 *) (p), (U8 *) (e), 0, 1), 0)) \
1857 : above_latin1))
b7d90381
KW
1858/* Like the above, but calls 'above_latin1(p)' to get the utf8 value.
1859 * 'above_latin1' can be a macro */
da8c1a98
KW
1860#define _generic_func_utf8_safe(classnum, above_latin1, p, e) \
1861 _generic_utf8_safe(classnum, p, e, above_latin1(p, e))
1862#define _generic_non_swash_utf8_safe(classnum, above_latin1, p, e) \
1863 _generic_utf8_safe(classnum, p, e, \
1864 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
1865 ? (_force_out_malformed_utf8_message( \
1866 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
1867 : above_latin1(p)))
4650c663 1868/* Like the above, but passes classnum to _isFOO_utf8(), instead of having an
cd500f2f 1869 * 'above_latin1' parameter */
da8c1a98
KW
1870#define _generic_swash_utf8_safe(classnum, p, e) \
1871_generic_utf8_safe(classnum, p, e, _is_utf8_FOO_with_len(classnum, p, e))
922e8cb4 1872
cc8ab7c0 1873/* Like the above, but should be used only when it is known that there are no
ff7ecfc3
KW
1874 * characters in the upper-Latin1 range (128-255 on ASCII platforms) which the
1875 * class is TRUE for. Hence it can skip the tests for this range.
1876 * 'above_latin1' should include its arguments */
da8c1a98
KW
1877#define _generic_utf8_safe_no_upper_latin1(classnum, p, e, above_latin1) \
1878 (__ASSERT_(_utf8_safe_assert(p, e)) \
1879 (UTF8_IS_INVARIANT(*(p))) \
1880 ? _generic_isCC(*(p), classnum) \
1881 : (UTF8_IS_DOWNGRADEABLE_START(*(p))) \
1882 ? 0 /* Note that doesn't check validity for latin1 */ \
1883 : above_latin1)
1884
84238efa
KW
1885/* NOTE that some of these macros have very similar ones in regcharclass.h.
1886 * For example, there is (at the time of this writing) an 'is_SPACE_utf8()'
1887 * there, differing in name only by an underscore from the one here
1888 * 'isSPACE_utf8(). The difference is that the ones here are probably more
1889 * efficient and smaller, using an O(1) array lookup for Latin1-range code
1890 * points; the regcharclass.h ones are implemented as a series of
1891 * "if-else-if-else ..." */
1892
34aeb2e9
KW
1893#define isALPHA_utf8(p) _generic_utf8(ALPHA, p)
1894#define isALPHANUMERIC_utf8(p) _generic_utf8(ALPHANUMERIC, p)
1895#define isASCII_utf8(p) _generic_utf8(ASCII, p)
1896#define isBLANK_utf8(p) _generic_utf8(BLANK, p)
1897#define isCNTRL_utf8(p) _generic_utf8(CNTRL, p)
1898#define isDIGIT_utf8(p) _generic_utf8(DIGIT, p)
1899#define isGRAPH_utf8(p) _generic_utf8(GRAPH, p)
1900#define isIDCONT_utf8(p) _generic_utf8(IDCONT, p)
1901#define isIDFIRST_utf8(p) _generic_utf8(IDFIRST, p)
1902#define isLOWER_utf8(p) _generic_utf8(LOWER, p)
1903#define isPRINT_utf8(p) _generic_utf8(PRINT, p)
1904#define isPSXSPC_utf8(p) _generic_utf8(PSXSPC, p)
1905#define isPUNCT_utf8(p) _generic_utf8(PUNCT, p)
1906#define isSPACE_utf8(p) _generic_utf8(SPACE, p)
1907#define isUPPER_utf8(p) _generic_utf8(UPPER, p)
1908#define isVERTWS_utf8(p) _generic_utf8(VERTSPACE, p)
1909#define isWORDCHAR_utf8(p) _generic_utf8(WORDCHAR, p)
1910#define isXDIGIT_utf8(p) _generic_utf8(XDIGIT, p)
e8fa43e2 1911
da8c1a98
KW
1912#define isALPHA_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_ALPHA, p, e)
1913#define isALPHANUMERIC_utf8_safe(p, e) \
1914 _generic_swash_utf8_safe(_CC_ALPHANUMERIC, p, e)
1915#define isASCII_utf8_safe(p, e) \
1916 /* Because ASCII is invariant under utf8, the non-utf8 macro \
1917 * works */ \
1918 (__ASSERT_(_utf8_safe_assert(p, e)) isASCII(*(p)))
1919#define isBLANK_utf8_safe(p, e) \
1920 _generic_non_swash_utf8_safe(_CC_BLANK, is_HORIZWS_high, p, e)
1921
e8fa43e2
KW
1922#ifdef EBCDIC
1923 /* Because all controls are UTF-8 invariants in EBCDIC, we can use this
1924 * more efficient macro instead of the more general one */
da8c1a98 1925# define isCNTRL_utf8_safe(p, e) \
56d02b8c 1926 (__ASSERT_(_utf8_safe_assert(p, e)) isCNTRL_L1(*(p)))
e8fa43e2 1927#else
da8c1a98 1928# define isCNTRL_utf8_safe(p, e) _generic_utf8_safe(_CC_CNTRL, p, e, 0)
e8fa43e2
KW
1929#endif
1930
da8c1a98
KW
1931#define isDIGIT_utf8_safe(p, e) \
1932 _generic_utf8_safe_no_upper_latin1(_CC_DIGIT, p, e, \
1933 _is_utf8_FOO_with_len(_CC_DIGIT, p, e))
1934#define isGRAPH_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_GRAPH, p, e)
1935#define isIDCONT_utf8_safe(p, e) _generic_func_utf8_safe(_CC_WORDCHAR, \
1936 _is_utf8_perl_idcont_with_len, p, e)
e5dcd934 1937
c11ff943
KW
1938/* To prevent S_scan_word in toke.c from hanging, we have to make sure that
1939 * IDFIRST is an alnum. See
c8362b00 1940 * http://rt.perl.org/rt3/Ticket/Display.html?id=74022 for more detail than you
f91dcd13
KW
1941 * ever wanted to know about. (In the ASCII range, there isn't a difference.)
1942 * This used to be not the XID version, but we decided to go with the more
1943 * modern Unicode definition */
da8c1a98
KW
1944#define isIDFIRST_utf8_safe(p, e) \
1945 _generic_func_utf8_safe(_CC_IDFIRST, \
1946 _is_utf8_perl_idstart_with_len, (U8 *) (p), (U8 *) (e))
1947
1948#define isLOWER_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_LOWER, p, e)
1949#define isPRINT_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_PRINT, p, e)
1950#define isPSXSPC_utf8_safe(p, e) isSPACE_utf8_safe(p, e)
1951#define isPUNCT_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_PUNCT, p, e)
1952#define isSPACE_utf8_safe(p, e) \
1953 _generic_non_swash_utf8_safe(_CC_SPACE, is_XPERLSPACE_high, p, e)
1954#define isUPPER_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_UPPER, p, e)
1955#define isVERTWS_utf8_safe(p, e) \
1956 _generic_non_swash_utf8_safe(_CC_VERTSPACE, is_VERTWS_high, p, e)
1957#define isWORDCHAR_utf8_safe(p, e) \
1958 _generic_swash_utf8_safe(_CC_WORDCHAR, p, e)
1959#define isXDIGIT_utf8_safe(p, e) \
1960 _generic_utf8_safe_no_upper_latin1(_CC_XDIGIT, p, e, \
1961 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
1962 ? (_force_out_malformed_utf8_message( \
1963 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
1964 : is_XDIGIT_high(p)))
a0ed51b3 1965
7d2b61dd 1966#define toFOLD_utf8(p,s,l) to_utf8_fold(p,s,l)
509fb054
KW
1967#define toLOWER_utf8(p,s,l) to_utf8_lower(p,s,l)
1968#define toTITLE_utf8(p,s,l) to_utf8_title(p,s,l)
1969#define toUPPER_utf8(p,s,l) to_utf8_upper(p,s,l)
2e8adce6 1970
567b353c 1971/* For internal core use only, subject to change */
607313a1
KW
1972#define _toFOLD_utf8_flags(p,e,s,l,f) _to_utf8_fold_flags (p,e,s,l,f, "", 0)
1973#define _toLOWER_utf8_flags(p,e,s,l,f) _to_utf8_lower_flags(p,e,s,l,f, "", 0)
1974#define _toTITLE_utf8_flags(p,e,s,l,f) _to_utf8_title_flags(p,e,s,l,f, "", 0)
1975#define _toUPPER_utf8_flags(p,e,s,l,f) _to_utf8_upper_flags(p,e,s,l,f, "", 0)
a1a5ec35
KW
1976
1977#define toFOLD_utf8_safe(p,e,s,l) _toFOLD_utf8_flags(p,e,s,l, FOLD_FLAGS_FULL)
1978#define toLOWER_utf8_safe(p,e,s,l) _toLOWER_utf8_flags(p,e,s,l, 0)
1979#define toTITLE_utf8_safe(p,e,s,l) _toTITLE_utf8_flags(p,e,s,l, 0)
1980#define toUPPER_utf8_safe(p,e,s,l) _toUPPER_utf8_flags(p,e,s,l, 0)
567b353c 1981
4650c663
KW
1982/* For internal core Perl use only: the base macros for defining macros like
1983 * isALPHA_LC_utf8. These are like _generic_utf8, but if the first code point
1984 * in 'p' is within the 0-255 range, it uses locale rules from the passed-in
1985 * 'macro' parameter */
34aeb2e9
KW
1986#define _generic_LC_utf8(name, p) _base_generic_utf8(name, name, p, 1)
1987
1988#define isALPHA_LC_utf8(p) _generic_LC_utf8(ALPHA, p)
1989#define isALPHANUMERIC_LC_utf8(p) _generic_LC_utf8(ALPHANUMERIC, p)
1990#define isASCII_LC_utf8(p) _generic_LC_utf8(ASCII, p)
1991#define isBLANK_LC_utf8(p) _generic_LC_utf8(BLANK, p)
1992#define isCNTRL_LC_utf8(p) _generic_LC_utf8(CNTRL, p)
1993#define isDIGIT_LC_utf8(p) _generic_LC_utf8(DIGIT, p)
1994#define isGRAPH_LC_utf8(p) _generic_LC_utf8(GRAPH, p)
1995#define isIDCONT_LC_utf8(p) _generic_LC_utf8(IDCONT, p)
1996#define isIDFIRST_LC_utf8(p) _generic_LC_utf8(IDFIRST, p)
1997#define isLOWER_LC_utf8(p) _generic_LC_utf8(LOWER, p)
1998#define isPRINT_LC_utf8(p) _generic_LC_utf8(PRINT, p)
1999#define isPSXSPC_LC_utf8(p) _generic_LC_utf8(PSXSPC, p)
2000#define isPUNCT_LC_utf8(p) _generic_LC_utf8(PUNCT, p)
2001#define isSPACE_LC_utf8(p) _generic_LC_utf8(SPACE, p)
2002#define isUPPER_LC_utf8(p) _generic_LC_utf8(UPPER, p)
2003#define isWORDCHAR_LC_utf8(p) _generic_LC_utf8(WORDCHAR, p)
2004#define isXDIGIT_LC_utf8(p) _generic_LC_utf8(XDIGIT, p)
2005
da8c1a98
KW
2006/* For internal core Perl use only: the base macros for defining macros like
2007 * isALPHA_LC_utf8_safe. These are like _generic_utf8, but if the first code
2008 * point in 'p' is within the 0-255 range, it uses locale rules from the
2009 * passed-in 'macro' parameter */
2010#define _generic_LC_utf8_safe(macro, p, e, above_latin1) \
2011 (__ASSERT_(_utf8_safe_assert(p, e)) \
2012 (UTF8_IS_INVARIANT(*(p))) \
2013 ? macro(*(p)) \
2014 : (UTF8_IS_DOWNGRADEABLE_START(*(p)) \
2015 ? ((LIKELY((e) - (p) > 1 && UTF8_IS_CONTINUATION(*((p)+1)))) \
2016 ? macro(EIGHT_BIT_UTF8_TO_NATIVE(*(p), *((p)+1))) \
2017 : (_force_out_malformed_utf8_message( \
2018 (U8 *) (p), (U8 *) (e), 0, 1), 0)) \
2019 : above_latin1))
2020
2021#define _generic_LC_swash_utf8_safe(macro, classnum, p, e) \
2022 _generic_LC_utf8_safe(macro, p, e, \
2023 _is_utf8_FOO_with_len(classnum, p, e))
2024
2025#define _generic_LC_func_utf8_safe(macro, above_latin1, p, e) \
2026 _generic_LC_utf8_safe(macro, p, e, above_latin1(p, e))
2027
2028#define _generic_LC_non_swash_utf8_safe(classnum, above_latin1, p, e) \
2029 _generic_LC_utf8_safe(classnum, p, e, \
2030 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
2031 ? (_force_out_malformed_utf8_message( \
2032 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
2033 : above_latin1(p)))
2034
2035#define isALPHANUMERIC_LC_utf8_safe(p, e) \
2036 _generic_LC_swash_utf8_safe(isALPHANUMERIC_LC, \
2037 _CC_ALPHANUMERIC, p, e)
2038#define isALPHA_LC_utf8_safe(p, e) \
2039 _generic_LC_swash_utf8_safe(isALPHA_LC, _CC_ALPHA, p, e)
2040#define isASCII_LC_utf8_safe(p, e) \
2041 (__ASSERT_(_utf8_safe_assert(p, e)) isASCII_LC(*(p)))
2042#define isBLANK_LC_utf8_safe(p, e) \
2043 _generic_LC_non_swash_utf8_safe(isBLANK_LC, is_HORIZWS_high, p, e)
2044#define isCNTRL_LC_utf8_safe(p, e) \
2045 _generic_LC_utf8_safe(isCNTRL_LC, p, e, 0)
2046#define isDIGIT_LC_utf8_safe(p, e) \
2047 _generic_LC_swash_utf8_safe(isDIGIT_LC, _CC_DIGIT, p, e)
2048#define isGRAPH_LC_utf8_safe(p, e) \
2049 _generic_LC_swash_utf8_safe(isGRAPH_LC, _CC_GRAPH, p, e)
2050#define isIDCONT_LC_utf8_safe(p, e) \
2051 _generic_LC_func_utf8_safe(isIDCONT_LC, \
2052 _is_utf8_perl_idcont_with_len, p, e)
2053#define isIDFIRST_LC_utf8_safe(p, e) \
2054 _generic_LC_func_utf8_safe(isIDFIRST_LC, \
2055 _is_utf8_perl_idstart_with_len, p, e)
2056#define isLOWER_LC_utf8_safe(p, e) \
2057 _generic_LC_swash_utf8_safe(isLOWER_LC, _CC_LOWER, p, e)
2058#define isPRINT_LC_utf8_safe(p, e) \
2059 _generic_LC_swash_utf8_safe(isPRINT_LC, _CC_PRINT, p, e)
2060#define isPSXSPC_LC_utf8_safe(p, e) isSPACE_LC_utf8_safe(p, e)
2061#define isPUNCT_LC_utf8_safe(p, e) \
2062 _generic_LC_swash_utf8_safe(isPUNCT_LC, _CC_PUNCT, p, e)
2063#define isSPACE_LC_utf8_safe(p, e) \
2064 _generic_LC_non_swash_utf8_safe(isSPACE_LC, is_XPERLSPACE_high, p, e)
2065#define isUPPER_LC_utf8_safe(p, e) \
2066 _generic_LC_swash_utf8_safe(isUPPER_LC, _CC_UPPER, p, e)
2067#define isWORDCHAR_LC_utf8_safe(p, e) \
2068 _generic_LC_swash_utf8_safe(isWORDCHAR_LC, _CC_WORDCHAR, p, e)
2069#define isXDIGIT_LC_utf8_safe(p, e) \
2070 _generic_LC_non_swash_utf8_safe(isXDIGIT_LC, is_XDIGIT_high, p, e)
aaa51d5e 2071
fbc19f27
KW
2072/* Macros for backwards compatibility and for completeness when the ASCII and
2073 * Latin1 values are identical */
b7d90381
KW
2074#define isALPHAU(c) isALPHA_L1(c)
2075#define isDIGIT_L1(c) isDIGIT_A(c)
2076#define isOCTAL(c) isOCTAL_A(c)
2077#define isOCTAL_L1(c) isOCTAL_A(c)
2078#define isXDIGIT_L1(c) isXDIGIT_A(c)
2079#define isALNUM(c) isWORDCHAR(c)
2080#define isALNUMU(c) isWORDCHAR_L1(c)
2081#define isALNUM_LC(c) isWORDCHAR_LC(c)
2082#define isALNUM_uni(c) isWORDCHAR_uni(c)
2e28f0b9 2083#define isALNUM_LC_uvchr(c) isWORDCHAR_LC_uvchr(c)
b7d90381
KW
2084#define isALNUM_utf8(p) isWORDCHAR_utf8(p)
2085#define isALNUM_LC_utf8(p) isWORDCHAR_LC_utf8(p)
2086#define isALNUMC_A(c) isALPHANUMERIC_A(c) /* Mnemonic: "C's alnum" */
2087#define isALNUMC_L1(c) isALPHANUMERIC_L1(c)
2088#define isALNUMC(c) isALPHANUMERIC(c)
2089#define isALNUMC_LC(c) isALPHANUMERIC_LC(c)
2090#define isALNUMC_uni(c) isALPHANUMERIC_uni(c)
15861f94 2091#define isALNUMC_LC_uvchr(c) isALPHANUMERIC_LC_uvchr(c)
b7d90381 2092#define isALNUMC_utf8(p) isALPHANUMERIC_utf8(p)
15861f94 2093#define isALNUMC_LC_utf8(p) isALPHANUMERIC_LC_utf8(p)
fbc19f27 2094
2bd1cbf6
KW
2095/* On EBCDIC platforms, CTRL-@ is 0, CTRL-A is 1, etc, just like on ASCII,
2096 * except that they don't necessarily mean the same characters, e.g. CTRL-D is
2097 * 4 on both systems, but that is EOT on ASCII; ST on EBCDIC.
2098 * '?' is special-cased on EBCDIC to APC, which is the control there that is
2099 * the outlier from the block that contains the other controls, just like
2100 * toCTRL('?') on ASCII yields DEL, the control that is the outlier from the C0
2101 * block. If it weren't special cased, it would yield a non-control.
88794300
KW
2102 * The conversion works both ways, so toCTRL('D') is 4, and toCTRL(4) is D,
2103 * etc. */
2bd1cbf6 2104#ifndef EBCDIC
75763b3a 2105# define toCTRL(c) (__ASSERT_(FITS_IN_8_BITS(c)) toUPPER(((U8)(c))) ^ 64)
2bd1cbf6 2106#else
75763b3a
KW
2107# define toCTRL(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
2108 ((isPRINT_A(c)) \
2109 ? (UNLIKELY((c) == '?') \
2110 ? QUESTION_MARK_CTRL \
2111 : (NATIVE_TO_LATIN1(toUPPER((U8) (c))) ^ 64)) \
2112 : (UNLIKELY((c) == QUESTION_MARK_CTRL) \
2113 ? '?' \
2114 : (LATIN1_TO_NATIVE(((U8) (c)) ^ 64)))))
2bd1cbf6 2115#endif
bbce6d69 2116
dea28490
JJ
2117/* Line numbers are unsigned, 32 bits. */
2118typedef U32 line_t;
e5dcd934 2119#define NOLINE ((line_t) 4294967295UL) /* = FFFFFFFF */
378cc40b 2120
91152fc1
DG
2121/* Helpful alias for version prescan */
2122#define is_LAX_VERSION(a,b) \
2123 (a != Perl_prescan_version(aTHX_ a, FALSE, b, NULL, NULL, NULL, NULL))
2124
2125#define is_STRICT_VERSION(a,b) \
2126 (a != Perl_prescan_version(aTHX_ a, TRUE, b, NULL, NULL, NULL, NULL))
2127
2128#define BADVERSION(a,b,c) \
2129 if (b) { \
2130 *b = c; \
2131 } \
2132 return a;
8c52afec 2133
cb27eebd
KW
2134/* Converts a character known to represent a hexadecimal digit (0-9, A-F, or
2135 * a-f) to its numeric value. READ_XDIGIT's argument is a string pointer,
2136 * which is advanced. The input is validated only by an assert() in DEBUGGING
2137 * builds. In both ASCII and EBCDIC the last 4 bits of the digits are 0-9; and
2138 * the last 4 bits of A-F and a-f are 1-6, so adding 9 yields 10-15 */
2139#define XDIGIT_VALUE(c) (__ASSERT_(isXDIGIT(c)) (0xf & (isDIGIT(c) \
2140 ? (c) \
2141 : ((c) + 9))))
902008b8
KW
2142#define READ_XDIGIT(s) (__ASSERT_(isXDIGIT(*s)) (0xf & (isDIGIT(*(s)) \
2143 ? (*(s)++) \
2144 : (*(s)++ + 9))))
95a59cab 2145
cb27eebd
KW
2146/* Converts a character known to represent an octal digit (0-7) to its numeric
2147 * value. The input is validated only by an assert() in DEBUGGING builds. In
2148 * both ASCII and EBCDIC the last 3 bits of the octal digits range from 0-7. */
2149#define OCTAL_VALUE(c) (__ASSERT_(isOCTAL(c)) (7 & (c)))
2150
305b8651
KW
2151/* Efficiently returns a boolean as to if two native characters are equivalent
2152 * case-insenstively. At least one of the characters must be one of [A-Za-z];
2153 * the ALPHA in the name is to remind you of that. This is asserted() in
2154 * DEBUGGING builds. Because [A-Za-z] are invariant under UTF-8, this macro
2155 * works (on valid input) for both non- and UTF-8-encoded bytes.
2156 *
2157 * When one of the inputs is a compile-time constant and gets folded by the
2158 * compiler, this reduces to an AND and a TEST. On both EBCDIC and ASCII
2159 * machines, 'A' and 'a' differ by a single bit; the same with the upper and
2160 * lower case of all other ASCII-range alphabetics. On ASCII platforms, they
96ca48da
KW
2161 * are 32 apart; on EBCDIC, they are 64. At compile time, this uses an
2162 * exclusive 'or' to find that bit and then inverts it to form a mask, with
2163 * just a single 0, in the bit position where the upper- and lowercase differ.
2164 * */
305b8651
KW
2165#define isALPHA_FOLD_EQ(c1, c2) \
2166 (__ASSERT_(isALPHA_A(c1) || isALPHA_A(c2)) \
2167 ((c1) & ~('A' ^ 'a')) == ((c2) & ~('A' ^ 'a')))
2168#define isALPHA_FOLD_NE(c1, c2) (! isALPHA_FOLD_EQ((c1), (c2)))
2169
8e84507e 2170/*
ccfc67b7
JH
2171=head1 Memory Management
2172
a02a5408 2173=for apidoc Am|void|Newx|void* ptr|int nitems|type
954c1994
GS
2174The XSUB-writer's interface to the C C<malloc> function.
2175
596f7718 2176Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2177
c5008215
JC
2178In 5.9.3, Newx() and friends replace the older New() API, and drops
2179the first parameter, I<x>, a debug aid which allowed callers to identify
37b8b4c9 2180themselves. This aid has been superseded by a new build option,
d10b4965 2181PERL_MEM_LOG (see L<perlhacktips/PERL_MEM_LOG>). The older API is still
c5008215
JC
2182there for use in XS modules supporting older perls.
2183
a02a5408 2184=for apidoc Am|void|Newxc|void* ptr|int nitems|type|cast
954c1994 2185The XSUB-writer's interface to the C C<malloc> function, with
fbe13c60 2186cast. See also C<L</Newx>>.
954c1994 2187
596f7718 2188Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2189
a02a5408 2190=for apidoc Am|void|Newxz|void* ptr|int nitems|type
954c1994 2191The XSUB-writer's interface to the C C<malloc> function. The allocated
fbe13c60 2192memory is zeroed with C<memzero>. See also C<L</Newx>>.
a02a5408 2193
596f7718 2194Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2195
954c1994
GS
2196=for apidoc Am|void|Renew|void* ptr|int nitems|type
2197The XSUB-writer's interface to the C C<realloc> function.
2198
596f7718 2199Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2200
954c1994
GS
2201=for apidoc Am|void|Renewc|void* ptr|int nitems|type|cast
2202The XSUB-writer's interface to the C C<realloc> function, with
2203cast.
2204
596f7718 2205Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2206
49b8b560 2207=for apidoc Am|void|Safefree|void* ptr
954c1994
GS
2208The XSUB-writer's interface to the C C<free> function.
2209
596f7718 2210This should B<ONLY> be used on memory obtained using L</"Newx"> and friends.
0d7b2759 2211
954c1994
GS
2212=for apidoc Am|void|Move|void* src|void* dest|int nitems|type
2213The XSUB-writer's interface to the C C<memmove> function. The C<src> is the
926bb54c 2214source, C<dest> is the destination, C<nitems> is the number of items, and
fbe13c60 2215C<type> is the type. Can do overlapping moves. See also C<L</Copy>>.
954c1994 2216
e90e2364 2217=for apidoc Am|void *|MoveD|void* src|void* dest|int nitems|type
796b6530 2218Like C<Move> but returns C<dest>. Useful
72d33970 2219for encouraging compilers to tail-call
e90e2364
NC
2220optimise.
2221
954c1994
GS
2222=for apidoc Am|void|Copy|void* src|void* dest|int nitems|type
2223The XSUB-writer's interface to the C C<memcpy> function. The C<src> is the
926bb54c 2224source, C<dest> is the destination, C<nitems> is the number of items, and
fbe13c60 2225C<type> is the type. May fail on overlapping copies. See also C<L</Move>>.
954c1994 2226
e90e2364
NC
2227=for apidoc Am|void *|CopyD|void* src|void* dest|int nitems|type
2228
796b6530 2229Like C<Copy> but returns C<dest>. Useful
72d33970 2230for encouraging compilers to tail-call
e90e2364
NC
2231optimise.
2232
954c1994
GS
2233=for apidoc Am|void|Zero|void* dest|int nitems|type
2234
2235The XSUB-writer's interface to the C C<memzero> function. The C<dest> is the
2236destination, C<nitems> is the number of items, and C<type> is the type.
2237
e90e2364
NC
2238=for apidoc Am|void *|ZeroD|void* dest|int nitems|type
2239
72d33970
FC
2240Like C<Zero> but returns dest. Useful
2241for encouraging compilers to tail-call
e90e2364
NC
2242optimise.
2243
da5d8dbb 2244=for apidoc Am|void|StructCopy|type *src|type *dest|type
4375e838 2245This is an architecture-independent macro to copy one structure to another.
954c1994 2246
7e337ee0
JH
2247=for apidoc Am|void|PoisonWith|void* dest|int nitems|type|U8 byte
2248
2249Fill up memory with a byte pattern (a byte repeated over and over
2250again) that hopefully catches attempts to access uninitialized memory.
2251
2252=for apidoc Am|void|PoisonNew|void* dest|int nitems|type
2253
2254PoisonWith(0xAB) for catching access to allocated but uninitialized memory.
2255
1c12ffb4 2256=for apidoc Am|void|PoisonFree|void* dest|int nitems|type
7e337ee0
JH
2257
2258PoisonWith(0xEF) for catching access to freed memory.
2259
9965345d
JH
2260=for apidoc Am|void|Poison|void* dest|int nitems|type
2261
7e337ee0 2262PoisonWith(0xEF) for catching access to freed memory.
9965345d
JH
2263
2264=cut */
954c1994 2265
561b68a9
SH
2266/* Maintained for backwards-compatibility only. Use newSV() instead. */
2267#ifndef PERL_CORE
ff06c60c 2268#define NEWSV(x,len) newSV(len)
561b68a9 2269#endif
ff06c60c 2270
b7112dce 2271#define MEM_SIZE_MAX ((MEM_SIZE)-1)
19a94d75 2272
a500027b 2273#define _PERL_STRLEN_ROUNDUP_UNCHECKED(n) (((n) - 1 + PERL_STRLEN_ROUNDUP_QUANTUM) & ~((MEM_SIZE)PERL_STRLEN_ROUNDUP_QUANTUM - 1))
e6bdf523 2274
27d5b266 2275#ifdef PERL_MALLOC_WRAP
e6bdf523
DM
2276
2277/* This expression will be constant-folded at compile time. It checks
2278 * whether or not the type of the count n is so small (e.g. U8 or U16, or
2279 * U32 on 64-bit systems) that there's no way a wrap-around could occur.
2280 * As well as avoiding the need for a run-time check in some cases, it's
2281 * designed to avoid compiler warnings like:
2282 * comparison is always false due to limited range of data type
73e8ff00
DM
2283 * It's mathematically equivalent to
2284 * max(n) * sizeof(t) > MEM_SIZE_MAX
e6bdf523
DM
2285 */
2286
2287# define _MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) \
445198b9
LM
2288 ( sizeof(MEM_SIZE) < sizeof(n) \
2289 || sizeof(t) > ((MEM_SIZE)1 << 8*(sizeof(MEM_SIZE) - sizeof(n))))
e6bdf523 2290
88f9f128 2291/* This is written in a slightly odd way to avoid various spurious
d98e5cde
DM
2292 * compiler warnings. We *want* to write the expression as
2293 * _MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) && (n > C)
2294 * (for some compile-time constant C), but even when the LHS
2295 * constant-folds to false at compile-time, g++ insists on emitting
2296 * warnings about the RHS (e.g. "comparison is always false"), so instead
2297 * we write it as
e6bdf523 2298 *
d98e5cde 2299 * (cond ? n : X) > C
88f9f128 2300 *
d98e5cde
DM
2301 * where X is a constant with X > C always false. Choosing a value for X
2302 * is tricky. If 0, some compilers will complain about 0 > C always being
2303 * false; if 1, Coverity complains when n happens to be the constant value
2304 * '1', that cond ? 1 : 1 has the same value on both branches; so use C
2305 * for X and hope that nothing else whines.
e6bdf523
DM
2306 */
2307
2308# define _MEM_WRAP_WILL_WRAP(n,t) \
88f9f128
DM
2309 ((_MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) ? (MEM_SIZE)(n) : \
2310 MEM_SIZE_MAX/sizeof(t)) > MEM_SIZE_MAX/sizeof(t))
e6bdf523
DM
2311
2312# define MEM_WRAP_CHECK(n,t) \
2313 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2314 && (croak_memory_wrap(),0))
2315
2316# define MEM_WRAP_CHECK_1(n,t,a) \
2317 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2318 && (Perl_croak_nocontext("%s",(a)),0))
2319
8b44ba4c 2320#define MEM_WRAP_CHECK_(n,t) MEM_WRAP_CHECK(n,t),
27d5b266 2321
a500027b 2322#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
2323#else
2324
410319be
NC
2325#define MEM_WRAP_CHECK(n,t)
2326#define MEM_WRAP_CHECK_1(n,t,a)
2327#define MEM_WRAP_CHECK_2(n,t,a,b)
8b44ba4c
NC
2328#define MEM_WRAP_CHECK_(n,t)
2329
a500027b 2330#define PERL_STRLEN_ROUNDUP(n) _PERL_STRLEN_ROUNDUP_UNCHECKED(n)
27d5b266 2331
1936d2a7 2332#endif
8b44ba4c 2333
fe4f188c 2334#ifdef PERL_MEM_LOG
46c6c7e2 2335/*
9f653bb5 2336 * If PERL_MEM_LOG is defined, all Newx()s, Renew()s, and Safefree()s
46c6c7e2
JH
2337 * go through functions, which are handy for debugging breakpoints, but
2338 * which more importantly get the immediate calling environment (file and
e352bcff
JH
2339 * line number, and C function name if available) passed in. This info can
2340 * then be used for logging the calls, for which one gets a sample
73d1d973 2341 * implementation unless -DPERL_MEM_LOG_NOIMPL is also defined.
3609ea0d 2342 *
46c6c7e2 2343 * Known problems:
94e892a6 2344 * - not all memory allocs get logged, only those
46c6c7e2 2345 * that go through Newx() and derivatives (while all
94e892a6 2346 * Safefrees do get logged)
46c6c7e2
JH
2347 * - __FILE__ and __LINE__ do not work everywhere
2348 * - __func__ or __FUNCTION__ even less so
2349 * - I think more goes on after the perlio frees but
2350 * the thing is that STDERR gets closed (as do all
2351 * the file descriptors)
2352 * - no deeper calling stack than the caller of the Newx()
2353 * or the kind, but do I look like a C reflection/introspection
2354 * utility to you?
2355 * - the function prototypes for the logging functions
2356 * probably should maybe be somewhere else than handy.h
2357 * - one could consider inlining (macrofying) the logging
2358 * for speed, but I am too lazy
2359 * - one could imagine recording the allocations in a hash,
2360 * (keyed by the allocation address?), and maintain that
2361 * through reallocs and frees, but how to do that without
2362 * any News() happening...?
73d1d973 2363 * - lots of -Ddefines to get useful/controllable output
b953482e 2364 * - lots of ENV reads
46c6c7e2
JH
2365 */
2366
0b0ab801 2367# ifdef PERL_CORE
73d1d973 2368# ifndef PERL_MEM_LOG_NOIMPL
0b0ab801
MHM
2369enum mem_log_type {
2370 MLT_ALLOC,
2371 MLT_REALLOC,
d7a2c63c
MHM
2372 MLT_FREE,
2373 MLT_NEW_SV,
2374 MLT_DEL_SV
0b0ab801
MHM
2375};
2376# endif
12754f92 2377# if defined(PERL_IN_SV_C) /* those are only used in sv.c */
d7a2c63c
MHM
2378void Perl_mem_log_new_sv(const SV *sv, const char *filename, const int linenumber, const char *funcname);
2379void Perl_mem_log_del_sv(const SV *sv, const char *filename, const int linenumber, const char *funcname);
12754f92 2380# endif
0b0ab801
MHM
2381# endif
2382
fe4f188c
JH
2383#endif
2384
2385#ifdef PERL_MEM_LOG
d1401ee9
MHM
2386#define MEM_LOG_ALLOC(n,t,a) Perl_mem_log_alloc(n,sizeof(t),STRINGIFY(t),a,__FILE__,__LINE__,FUNCTION__)
2387#define MEM_LOG_REALLOC(n,t,v,a) Perl_mem_log_realloc(n,sizeof(t),STRINGIFY(t),v,a,__FILE__,__LINE__,FUNCTION__)
46c6c7e2 2388#define MEM_LOG_FREE(a) Perl_mem_log_free(a,__FILE__,__LINE__,FUNCTION__)
fe4f188c
JH
2389#endif
2390
2391#ifndef MEM_LOG_ALLOC
2392#define MEM_LOG_ALLOC(n,t,a) (a)
2393#endif
2394#ifndef MEM_LOG_REALLOC
2395#define MEM_LOG_REALLOC(n,t,v,a) (a)
2396#endif
2397#ifndef MEM_LOG_FREE
2398#define MEM_LOG_FREE(a) (a)
2399#endif
2400
d1401ee9
MHM
2401#define Newx(v,n,t) (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_ALLOC(n,t,safemalloc((MEM_SIZE)((n)*sizeof(t))))))
2402#define Newxc(v,n,t,c) (v = (MEM_WRAP_CHECK_(n,t) (c*)MEM_LOG_ALLOC(n,t,safemalloc((MEM_SIZE)((n)*sizeof(t))))))
2403#define Newxz(v,n,t) (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_ALLOC(n,t,safecalloc((n),sizeof(t)))))
a6f6820f
NC
2404
2405#ifndef PERL_CORE
a02a5408
JC
2406/* pre 5.9.x compatibility */
2407#define New(x,v,n,t) Newx(v,n,t)
2408#define Newc(x,v,n,t,c) Newxc(v,n,t,c)
4541904d 2409#define Newz(x,v,n,t) Newxz(v,n,t)
a6f6820f 2410#endif
a02a5408 2411
ff68c719 2412#define Renew(v,n,t) \
d1401ee9 2413 (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_REALLOC(n,t,v,saferealloc((Malloc_t)(v),(MEM_SIZE)((n)*sizeof(t))))))
ff68c719 2414#define Renewc(v,n,t,c) \
d1401ee9 2415 (v = (MEM_WRAP_CHECK_(n,t) (c*)MEM_LOG_REALLOC(n,t,v,saferealloc((Malloc_t)(v),(MEM_SIZE)((n)*sizeof(t))))))
94010e71
NC
2416
2417#ifdef PERL_POISON
2418#define Safefree(d) \
06c0cc96 2419 ((d) ? (void)(safefree(MEM_LOG_FREE((Malloc_t)(d))), Poison(&(d), 1, Malloc_t)) : (void) 0)
94010e71 2420#else
fe4f188c 2421#define Safefree(d) safefree(MEM_LOG_FREE((Malloc_t)(d)))
94010e71 2422#endif
55497cff 2423
dbb57106
YO
2424/* assert that a valid ptr has been supplied - use this instead of assert(ptr) *
2425 * as it handles cases like constant string arguments without throwing warnings *
2426 * the cast is required, as is the inequality check, to avoid warnings */
45908e4d 2427#define perl_assert_ptr(p) assert( ((void*)(p)) != 0 )
55497cff 2428
45908e4d
YO
2429
2430#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)))
2431#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)))
2432#define Zero(d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), (void)memzero((char*)(d), (n) * sizeof(t)))
2433
bdd1531d 2434/* Like above, but returns a pointer to 'd' */
45908e4d
YO
2435#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)))
2436#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 2437#define ZeroD(d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), memzero((char*)(d), (n) * sizeof(t)))
e90e2364 2438
7e337ee0
JH
2439#define PoisonWith(d,n,t,b) (MEM_WRAP_CHECK_(n,t) (void)memset((char*)(d), (U8)(b), (n) * sizeof(t)))
2440#define PoisonNew(d,n,t) PoisonWith(d,n,t,0xAB)
2441#define PoisonFree(d,n,t) PoisonWith(d,n,t,0xEF)
2442#define Poison(d,n,t) PoisonFree(d,n,t)
27d5b266 2443
caa674f3
DD
2444#ifdef PERL_POISON
2445# define PERL_POISON_EXPR(x) x
2446#else
2447# define PERL_POISON_EXPR(x)
2448#endif
2449
ff68c719 2450#define StructCopy(s,d,t) (*((t*)(d)) = *((t*)(s)))
2cc61e15 2451
c3caa5c3
JH
2452/* C_ARRAY_LENGTH is the number of elements in the C array (so you
2453 * want your zero-based indices to be less than but not equal to).
2454 *
2455 * C_ARRAY_END is one past the last: half-open/half-closed range,
2456 * not last-inclusive range. */
622913ab 2457#define C_ARRAY_LENGTH(a) (sizeof(a)/sizeof((a)[0]))
c3caa5c3 2458#define C_ARRAY_END(a) ((a) + C_ARRAY_LENGTH(a))
622913ab 2459
2cc61e15
DD
2460#ifdef NEED_VA_COPY
2461# ifdef va_copy
2462# define Perl_va_copy(s, d) va_copy(d, s)
07798b17
AC
2463# elif defined(__va_copy)
2464# define Perl_va_copy(s, d) __va_copy(d, s)
2cc61e15 2465# else
07798b17 2466# define Perl_va_copy(s, d) Copy(s, d, 1, va_list)
2cc61e15
DD
2467# endif
2468#endif
2469
472d47bc
SB
2470/* convenience debug macros */
2471#ifdef USE_ITHREADS
2472#define pTHX_FORMAT "Perl interpreter: 0x%p"
2473#define pTHX__FORMAT ", Perl interpreter: 0x%p"
f54cb97a
AL
2474#define pTHX_VALUE_ (void *)my_perl,
2475#define pTHX_VALUE (void *)my_perl
2476#define pTHX__VALUE_ ,(void *)my_perl,
2477#define pTHX__VALUE ,(void *)my_perl
472d47bc 2478#else
3609ea0d 2479#define pTHX_FORMAT
472d47bc 2480#define pTHX__FORMAT
3609ea0d 2481#define pTHX_VALUE_
472d47bc 2482#define pTHX_VALUE
3609ea0d 2483#define pTHX__VALUE_
472d47bc
SB
2484#define pTHX__VALUE
2485#endif /* USE_ITHREADS */
3609ea0d 2486
2acdbac1
NC
2487/* Perl_deprecate was not part of the public API, and did not have a deprecate()
2488 shortcut macro defined without -DPERL_CORE. Neither codesearch.google.com nor
2489 CPAN::Unpack show any users outside the core. */
2490#ifdef PERL_CORE
dc6e8de0
A
2491# define deprecate(s) Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2492 "Use of " s " is deprecated")
c9680906
A
2493# define deprecate_disappears_in(when,message) \
2494 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2495 message ", and will disappear in Perl " when)
ac641426
A
2496# define deprecate_fatal_in(when,message) \
2497 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2498 message ". Its use will be fatal in Perl " when)
2acdbac1
NC
2499#endif
2500
dfff4baf
BF
2501/* Internal macros to deal with gids and uids */
2502#ifdef PERL_CORE
2503
2504# if Uid_t_size > IVSIZE
2505# define sv_setuid(sv, uid) sv_setnv((sv), (NV)(uid))
2506# define SvUID(sv) SvNV(sv)
07798b17
AC
2507# elif Uid_t_sign <= 0
2508# define sv_setuid(sv, uid) sv_setiv((sv), (IV)(uid))
2509# define SvUID(sv) SvIV(sv)
dfff4baf 2510# else
07798b17
AC
2511# define sv_setuid(sv, uid) sv_setuv((sv), (UV)(uid))
2512# define SvUID(sv) SvUV(sv)
dfff4baf
BF
2513# endif /* Uid_t_size */
2514
2515# if Gid_t_size > IVSIZE
2516# define sv_setgid(sv, gid) sv_setnv((sv), (NV)(gid))
2517# define SvGID(sv) SvNV(sv)
07798b17
AC
2518# elif Gid_t_sign <= 0
2519# define sv_setgid(sv, gid) sv_setiv((sv), (IV)(gid))
2520# define SvGID(sv) SvIV(sv)
dfff4baf 2521# else
07798b17
AC
2522# define sv_setgid(sv, gid) sv_setuv((sv), (UV)(gid))
2523# define SvGID(sv) SvUV(sv)
dfff4baf
BF
2524# endif /* Gid_t_size */
2525
2526#endif
2527
6a5bc5ac 2528#endif /* PERL_HANDY_H_ */
9d745869 2529
e9a8c099 2530/*
14d04a33 2531 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 2532 */