This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
better document macros taking literal strings
[perl5.git] / handy.h
CommitLineData
a0d0e21e 1/* handy.h
a687059c 2 *
1129b882 3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1999, 2000,
da5d8dbb 4 * 2001, 2002, 2004, 2005, 2006, 2007, 2008, 2012 by Larry Wall and others
a687059c 5 *
6e21c824
LW
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
8d063cd8 8 *
8d063cd8
LW
9 */
10
4650c663
KW
11/* IMPORTANT NOTE: Everything whose name begins with an underscore is for
12 * internal core Perl use only. */
13
6a5bc5ac
KW
14#ifndef PERL_HANDY_H_ /* Guard against nested #inclusion */
15#define PERL_HANDY_H_
9d745869 16
24792b8d
NC
17#ifndef PERL_CORE
18# define Null(type) ((type)NULL)
954c1994
GS
19
20/*
ccfc67b7 21=head1 Handy Values
954c1994 22
3609ea0d 23=for apidoc AmU||Nullch
72d33970
FC
24Null character pointer. (No longer available when C<PERL_CORE> is
25defined.)
2307c6d0 26
954c1994 27=for apidoc AmU||Nullsv
72d33970 28Null SV pointer. (No longer available when C<PERL_CORE> is defined.)
954c1994
GS
29
30=cut
31*/
32
24792b8d
NC
33# define Nullch Null(char*)
34# define Nullfp Null(PerlIO*)
35# define Nullsv Null(SV*)
36#endif
8d063cd8 37
641d3f0b
PP
38#ifdef TRUE
39#undef TRUE
40#endif
41#ifdef FALSE
42#undef FALSE
43#endif
44#define TRUE (1)
45#define FALSE (0)
46
cf3f0ffb
DM
47/* The MUTABLE_*() macros cast pointers to the types shown, in such a way
48 * (compiler permitting) that casting away const-ness will give a warning;
49 * e.g.:
50 *
51 * const SV *sv = ...;
52 * AV *av1 = (AV*)sv; <== BAD: the const has been silently cast away
53 * AV *av2 = MUTABLE_AV(sv); <== GOOD: it may warn
54 */
55
b1bc3f34 56#if defined(__GNUC__) && !defined(PERL_GCC_BRACE_GROUPS_FORBIDDEN)
f065df04 57# define MUTABLE_PTR(p) ({ void *_p = (p); _p; })
b1bc3f34
NC
58#else
59# define MUTABLE_PTR(p) ((void *) (p))
60#endif
61
a062e10d 62#define MUTABLE_AV(p) ((AV *)MUTABLE_PTR(p))
ea726b52 63#define MUTABLE_CV(p) ((CV *)MUTABLE_PTR(p))
159b6efe 64#define MUTABLE_GV(p) ((GV *)MUTABLE_PTR(p))
dbebbdb4 65#define MUTABLE_HV(p) ((HV *)MUTABLE_PTR(p))
a45c7426 66#define MUTABLE_IO(p) ((IO *)MUTABLE_PTR(p))
b1bc3f34 67#define MUTABLE_SV(p) ((SV *)MUTABLE_PTR(p))
27d4fb96 68
f789f6a4 69#if defined(I_STDBOOL) && !defined(PERL_BOOL_AS_CHAR)
bd31be4b
NC
70# include <stdbool.h>
71# ifndef HAS_BOOL
72# define HAS_BOOL 1
73# endif
74#endif
75
8e84507e 76/* bool is built-in for g++-2.6.3 and later, which might be used
c1d22f6b
GS
77 for extensions. <_G_config.h> defines _G_HAVE_BOOL, but we can't
78 be sure _G_config.h will be included before this file. _G_config.h
8e84507e 79 also defines _G_HAVE_BOOL for both gcc and g++, but only g++
c1d22f6b
GS
80 actually has bool. Hence, _G_HAVE_BOOL is pretty useless for us.
81 g++ can be identified by __GNUG__.
82 Andy Dougherty February 2000
5d94fbed 83*/
3609ea0d 84#ifdef __GNUG__ /* GNU g++ has bool built-in */
f789f6a4 85# ifndef PERL_BOOL_AS_CHAR
5d94fbed 86# ifndef HAS_BOOL
c1d22f6b 87# define HAS_BOOL 1
5d94fbed 88# endif
f789f6a4 89# endif
5d94fbed
AD
90#endif
91
92#ifndef HAS_BOOL
f789f6a4
FC
93# ifdef bool
94# undef bool
95# endif
70d5cb32 96# define bool char
c1d22f6b 97# define HAS_BOOL 1
a687059c 98#endif
0d3e774c 99
236415e2
KW
100/* cast-to-bool. A simple (bool) cast may not do the right thing: if bool is
101 * defined as char for example, then the cast from int is
102 * implementation-defined (bool)!!(cbool) in a ternary triggers a bug in xlc on
103 * AIX */
18f5643b 104#define cBOOL(cbool) ((cbool) ? (bool)1 : (bool)0)
f2338a2e 105
46c6c7e2 106/* Try to figure out __func__ or __FUNCTION__ equivalent, if any.
e352bcff
JH
107 * XXX Should really be a Configure probe, with HAS__FUNCTION__
108 * and FUNCTION__ as results.
109 * XXX Similarly, a Configure probe for __FILE__ and __LINE__ is needed. */
46c6c7e2
JH
110#if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L) || (defined(__SUNPRO_C)) /* C99 or close enough. */
111# define FUNCTION__ __func__
07798b17
AC
112#elif (defined(USING_MSVC6)) || /* MSVC6 has neither __func__ nor __FUNCTION and no good workarounds, either. */ \
113 (defined(__DECC_VER)) /* Tru64 or VMS, and strict C89 being used, but not modern enough cc (in Tur64, -c99 not known, only -std1). */
114# define FUNCTION__ ""
46c6c7e2 115#else
07798b17 116# define FUNCTION__ __FUNCTION__ /* Common extension. */
46c6c7e2
JH
117#endif
118
27d4fb96
PP
119/* XXX A note on the perl source internal type system. The
120 original intent was that I32 be *exactly* 32 bits.
121
122 Currently, we only guarantee that I32 is *at least* 32 bits.
123 Specifically, if int is 64 bits, then so is I32. (This is the case
124 for the Cray.) This has the advantage of meshing nicely with
125 standard library calls (where we pass an I32 and the library is
126 expecting an int), but the disadvantage that an I32 is not 32 bits.
127 Andy Dougherty August 1996
24fef2a7 128
dc45a647
MB
129 There is no guarantee that there is *any* integral type with
130 exactly 32 bits. It is perfectly legal for a system to have
131 sizeof(short) == sizeof(int) == sizeof(long) == 8.
693762b4 132
dc45a647
MB
133 Similarly, there is no guarantee that I16 and U16 have exactly 16
134 bits.
693762b4 135
8e84507e
NIS
136 For dealing with issues that may arise from various 32/64-bit
137 systems, we will ask Configure to check out
8175356b 138
3609ea0d
JH
139 SHORTSIZE == sizeof(short)
140 INTSIZE == sizeof(int)
141 LONGSIZE == sizeof(long)
dc45a647 142 LONGLONGSIZE == sizeof(long long) (if HAS_LONG_LONG)
3609ea0d 143 PTRSIZE == sizeof(void *)
dc45a647
MB
144 DOUBLESIZE == sizeof(double)
145 LONG_DOUBLESIZE == sizeof(long double) (if HAS_LONG_DOUBLE).
8175356b 146
27d4fb96
PP
147*/
148
69512466
JH
149#ifdef I_INTTYPES /* e.g. Linux has int64_t without <inttypes.h> */
150# include <inttypes.h>
dd0eed91
JH
151# ifdef INT32_MIN_BROKEN
152# undef INT32_MIN
153# define INT32_MIN (-2147483647-1)
154# endif
155# ifdef INT64_MIN_BROKEN
156# undef INT64_MIN
157# define INT64_MIN (-9223372036854775807LL-1)
158# endif
69512466
JH
159#endif
160
8175356b
JH
161typedef I8TYPE I8;
162typedef U8TYPE U8;
163typedef I16TYPE I16;
164typedef U16TYPE U16;
165typedef I32TYPE I32;
166typedef U32TYPE U32;
16d89be8
JH
167
168#ifdef HAS_QUAD
8175356b
JH
169typedef I64TYPE I64;
170typedef U64TYPE U64;
16d89be8 171#endif
8175356b 172
9c39b21c
JH
173/* INT64_C/UINT64_C are C99 from <stdint.h> (so they will not be
174 * available in strict C89 mode), but they are nice, so let's define
175 * them if necessary. */
0d3bd0a9 176#if defined(HAS_QUAD)
9c39b21c
JH
177# undef PeRl_INT64_C
178# undef PeRl_UINT64_C
179/* Prefer the native integer types (int and long) over long long
180 * (which is not C89) and Win32-specific __int64. */
181# if QUADKIND == QUAD_IS_INT && INTSIZE == 8
182# define PeRl_INT64_C(c) (c)
183# define PeRl_UINT64_C(c) CAT2(c,U)
184# endif
185# if QUADKIND == QUAD_IS_LONG && LONGSIZE == 8
186# define PeRl_INT64_C(c) CAT2(c,L)
187# define PeRl_UINT64_C(c) CAT2(c,UL)
188# endif
189# if QUADKIND == QUAD_IS_LONG_LONG && defined(HAS_LONG_LONG)
190# define PeRl_INT64_C(c) CAT2(c,LL)
191# define PeRl_UINT64_C(c) CAT2(c,ULL)
192# endif
193# if QUADKIND == QUAD_IS___INT64
194# define PeRl_INT64_C(c) CAT2(c,I64)
195# define PeRl_UINT64_C(c) CAT2(c,UI64)
196# endif
197# ifndef PeRl_INT64_C
702c92eb
JH
198# define PeRl_INT64_C(c) ((I64)(c)) /* last resort */
199# define PeRl_UINT64_C(c) ((U64)(c))
9c39b21c
JH
200# endif
201/* In OS X the INT64_C/UINT64_C are defined with LL/ULL, which will
202 * not fly with C89-pedantic gcc, so let's undefine them first so that
203 * we can redefine them with our native integer preferring versions. */
8eafc6eb 204# if defined(PERL_DARWIN) && defined(PERL_GCC_PEDANTIC)
9c39b21c
JH
205# undef INT64_C
206# undef UINT64_C
207# endif
208# ifndef INT64_C
209# define INT64_C(c) PeRl_INT64_C(c)
210# endif
211# ifndef UINT64_C
212# define UINT64_C(c) PeRl_UINT64_C(c)
213# endif
e8c95190 214#endif
dc750f2e 215
d8668976 216#if defined(UINT8_MAX) && defined(INT16_MAX) && defined(INT32_MAX)
5ff3f7a4 217
5ff3f7a4
GS
218/* I8_MAX and I8_MIN constants are not defined, as I8 is an ambiguous type.
219 Please search CHAR_MAX in perl.h for further details. */
220#define U8_MAX UINT8_MAX
221#define U8_MIN UINT8_MIN
222
5ff3f7a4
GS
223#define I16_MAX INT16_MAX
224#define I16_MIN INT16_MIN
225#define U16_MAX UINT16_MAX
226#define U16_MIN UINT16_MIN
227
5ff3f7a4
GS
228#define I32_MAX INT32_MAX
229#define I32_MIN INT32_MIN
0e983133
GS
230#ifndef UINT32_MAX_BROKEN /* e.g. HP-UX with gcc messes this up */
231# define U32_MAX UINT32_MAX
232#else
233# define U32_MAX 4294967295U
234#endif
5ff3f7a4
GS
235#define U32_MIN UINT32_MIN
236
237#else
238
5c9fa16e
KA
239/* I8_MAX and I8_MIN constants are not defined, as I8 is an ambiguous type.
240 Please search CHAR_MAX in perl.h for further details. */
27d4fb96
PP
241#define U8_MAX PERL_UCHAR_MAX
242#define U8_MIN PERL_UCHAR_MIN
79072805 243
27d4fb96
PP
244#define I16_MAX PERL_SHORT_MAX
245#define I16_MIN PERL_SHORT_MIN
246#define U16_MAX PERL_USHORT_MAX
247#define U16_MIN PERL_USHORT_MIN
79072805 248
c4f23d77 249#if LONGSIZE > 4
27d4fb96
PP
250# define I32_MAX PERL_INT_MAX
251# define I32_MIN PERL_INT_MIN
252# define U32_MAX PERL_UINT_MAX
253# define U32_MIN PERL_UINT_MIN
79072805 254#else
27d4fb96
PP
255# define I32_MAX PERL_LONG_MAX
256# define I32_MIN PERL_LONG_MIN
257# define U32_MAX PERL_ULONG_MAX
258# define U32_MIN PERL_ULONG_MIN
79072805
LW
259#endif
260
5ff3f7a4
GS
261#endif
262
464decb6
KW
263/* log(2) (i.e., log base 10 of 2) is pretty close to 0.30103, just in case
264 * anyone is grepping for it */
265#define BIT_DIGITS(N) (((N)*146)/485 + 1) /* log10(2) =~ 146/485 */
fc36a67e
PP
266#define TYPE_DIGITS(T) BIT_DIGITS(sizeof(T) * 8)
267#define TYPE_CHARS(T) (TYPE_DIGITS(T) + 2) /* sign, NUL */
268
88794300 269/* Unused by core; should be deprecated */
ff68c719 270#define Ctl(ch) ((ch) & 037)
8d063cd8 271
98fce2a4
KW
272#if defined(PERL_CORE) || defined(PERL_EXT)
273# ifndef MIN
274# define MIN(a,b) ((a) < (b) ? (a) : (b))
275# endif
276# ifndef MAX
277# define MAX(a,b) ((a) > (b) ? (a) : (b))
278# endif
279#endif
280
4650c663
KW
281/* This is a helper macro to avoid preprocessor issues, replaced by nothing
282 * unless under DEBUGGING, where it expands to an assert of its argument,
283 * followed by a comma (hence the comma operator). If we just used a straight
3e94db23
JH
284 * assert(), we would get a comma with nothing before it when not DEBUGGING.
285 *
286 * We also use empty definition under Coverity since the __ASSERT__
287 * checks often check for things that Really Cannot Happen, and Coverity
288 * detects that and gets all excited. */
289
290#if defined(DEBUGGING) && !defined(__COVERITY__)
0f092d08
KW
291# define __ASSERT_(statement) assert(statement),
292#else
293# define __ASSERT_(statement)
294#endif
295
3fe05580
MHM
296/*
297=head1 SV-Body Allocation
298
1568d13a
Z
299=for apidoc Ama|SV*|newSVpvs|"literal string" s
300Like C<newSVpvn>, but takes a literal string instead of a
30a15352 301string/length pair.
3fe05580 302
1568d13a
Z
303=for apidoc Ama|SV*|newSVpvs_flags|"literal string" s|U32 flags
304Like C<newSVpvn_flags>, but takes a literal string instead of
30a15352 305a string/length pair.
84bafc02 306
1568d13a
Z
307=for apidoc Ama|SV*|newSVpvs_share|"literal string" s
308Like C<newSVpvn_share>, but takes a literal string instead of
30a15352 309a string/length pair and omits the hash parameter.
3fe05580 310
1568d13a
Z
311=for apidoc Am|void|sv_catpvs_flags|SV* sv|"literal string" s|I32 flags
312Like C<sv_catpvn_flags>, but takes a literal string instead
30a15352 313of a string/length pair.
9dcc53ea 314
1568d13a
Z
315=for apidoc Am|void|sv_catpvs_nomg|SV* sv|"literal string" s
316Like C<sv_catpvn_nomg>, but takes a literal string instead of
0c395ea5 317a string/length pair.
9dcc53ea 318
1568d13a
Z
319=for apidoc Am|void|sv_catpvs|SV* sv|"literal string" s
320Like C<sv_catpvn>, but takes a literal string instead of a
0c395ea5 321string/length pair.
3fe05580 322
1568d13a
Z
323=for apidoc Am|void|sv_catpvs_mg|SV* sv|"literal string" s
324Like C<sv_catpvn_mg>, but takes a literal string instead of a
9dcc53ea
Z
325string/length pair.
326
1568d13a
Z
327=for apidoc Am|void|sv_setpvs|SV* sv|"literal string" s
328Like C<sv_setpvn>, but takes a literal string instead of a
0c395ea5 329string/length pair.
3fe05580 330
1568d13a
Z
331=for apidoc Am|void|sv_setpvs_mg|SV* sv|"literal string" s
332Like C<sv_setpvn_mg>, but takes a literal string instead of a
9dcc53ea
Z
333string/length pair.
334
1568d13a
Z
335=for apidoc Am|SV *|sv_setref_pvs|"literal string" s
336Like C<sv_setref_pvn>, but takes a literal string instead of
0c395ea5 337a string/length pair.
9dcc53ea 338
3fe05580
MHM
339=head1 Memory Management
340
1568d13a
Z
341=for apidoc Ama|char*|savepvs|"literal string" s
342Like C<savepvn>, but takes a literal string instead of a
30a15352 343string/length pair.
3fe05580 344
1568d13a 345=for apidoc Ama|char*|savesharedpvs|"literal string" s
9dcc53ea
Z
346A version of C<savepvs()> which allocates the duplicate string in memory
347which is shared between threads.
348
3fe05580
MHM
349=head1 GV Functions
350
1568d13a
Z
351=for apidoc Am|HV*|gv_stashpvs|"literal string" name|I32 create
352Like C<gv_stashpvn>, but takes a literal string instead of a
0c395ea5 353string/length pair.
3fe05580
MHM
354
355=head1 Hash Manipulation Functions
356
1568d13a
Z
357=for apidoc Am|SV**|hv_fetchs|HV* tb|"literal string" key|I32 lval
358Like C<hv_fetch>, but takes a literal string instead of a
0c395ea5 359string/length pair.
3fe05580 360
1568d13a
Z
361=for apidoc Am|SV**|hv_stores|HV* tb|"literal string" key|NULLOK SV* val
362Like C<hv_store>, but takes a literal string instead of a
0c395ea5 363string/length pair
3fe05580
MHM
364and omits the hash parameter.
365
510966aa
Z
366=head1 Lexer interface
367
1568d13a 368=for apidoc Amx|void|lex_stuff_pvs|"literal string" pv|U32 flags
510966aa 369
1568d13a 370Like L</lex_stuff_pvn>, but takes a literal string instead of
0c395ea5 371a string/length pair.
510966aa 372
3fe05580
MHM
373=cut
374*/
375
926bb54c
KW
376/* concatenating with "" ensures that only literal strings are accepted as
377 * argument */
26bfd924 378#define STR_WITH_LEN(s) ("" s ""), (sizeof(s)-1)
2efa8cc7 379
926bb54c
KW
380/* note that STR_WITH_LEN() can't be used as argument to macros or functions
381 * that under some configurations might be macros, which means that it requires
382 * the full Perl_xxx(aTHX_ ...) form for any API calls where it's used.
ba3a79e7
GA
383 */
384
385/* STR_WITH_LEN() shortcuts */
386#define newSVpvs(str) Perl_newSVpvn(aTHX_ STR_WITH_LEN(str))
84bafc02
NC
387#define newSVpvs_flags(str,flags) \
388 Perl_newSVpvn_flags(aTHX_ STR_WITH_LEN(str), flags)
ba3a79e7 389#define newSVpvs_share(str) Perl_newSVpvn_share(aTHX_ STR_WITH_LEN(str), 0)
9dcc53ea
Z
390#define sv_catpvs_flags(sv, str, flags) \
391 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), flags)
392#define sv_catpvs_nomg(sv, str) \
393 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), 0)
394#define sv_catpvs(sv, str) \
395 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), SV_GMAGIC)
396#define sv_catpvs_mg(sv, str) \
397 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), SV_GMAGIC|SV_SMAGIC)
3fe05580 398#define sv_setpvs(sv, str) Perl_sv_setpvn(aTHX_ sv, STR_WITH_LEN(str))
9dcc53ea
Z
399#define sv_setpvs_mg(sv, str) Perl_sv_setpvn_mg(aTHX_ sv, STR_WITH_LEN(str))
400#define sv_setref_pvs(rv, classname, str) \
401 Perl_sv_setref_pvn(aTHX_ rv, classname, STR_WITH_LEN(str))
ba3a79e7 402#define savepvs(str) Perl_savepvn(aTHX_ STR_WITH_LEN(str))
9dcc53ea
Z
403#define savesharedpvs(str) Perl_savesharedpvn(aTHX_ STR_WITH_LEN(str))
404#define gv_stashpvs(str, create) \
405 Perl_gv_stashpvn(aTHX_ STR_WITH_LEN(str), create)
406#define gv_fetchpvs(namebeg, add, sv_type) \
407 Perl_gv_fetchpvn_flags(aTHX_ STR_WITH_LEN(namebeg), add, sv_type)
408#define gv_fetchpvn(namebeg, len, add, sv_type) \
409 Perl_gv_fetchpvn_flags(aTHX_ namebeg, len, add, sv_type)
410#define sv_catxmlpvs(dsv, str, utf8) \
411 Perl_sv_catxmlpvn(aTHX_ dsv, STR_WITH_LEN(str), utf8)
4ac46235 412
ba3a79e7 413
510966aa
Z
414#define lex_stuff_pvs(pv,flags) Perl_lex_stuff_pvn(aTHX_ STR_WITH_LEN(pv), flags)
415
b96d8cd9
NC
416#define get_cvs(str, flags) \
417 Perl_get_cvn_flags(aTHX_ STR_WITH_LEN(str), (flags))
5c1737d1 418
954c1994 419/*
ccfc67b7
JH
420=head1 Miscellaneous Functions
421
954c1994 422=for apidoc Am|bool|strNE|char* s1|char* s2
dc6b0978
KW
423Test two C<NUL>-terminated strings to see if they are different. Returns true
424or false.
954c1994
GS
425
426=for apidoc Am|bool|strEQ|char* s1|char* s2
dc6b0978
KW
427Test two C<NUL>-terminated strings to see if they are equal. Returns true or
428false.
954c1994
GS
429
430=for apidoc Am|bool|strLT|char* s1|char* s2
dc6b0978
KW
431Test two C<NUL>-terminated strings to see if the first, C<s1>, is less than the
432second, C<s2>. Returns true or false.
954c1994
GS
433
434=for apidoc Am|bool|strLE|char* s1|char* s2
dc6b0978
KW
435Test two C<NUL>-terminated strings to see if the first, C<s1>, is less than or
436equal to the second, C<s2>. Returns true or false.
954c1994
GS
437
438=for apidoc Am|bool|strGT|char* s1|char* s2
dc6b0978
KW
439Test two C<NUL>-terminated strings to see if the first, C<s1>, is greater than
440the second, C<s2>. Returns true or false.
954c1994
GS
441
442=for apidoc Am|bool|strGE|char* s1|char* s2
dc6b0978
KW
443Test two C<NUL>-terminated strings to see if the first, C<s1>, is greater than
444or equal to the second, C<s2>. Returns true or false.
954c1994
GS
445
446=for apidoc Am|bool|strnNE|char* s1|char* s2|STRLEN len
dc6b0978
KW
447Test two C<NUL>-terminated strings to see if they are different. The C<len>
448parameter indicates the number of bytes to compare. Returns true or false. (A
954c1994
GS
449wrapper for C<strncmp>).
450
451=for apidoc Am|bool|strnEQ|char* s1|char* s2|STRLEN len
dc6b0978
KW
452Test two C<NUL>-terminated strings to see if they are equal. The C<len>
453parameter indicates the number of bytes to compare. Returns true or false. (A
454wrapper for C<strncmp>).
954c1994 455
bd18bd40
KW
456=for apidoc Am|bool|memEQ|char* s1|char* s2|STRLEN len
457Test two buffers (which may contain embedded C<NUL> characters, to see if they
458are equal. The C<len> parameter indicates the number of bytes to compare.
459Returns zero if equal, or non-zero if non-equal.
460
461=for apidoc Am|bool|memNE|char* s1|char* s2|STRLEN len
462Test two buffers (which may contain embedded C<NUL> characters, to see if they
463are not equal. The C<len> parameter indicates the number of bytes to compare.
464Returns zero if non-equal, or non-zero if equal.
465
954c1994 466=cut
fc169e00
KW
467
468New macros should use the following conventions for their names (which are
469based on the underlying C library functions):
470
471 (mem | str n? ) (EQ | NE | LT | GT | GE | (( BEGIN | END ) P? )) l? s?
472
473 Each has two main parameters, string-like operands that are compared
474 against each other, as specified by the macro name. Some macros may
475 additionally have one or potentially even two length parameters. If a length
476 parameter applies to both string parameters, it will be positioned third;
477 otherwise any length parameter immediately follows the string parameter it
478 applies to.
479
480 If the prefix to the name is 'str', the string parameter is a pointer to a C
481 language string. Such a string does not contain embedded NUL bytes; its
482 length may be unknown, but can be calculated by C<strlen()>, since it is
483 terminated by a NUL, which isn't included in its length.
484
485 The optional 'n' following 'str' means that that there is a third parameter,
486 giving the maximum number of bytes to look at in each string. Even if both
487 strings are longer than the length parameter, those extra bytes will be
488 unexamined.
489
490 The 's' suffix means that the 2nd byte string parameter is a literal C
491 double-quoted string. Its length will automatically be calculated by the
492 macro, so no length parameter will ever be needed for it.
493
494 If the prefix is 'mem', the string parameters don't have to be C strings;
495 they may contain embedded NUL bytes, do not necessarily have a terminating
496 NUL, and their lengths can be known only through other means, which in
497 practice are additional parameter(s) passed to the function. All 'mem'
498 functions have at least one length parameter. Barring any 'l' or 's' suffix,
499 there is a single length parameter, in position 3, which applies to both
500 string parameters. The 's' suffix means, as described above, that the 2nd
501 string is a literal double-quoted C string (hence its length is calculated by
502 the macro, and the length parameter to the function applies just to the first
503 string parameter, and hence is positioned just after it). An 'l' suffix
504 means that the 2nd string parameter has its own length parameter, and the
505 signature will look like memFOOl(s1, l1, s2, l2).
506
507 BEGIN (and END) are for testing if the 2nd string is an initial (or final)
508 substring of the 1st string. 'P' if present indicates that the substring
509 must be a "proper" one in tha mathematical sense that the first one must be
510 strictly larger than the 2nd.
511
954c1994
GS
512*/
513
62946e08 514
75400963
KW
515#define strNE(s1,s2) (strcmp(s1,s2) != 0)
516#define strEQ(s1,s2) (strcmp(s1,s2) == 0)
8d063cd8
LW
517#define strLT(s1,s2) (strcmp(s1,s2) < 0)
518#define strLE(s1,s2) (strcmp(s1,s2) <= 0)
519#define strGT(s1,s2) (strcmp(s1,s2) > 0)
520#define strGE(s1,s2) (strcmp(s1,s2) >= 0)
62946e08 521
75400963
KW
522#define strnNE(s1,s2,l) (strncmp(s1,s2,l) != 0)
523#define strnEQ(s1,s2,l) (strncmp(s1,s2,l) == 0)
378cc40b 524
75400963
KW
525#define memNE(s1,s2,l) (memcmp(s1,s2,l) != 0)
526#define memEQ(s1,s2,l) (memcmp(s1,s2,l) == 0)
36477c24 527
085b7534 528/* memEQ and memNE where second comparand is a string constant */
568a785a 529#define memEQs(s1, l, s2) \
777fa2cb 530 (((sizeof(s2)-1) == (l)) && memEQ((s1), ("" s2 ""), (sizeof(s2)-1)))
5f50c6c9 531#define memNEs(s1, l, s2) (! memEQs(s1, l, s2))
568a785a 532
fdbb9a7c
KW
533/* Keep these private until we decide it was a good idea */
534#if defined(PERL_CORE) || defined(PERL_EXT) || defined(PERL_EXT_POSIX)
535
536#define strBEGINs(s1,s2) (strncmp(s1,"" s2 "", sizeof(s2)-1) == 0)
537
bdb7e3f0
KW
538#define memBEGINs(s1, l, s2) \
539 ( (l) >= sizeof(s2) - 1 \
540 && memEQ(s1, "" s2 "", sizeof(s2)-1))
de627158
KW
541#define memBEGINPs(s1, l, s2) \
542 ( (l) > sizeof(s2) - 1 \
543 && memEQ(s1, "" s2 "", sizeof(s2)-1))
bdb7e3f0
KW
544#define memENDs(s1, l, s2) \
545 ( (l) >= sizeof(s2) - 1 \
546 && memEQ(s1 + (l) - (sizeof(s2) - 1), "" s2 "", sizeof(s2)-1))
b80f8424
KW
547#define memENDPs(s1, l, s2) \
548 ( (l) > sizeof(s2) \
549 && memEQ(s1 + (l) - (sizeof(s2) - 1), "" s2 "", sizeof(s2)-1))
fdbb9a7c 550#endif /* End of making macros private */
bdb7e3f0 551
062b6850
KW
552#define memLT(s1,s2,l) (memcmp(s1,s2,l) < 0)
553#define memLE(s1,s2,l) (memcmp(s1,s2,l) <= 0)
554#define memGT(s1,s2,l) (memcmp(s1,s2,l) > 0)
555#define memGE(s1,s2,l) (memcmp(s1,s2,l) >= 0)
556
bbce6d69
PP
557/*
558 * Character classes.
559 *
560 * Unfortunately, the introduction of locales means that we
561 * can't trust isupper(), etc. to tell the truth. And when
562 * it comes to /\w+/ with tainting enabled, we *must* be able
563 * to trust our character classes.
564 *
565 * Therefore, the default tests in the text of Perl will be
566 * independent of locale. Any code that wants to depend on
567 * the current locale will use the tests that begin with "lc".
568 */
569
2304df62
AD
570#ifdef HAS_SETLOCALE /* XXX Is there a better test for this? */
571# ifndef CTYPE256
572# define CTYPE256
573# endif
574#endif
575
954c1994 576/*
ccfc67b7 577
dcccc8ff 578=head1 Character classification
243effed
KW
579This section is about functions (really macros) that classify characters
580into types, such as punctuation versus alphabetic, etc. Most of these are
581analogous to regular expression character classes. (See
582L<perlrecharclass/POSIX Character Classes>.) There are several variants for
583each class. (Not all macros have all variants; each item below lists the
584ones valid for it.) None are affected by C<use bytes>, and only the ones
585with C<LC> in the name are affected by the current locale.
586
587The base function, e.g., C<isALPHA()>, takes an octet (either a C<char> or a
588C<U8>) as input and returns a boolean as to whether or not the character
6aff1f14
KW
589represented by that octet is (or on non-ASCII platforms, corresponds to) an
590ASCII character in the named class based on platform, Unicode, and Perl rules.
591If the input is a number that doesn't fit in an octet, FALSE is returned.
243effed 592
c98722a4 593Variant C<isI<FOO>_A> (e.g., C<isALPHA_A()>) is identical to the base function
550da823
KW
594with no suffix C<"_A">. This variant is used to emphasize by its name that
595only ASCII-range characters can return TRUE.
4b9734bf 596
d60679e1 597Variant C<isI<FOO>_L1> imposes the Latin-1 (or EBCDIC equivalent) character set
4b9734bf
KW
598onto the platform. That is, the code points that are ASCII are unaffected,
599since ASCII is a subset of Latin-1. But the non-ASCII code points are treated
600as if they are Latin-1 characters. For example, C<isWORDCHAR_L1()> will return
601true when called with the code point 0xDF, which is a word character in both
4650c663 602ASCII and EBCDIC (though it represents different characters in each).
243effed 603
c98722a4 604Variant C<isI<FOO>_uvchr> is like the C<isI<FOO>_L1> variant, but accepts any UV code
243effed
KW
605point as input. If the code point is larger than 255, Unicode rules are used
606to determine if it is in the character class. For example,
d0da05db 607C<isWORDCHAR_uvchr(0x100)> returns TRUE, since 0x100 is LATIN CAPITAL LETTER A
6aff1f14 608WITH MACRON in Unicode, and is a word character.
243effed 609
c98722a4 610Variant C<isI<FOO>_utf8_safe> is like C<isI<FOO>_uvchr>, but is used for UTF-8
da8c1a98
KW
611encoded strings. Each call classifies one character, even if the string
612contains many. This variant takes two parameters. The first, C<p>, is a
613pointer to the first byte of the character to be classified. (Recall that it
614may take more than one byte to represent a character in UTF-8 strings.) The
615second parameter, C<e>, points to anywhere in the string beyond the first
616character, up to one byte past the end of the entire string. The suffix
617C<_safe> in the function's name indicates that it will not attempt to read
618beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is true (this
619is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the input
620character is malformed in some way, the program may croak, or the function may
621return FALSE, at the discretion of the implementation, and subject to change in
622future releases.
623
c98722a4 624Variant C<isI<FOO>_utf8> is like C<isI<FOO>_utf8_safe>, but takes just a single
da8c1a98 625parameter, C<p>, which has the same meaning as the corresponding parameter does
c98722a4 626in C<isI<FOO>_utf8_safe>. The function therefore can't check if it is reading
34aeb2e9 627beyond the end of the string. Starting in Perl v5.30, it will take a second
c98722a4 628parameter, becoming a synonym for C<isI<FOO>_utf8_safe>. At that time every
34aeb2e9 629program that uses it will have to be changed to successfully compile. In the
c98722a4 630meantime, the first runtime call to C<isI<FOO>_utf8> from each call point in the
34aeb2e9 631program will raise a deprecation warning, enabled by default. You can convert
c98722a4 632your program now to use C<isI<FOO>_utf8_safe>, and avoid the warnings, and get an
34aeb2e9
KW
633extra measure of protection, or you can wait until v5.30, when you'll be forced
634to add the C<e> parameter.
243effed 635
c98722a4 636Variant C<isI<FOO>_LC> is like the C<isI<FOO>_A> and C<isI<FOO>_L1> variants, but the
31f05a37
KW
637result is based on the current locale, which is what C<LC> in the name stands
638for. If Perl can determine that the current locale is a UTF-8 locale, it uses
639the published Unicode rules; otherwise, it uses the C library function that
640gives the named classification. For example, C<isDIGIT_LC()> when not in a
641UTF-8 locale returns the result of calling C<isdigit()>. FALSE is always
1a83413c
KW
642returned if the input won't fit into an octet. On some platforms where the C
643library function is known to be defective, Perl changes its result to follow
644the POSIX standard's rules.
243effed 645
c98722a4
KW
646Variant C<isI<FOO>_LC_uvchr> is like C<isI<FOO>_LC>, but is defined on any UV. It
647returns the same as C<isI<FOO>_LC> for input code points less than 256, and
243effed
KW
648returns the hard-coded, not-affected-by-locale, Unicode results for larger ones.
649
c98722a4 650Variant C<isI<FOO>_LC_utf8_safe> is like C<isI<FOO>_LC_uvchr>, but is used for UTF-8
da8c1a98
KW
651encoded strings. Each call classifies one character, even if the string
652contains many. This variant takes two parameters. The first, C<p>, is a
653pointer to the first byte of the character to be classified. (Recall that it
654may take more than one byte to represent a character in UTF-8 strings.) The
655second parameter, C<e>, points to anywhere in the string beyond the first
656character, up to one byte past the end of the entire string. The suffix
657C<_safe> in the function's name indicates that it will not attempt to read
658beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is true (this
659is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the input
660character is malformed in some way, the program may croak, or the function may
661return FALSE, at the discretion of the implementation, and subject to change in
662future releases.
663
c98722a4 664Variant C<isI<FOO>_LC_utf8> is like C<isI<FOO>_LC_utf8_safe>, but takes just a single
da8c1a98 665parameter, C<p>, which has the same meaning as the corresponding parameter does
c98722a4 666in C<isI<FOO>_LC_utf8_safe>. The function therefore can't check if it is reading
34aeb2e9 667beyond the end of the string. Starting in Perl v5.30, it will take a second
c98722a4 668parameter, becoming a synonym for C<isI<FOO>_LC_utf8_safe>. At that time every
34aeb2e9 669program that uses it will have to be changed to successfully compile. In the
c98722a4 670meantime, the first runtime call to C<isI<FOO>_LC_utf8> from each call point in
34aeb2e9 671the program will raise a deprecation warning, enabled by default. You can
c98722a4 672convert your program now to use C<isI<FOO>_LC_utf8_safe>, and avoid the warnings,
34aeb2e9
KW
673and get an extra measure of protection, or you can wait until v5.30, when
674you'll be forced to add the C<e> parameter.
ccfc67b7 675
954c1994 676=for apidoc Am|bool|isALPHA|char ch
2787a470 677Returns a boolean indicating whether the specified character is an
6aff1f14 678alphabetic character, analogous to C<m/[[:alpha:]]/>.
dcccc8ff
KW
679See the L<top of this section|/Character classification> for an explanation of
680variants
da8c1a98
KW
681C<isALPHA_A>, C<isALPHA_L1>, C<isALPHA_uvchr>, C<isALPHA_utf8_safe>,
682C<isALPHA_LC>, C<isALPHA_LC_uvchr>, and C<isALPHA_LC_utf8_safe>.
8a58bdcf 683
15861f94
KW
684=for apidoc Am|bool|isALPHANUMERIC|char ch
685Returns a boolean indicating whether the specified character is a either an
6aff1f14 686alphabetic character or decimal digit, analogous to C<m/[[:alnum:]]/>.
dcccc8ff
KW
687See the L<top of this section|/Character classification> for an explanation of
688variants
d0da05db 689C<isALPHANUMERIC_A>, C<isALPHANUMERIC_L1>, C<isALPHANUMERIC_uvchr>,
da8c1a98
KW
690C<isALPHANUMERIC_utf8_safe>, C<isALPHANUMERIC_LC>, C<isALPHANUMERIC_LC_uvchr>,
691and C<isALPHANUMERIC_LC_utf8_safe>.
15861f94 692
8a58bdcf
KW
693=for apidoc Am|bool|isASCII|char ch
694Returns a boolean indicating whether the specified character is one of the 128
243effed 695characters in the ASCII character set, analogous to C<m/[[:ascii:]]/>.
e5ad6aba 696On non-ASCII platforms, it returns TRUE iff this
8a58bdcf
KW
697character corresponds to an ASCII character. Variants C<isASCII_A()> and
698C<isASCII_L1()> are identical to C<isASCII()>.
dcccc8ff
KW
699See the L<top of this section|/Character classification> for an explanation of
700variants
da8c1a98
KW
701C<isASCII_uvchr>, C<isASCII_utf8_safe>, C<isASCII_LC>, C<isASCII_LC_uvchr>, and
702C<isASCII_LC_utf8_safe>. Note, however, that some platforms do not have the C
243effed
KW
703library routine C<isascii()>. In these cases, the variants whose names contain
704C<LC> are the same as the corresponding ones without.
705
d98532ea
KW
706Also note, that because all ASCII characters are UTF-8 invariant (meaning they
707have the exact same representation (always a single byte) whether encoded in
708UTF-8 or not), C<isASCII> will give the correct results when called with any
da8c1a98
KW
709byte in any string encoded or not in UTF-8. And similarly C<isASCII_utf8_safe>
710will work properly on any string encoded or not in UTF-8.
d98532ea 711
243effed
KW
712=for apidoc Am|bool|isBLANK|char ch
713Returns a boolean indicating whether the specified character is a
6aff1f14 714character considered to be a blank, analogous to C<m/[[:blank:]]/>.
dcccc8ff
KW
715See the L<top of this section|/Character classification> for an explanation of
716variants
da8c1a98
KW
717C<isBLANK_A>, C<isBLANK_L1>, C<isBLANK_uvchr>, C<isBLANK_utf8_safe>,
718C<isBLANK_LC>, C<isBLANK_LC_uvchr>, and C<isBLANK_LC_utf8_safe>. Note,
719however, that some platforms do not have the C library routine
720C<isblank()>. In these cases, the variants whose names contain C<LC> are
721the same as the corresponding ones without.
243effed
KW
722
723=for apidoc Am|bool|isCNTRL|char ch
724Returns a boolean indicating whether the specified character is a
6aff1f14 725control character, analogous to C<m/[[:cntrl:]]/>.
dcccc8ff
KW
726See the L<top of this section|/Character classification> for an explanation of
727variants
da8c1a98
KW
728C<isCNTRL_A>, C<isCNTRL_L1>, C<isCNTRL_uvchr>, C<isCNTRL_utf8_safe>,
729C<isCNTRL_LC>, C<isCNTRL_LC_uvchr>, and C<isCNTRL_LC_utf8_safe> On EBCDIC
730platforms, you almost always want to use the C<isCNTRL_L1> variant.
954c1994 731
954c1994 732=for apidoc Am|bool|isDIGIT|char ch
2787a470 733Returns a boolean indicating whether the specified character is a
6aff1f14 734digit, analogous to C<m/[[:digit:]]/>.
8a58bdcf 735Variants C<isDIGIT_A> and C<isDIGIT_L1> are identical to C<isDIGIT>.
dcccc8ff
KW
736See the L<top of this section|/Character classification> for an explanation of
737variants
da8c1a98
KW
738C<isDIGIT_uvchr>, C<isDIGIT_utf8_safe>, C<isDIGIT_LC>, C<isDIGIT_LC_uvchr>, and
739C<isDIGIT_LC_utf8_safe>.
243effed
KW
740
741=for apidoc Am|bool|isGRAPH|char ch
742Returns a boolean indicating whether the specified character is a
6aff1f14 743graphic character, analogous to C<m/[[:graph:]]/>.
dcccc8ff 744See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
745variants C<isGRAPH_A>, C<isGRAPH_L1>, C<isGRAPH_uvchr>, C<isGRAPH_utf8_safe>,
746C<isGRAPH_LC>, C<isGRAPH_LC_uvchr>, and C<isGRAPH_LC_utf8_safe>.
954c1994 747
0c82b6df 748=for apidoc Am|bool|isLOWER|char ch
2787a470 749Returns a boolean indicating whether the specified character is a
6aff1f14 750lowercase character, analogous to C<m/[[:lower:]]/>.
dcccc8ff
KW
751See the L<top of this section|/Character classification> for an explanation of
752variants
da8c1a98
KW
753C<isLOWER_A>, C<isLOWER_L1>, C<isLOWER_uvchr>, C<isLOWER_utf8_safe>,
754C<isLOWER_LC>, C<isLOWER_LC_uvchr>, and C<isLOWER_LC_utf8_safe>.
0c82b6df 755
c99e91e9 756=for apidoc Am|bool|isOCTAL|char ch
2787a470 757Returns a boolean indicating whether the specified character is an
6aff1f14 758octal digit, [0-7].
243effed
KW
759The only two variants are C<isOCTAL_A> and C<isOCTAL_L1>; each is identical to
760C<isOCTAL>.
761
762=for apidoc Am|bool|isPUNCT|char ch
763Returns a boolean indicating whether the specified character is a
6aff1f14
KW
764punctuation character, analogous to C<m/[[:punct:]]/>.
765Note that the definition of what is punctuation isn't as
243effed
KW
766straightforward as one might desire. See L<perlrecharclass/POSIX Character
767Classes> for details.
dcccc8ff 768See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
769variants C<isPUNCT_A>, C<isPUNCT_L1>, C<isPUNCT_uvchr>, C<isPUNCT_utf8_safe>,
770C<isPUNCT_LC>, C<isPUNCT_LC_uvchr>, and C<isPUNCT_LC_utf8_safe>.
c99e91e9 771
0c82b6df 772=for apidoc Am|bool|isSPACE|char ch
2787a470 773Returns a boolean indicating whether the specified character is a
6aff1f14 774whitespace character. This is analogous
398d098a 775to what C<m/\s/> matches in a regular expression. Starting in Perl 5.18
779cf272 776this also matches what C<m/[[:space:]]/> does. Prior to 5.18, only the
398d098a
KW
777locale forms of this macro (the ones with C<LC> in their names) matched
778precisely what C<m/[[:space:]]/> does. In those releases, the only difference,
779in the non-locale variants, was that C<isSPACE()> did not match a vertical tab.
780(See L</isPSXSPC> for a macro that matches a vertical tab in all releases.)
dcccc8ff
KW
781See the L<top of this section|/Character classification> for an explanation of
782variants
da8c1a98
KW
783C<isSPACE_A>, C<isSPACE_L1>, C<isSPACE_uvchr>, C<isSPACE_utf8_safe>,
784C<isSPACE_LC>, C<isSPACE_LC_uvchr>, and C<isSPACE_LC_utf8_safe>.
0c82b6df 785
398d098a
KW
786=for apidoc Am|bool|isPSXSPC|char ch
787(short for Posix Space)
779cf272
KW
788Starting in 5.18, this is identical in all its forms to the
789corresponding C<isSPACE()> macros.
398d098a
KW
790The locale forms of this macro are identical to their corresponding
791C<isSPACE()> forms in all Perl releases. In releases prior to 5.18, the
792non-locale forms differ from their C<isSPACE()> forms only in that the
793C<isSPACE()> forms don't match a Vertical Tab, and the C<isPSXSPC()> forms do.
794Otherwise they are identical. Thus this macro is analogous to what
795C<m/[[:space:]]/> matches in a regular expression.
dcccc8ff 796See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
797variants C<isPSXSPC_A>, C<isPSXSPC_L1>, C<isPSXSPC_uvchr>, C<isPSXSPC_utf8_safe>,
798C<isPSXSPC_LC>, C<isPSXSPC_LC_uvchr>, and C<isPSXSPC_LC_utf8_safe>.
398d098a 799
954c1994 800=for apidoc Am|bool|isUPPER|char ch
2787a470 801Returns a boolean indicating whether the specified character is an
6aff1f14 802uppercase character, analogous to C<m/[[:upper:]]/>.
dcccc8ff 803See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
804variants C<isUPPER_A>, C<isUPPER_L1>, C<isUPPER_uvchr>, C<isUPPER_utf8_safe>,
805C<isUPPER_LC>, C<isUPPER_LC_uvchr>, and C<isUPPER_LC_utf8_safe>.
954c1994 806
243effed 807=for apidoc Am|bool|isPRINT|char ch
8eea39dd 808Returns a boolean indicating whether the specified character is a
6aff1f14 809printable character, analogous to C<m/[[:print:]]/>.
dcccc8ff
KW
810See the L<top of this section|/Character classification> for an explanation of
811variants
da8c1a98
KW
812C<isPRINT_A>, C<isPRINT_L1>, C<isPRINT_uvchr>, C<isPRINT_utf8_safe>,
813C<isPRINT_LC>, C<isPRINT_LC_uvchr>, and C<isPRINT_LC_utf8_safe>.
243effed
KW
814
815=for apidoc Am|bool|isWORDCHAR|char ch
816Returns a boolean indicating whether the specified character is a character
817that is a word character, analogous to what C<m/\w/> and C<m/[[:word:]]/> match
818in a regular expression. A word character is an alphabetic character, a
819decimal digit, a connecting punctuation character (such as an underscore), or
820a "mark" character that attaches to one of those (like some sort of accent).
821C<isALNUM()> is a synonym provided for backward compatibility, even though a
822word character includes more than the standard C language meaning of
823alphanumeric.
dcccc8ff 824See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
825variants C<isWORDCHAR_A>, C<isWORDCHAR_L1>, C<isWORDCHAR_uvchr>, and
826C<isWORDCHAR_utf8_safe>. C<isWORDCHAR_LC>, C<isWORDCHAR_LC_uvchr>, and
827C<isWORDCHAR_LC_utf8_safe> are also as described there, but additionally
828include the platform's native underscore.
8a58bdcf
KW
829
830=for apidoc Am|bool|isXDIGIT|char ch
831Returns a boolean indicating whether the specified character is a hexadecimal
243effed
KW
832digit. In the ASCII range these are C<[0-9A-Fa-f]>. Variants C<isXDIGIT_A()>
833and C<isXDIGIT_L1()> are identical to C<isXDIGIT()>.
dcccc8ff
KW
834See the L<top of this section|/Character classification> for an explanation of
835variants
da8c1a98
KW
836C<isXDIGIT_uvchr>, C<isXDIGIT_utf8_safe>, C<isXDIGIT_LC>, C<isXDIGIT_LC_uvchr>,
837and C<isXDIGIT_LC_utf8_safe>.
243effed 838
3c3ecf18
KW
839=for apidoc Am|bool|isIDFIRST|char ch
840Returns a boolean indicating whether the specified character can be the first
841character of an identifier. This is very close to, but not quite the same as
842the official Unicode property C<XID_Start>. The difference is that this
843returns true only if the input character also matches L</isWORDCHAR>.
dcccc8ff
KW
844See the L<top of this section|/Character classification> for an explanation of
845variants
da8c1a98
KW
846C<isIDFIRST_A>, C<isIDFIRST_L1>, C<isIDFIRST_uvchr>, C<isIDFIRST_utf8_safe>,
847C<isIDFIRST_LC>, C<isIDFIRST_LC_uvchr>, and C<isIDFIRST_LC_utf8_safe>.
3c3ecf18
KW
848
849=for apidoc Am|bool|isIDCONT|char ch
850Returns a boolean indicating whether the specified character can be the
851second or succeeding character of an identifier. This is very close to, but
852not quite the same as the official Unicode property C<XID_Continue>. The
853difference is that this returns true only if the input character also matches
dcccc8ff
KW
854L</isWORDCHAR>. See the L<top of this section|/Character classification> for
855an
d0da05db 856explanation of variants C<isIDCONT_A>, C<isIDCONT_L1>, C<isIDCONT_uvchr>,
da8c1a98
KW
857C<isIDCONT_utf8_safe>, C<isIDCONT_LC>, C<isIDCONT_LC_uvchr>, and
858C<isIDCONT_LC_utf8_safe>.
3c3ecf18 859
243effed 860=head1 Miscellaneous Functions
8eea39dd 861
95a59cab 862=for apidoc Am|U8|READ_XDIGIT|char str*
243effed 863Returns the value of an ASCII-range hex digit and advances the string pointer.
95a59cab
YO
864Behaviour is only well defined when isXDIGIT(*str) is true.
865
e7c1e6c1 866=head1 Character case changing
21da7284
KW
867Perl uses "full" Unicode case mappings. This means that converting a single
868character to another case may result in a sequence of more than one character.
869For example, the uppercase of C<E<223>> (LATIN SMALL LETTER SHARP S) is the two
870character sequence C<SS>. This presents some complications The lowercase of
871all characters in the range 0..255 is a single character, and thus
872C<L</toLOWER_L1>> is furnished. But, C<toUPPER_L1> can't exist, as it couldn't
873return a valid result for all legal inputs. Instead C<L</toUPPER_uvchr>> has
874an API that does allow every possible legal result to be returned.) Likewise
875no other function that is crippled by not being able to give the correct
876results for the full range of possible inputs has been implemented here.
e7c1e6c1 877
1f607577
KW
878=for apidoc Am|U8|toUPPER|U8 ch
879Converts the specified character to uppercase. If the input is anything but an
880ASCII lowercase character, that input character itself is returned. Variant
c753c8d3 881C<toUPPER_A> is equivalent.
954c1994 882
d0da05db
KW
883=for apidoc Am|UV|toUPPER_uvchr|UV cp|U8* s|STRLEN* lenp
884Converts the code point C<cp> to its uppercase version, and
885stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
886point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
887that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
888bytes since the uppercase version may be longer than the original character.
889
890The first code point of the uppercased version is returned
21da7284
KW
891(but note, as explained at L<the top of this section|/Character case
892changing>, that there may be more.)
1f607577 893
a239b1e2
KW
894=for apidoc Am|UV|toUPPER_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
895Converts the first UTF-8 encoded character in the sequence starting at C<p> and
896extending no further than S<C<e - 1>> to its uppercase version, and
1f607577
KW
897stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
898that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
899bytes since the uppercase version may be longer than the original character.
900
901The first code point of the uppercased version is returned
21da7284
KW
902(but note, as explained at L<the top of this section|/Character case
903changing>, that there may be more).
1f607577 904
a239b1e2
KW
905The suffix C<_safe> in the function's name indicates that it will not attempt
906to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
907true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
908input character is malformed in some way, the program may croak, or the
909function may return the REPLACEMENT CHARACTER, at the discretion of the
910implementation, and subject to change in future releases.
911
912=for apidoc Am|UV|toUPPER_utf8|U8* p|U8* s|STRLEN* lenp
913This is like C<L</toUPPER_utf8_safe>>, but doesn't have the C<e>
914parameter The function therefore can't check if it is reading
607313a1
KW
915beyond the end of the string. Starting in Perl v5.30, it will take the C<e>
916parameter, becoming a synonym for C<toUPPER_utf8_safe>. At that time every
917program that uses it will have to be changed to successfully compile. In the
918meantime, the first runtime call to C<toUPPER_utf8> from each call point in the
919program will raise a deprecation warning, enabled by default. You can convert
920your program now to use C<toUPPER_utf8_safe>, and avoid the warnings, and get an
921extra measure of protection, or you can wait until v5.30, when you'll be forced
922to add the C<e> parameter.
1f607577 923
25200305
KW
924=for apidoc Am|U8|toFOLD|U8 ch
925Converts the specified character to foldcase. If the input is anything but an
926ASCII uppercase character, that input character itself is returned. Variant
927C<toFOLD_A> is equivalent. (There is no equivalent C<to_FOLD_L1> for the full
d0da05db 928Latin1 range, as the full generality of L</toFOLD_uvchr> is needed there.)
25200305 929
d0da05db
KW
930=for apidoc Am|UV|toFOLD_uvchr|UV cp|U8* s|STRLEN* lenp
931Converts the code point C<cp> to its foldcase version, and
932stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
933point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
934that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
935bytes since the foldcase version may be longer than the original character.
936
937The first code point of the foldcased version is returned
21da7284
KW
938(but note, as explained at L<the top of this section|/Character case
939changing>, that there may be more).
1f607577 940
a239b1e2
KW
941=for apidoc Am|UV|toFOLD_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
942Converts the first UTF-8 encoded character in the sequence starting at C<p> and
943extending no further than S<C<e - 1>> to its foldcase version, and
1f607577
KW
944stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
945that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
946bytes since the foldcase version may be longer than the original character.
947
948The first code point of the foldcased version is returned
21da7284
KW
949(but note, as explained at L<the top of this section|/Character case
950changing>, that there may be more).
1f607577 951
a239b1e2
KW
952The suffix C<_safe> in the function's name indicates that it will not attempt
953to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
954true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
955input character is malformed in some way, the program may croak, or the
956function may return the REPLACEMENT CHARACTER, at the discretion of the
957implementation, and subject to change in future releases.
958
959=for apidoc Am|UV|toFOLD_utf8|U8* p|U8* s|STRLEN* lenp
960This is like C<L</toFOLD_utf8_safe>>, but doesn't have the C<e>
961parameter The function therefore can't check if it is reading
607313a1
KW
962beyond the end of the string. Starting in Perl v5.30, it will take the C<e>
963parameter, becoming a synonym for C<toFOLD_utf8_safe>. At that time every
964program that uses it will have to be changed to successfully compile. In the
965meantime, the first runtime call to C<toFOLD_utf8> from each call point in the
966program will raise a deprecation warning, enabled by default. You can convert
967your program now to use C<toFOLD_utf8_safe>, and avoid the warnings, and get an
968extra measure of protection, or you can wait until v5.30, when you'll be forced
969to add the C<e> parameter.
1f607577
KW
970
971=for apidoc Am|U8|toLOWER|U8 ch
972Converts the specified character to lowercase. If the input is anything but an
973ASCII uppercase character, that input character itself is returned. Variant
c753c8d3 974C<toLOWER_A> is equivalent.
954c1994 975
1f607577 976=for apidoc Am|U8|toLOWER_L1|U8 ch
b7d90381
KW
977Converts the specified Latin1 character to lowercase. The results are
978undefined if the input doesn't fit in a byte.
1f607577
KW
979
980=for apidoc Am|U8|toLOWER_LC|U8 ch
981Converts the specified character to lowercase using the current locale's rules,
982if possible; otherwise returns the input character itself.
983
d0da05db
KW
984=for apidoc Am|UV|toLOWER_uvchr|UV cp|U8* s|STRLEN* lenp
985Converts the code point C<cp> to its lowercase version, and
986stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
987point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
988that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
989bytes since the lowercase version may be longer than the original character.
990
991The first code point of the lowercased version is returned
21da7284
KW
992(but note, as explained at L<the top of this section|/Character case
993changing>, that there may be more).
1f607577 994
a239b1e2
KW
995
996=for apidoc Am|UV|toLOWER_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
997Converts the first UTF-8 encoded character in the sequence starting at C<p> and
998extending no further than S<C<e - 1>> to its lowercase version, and
1f607577
KW
999stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
1000that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
1001bytes since the lowercase version may be longer than the original character.
1002
1003The first code point of the lowercased version is returned
21da7284
KW
1004(but note, as explained at L<the top of this section|/Character case
1005changing>, that there may be more).
1f607577 1006
a239b1e2
KW
1007The suffix C<_safe> in the function's name indicates that it will not attempt
1008to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
1009true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
1010input character is malformed in some way, the program may croak, or the
1011function may return the REPLACEMENT CHARACTER, at the discretion of the
1012implementation, and subject to change in future releases.
1013
1014=for apidoc Am|UV|toLOWER_utf8|U8* p|U8* s|STRLEN* lenp
1015This is like C<L</toLOWER_utf8_safe>>, but doesn't have the C<e>
1016parameter The function therefore can't check if it is reading
607313a1
KW
1017beyond the end of the string. Starting in Perl v5.30, it will take the C<e>
1018parameter, becoming a synonym for C<toLOWER_utf8_safe>. At that time every
1019program that uses it will have to be changed to successfully compile. In the
1020meantime, the first runtime call to C<toLOWER_utf8> from each call point in the
1021program will raise a deprecation warning, enabled by default. You can convert
1022your program now to use C<toLOWER_utf8_safe>, and avoid the warnings, and get an
1023extra measure of protection, or you can wait until v5.30, when you'll be forced
1024to add the C<e> parameter.
1f607577 1025
25200305
KW
1026=for apidoc Am|U8|toTITLE|U8 ch
1027Converts the specified character to titlecase. If the input is anything but an
1028ASCII lowercase character, that input character itself is returned. Variant
b7d90381 1029C<toTITLE_A> is equivalent. (There is no C<toTITLE_L1> for the full Latin1
d0da05db 1030range, as the full generality of L</toTITLE_uvchr> is needed there. Titlecase is
b7d90381 1031not a concept used in locale handling, so there is no functionality for that.)
25200305 1032
d0da05db
KW
1033=for apidoc Am|UV|toTITLE_uvchr|UV cp|U8* s|STRLEN* lenp
1034Converts the code point C<cp> to its titlecase version, and
1035stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
1036point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
1037that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
1038bytes since the titlecase version may be longer than the original character.
1039
1040The first code point of the titlecased version is returned
21da7284
KW
1041(but note, as explained at L<the top of this section|/Character case
1042changing>, that there may be more).
1f607577 1043
a239b1e2
KW
1044=for apidoc Am|UV|toTITLE_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
1045Converts the first UTF-8 encoded character in the sequence starting at C<p> and
1046extending no further than S<C<e - 1>> to its titlecase version, and
1f607577
KW
1047stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
1048that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
1049bytes since the titlecase version may be longer than the original character.
1050
1051The first code point of the titlecased version is returned
21da7284
KW
1052(but note, as explained at L<the top of this section|/Character case
1053changing>, that there may be more).
1f607577 1054
a239b1e2
KW
1055The suffix C<_safe> in the function's name indicates that it will not attempt
1056to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
1057true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
1058input character is malformed in some way, the program may croak, or the
1059function may return the REPLACEMENT CHARACTER, at the discretion of the
1060implementation, and subject to change in future releases.
1061
1062=for apidoc Am|UV|toTITLE_utf8|U8* p|U8* s|STRLEN* lenp
1063This is like C<L</toLOWER_utf8_safe>>, but doesn't have the C<e>
1064parameter The function therefore can't check if it is reading
607313a1
KW
1065beyond the end of the string. Starting in Perl v5.30, it will take the C<e>
1066parameter, becoming a synonym for C<toTITLE_utf8_safe>. At that time every
1067program that uses it will have to be changed to successfully compile. In the
1068meantime, the first runtime call to C<toTITLE_utf8> from each call point in the
1069program will raise a deprecation warning, enabled by default. You can convert
1070your program now to use C<toTITLE_utf8_safe>, and avoid the warnings, and get an
1071extra measure of protection, or you can wait until v5.30, when you'll be forced
1072to add the C<e> parameter.
1f607577 1073
954c1994 1074=cut
353c9b6f 1075
d0da05db 1076XXX Still undocumented isVERTWS_uvchr and _utf8; it's unclear what their names
1e222e4f
KW
1077really should be. Also toUPPER_LC and toFOLD_LC, which are subject to change,
1078and aren't general purpose as they don't work on U+DF, and assert against that.
243effed 1079
8a58bdcf 1080Note that these macros are repeated in Devel::PPPort, so should also be
62fa66b6
KW
1081patched there. The file as of this writing is cpan/Devel-PPPort/parts/inc/misc
1082
954c1994
GS
1083*/
1084
7c062697
KW
1085/* Specify the widest unsigned type on the platform. Use U64TYPE because U64
1086 * is known only in the perl core, and this macro can be called from outside
1087 * that */
1088#ifdef HAS_QUAD
1089# define WIDEST_UTYPE U64TYPE
1090#else
1091# define WIDEST_UTYPE U32
1092#endif
1093
3912bc88
KW
1094/* FITS_IN_8_BITS(c) returns true if c doesn't have a bit set other than in
1095 * the lower 8. It is designed to be hopefully bomb-proof, making sure that no
1096 * bits of information are lost even on a 64-bit machine, but to get the
1097 * compiler to optimize it out if possible. This is because Configure makes
1098 * sure that the machine has an 8-bit byte, so if c is stored in a byte, the
1099 * sizeof() guarantees that this evaluates to a constant true at compile time.
7e75d1a1
JH
1100 *
1101 * For Coverity, be always true, because otherwise Coverity thinks
1102 * it finds several expressions that are always true, independent
1103 * of operands. Well, they are, but that is kind of the point.
220c71bf 1104 */
7e75d1a1 1105#ifndef __COVERITY__
6c5b02ac
KW
1106 /* The '| 0' part ensures a compiler error if c is not integer (like e.g., a
1107 * pointer) */
1108#define FITS_IN_8_BITS(c) ( (sizeof(c) == 1) \
ace3ad0f 1109 || !(((WIDEST_UTYPE)((c) | 0)) & ~0xFF))
7e75d1a1
JH
1110#else
1111#define FITS_IN_8_BITS(c) (1)
1112#endif
cf301eb7 1113
41f43cc2 1114#ifdef EBCDIC
b6340bd0 1115# ifndef _ALL_SOURCE
0852beac
KW
1116 /* The native libc isascii() et.al. functions return the wrong results
1117 * on at least z/OS unless this is defined. */
b6340bd0
KW
1118# error _ALL_SOURCE should probably be defined
1119# endif
41f43cc2 1120#else
0852beac
KW
1121 /* There is a simple definition of ASCII for ASCII platforms. But the
1122 * EBCDIC one isn't so simple, so is defined using table look-up like the
9c903d59 1123 * other macros below.
3f3c579d
KW
1124 *
1125 * The cast here is used instead of '(c) >= 0', because some compilers emit
1126 * a warning that that test is always true when the parameter is an
1127 * unsigned type. khw supposes that it could be written as
1128 * && ((c) == '\0' || (c) > 0)
1129 * to avoid the message, but the cast will likely avoid extra branches even
1130 * with stupid compilers.
1131 *
1132 * The '| 0' part ensures a compiler error if c is not integer (like e.g.,
1133 * a pointer) */
9c903d59 1134# define isASCII(c) ((WIDEST_UTYPE)((c) | 0) < 128)
41f43cc2
KW
1135#endif
1136
38694112
KW
1137/* Take the eight possible bit patterns of the lower 3 bits and you get the
1138 * lower 3 bits of the 8 octal digits, in both ASCII and EBCDIC, so those bits
1139 * can be ignored. If the rest match '0', we have an octal */
1140#define isOCTAL_A(c) (((WIDEST_UTYPE)((c) | 0) & ~7) == '0')
c2da0b36 1141
9fb1bf9d 1142#ifdef H_PERL /* If have access to perl.h, lookup in its table */
f4cdb42c 1143
a500dc72
KW
1144/* Character class numbers. For internal core Perl use only. The ones less
1145 * than 32 are used in PL_charclass[] and the ones up through the one that
1146 * corresponds to <_HIGHEST_REGCOMP_DOT_H_SYNC> are used by regcomp.h and
1147 * related files. PL_charclass ones use names used in l1_char_class_tab.h but
1148 * their actual definitions are here. If that file has a name not used here,
1149 * it won't compile.
1709d539
KW
1150 *
1151 * The first group of these is ordered in what I (khw) estimate to be the
31c7f561
KW
1152 * frequency of their use. This gives a slight edge to exiting a loop earlier
1153 * (in reginclass() in regexec.c) */
1709d539
KW
1154# define _CC_WORDCHAR 0 /* \w and [:word:] */
1155# define _CC_DIGIT 1 /* \d and [:digit:] */
1156# define _CC_ALPHA 2 /* [:alpha:] */
1157# define _CC_LOWER 3 /* [:lower:] */
1158# define _CC_UPPER 4 /* [:upper:] */
1159# define _CC_PUNCT 5 /* [:punct:] */
1160# define _CC_PRINT 6 /* [:print:] */
15861f94 1161# define _CC_ALPHANUMERIC 7 /* [:alnum:] */
1709d539 1162# define _CC_GRAPH 8 /* [:graph:] */
359b005e 1163# define _CC_CASED 9 /* [:lower:] or [:upper:] under /i */
1709d539 1164
b0d691b2 1165#define _FIRST_NON_SWASH_CC 10
1709d539
KW
1166/* The character classes above are implemented with swashes. The second group
1167 * (just below) contains the ones implemented without. These are also sorted
1168 * in rough order of the frequency of their use, except that \v should be last,
1169 * as it isn't a real Posix character class, and some (small) inefficiencies in
1170 * regular expression handling would be introduced by putting it in the middle
1171 * of those that are. Also, cntrl and ascii come after the others as it may be
1172 * useful to group these which have no members that match above Latin1, (or
1173 * above ASCII in the latter case) */
1174
779cf272 1175# define _CC_SPACE 10 /* \s, [:space:] */
34aeb2e9
KW
1176# define _CC_PSXSPC _CC_SPACE /* XXX Temporary, can be removed
1177 when the deprecated isFOO_utf8()
1178 functions are removed */
b0d691b2
KW
1179# define _CC_BLANK 11 /* [:blank:] */
1180# define _CC_XDIGIT 12 /* [:xdigit:] */
779cf272
KW
1181# define _CC_CNTRL 13 /* [:cntrl:] */
1182# define _CC_ASCII 14 /* [:ascii:] */
1183# define _CC_VERTSPACE 15 /* \v */
1709d539 1184
a0947d7b
KW
1185# define _HIGHEST_REGCOMP_DOT_H_SYNC _CC_VERTSPACE
1186
1709d539 1187/* The members of the third group below do not need to be coordinated with data
3ffc8c70 1188 * structures in regcomp.[ch] and regexec.c. */
779cf272
KW
1189# define _CC_IDFIRST 16
1190# define _CC_CHARNAME_CONT 17
1191# define _CC_NONLATIN1_FOLD 18
1192# define _CC_NONLATIN1_SIMPLE_FOLD 19
1193# define _CC_QUOTEMETA 20
1194# define _CC_NON_FINAL_FOLD 21
1195# define _CC_IS_IN_SOME_FOLD 22
1196# define _CC_MNEMONIC_CNTRL 23
073c22b3 1197
34aeb2e9
KW
1198# define _CC_IDCONT 24 /* XXX Temporary, can be removed when the deprecated
1199 isFOO_utf8() functions are removed */
1200
073c22b3
KW
1201/* This next group is only used on EBCDIC platforms, so theoretically could be
1202 * shared with something entirely different that's only on ASCII platforms */
5d5376e2
KW
1203# define _CC_UTF8_START_BYTE_IS_FOR_AT_LEAST_SURROGATE 28
1204# define _CC_UTF8_IS_START 29
1205# define _CC_UTF8_IS_DOWNGRADEABLE_START 30
1206# define _CC_UTF8_IS_CONTINUATION 31
1207/* Unused: 24-27
f4cdb42c
KW
1208 * If more bits are needed, one could add a second word for non-64bit
1209 * QUAD_IS_INT systems, using some #ifdefs to distinguish between having a 2nd
37ede926
KW
1210 * word or not. The IS_IN_SOME_FOLD bit is the most easily expendable, as it
1211 * is used only for optimization (as of this writing), and differs in the
1212 * Latin1 range from the ALPHA bit only in two relatively unimportant
a500dc72 1213 * characters: the masculine and feminine ordinal indicators, so removing it
073c22b3
KW
1214 * would just cause /i regexes which match them to run less efficiently.
1215 * Similarly the EBCDIC-only bits are used just for speed, and could be
1216 * replaced by other means */
96ac0975 1217
3a371f2f
KW
1218#if defined(PERL_CORE) || defined(PERL_EXT)
1219/* An enum version of the character class numbers, to help compilers
1220 * optimize */
1221typedef enum {
3a371f2f 1222 _CC_ENUM_ALPHA = _CC_ALPHA,
e8d596e0
KW
1223 _CC_ENUM_ALPHANUMERIC = _CC_ALPHANUMERIC,
1224 _CC_ENUM_ASCII = _CC_ASCII,
1225 _CC_ENUM_BLANK = _CC_BLANK,
b0d691b2 1226 _CC_ENUM_CASED = _CC_CASED,
e8d596e0 1227 _CC_ENUM_CNTRL = _CC_CNTRL,
3a371f2f
KW
1228 _CC_ENUM_DIGIT = _CC_DIGIT,
1229 _CC_ENUM_GRAPH = _CC_GRAPH,
1230 _CC_ENUM_LOWER = _CC_LOWER,
1231 _CC_ENUM_PRINT = _CC_PRINT,
1232 _CC_ENUM_PUNCT = _CC_PUNCT,
e8d596e0 1233 _CC_ENUM_SPACE = _CC_SPACE,
3a371f2f 1234 _CC_ENUM_UPPER = _CC_UPPER,
e8d596e0 1235 _CC_ENUM_VERTSPACE = _CC_VERTSPACE,
3a371f2f 1236 _CC_ENUM_WORDCHAR = _CC_WORDCHAR,
e8d596e0 1237 _CC_ENUM_XDIGIT = _CC_XDIGIT
3a371f2f
KW
1238} _char_class_number;
1239#endif
1240
63c61c3f 1241#define POSIX_SWASH_COUNT _FIRST_NON_SWASH_CC
86f72d56 1242#define POSIX_CC_COUNT (_HIGHEST_REGCOMP_DOT_H_SYNC + 1)
63c61c3f 1243
b7d90381
KW
1244#if defined(PERL_IN_UTF8_C) \
1245 || defined(PERL_IN_REGCOMP_C) \
1246 || defined(PERL_IN_REGEXEC_C)
922e8cb4 1247# if _CC_WORDCHAR != 0 || _CC_DIGIT != 1 || _CC_ALPHA != 2 || _CC_LOWER != 3 \
b0d691b2
KW
1248 || _CC_UPPER != 4 || _CC_PUNCT != 5 || _CC_PRINT != 6 \
1249 || _CC_ALPHANUMERIC != 7 || _CC_GRAPH != 8 || _CC_CASED != 9
922e8cb4
KW
1250 #error Need to adjust order of swash_property_names[]
1251# endif
1252
b44daeeb
KW
1253/* This is declared static in each of the few files that this is #defined for
1254 * to keep them from being publicly accessible. Hence there is a small amount
1255 * of wasted space */
1256
922e8cb4
KW
1257static const char* const swash_property_names[] = {
1258 "XPosixWord",
1259 "XPosixDigit",
1260 "XPosixAlpha",
1261 "XPosixLower",
1262 "XPosixUpper",
1263 "XPosixPunct",
1264 "XPosixPrint",
1265 "XPosixAlnum",
b0d691b2
KW
1266 "XPosixGraph",
1267 "Cased"
922e8cb4
KW
1268};
1269#endif
1270
6635f04f 1271START_EXTERN_C
96ac0975
NC
1272# ifdef DOINIT
1273EXTCONST U32 PL_charclass[] = {
1274# include "l1_char_class_tab.h"
1275};
1276
1277# else /* ! DOINIT */
1278EXTCONST U32 PL_charclass[];
1279# endif
6635f04f 1280END_EXTERN_C
96ac0975 1281
265c1f46 1282 /* The 1U keeps Solaris from griping when shifting sets the uppermost bit */
430b7c70 1283# define _CC_mask(classnum) (1U << (classnum))
4650c663
KW
1284
1285 /* For internal core Perl use only: the base macro for defining macros like
1286 * isALPHA */
ff7ecfc3 1287# define _generic_isCC(c, classnum) cBOOL(FITS_IN_8_BITS(c) \
f4cd282c 1288 && (PL_charclass[(U8) (c)] & _CC_mask(classnum)))
4eeeb416 1289
f4cdb42c
KW
1290 /* The mask for the _A versions of the macros; it just adds in the bit for
1291 * ASCII. */
1292# define _CC_mask_A(classnum) (_CC_mask(classnum) | _CC_mask(_CC_ASCII))
1293
4650c663
KW
1294 /* For internal core Perl use only: the base macro for defining macros like
1295 * isALPHA_A. The foo_A version makes sure that both the desired bit and
1296 * the ASCII bit are present */
b7d90381
KW
1297# define _generic_isCC_A(c, classnum) (FITS_IN_8_BITS(c) \
1298 && ((PL_charclass[(U8) (c)] & _CC_mask_A(classnum)) \
1299 == _CC_mask_A(classnum)))
f4cdb42c 1300
f4cdb42c 1301# define isALPHA_A(c) _generic_isCC_A(c, _CC_ALPHA)
15861f94 1302# define isALPHANUMERIC_A(c) _generic_isCC_A(c, _CC_ALPHANUMERIC)
f4cdb42c
KW
1303# define isBLANK_A(c) _generic_isCC_A(c, _CC_BLANK)
1304# define isCNTRL_A(c) _generic_isCC_A(c, _CC_CNTRL)
96ca48da 1305# define isDIGIT_A(c) _generic_isCC(c, _CC_DIGIT) /* No non-ASCII digits */
f4cdb42c
KW
1306# define isGRAPH_A(c) _generic_isCC_A(c, _CC_GRAPH)
1307# define isLOWER_A(c) _generic_isCC_A(c, _CC_LOWER)
1308# define isPRINT_A(c) _generic_isCC_A(c, _CC_PRINT)
f4cdb42c
KW
1309# define isPUNCT_A(c) _generic_isCC_A(c, _CC_PUNCT)
1310# define isSPACE_A(c) _generic_isCC_A(c, _CC_SPACE)
1311# define isUPPER_A(c) _generic_isCC_A(c, _CC_UPPER)
1312# define isWORDCHAR_A(c) _generic_isCC_A(c, _CC_WORDCHAR)
b7d90381
KW
1313# define isXDIGIT_A(c) _generic_isCC(c, _CC_XDIGIT) /* No non-ASCII xdigits
1314 */
d95f8b6a 1315# define isIDFIRST_A(c) _generic_isCC_A(c, _CC_IDFIRST)
3ded5eb0
KW
1316# define isALPHA_L1(c) _generic_isCC(c, _CC_ALPHA)
1317# define isALPHANUMERIC_L1(c) _generic_isCC(c, _CC_ALPHANUMERIC)
1318# define isBLANK_L1(c) _generic_isCC(c, _CC_BLANK)
1319
1320 /* continuation character for legal NAME in \N{NAME} */
1321# define isCHARNAME_CONT(c) _generic_isCC(c, _CC_CHARNAME_CONT)
1322
1323# define isCNTRL_L1(c) _generic_isCC(c, _CC_CNTRL)
1324# define isGRAPH_L1(c) _generic_isCC(c, _CC_GRAPH)
1325# define isLOWER_L1(c) _generic_isCC(c, _CC_LOWER)
1326# define isPRINT_L1(c) _generic_isCC(c, _CC_PRINT)
b7d90381 1327# define isPSXSPC_L1(c) isSPACE_L1(c)
3ded5eb0
KW
1328# define isPUNCT_L1(c) _generic_isCC(c, _CC_PUNCT)
1329# define isSPACE_L1(c) _generic_isCC(c, _CC_SPACE)
1330# define isUPPER_L1(c) _generic_isCC(c, _CC_UPPER)
1331# define isWORDCHAR_L1(c) _generic_isCC(c, _CC_WORDCHAR)
1332# define isIDFIRST_L1(c) _generic_isCC(c, _CC_IDFIRST)
f4cdb42c 1333
0852beac
KW
1334# ifdef EBCDIC
1335# define isASCII(c) _generic_isCC(c, _CC_ASCII)
1336# endif
1337
f12c0118
KW
1338 /* Participates in a single-character fold with a character above 255 */
1339# 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)))
1340
1341 /* Like the above, but also can be part of a multi-char fold */
f4cd282c 1342# 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 1343
4eeeb416 1344# define _isQUOTEMETA(c) _generic_isCC(c, _CC_QUOTEMETA)
26faadbd 1345# define _IS_NON_FINAL_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
b7d90381 1346 _generic_isCC(c, _CC_NON_FINAL_FOLD)
37ede926 1347# define _IS_IN_SOME_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
b7d90381 1348 _generic_isCC(c, _CC_IS_IN_SOME_FOLD)
8e35b056
KW
1349# define _IS_MNEMONIC_CNTRL_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
1350 _generic_isCC(c, _CC_MNEMONIC_CNTRL)
687c8d01 1351#else /* else we don't have perl.h H_PERL */
3ded5eb0
KW
1352
1353 /* If we don't have perl.h, we are compiling a utility program. Below we
1354 * hard-code various macro definitions that wouldn't otherwise be available
fc273927 1355 * to it. Most are coded based on first principles. These are written to
74665a89 1356 * avoid EBCDIC vs. ASCII #ifdef's as much as possible. */
0852beac
KW
1357# define isDIGIT_A(c) ((c) <= '9' && (c) >= '0')
1358# define isBLANK_A(c) ((c) == ' ' || (c) == '\t')
74665a89
KW
1359# define isSPACE_A(c) (isBLANK_A(c) \
1360 || (c) == '\n' \
1361 || (c) == '\r' \
1362 || (c) == '\v' \
0852beac 1363 || (c) == '\f')
74665a89
KW
1364 /* On EBCDIC, there are gaps between 'i' and 'j'; 'r' and 's'. Same for
1365 * uppercase. The tests for those aren't necessary on ASCII, but hurt only
1366 * performance (if optimization isn't on), and allow the same code to be
1367 * used for both platform types */
1368# define isLOWER_A(c) ((c) >= 'a' && (c) <= 'z' \
1369 && ( (c) <= 'i' \
1370 || ((c) >= 'j' && (c) <= 'r') \
1371 || (c) >= 's'))
1372# define isUPPER_A(c) ((c) >= 'A' && (c) <= 'Z' \
1373 && ( (c) <= 'I' \
1374 || ((c) >= 'J' && (c) <= 'R') \
1375 || (c) >= 'S'))
a4d7a999
KW
1376# define isALPHA_A(c) (isUPPER_A(c) || isLOWER_A(c))
1377# define isALPHANUMERIC_A(c) (isALPHA_A(c) || isDIGIT_A(c))
3ded5eb0 1378# define isWORDCHAR_A(c) (isALPHANUMERIC_A(c) || (c) == '_')
0852beac 1379# define isIDFIRST_A(c) (isALPHA_A(c) || (c) == '_')
74665a89
KW
1380# define isXDIGIT_A(c) (isDIGIT_A(c) \
1381 || ((c) >= 'a' && (c) <= 'f') \
a4d7a999 1382 || ((c) <= 'F' && (c) >= 'A'))
74665a89
KW
1383# define isPUNCT_A(c) ((c) == '-' || (c) == '!' || (c) == '"' \
1384 || (c) == '#' || (c) == '$' || (c) == '%' \
1385 || (c) == '&' || (c) == '\'' || (c) == '(' \
1386 || (c) == ')' || (c) == '*' || (c) == '+' \
1387 || (c) == ',' || (c) == '.' || (c) == '/' \
1388 || (c) == ':' || (c) == ';' || (c) == '<' \
1389 || (c) == '=' || (c) == '>' || (c) == '?' \
1390 || (c) == '@' || (c) == '[' || (c) == '\\' \
1391 || (c) == ']' || (c) == '^' || (c) == '_' \
1392 || (c) == '`' || (c) == '{' || (c) == '|' \
1393 || (c) == '}' || (c) == '~')
1394# define isGRAPH_A(c) (isALPHANUMERIC_A(c) || isPUNCT_A(c))
1395# define isPRINT_A(c) (isGRAPH_A(c) || (c) == ' ')
3ded5eb0 1396
0852beac 1397# ifdef EBCDIC
74665a89
KW
1398 /* The below is accurate for the 3 EBCDIC code pages traditionally
1399 * supported by perl. The only difference between them in the controls
1400 * is the position of \n, and that is represented symbolically below */
1401# define isCNTRL_A(c) ((c) == '\0' || (c) == '\a' || (c) == '\b' \
1402 || (c) == '\f' || (c) == '\n' || (c) == '\r' \
1403 || (c) == '\t' || (c) == '\v' \
1404 || ((c) <= 3 && (c) >= 1) /* SOH, STX, ETX */ \
1405 || (c) == 7 /* U+7F DEL */ \
1406 || ((c) <= 0x13 && (c) >= 0x0E) /* SO, SI */ \
1407 /* DLE, DC[1-3] */ \
1408 || (c) == 0x18 /* U+18 CAN */ \
1409 || (c) == 0x19 /* U+19 EOM */ \
1410 || ((c) <= 0x1F && (c) >= 0x1C) /* [FGRU]S */ \
1411 || (c) == 0x26 /* U+17 ETB */ \
1412 || (c) == 0x27 /* U+1B ESC */ \
1413 || (c) == 0x2D /* U+05 ENQ */ \
1414 || (c) == 0x2E /* U+06 ACK */ \
1415 || (c) == 0x32 /* U+16 SYN */ \
1416 || (c) == 0x37 /* U+04 EOT */ \
1417 || (c) == 0x3C /* U+14 DC4 */ \
1418 || (c) == 0x3D /* U+15 NAK */ \
1419 || (c) == 0x3F)/* U+1A SUB */
0852beac 1420# define isASCII(c) (isCNTRL_A(c) || isPRINT_A(c))
74665a89
KW
1421# else /* isASCII is already defined for ASCII platforms, so can use that to
1422 define isCNTRL */
1423# define isCNTRL_A(c) (isASCII(c) && ! isPRINT_A(c))
0852beac
KW
1424# endif
1425
3ffc8c70 1426 /* The _L1 macros may be unnecessary for the utilities; I (khw) added them
caa94d35
KW
1427 * during debugging, and it seems best to keep them. We may be called
1428 * without NATIVE_TO_LATIN1 being defined. On ASCII platforms, it doesn't
1429 * do anything anyway, so make it not a problem */
1430# if ! defined(EBCDIC) && ! defined(NATIVE_TO_LATIN1)
1431# define NATIVE_TO_LATIN1(ch) (ch)
1432# endif
3ded5eb0
KW
1433# define isALPHA_L1(c) (isUPPER_L1(c) || isLOWER_L1(c))
1434# define isALPHANUMERIC_L1(c) (isALPHA_L1(c) || isDIGIT_A(c))
1435# define isBLANK_L1(c) (isBLANK_A(c) \
1436 || (FITS_IN_8_BITS(c) \
1437 && NATIVE_TO_LATIN1((U8) c) == 0xA0))
1438# define isCNTRL_L1(c) (FITS_IN_8_BITS(c) && (! isPRINT_L1(c)))
1439# define isGRAPH_L1(c) (isPRINT_L1(c) && (! isBLANK_L1(c)))
1440# define isLOWER_L1(c) (isLOWER_A(c) \
1441 || (FITS_IN_8_BITS(c) \
ae683a5f 1442 && (( NATIVE_TO_LATIN1((U8) c) >= 0xDF \
3ded5eb0
KW
1443 && NATIVE_TO_LATIN1((U8) c) != 0xF7) \
1444 || NATIVE_TO_LATIN1((U8) c) == 0xAA \
1445 || NATIVE_TO_LATIN1((U8) c) == 0xBA \
1446 || NATIVE_TO_LATIN1((U8) c) == 0xB5)))
1447# define isPRINT_L1(c) (isPRINT_A(c) \
1448 || (FITS_IN_8_BITS(c) \
1449 && NATIVE_TO_LATIN1((U8) c) >= 0xA0))
3ded5eb0
KW
1450# define isPUNCT_L1(c) (isPUNCT_A(c) \
1451 || (FITS_IN_8_BITS(c) \
ae683a5f 1452 && ( NATIVE_TO_LATIN1((U8) c) == 0xA1 \
3ded5eb0
KW
1453 || NATIVE_TO_LATIN1((U8) c) == 0xA7 \
1454 || NATIVE_TO_LATIN1((U8) c) == 0xAB \
1455 || NATIVE_TO_LATIN1((U8) c) == 0xB6 \
1456 || NATIVE_TO_LATIN1((U8) c) == 0xB7 \
1457 || NATIVE_TO_LATIN1((U8) c) == 0xBB \
1458 || NATIVE_TO_LATIN1((U8) c) == 0xBF)))
1459# define isSPACE_L1(c) (isSPACE_A(c) \
1460 || (FITS_IN_8_BITS(c) \
ae683a5f 1461 && ( NATIVE_TO_LATIN1((U8) c) == 0x85 \
3ded5eb0
KW
1462 || NATIVE_TO_LATIN1((U8) c) == 0xA0)))
1463# define isUPPER_L1(c) (isUPPER_A(c) \
1464 || (FITS_IN_8_BITS(c) \
ae683a5f 1465 && ( NATIVE_TO_LATIN1((U8) c) >= 0xC0 \
3ded5eb0
KW
1466 && NATIVE_TO_LATIN1((U8) c) <= 0xDE \
1467 && NATIVE_TO_LATIN1((U8) c) != 0xD7)))
1468# define isWORDCHAR_L1(c) (isIDFIRST_L1(c) || isDIGIT_A(c))
1469# define isIDFIRST_L1(c) (isALPHA_L1(c) || NATIVE_TO_LATIN1(c) == '_')
1470# define isCHARNAME_CONT(c) (isWORDCHAR_L1(c) \
1471 || isBLANK_L1(c) \
1472 || (c) == '-' \
1473 || (c) == '(' \
1474 || (c) == ')')
1475 /* The following are not fully accurate in the above-ASCII range. I (khw)
1476 * don't think it's necessary to be so for the purposes where this gets
1477 * compiled */
1478# define _isQUOTEMETA(c) (FITS_IN_8_BITS(c) && ! isWORDCHAR_L1(c))
1479# define _IS_IN_SOME_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) isALPHA_L1(c)
1480
1481 /* And these aren't accurate at all. They are useful only for above
1482 * Latin1, which utilities and bootstrapping don't deal with */
1483# define _IS_NON_FINAL_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) 0
6838b41e 1484# define _HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) 0
3ded5eb0
KW
1485# define _HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) 0
1486
1487 /* Many of the macros later in this file are defined in terms of these. By
1488 * implementing them with a function, which converts the class number into
1489 * a call to the desired macro, all of the later ones work. However, that
1490 * function won't be actually defined when building a utility program (no
1491 * perl.h), and so a compiler error will be generated if one is attempted
1492 * to be used. And the above-Latin1 code points require Unicode tables to
1493 * be present, something unlikely to be the case when bootstrapping */
1494# define _generic_isCC(c, classnum) \
1495 (FITS_IN_8_BITS(c) && S_bootstrap_ctype((U8) (c), (classnum), TRUE))
1496# define _generic_isCC_A(c, classnum) \
1497 (FITS_IN_8_BITS(c) && S_bootstrap_ctype((U8) (c), (classnum), FALSE))
687c8d01 1498#endif /* End of no perl.h H_PERL */
8a58bdcf 1499
e66b99e9
KW
1500#define isALPHANUMERIC(c) isALPHANUMERIC_A(c)
1501#define isALPHA(c) isALPHA_A(c)
0852beac
KW
1502#define isASCII_A(c) isASCII(c)
1503#define isASCII_L1(c) isASCII(c)
e66b99e9
KW
1504#define isBLANK(c) isBLANK_A(c)
1505#define isCNTRL(c) isCNTRL_A(c)
1506#define isDIGIT(c) isDIGIT_A(c)
1507#define isGRAPH(c) isGRAPH_A(c)
1508#define isIDFIRST(c) isIDFIRST_A(c)
1509#define isLOWER(c) isLOWER_A(c)
1510#define isPRINT(c) isPRINT_A(c)
779cf272 1511#define isPSXSPC_A(c) isSPACE_A(c)
e66b99e9 1512#define isPSXSPC(c) isPSXSPC_A(c)
779cf272 1513#define isPSXSPC_L1(c) isSPACE_L1(c)
e66b99e9
KW
1514#define isPUNCT(c) isPUNCT_A(c)
1515#define isSPACE(c) isSPACE_A(c)
1516#define isUPPER(c) isUPPER_A(c)
1517#define isWORDCHAR(c) isWORDCHAR_A(c)
1518#define isXDIGIT(c) isXDIGIT_A(c)
1519
1520/* ASCII casing. These could also be written as
1521 #define toLOWER(c) (isASCII(c) ? toLOWER_LATIN1(c) : (c))
1522 #define toUPPER(c) (isASCII(c) ? toUPPER_LATIN1_MOD(c) : (c))
1523 which uses table lookup and mask instead of subtraction. (This would
1524 work because the _MOD does not apply in the ASCII range) */
68067e4e
DM
1525#define toLOWER(c) (isUPPER(c) ? (U8)((c) + ('a' - 'A')) : (c))
1526#define toUPPER(c) (isLOWER(c) ? (U8)((c) - ('a' - 'A')) : (c))
bbce6d69 1527
25200305
KW
1528/* In the ASCII range, these are equivalent to what they're here defined to be.
1529 * But by creating these definitions, other code doesn't have to be aware of
1530 * this detail */
1531#define toFOLD(c) toLOWER(c)
25200305
KW
1532#define toTITLE(c) toUPPER(c)
1533
c753c8d3
KW
1534#define toLOWER_A(c) toLOWER(c)
1535#define toUPPER_A(c) toUPPER(c)
25200305
KW
1536#define toFOLD_A(c) toFOLD(c)
1537#define toTITLE_A(c) toTITLE(c)
1a0901db 1538
4650c663 1539/* Use table lookup for speed; returns the input itself if is out-of-range */
b2bf251f 1540#define toLOWER_LATIN1(c) ((! FITS_IN_8_BITS(c)) \
8e7c6e7d 1541 ? (c) \
f4cd282c 1542 : PL_latin1_lc[ (U8) (c) ])
c753c8d3
KW
1543#define toLOWER_L1(c) toLOWER_LATIN1(c) /* Synonym for consistency */
1544
1a0901db 1545/* Modified uc. Is correct uc except for three non-ascii chars which are
4650c663
KW
1546 * all mapped to one of them, and these need special handling; returns the
1547 * input itself if is out-of-range */
b2bf251f 1548#define toUPPER_LATIN1_MOD(c) ((! FITS_IN_8_BITS(c)) \
8e7c6e7d 1549 ? (c) \
f4cd282c 1550 : PL_mod_latin1_uc[ (U8) (c) ])
31f05a37 1551#define IN_UTF8_CTYPE_LOCALE PL_in_utf8_CTYPE_locale
84061b6a 1552
beab9ebe
KW
1553/* Use foo_LC_uvchr() instead of these for beyond the Latin1 range */
1554
1555/* For internal core Perl use only: the base macro for defining macros like
1556 * isALPHA_LC, which uses the current LC_CTYPE locale. 'c' is the code point
31f05a37
KW
1557 * (0-255) to check. In a UTF-8 locale, the result is the same as calling
1558 * isFOO_L1(); the 'utf8_locale_classnum' parameter is something like
1559 * _CC_UPPER, which gives the class number for doing this. For non-UTF-8
1560 * locales, the code to actually do the test this is passed in 'non_utf8'. If
1561 * 'c' is above 255, 0 is returned. For accessing the full range of possible
1562 * code points under locale rules, use the macros based on _generic_LC_uvchr
1563 * instead of this. */
beab9ebe
KW
1564#define _generic_LC_base(c, utf8_locale_classnum, non_utf8) \
1565 (! FITS_IN_8_BITS(c) \
1566 ? 0 \
31f05a37
KW
1567 : IN_UTF8_CTYPE_LOCALE \
1568 ? cBOOL(PL_charclass[(U8) (c)] & _CC_mask(utf8_locale_classnum)) \
beab9ebe
KW
1569 : cBOOL(non_utf8))
1570
1571/* For internal core Perl use only: a helper macro for defining macros like
1572 * isALPHA_LC. 'c' is the code point (0-255) to check. The function name to
1573 * actually do this test is passed in 'non_utf8_func', which is called on 'c',
1574 * casting 'c' to the macro _LC_CAST, which should not be parenthesized. See
1575 * _generic_LC_base for more info */
1576#define _generic_LC(c, utf8_locale_classnum, non_utf8_func) \
1577 _generic_LC_base(c,utf8_locale_classnum, \
1578 non_utf8_func( (_LC_CAST) (c)))
1579
1580/* For internal core Perl use only: like _generic_LC, but also returns TRUE if
1581 * 'c' is the platform's native underscore character */
1582#define _generic_LC_underscore(c,utf8_locale_classnum,non_utf8_func) \
1583 _generic_LC_base(c, utf8_locale_classnum, \
1584 (non_utf8_func( (_LC_CAST) (c)) \
1585 || (char)(c) == '_'))
1586
1587/* These next three are also for internal core Perl use only: case-change
1588 * helper macros */
1589#define _generic_toLOWER_LC(c, function, cast) (! FITS_IN_8_BITS(c) \
1590 ? (c) \
31f05a37
KW
1591 : (IN_UTF8_CTYPE_LOCALE) \
1592 ? PL_latin1_lc[ (U8) (c) ] \
d2a8cf7f 1593 : (cast)function((cast)(c)))
beab9ebe 1594
31f05a37
KW
1595/* Note that the result can be larger than a byte in a UTF-8 locale. It
1596 * returns a single value, so can't adequately return the upper case of LATIN
1597 * SMALL LETTER SHARP S in a UTF-8 locale (which should be a string of two
1598 * values "SS"); instead it asserts against that under DEBUGGING, and
1599 * otherwise returns its input */
beab9ebe
KW
1600#define _generic_toUPPER_LC(c, function, cast) \
1601 (! FITS_IN_8_BITS(c) \
1602 ? (c) \
31f05a37 1603 : ((! IN_UTF8_CTYPE_LOCALE) \
b7d90381 1604 ? (cast)function((cast)(c)) \
31f05a37
KW
1605 : ((((U8)(c)) == MICRO_SIGN) \
1606 ? GREEK_CAPITAL_LETTER_MU \
1607 : ((((U8)(c)) == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS) \
1608 ? LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS \
1609 : ((((U8)(c)) == LATIN_SMALL_LETTER_SHARP_S) \
1610 ? (__ASSERT_(0) (c)) \
1611 : PL_mod_latin1_uc[ (U8) (c) ])))))
1612
1613/* Note that the result can be larger than a byte in a UTF-8 locale. It
1614 * returns a single value, so can't adequately return the fold case of LATIN
1615 * SMALL LETTER SHARP S in a UTF-8 locale (which should be a string of two
1616 * values "ss"); instead it asserts against that under DEBUGGING, and
1617 * otherwise returns its input */
beab9ebe 1618#define _generic_toFOLD_LC(c, function, cast) \
4d7db1b9 1619 ((UNLIKELY((c) == MICRO_SIGN) && IN_UTF8_CTYPE_LOCALE) \
31f05a37 1620 ? GREEK_SMALL_LETTER_MU \
4d7db1b9
KW
1621 : (__ASSERT_(! IN_UTF8_CTYPE_LOCALE \
1622 || (c) != LATIN_SMALL_LETTER_SHARP_S) \
1623 _generic_toLOWER_LC(c, function, cast)))
beab9ebe 1624
84061b6a 1625/* Use the libc versions for these if available. */
f05550c0 1626#if defined(HAS_ISASCII)
84061b6a
KW
1627# define isASCII_LC(c) (FITS_IN_8_BITS(c) && isascii( (U8) (c)))
1628#else
1629# define isASCII_LC(c) isASCII(c)
1630#endif
1631
f05550c0 1632#if defined(HAS_ISBLANK)
84061b6a 1633# define isBLANK_LC(c) _generic_LC(c, _CC_BLANK, isblank)
95f34b6f 1634#else /* Unlike isASCII, varies if in a UTF-8 locale */
278b2b56 1635# define isBLANK_LC(c) ((IN_UTF8_CTYPE_LOCALE) ? isBLANK_L1(c) : isBLANK(c))
84061b6a
KW
1636#endif
1637
f05550c0 1638#define _LC_CAST U8
bbce6d69 1639
f05550c0 1640#ifdef WIN32
375f5f06
KW
1641 /* The Windows functions don't bother to follow the POSIX standard, which
1642 * for example says that something can't both be a printable and a control.
1643 * But Windows treats the \t control as a printable, and does such things
1644 * as making superscripts into both digits and punctuation. This tames
1645 * these flaws by assuming that the definitions of both controls and space
1646 * are correct, and then making sure that other definitions don't have
1647 * weirdnesses, by making sure that isalnum() isn't also ispunct(), etc.
1648 * Not all possible weirdnesses are checked for, just the ones that were
1649 * detected on actual Microsoft code pages */
1650
b7d90381
KW
1651# define isCNTRL_LC(c) _generic_LC(c, _CC_CNTRL, iscntrl)
1652# define isSPACE_LC(c) _generic_LC(c, _CC_SPACE, isspace)
1653
1654# define isALPHA_LC(c) (_generic_LC(c, _CC_ALPHA, isalpha) \
1655 && isALPHANUMERIC_LC(c))
1656# define isALPHANUMERIC_LC(c) (_generic_LC(c, _CC_ALPHANUMERIC, isalnum) && \
1657 ! isPUNCT_LC(c))
1658# define isDIGIT_LC(c) (_generic_LC(c, _CC_DIGIT, isdigit) && \
1659 isALPHANUMERIC_LC(c))
1660# define isGRAPH_LC(c) (_generic_LC(c, _CC_GRAPH, isgraph) && isPRINT_LC(c))
1661# define isIDFIRST_LC(c) (((c) == '_') \
1662 || (_generic_LC(c, _CC_IDFIRST, isalpha) && ! isPUNCT_LC(c)))
1663# define isLOWER_LC(c) (_generic_LC(c, _CC_LOWER, islower) && isALPHA_LC(c))
1664# define isPRINT_LC(c) (_generic_LC(c, _CC_PRINT, isprint) && ! isCNTRL_LC(c))
1665# define isPUNCT_LC(c) (_generic_LC(c, _CC_PUNCT, ispunct) && ! isCNTRL_LC(c))
1666# define isUPPER_LC(c) (_generic_LC(c, _CC_UPPER, isupper) && isALPHA_LC(c))
f05550c0 1667# define isWORDCHAR_LC(c) (((c) == '_') || isALPHANUMERIC_LC(c))
b7d90381
KW
1668# define isXDIGIT_LC(c) (_generic_LC(c, _CC_XDIGIT, isxdigit) \
1669 && isALPHANUMERIC_LC(c))
f05550c0
BF
1670
1671# define toLOWER_LC(c) _generic_toLOWER_LC((c), tolower, U8)
1672# define toUPPER_LC(c) _generic_toUPPER_LC((c), toupper, U8)
1673# define toFOLD_LC(c) _generic_toFOLD_LC((c), tolower, U8)
1674
1675#elif defined(CTYPE256) || (!defined(isascii) && !defined(HAS_ISASCII))
4650c663 1676 /* For most other platforms */
beab9ebe 1677
f05550c0
BF
1678# define isALPHA_LC(c) _generic_LC(c, _CC_ALPHA, isalpha)
1679# define isALPHANUMERIC_LC(c) _generic_LC(c, _CC_ALPHANUMERIC, isalnum)
1680# define isCNTRL_LC(c) _generic_LC(c, _CC_CNTRL, iscntrl)
1681# define isDIGIT_LC(c) _generic_LC(c, _CC_DIGIT, isdigit)
1682# define isGRAPH_LC(c) _generic_LC(c, _CC_GRAPH, isgraph)
1683# define isIDFIRST_LC(c) _generic_LC_underscore(c, _CC_IDFIRST, isalpha)
1684# define isLOWER_LC(c) _generic_LC(c, _CC_LOWER, islower)
1685# define isPRINT_LC(c) _generic_LC(c, _CC_PRINT, isprint)
1686# define isPUNCT_LC(c) _generic_LC(c, _CC_PUNCT, ispunct)
1687# define isSPACE_LC(c) _generic_LC(c, _CC_SPACE, isspace)
1688# define isUPPER_LC(c) _generic_LC(c, _CC_UPPER, isupper)
1689# define isWORDCHAR_LC(c) _generic_LC_underscore(c, _CC_WORDCHAR, isalnum)
1690# define isXDIGIT_LC(c) _generic_LC(c, _CC_XDIGIT, isxdigit)
1691
1692
1693# define toLOWER_LC(c) _generic_toLOWER_LC((c), tolower, U8)
1694# define toUPPER_LC(c) _generic_toUPPER_LC((c), toupper, U8)
1695# define toFOLD_LC(c) _generic_toFOLD_LC((c), tolower, U8)
1696
1697#else /* The final fallback position */
1698
b7d90381
KW
1699# define isALPHA_LC(c) (isascii(c) && isalpha(c))
1700# define isALPHANUMERIC_LC(c) (isascii(c) && isalnum(c))
1701# define isCNTRL_LC(c) (isascii(c) && iscntrl(c))
1702# define isDIGIT_LC(c) (isascii(c) && isdigit(c))
1703# define isGRAPH_LC(c) (isascii(c) && isgraph(c))
f05550c0 1704# define isIDFIRST_LC(c) (isascii(c) && (isalpha(c) || (c) == '_'))
b7d90381
KW
1705# define isLOWER_LC(c) (isascii(c) && islower(c))
1706# define isPRINT_LC(c) (isascii(c) && isprint(c))
1707# define isPUNCT_LC(c) (isascii(c) && ispunct(c))
1708# define isSPACE_LC(c) (isascii(c) && isspace(c))
1709# define isUPPER_LC(c) (isascii(c) && isupper(c))
f05550c0 1710# define isWORDCHAR_LC(c) (isascii(c) && (isalnum(c) || (c) == '_'))
b7d90381 1711# define isXDIGIT_LC(c) (isascii(c) && isxdigit(c))
f05550c0
BF
1712
1713# define toLOWER_LC(c) (isascii(c) ? tolower(c) : (c))
1714# define toUPPER_LC(c) (isascii(c) ? toupper(c) : (c))
1715# define toFOLD_LC(c) (isascii(c) ? tolower(c) : (c))
bbce6d69 1716
f05550c0 1717#endif
55204971 1718
eba68aa0
KW
1719#define isIDCONT(c) isWORDCHAR(c)
1720#define isIDCONT_A(c) isWORDCHAR_A(c)
1721#define isIDCONT_L1(c) isWORDCHAR_L1(c)
1722#define isIDCONT_LC(c) isWORDCHAR_LC(c)
13380643 1723#define isPSXSPC_LC(c) isSPACE_LC(c)
aaa51d5e 1724
4650c663 1725/* For internal core Perl use only: the base macros for defining macros like
d0da05db
KW
1726 * isALPHA_uvchr. 'c' is the code point to check. 'classnum' is the POSIX class
1727 * number defined earlier in this file. _generic_uvchr() is used for POSIX
4650c663
KW
1728 * classes where there is a macro or function 'above_latin1' that takes the
1729 * single argument 'c' and returns the desired value. These exist for those
1730 * classes which have simple definitions, avoiding the overhead of a hash
d0da05db 1731 * lookup or inversion list binary search. _generic_swash_uvchr() can be used
4650c663 1732 * for classes where that overhead is faster than a direct lookup.
d0da05db 1733 * _generic_uvchr() won't compile if 'c' isn't unsigned, as it won't match the
4650c663
KW
1734 * 'above_latin1' prototype. _generic_isCC() macro does bounds checking, so
1735 * have duplicate checks here, so could create versions of the macros that
1736 * don't, but experiments show that gcc optimizes them out anyway. */
66c17564
KW
1737
1738/* Note that all ignore 'use bytes' */
1e222e4f
KW
1739#define _generic_uvchr(classnum, above_latin1, c) ((c) < 256 \
1740 ? _generic_isCC(c, classnum) \
cd500f2f 1741 : above_latin1(c))
1e222e4f
KW
1742#define _generic_swash_uvchr(classnum, c) ((c) < 256 \
1743 ? _generic_isCC(c, classnum) \
922e8cb4 1744 : _is_uni_FOO(classnum, c))
d0da05db
KW
1745#define isALPHA_uvchr(c) _generic_swash_uvchr(_CC_ALPHA, c)
1746#define isALPHANUMERIC_uvchr(c) _generic_swash_uvchr(_CC_ALPHANUMERIC, c)
1747#define isASCII_uvchr(c) isASCII(c)
1748#define isBLANK_uvchr(c) _generic_uvchr(_CC_BLANK, is_HORIZWS_cp_high, c)
1749#define isCNTRL_uvchr(c) isCNTRL_L1(c) /* All controls are in Latin1 */
1750#define isDIGIT_uvchr(c) _generic_swash_uvchr(_CC_DIGIT, c)
1751#define isGRAPH_uvchr(c) _generic_swash_uvchr(_CC_GRAPH, c)
1e222e4f
KW
1752#define isIDCONT_uvchr(c) \
1753 _generic_uvchr(_CC_WORDCHAR, _is_uni_perl_idcont, c)
1754#define isIDFIRST_uvchr(c) \
1755 _generic_uvchr(_CC_IDFIRST, _is_uni_perl_idstart, c)
d0da05db
KW
1756#define isLOWER_uvchr(c) _generic_swash_uvchr(_CC_LOWER, c)
1757#define isPRINT_uvchr(c) _generic_swash_uvchr(_CC_PRINT, c)
1758
1759#define isPUNCT_uvchr(c) _generic_swash_uvchr(_CC_PUNCT, c)
1760#define isSPACE_uvchr(c) _generic_uvchr(_CC_SPACE, is_XPERLSPACE_cp_high, c)
1761#define isPSXSPC_uvchr(c) isSPACE_uvchr(c)
1762
1763#define isUPPER_uvchr(c) _generic_swash_uvchr(_CC_UPPER, c)
1764#define isVERTWS_uvchr(c) _generic_uvchr(_CC_VERTSPACE, is_VERTWS_cp_high, c)
1765#define isWORDCHAR_uvchr(c) _generic_swash_uvchr(_CC_WORDCHAR, c)
1766#define isXDIGIT_uvchr(c) _generic_uvchr(_CC_XDIGIT, is_XDIGIT_cp_high, c)
1767
1768#define toFOLD_uvchr(c,s,l) to_uni_fold(c,s,l)
1769#define toLOWER_uvchr(c,s,l) to_uni_lower(c,s,l)
1770#define toTITLE_uvchr(c,s,l) to_uni_title(c,s,l)
1771#define toUPPER_uvchr(c,s,l) to_uni_upper(c,s,l)
1772
1773/* For backwards compatibility, even though '_uni' should mean official Unicode
1774 * code points, in Perl it means native for those below 256 */
1775#define isALPHA_uni(c) isALPHA_uvchr(c)
1776#define isALPHANUMERIC_uni(c) isALPHANUMERIC_uvchr(c)
1777#define isASCII_uni(c) isASCII_uvchr(c)
1778#define isBLANK_uni(c) isBLANK_uvchr(c)
1779#define isCNTRL_uni(c) isCNTRL_uvchr(c)
1780#define isDIGIT_uni(c) isDIGIT_uvchr(c)
1781#define isGRAPH_uni(c) isGRAPH_uvchr(c)
1782#define isIDCONT_uni(c) isIDCONT_uvchr(c)
1783#define isIDFIRST_uni(c) isIDFIRST_uvchr(c)
1784#define isLOWER_uni(c) isLOWER_uvchr(c)
1785#define isPRINT_uni(c) isPRINT_uvchr(c)
1786#define isPUNCT_uni(c) isPUNCT_uvchr(c)
1787#define isSPACE_uni(c) isSPACE_uvchr(c)
1788#define isPSXSPC_uni(c) isPSXSPC_uvchr(c)
1789#define isUPPER_uni(c) isUPPER_uvchr(c)
1790#define isVERTWS_uni(c) isVERTWS_uvchr(c)
1791#define isWORDCHAR_uni(c) isWORDCHAR_uvchr(c)
1792#define isXDIGIT_uni(c) isXDIGIT_uvchr(c)
1793#define toFOLD_uni(c,s,l) toFOLD_uvchr(c,s,l)
1794#define toLOWER_uni(c,s,l) toLOWER_uvchr(c,s,l)
1795#define toTITLE_uni(c,s,l) toTITLE_uvchr(c,s,l)
1796#define toUPPER_uni(c,s,l) toUPPER_uvchr(c,s,l)
a0ed51b3 1797
4650c663
KW
1798/* For internal core Perl use only: the base macros for defining macros like
1799 * isALPHA_LC_uvchr. These are like isALPHA_LC, but the input can be any code
d0da05db 1800 * point, not just 0-255. Like _generic_uvchr, there are two versions, one for
4650c663 1801 * simple class definitions; the other for more complex. These are like
d0da05db 1802 * _generic_uvchr, so see it for more info. */
cd500f2f
KW
1803#define _generic_LC_uvchr(latin1, above_latin1, c) \
1804 (c < 256 ? latin1(c) : above_latin1(c))
1805#define _generic_LC_swash_uvchr(latin1, classnum, c) \
1806 (c < 256 ? latin1(c) : _is_uni_FOO(classnum, c))
1807
1808#define isALPHA_LC_uvchr(c) _generic_LC_swash_uvchr(isALPHA_LC, _CC_ALPHA, c)
1809#define isALPHANUMERIC_LC_uvchr(c) _generic_LC_swash_uvchr(isALPHANUMERIC_LC, \
1810 _CC_ALPHANUMERIC, c)
b7d90381
KW
1811#define isASCII_LC_uvchr(c) isASCII_LC(c)
1812#define isBLANK_LC_uvchr(c) _generic_LC_uvchr(isBLANK_LC, \
1813 is_HORIZWS_cp_high, c)
feeab5a9 1814#define isCNTRL_LC_uvchr(c) (c < 256 ? isCNTRL_LC(c) : 0)
cd500f2f
KW
1815#define isDIGIT_LC_uvchr(c) _generic_LC_swash_uvchr(isDIGIT_LC, _CC_DIGIT, c)
1816#define isGRAPH_LC_uvchr(c) _generic_LC_swash_uvchr(isGRAPH_LC, _CC_GRAPH, c)
b7d90381 1817#define isIDCONT_LC_uvchr(c) _generic_LC_uvchr(isIDCONT_LC, \
eba68aa0 1818 _is_uni_perl_idcont, c)
b7d90381 1819#define isIDFIRST_LC_uvchr(c) _generic_LC_uvchr(isIDFIRST_LC, \
cd500f2f
KW
1820 _is_uni_perl_idstart, c)
1821#define isLOWER_LC_uvchr(c) _generic_LC_swash_uvchr(isLOWER_LC, _CC_LOWER, c)
1822#define isPRINT_LC_uvchr(c) _generic_LC_swash_uvchr(isPRINT_LC, _CC_PRINT, c)
b7d90381 1823#define isPSXSPC_LC_uvchr(c) isSPACE_LC_uvchr(c)
cd500f2f 1824#define isPUNCT_LC_uvchr(c) _generic_LC_swash_uvchr(isPUNCT_LC, _CC_PUNCT, c)
b7d90381 1825#define isSPACE_LC_uvchr(c) _generic_LC_uvchr(isSPACE_LC, \
509fb054 1826 is_XPERLSPACE_cp_high, c)
cd500f2f 1827#define isUPPER_LC_uvchr(c) _generic_LC_swash_uvchr(isUPPER_LC, _CC_UPPER, c)
b7d90381 1828#define isWORDCHAR_LC_uvchr(c) _generic_LC_swash_uvchr(isWORDCHAR_LC, \
cd500f2f 1829 _CC_WORDCHAR, c)
b7d90381
KW
1830#define isXDIGIT_LC_uvchr(c) _generic_LC_uvchr(isXDIGIT_LC, \
1831 is_XDIGIT_cp_high, c)
e712593e 1832
b7d90381 1833#define isBLANK_LC_uni(c) isBLANK_LC_uvchr(UNI_TO_NATIVE(c))
aaa51d5e 1834
4650c663
KW
1835/* For internal core Perl use only: the base macros for defining macros like
1836 * isALPHA_utf8. These are like the earlier defined macros, but take an input
1837 * UTF-8 encoded string 'p'. If the input is in the Latin1 range, use
1838 * the Latin1 macro 'classnum' on 'p'. Otherwise use the value given by the
1839 * 'utf8' parameter. This relies on the fact that ASCII characters have the
1840 * same representation whether utf8 or not. Note that it assumes that the utf8
1841 * has been validated, and ignores 'use bytes' */
34aeb2e9
KW
1842#define _base_generic_utf8(enum_name, name, p, use_locale ) \
1843 _is_utf8_FOO(CAT2(_CC_, enum_name), \
1844 (const U8 *) p, \
1845 "is" STRINGIFY(name) "_utf8", \
1846 "is" STRINGIFY(name) "_utf8_safe", \
1847 1, use_locale, __FILE__,__LINE__)
1848
1849#define _generic_utf8(name, p) _base_generic_utf8(name, name, p, 0)
da8c1a98
KW
1850
1851/* The "_safe" macros make sure that we don't attempt to read beyond 'e', but
1852 * they don't otherwise go out of their way to look for malformed UTF-8. If
1853 * they can return accurate results without knowing if the input is otherwise
1854 * malformed, they do so. For example isASCII is accurate in spite of any
1855 * non-length malformations because it looks only at a single byte. Likewise
1856 * isDIGIT looks just at the first byte for code points 0-255, as all UTF-8
1857 * variant ones return FALSE. But, if the input has to be well-formed in order
1858 * for the results to be accurate, the macros will test and if malformed will
1859 * call a routine to die
1860 *
1861 * Except for toke.c, the macros do assume that e > p, asserting that on
1862 * DEBUGGING builds. Much code that calls these depends on this being true,
1863 * for other reasons. toke.c is treated specially as using the regular
1864 * assertion breaks it in many ways. All strings that these operate on there
1865 * are supposed to have an extra NUL character at the end, so that *e = \0. A
1866 * bunch of code in toke.c assumes that this is true, so the assertion allows
1867 * for that */
1868#ifdef PERL_IN_TOKE_C
1869# define _utf8_safe_assert(p,e) ((e) > (p) || ((e) == (p) && *(p) == '\0'))
1870#else
1871# define _utf8_safe_assert(p,e) ((e) > (p))
1872#endif
1873
1874#define _generic_utf8_safe(classnum, p, e, above_latin1) \
1875 (__ASSERT_(_utf8_safe_assert(p, e)) \
1876 (UTF8_IS_INVARIANT(*(p))) \
1877 ? _generic_isCC(*(p), classnum) \
1878 : (UTF8_IS_DOWNGRADEABLE_START(*(p)) \
1879 ? ((LIKELY((e) - (p) > 1 && UTF8_IS_CONTINUATION(*((p)+1)))) \
1880 ? _generic_isCC(EIGHT_BIT_UTF8_TO_NATIVE(*(p), *((p)+1 )), \
1881 classnum) \
1882 : (_force_out_malformed_utf8_message( \
1883 (U8 *) (p), (U8 *) (e), 0, 1), 0)) \
1884 : above_latin1))
b7d90381
KW
1885/* Like the above, but calls 'above_latin1(p)' to get the utf8 value.
1886 * 'above_latin1' can be a macro */
da8c1a98
KW
1887#define _generic_func_utf8_safe(classnum, above_latin1, p, e) \
1888 _generic_utf8_safe(classnum, p, e, above_latin1(p, e))
1889#define _generic_non_swash_utf8_safe(classnum, above_latin1, p, e) \
1890 _generic_utf8_safe(classnum, p, e, \
1891 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
1892 ? (_force_out_malformed_utf8_message( \
1893 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
1894 : above_latin1(p)))
4650c663 1895/* Like the above, but passes classnum to _isFOO_utf8(), instead of having an
cd500f2f 1896 * 'above_latin1' parameter */
da8c1a98
KW
1897#define _generic_swash_utf8_safe(classnum, p, e) \
1898_generic_utf8_safe(classnum, p, e, _is_utf8_FOO_with_len(classnum, p, e))
922e8cb4 1899
cc8ab7c0 1900/* Like the above, but should be used only when it is known that there are no
ff7ecfc3
KW
1901 * characters in the upper-Latin1 range (128-255 on ASCII platforms) which the
1902 * class is TRUE for. Hence it can skip the tests for this range.
1903 * 'above_latin1' should include its arguments */
da8c1a98
KW
1904#define _generic_utf8_safe_no_upper_latin1(classnum, p, e, above_latin1) \
1905 (__ASSERT_(_utf8_safe_assert(p, e)) \
1906 (UTF8_IS_INVARIANT(*(p))) \
1907 ? _generic_isCC(*(p), classnum) \
1908 : (UTF8_IS_DOWNGRADEABLE_START(*(p))) \
1909 ? 0 /* Note that doesn't check validity for latin1 */ \
1910 : above_latin1)
1911
84238efa
KW
1912/* NOTE that some of these macros have very similar ones in regcharclass.h.
1913 * For example, there is (at the time of this writing) an 'is_SPACE_utf8()'
1914 * there, differing in name only by an underscore from the one here
1915 * 'isSPACE_utf8(). The difference is that the ones here are probably more
1916 * efficient and smaller, using an O(1) array lookup for Latin1-range code
1917 * points; the regcharclass.h ones are implemented as a series of
1918 * "if-else-if-else ..." */
1919
34aeb2e9
KW
1920#define isALPHA_utf8(p) _generic_utf8(ALPHA, p)
1921#define isALPHANUMERIC_utf8(p) _generic_utf8(ALPHANUMERIC, p)
1922#define isASCII_utf8(p) _generic_utf8(ASCII, p)
1923#define isBLANK_utf8(p) _generic_utf8(BLANK, p)
1924#define isCNTRL_utf8(p) _generic_utf8(CNTRL, p)
1925#define isDIGIT_utf8(p) _generic_utf8(DIGIT, p)
1926#define isGRAPH_utf8(p) _generic_utf8(GRAPH, p)
1927#define isIDCONT_utf8(p) _generic_utf8(IDCONT, p)
1928#define isIDFIRST_utf8(p) _generic_utf8(IDFIRST, p)
1929#define isLOWER_utf8(p) _generic_utf8(LOWER, p)
1930#define isPRINT_utf8(p) _generic_utf8(PRINT, p)
1931#define isPSXSPC_utf8(p) _generic_utf8(PSXSPC, p)
1932#define isPUNCT_utf8(p) _generic_utf8(PUNCT, p)
1933#define isSPACE_utf8(p) _generic_utf8(SPACE, p)
1934#define isUPPER_utf8(p) _generic_utf8(UPPER, p)
1935#define isVERTWS_utf8(p) _generic_utf8(VERTSPACE, p)
1936#define isWORDCHAR_utf8(p) _generic_utf8(WORDCHAR, p)
1937#define isXDIGIT_utf8(p) _generic_utf8(XDIGIT, p)
e8fa43e2 1938
da8c1a98
KW
1939#define isALPHA_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_ALPHA, p, e)
1940#define isALPHANUMERIC_utf8_safe(p, e) \
1941 _generic_swash_utf8_safe(_CC_ALPHANUMERIC, p, e)
1942#define isASCII_utf8_safe(p, e) \
1943 /* Because ASCII is invariant under utf8, the non-utf8 macro \
1944 * works */ \
1945 (__ASSERT_(_utf8_safe_assert(p, e)) isASCII(*(p)))
1946#define isBLANK_utf8_safe(p, e) \
1947 _generic_non_swash_utf8_safe(_CC_BLANK, is_HORIZWS_high, p, e)
1948
e8fa43e2
KW
1949#ifdef EBCDIC
1950 /* Because all controls are UTF-8 invariants in EBCDIC, we can use this
1951 * more efficient macro instead of the more general one */
da8c1a98 1952# define isCNTRL_utf8_safe(p, e) \
56d02b8c 1953 (__ASSERT_(_utf8_safe_assert(p, e)) isCNTRL_L1(*(p)))
e8fa43e2 1954#else
da8c1a98 1955# define isCNTRL_utf8_safe(p, e) _generic_utf8_safe(_CC_CNTRL, p, e, 0)
e8fa43e2
KW
1956#endif
1957
da8c1a98
KW
1958#define isDIGIT_utf8_safe(p, e) \
1959 _generic_utf8_safe_no_upper_latin1(_CC_DIGIT, p, e, \
1960 _is_utf8_FOO_with_len(_CC_DIGIT, p, e))
1961#define isGRAPH_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_GRAPH, p, e)
1962#define isIDCONT_utf8_safe(p, e) _generic_func_utf8_safe(_CC_WORDCHAR, \
1963 _is_utf8_perl_idcont_with_len, p, e)
e5dcd934 1964
c11ff943
KW
1965/* To prevent S_scan_word in toke.c from hanging, we have to make sure that
1966 * IDFIRST is an alnum. See
c8362b00 1967 * http://rt.perl.org/rt3/Ticket/Display.html?id=74022 for more detail than you
f91dcd13
KW
1968 * ever wanted to know about. (In the ASCII range, there isn't a difference.)
1969 * This used to be not the XID version, but we decided to go with the more
1970 * modern Unicode definition */
da8c1a98
KW
1971#define isIDFIRST_utf8_safe(p, e) \
1972 _generic_func_utf8_safe(_CC_IDFIRST, \
1973 _is_utf8_perl_idstart_with_len, (U8 *) (p), (U8 *) (e))
1974
1975#define isLOWER_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_LOWER, p, e)
1976#define isPRINT_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_PRINT, p, e)
1977#define isPSXSPC_utf8_safe(p, e) isSPACE_utf8_safe(p, e)
1978#define isPUNCT_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_PUNCT, p, e)
1979#define isSPACE_utf8_safe(p, e) \
1980 _generic_non_swash_utf8_safe(_CC_SPACE, is_XPERLSPACE_high, p, e)
1981#define isUPPER_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_UPPER, p, e)
1982#define isVERTWS_utf8_safe(p, e) \
1983 _generic_non_swash_utf8_safe(_CC_VERTSPACE, is_VERTWS_high, p, e)
1984#define isWORDCHAR_utf8_safe(p, e) \
1985 _generic_swash_utf8_safe(_CC_WORDCHAR, p, e)
1986#define isXDIGIT_utf8_safe(p, e) \
1987 _generic_utf8_safe_no_upper_latin1(_CC_XDIGIT, p, e, \
1988 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
1989 ? (_force_out_malformed_utf8_message( \
1990 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
1991 : is_XDIGIT_high(p)))
a0ed51b3 1992
7d2b61dd 1993#define toFOLD_utf8(p,s,l) to_utf8_fold(p,s,l)
509fb054
KW
1994#define toLOWER_utf8(p,s,l) to_utf8_lower(p,s,l)
1995#define toTITLE_utf8(p,s,l) to_utf8_title(p,s,l)
1996#define toUPPER_utf8(p,s,l) to_utf8_upper(p,s,l)
2e8adce6 1997
567b353c 1998/* For internal core use only, subject to change */
607313a1
KW
1999#define _toFOLD_utf8_flags(p,e,s,l,f) _to_utf8_fold_flags (p,e,s,l,f, "", 0)
2000#define _toLOWER_utf8_flags(p,e,s,l,f) _to_utf8_lower_flags(p,e,s,l,f, "", 0)
2001#define _toTITLE_utf8_flags(p,e,s,l,f) _to_utf8_title_flags(p,e,s,l,f, "", 0)
2002#define _toUPPER_utf8_flags(p,e,s,l,f) _to_utf8_upper_flags(p,e,s,l,f, "", 0)
a1a5ec35
KW
2003
2004#define toFOLD_utf8_safe(p,e,s,l) _toFOLD_utf8_flags(p,e,s,l, FOLD_FLAGS_FULL)
2005#define toLOWER_utf8_safe(p,e,s,l) _toLOWER_utf8_flags(p,e,s,l, 0)
2006#define toTITLE_utf8_safe(p,e,s,l) _toTITLE_utf8_flags(p,e,s,l, 0)
2007#define toUPPER_utf8_safe(p,e,s,l) _toUPPER_utf8_flags(p,e,s,l, 0)
567b353c 2008
4650c663
KW
2009/* For internal core Perl use only: the base macros for defining macros like
2010 * isALPHA_LC_utf8. These are like _generic_utf8, but if the first code point
2011 * in 'p' is within the 0-255 range, it uses locale rules from the passed-in
2012 * 'macro' parameter */
34aeb2e9
KW
2013#define _generic_LC_utf8(name, p) _base_generic_utf8(name, name, p, 1)
2014
2015#define isALPHA_LC_utf8(p) _generic_LC_utf8(ALPHA, p)
2016#define isALPHANUMERIC_LC_utf8(p) _generic_LC_utf8(ALPHANUMERIC, p)
2017#define isASCII_LC_utf8(p) _generic_LC_utf8(ASCII, p)
2018#define isBLANK_LC_utf8(p) _generic_LC_utf8(BLANK, p)
2019#define isCNTRL_LC_utf8(p) _generic_LC_utf8(CNTRL, p)
2020#define isDIGIT_LC_utf8(p) _generic_LC_utf8(DIGIT, p)
2021#define isGRAPH_LC_utf8(p) _generic_LC_utf8(GRAPH, p)
2022#define isIDCONT_LC_utf8(p) _generic_LC_utf8(IDCONT, p)
2023#define isIDFIRST_LC_utf8(p) _generic_LC_utf8(IDFIRST, p)
2024#define isLOWER_LC_utf8(p) _generic_LC_utf8(LOWER, p)
2025#define isPRINT_LC_utf8(p) _generic_LC_utf8(PRINT, p)
2026#define isPSXSPC_LC_utf8(p) _generic_LC_utf8(PSXSPC, p)
2027#define isPUNCT_LC_utf8(p) _generic_LC_utf8(PUNCT, p)
2028#define isSPACE_LC_utf8(p) _generic_LC_utf8(SPACE, p)
2029#define isUPPER_LC_utf8(p) _generic_LC_utf8(UPPER, p)
2030#define isWORDCHAR_LC_utf8(p) _generic_LC_utf8(WORDCHAR, p)
2031#define isXDIGIT_LC_utf8(p) _generic_LC_utf8(XDIGIT, p)
2032
da8c1a98
KW
2033/* For internal core Perl use only: the base macros for defining macros like
2034 * isALPHA_LC_utf8_safe. These are like _generic_utf8, but if the first code
2035 * point in 'p' is within the 0-255 range, it uses locale rules from the
2036 * passed-in 'macro' parameter */
2037#define _generic_LC_utf8_safe(macro, p, e, above_latin1) \
2038 (__ASSERT_(_utf8_safe_assert(p, e)) \
2039 (UTF8_IS_INVARIANT(*(p))) \
2040 ? macro(*(p)) \
2041 : (UTF8_IS_DOWNGRADEABLE_START(*(p)) \
2042 ? ((LIKELY((e) - (p) > 1 && UTF8_IS_CONTINUATION(*((p)+1)))) \
2043 ? macro(EIGHT_BIT_UTF8_TO_NATIVE(*(p), *((p)+1))) \
2044 : (_force_out_malformed_utf8_message( \
2045 (U8 *) (p), (U8 *) (e), 0, 1), 0)) \
2046 : above_latin1))
2047
2048#define _generic_LC_swash_utf8_safe(macro, classnum, p, e) \
2049 _generic_LC_utf8_safe(macro, p, e, \
2050 _is_utf8_FOO_with_len(classnum, p, e))
2051
2052#define _generic_LC_func_utf8_safe(macro, above_latin1, p, e) \
2053 _generic_LC_utf8_safe(macro, p, e, above_latin1(p, e))
2054
2055#define _generic_LC_non_swash_utf8_safe(classnum, above_latin1, p, e) \
2056 _generic_LC_utf8_safe(classnum, p, e, \
2057 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
2058 ? (_force_out_malformed_utf8_message( \
2059 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
2060 : above_latin1(p)))
2061
2062#define isALPHANUMERIC_LC_utf8_safe(p, e) \
2063 _generic_LC_swash_utf8_safe(isALPHANUMERIC_LC, \
2064 _CC_ALPHANUMERIC, p, e)
2065#define isALPHA_LC_utf8_safe(p, e) \
2066 _generic_LC_swash_utf8_safe(isALPHA_LC, _CC_ALPHA, p, e)
2067#define isASCII_LC_utf8_safe(p, e) \
2068 (__ASSERT_(_utf8_safe_assert(p, e)) isASCII_LC(*(p)))
2069#define isBLANK_LC_utf8_safe(p, e) \
2070 _generic_LC_non_swash_utf8_safe(isBLANK_LC, is_HORIZWS_high, p, e)
2071#define isCNTRL_LC_utf8_safe(p, e) \
2072 _generic_LC_utf8_safe(isCNTRL_LC, p, e, 0)
2073#define isDIGIT_LC_utf8_safe(p, e) \
2074 _generic_LC_swash_utf8_safe(isDIGIT_LC, _CC_DIGIT, p, e)
2075#define isGRAPH_LC_utf8_safe(p, e) \
2076 _generic_LC_swash_utf8_safe(isGRAPH_LC, _CC_GRAPH, p, e)
2077#define isIDCONT_LC_utf8_safe(p, e) \
2078 _generic_LC_func_utf8_safe(isIDCONT_LC, \
2079 _is_utf8_perl_idcont_with_len, p, e)
2080#define isIDFIRST_LC_utf8_safe(p, e) \
2081 _generic_LC_func_utf8_safe(isIDFIRST_LC, \
2082 _is_utf8_perl_idstart_with_len, p, e)
2083#define isLOWER_LC_utf8_safe(p, e) \
2084 _generic_LC_swash_utf8_safe(isLOWER_LC, _CC_LOWER, p, e)
2085#define isPRINT_LC_utf8_safe(p, e) \
2086 _generic_LC_swash_utf8_safe(isPRINT_LC, _CC_PRINT, p, e)
2087#define isPSXSPC_LC_utf8_safe(p, e) isSPACE_LC_utf8_safe(p, e)
2088#define isPUNCT_LC_utf8_safe(p, e) \
2089 _generic_LC_swash_utf8_safe(isPUNCT_LC, _CC_PUNCT, p, e)
2090#define isSPACE_LC_utf8_safe(p, e) \
2091 _generic_LC_non_swash_utf8_safe(isSPACE_LC, is_XPERLSPACE_high, p, e)
2092#define isUPPER_LC_utf8_safe(p, e) \
2093 _generic_LC_swash_utf8_safe(isUPPER_LC, _CC_UPPER, p, e)
2094#define isWORDCHAR_LC_utf8_safe(p, e) \
2095 _generic_LC_swash_utf8_safe(isWORDCHAR_LC, _CC_WORDCHAR, p, e)
2096#define isXDIGIT_LC_utf8_safe(p, e) \
2097 _generic_LC_non_swash_utf8_safe(isXDIGIT_LC, is_XDIGIT_high, p, e)
aaa51d5e 2098
fbc19f27
KW
2099/* Macros for backwards compatibility and for completeness when the ASCII and
2100 * Latin1 values are identical */
b7d90381
KW
2101#define isALPHAU(c) isALPHA_L1(c)
2102#define isDIGIT_L1(c) isDIGIT_A(c)
2103#define isOCTAL(c) isOCTAL_A(c)
2104#define isOCTAL_L1(c) isOCTAL_A(c)
2105#define isXDIGIT_L1(c) isXDIGIT_A(c)
2106#define isALNUM(c) isWORDCHAR(c)
2107#define isALNUMU(c) isWORDCHAR_L1(c)
2108#define isALNUM_LC(c) isWORDCHAR_LC(c)
2109#define isALNUM_uni(c) isWORDCHAR_uni(c)
2e28f0b9 2110#define isALNUM_LC_uvchr(c) isWORDCHAR_LC_uvchr(c)
b7d90381
KW
2111#define isALNUM_utf8(p) isWORDCHAR_utf8(p)
2112#define isALNUM_LC_utf8(p) isWORDCHAR_LC_utf8(p)
2113#define isALNUMC_A(c) isALPHANUMERIC_A(c) /* Mnemonic: "C's alnum" */
2114#define isALNUMC_L1(c) isALPHANUMERIC_L1(c)
2115#define isALNUMC(c) isALPHANUMERIC(c)
2116#define isALNUMC_LC(c) isALPHANUMERIC_LC(c)
2117#define isALNUMC_uni(c) isALPHANUMERIC_uni(c)
15861f94 2118#define isALNUMC_LC_uvchr(c) isALPHANUMERIC_LC_uvchr(c)
b7d90381 2119#define isALNUMC_utf8(p) isALPHANUMERIC_utf8(p)
15861f94 2120#define isALNUMC_LC_utf8(p) isALPHANUMERIC_LC_utf8(p)
fbc19f27 2121
2bd1cbf6
KW
2122/* On EBCDIC platforms, CTRL-@ is 0, CTRL-A is 1, etc, just like on ASCII,
2123 * except that they don't necessarily mean the same characters, e.g. CTRL-D is
2124 * 4 on both systems, but that is EOT on ASCII; ST on EBCDIC.
2125 * '?' is special-cased on EBCDIC to APC, which is the control there that is
2126 * the outlier from the block that contains the other controls, just like
2127 * toCTRL('?') on ASCII yields DEL, the control that is the outlier from the C0
2128 * block. If it weren't special cased, it would yield a non-control.
88794300
KW
2129 * The conversion works both ways, so toCTRL('D') is 4, and toCTRL(4) is D,
2130 * etc. */
2bd1cbf6 2131#ifndef EBCDIC
75763b3a 2132# define toCTRL(c) (__ASSERT_(FITS_IN_8_BITS(c)) toUPPER(((U8)(c))) ^ 64)
2bd1cbf6 2133#else
75763b3a
KW
2134# define toCTRL(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
2135 ((isPRINT_A(c)) \
2136 ? (UNLIKELY((c) == '?') \
2137 ? QUESTION_MARK_CTRL \
2138 : (NATIVE_TO_LATIN1(toUPPER((U8) (c))) ^ 64)) \
2139 : (UNLIKELY((c) == QUESTION_MARK_CTRL) \
2140 ? '?' \
2141 : (LATIN1_TO_NATIVE(((U8) (c)) ^ 64)))))
2bd1cbf6 2142#endif
bbce6d69 2143
dea28490
JJ
2144/* Line numbers are unsigned, 32 bits. */
2145typedef U32 line_t;
e5dcd934 2146#define NOLINE ((line_t) 4294967295UL) /* = FFFFFFFF */
378cc40b 2147
91152fc1
DG
2148/* Helpful alias for version prescan */
2149#define is_LAX_VERSION(a,b) \
2150 (a != Perl_prescan_version(aTHX_ a, FALSE, b, NULL, NULL, NULL, NULL))
2151
2152#define is_STRICT_VERSION(a,b) \
2153 (a != Perl_prescan_version(aTHX_ a, TRUE, b, NULL, NULL, NULL, NULL))
2154
2155#define BADVERSION(a,b,c) \
2156 if (b) { \
2157 *b = c; \
2158 } \
2159 return a;
8c52afec 2160
cb27eebd
KW
2161/* Converts a character known to represent a hexadecimal digit (0-9, A-F, or
2162 * a-f) to its numeric value. READ_XDIGIT's argument is a string pointer,
2163 * which is advanced. The input is validated only by an assert() in DEBUGGING
2164 * builds. In both ASCII and EBCDIC the last 4 bits of the digits are 0-9; and
2165 * the last 4 bits of A-F and a-f are 1-6, so adding 9 yields 10-15 */
2166#define XDIGIT_VALUE(c) (__ASSERT_(isXDIGIT(c)) (0xf & (isDIGIT(c) \
2167 ? (c) \
2168 : ((c) + 9))))
902008b8
KW
2169#define READ_XDIGIT(s) (__ASSERT_(isXDIGIT(*s)) (0xf & (isDIGIT(*(s)) \
2170 ? (*(s)++) \
2171 : (*(s)++ + 9))))
95a59cab 2172
cb27eebd
KW
2173/* Converts a character known to represent an octal digit (0-7) to its numeric
2174 * value. The input is validated only by an assert() in DEBUGGING builds. In
2175 * both ASCII and EBCDIC the last 3 bits of the octal digits range from 0-7. */
2176#define OCTAL_VALUE(c) (__ASSERT_(isOCTAL(c)) (7 & (c)))
2177
305b8651
KW
2178/* Efficiently returns a boolean as to if two native characters are equivalent
2179 * case-insenstively. At least one of the characters must be one of [A-Za-z];
2180 * the ALPHA in the name is to remind you of that. This is asserted() in
2181 * DEBUGGING builds. Because [A-Za-z] are invariant under UTF-8, this macro
2182 * works (on valid input) for both non- and UTF-8-encoded bytes.
2183 *
2184 * When one of the inputs is a compile-time constant and gets folded by the
2185 * compiler, this reduces to an AND and a TEST. On both EBCDIC and ASCII
2186 * machines, 'A' and 'a' differ by a single bit; the same with the upper and
2187 * lower case of all other ASCII-range alphabetics. On ASCII platforms, they
96ca48da
KW
2188 * are 32 apart; on EBCDIC, they are 64. At compile time, this uses an
2189 * exclusive 'or' to find that bit and then inverts it to form a mask, with
2190 * just a single 0, in the bit position where the upper- and lowercase differ.
2191 * */
305b8651
KW
2192#define isALPHA_FOLD_EQ(c1, c2) \
2193 (__ASSERT_(isALPHA_A(c1) || isALPHA_A(c2)) \
2194 ((c1) & ~('A' ^ 'a')) == ((c2) & ~('A' ^ 'a')))
2195#define isALPHA_FOLD_NE(c1, c2) (! isALPHA_FOLD_EQ((c1), (c2)))
2196
8e84507e 2197/*
ccfc67b7
JH
2198=head1 Memory Management
2199
a02a5408 2200=for apidoc Am|void|Newx|void* ptr|int nitems|type
954c1994
GS
2201The XSUB-writer's interface to the C C<malloc> function.
2202
596f7718 2203Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2204
c5008215
JC
2205In 5.9.3, Newx() and friends replace the older New() API, and drops
2206the first parameter, I<x>, a debug aid which allowed callers to identify
37b8b4c9 2207themselves. This aid has been superseded by a new build option,
d10b4965 2208PERL_MEM_LOG (see L<perlhacktips/PERL_MEM_LOG>). The older API is still
c5008215
JC
2209there for use in XS modules supporting older perls.
2210
a02a5408 2211=for apidoc Am|void|Newxc|void* ptr|int nitems|type|cast
954c1994 2212The XSUB-writer's interface to the C C<malloc> function, with
fbe13c60 2213cast. See also C<L</Newx>>.
954c1994 2214
596f7718 2215Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2216
a02a5408 2217=for apidoc Am|void|Newxz|void* ptr|int nitems|type
954c1994 2218The XSUB-writer's interface to the C C<malloc> function. The allocated
fbe13c60 2219memory is zeroed with C<memzero>. See also C<L</Newx>>.
a02a5408 2220
596f7718 2221Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2222
954c1994
GS
2223=for apidoc Am|void|Renew|void* ptr|int nitems|type
2224The XSUB-writer's interface to the C C<realloc> function.
2225
596f7718 2226Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2227
954c1994
GS
2228=for apidoc Am|void|Renewc|void* ptr|int nitems|type|cast
2229The XSUB-writer's interface to the C C<realloc> function, with
2230cast.
2231
596f7718 2232Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2233
49b8b560 2234=for apidoc Am|void|Safefree|void* ptr
954c1994
GS
2235The XSUB-writer's interface to the C C<free> function.
2236
596f7718 2237This should B<ONLY> be used on memory obtained using L</"Newx"> and friends.
0d7b2759 2238
954c1994
GS
2239=for apidoc Am|void|Move|void* src|void* dest|int nitems|type
2240The XSUB-writer's interface to the C C<memmove> function. The C<src> is the
926bb54c 2241source, C<dest> is the destination, C<nitems> is the number of items, and
fbe13c60 2242C<type> is the type. Can do overlapping moves. See also C<L</Copy>>.
954c1994 2243
e90e2364 2244=for apidoc Am|void *|MoveD|void* src|void* dest|int nitems|type
796b6530 2245Like C<Move> but returns C<dest>. Useful
72d33970 2246for encouraging compilers to tail-call
e90e2364
NC
2247optimise.
2248
954c1994
GS
2249=for apidoc Am|void|Copy|void* src|void* dest|int nitems|type
2250The XSUB-writer's interface to the C C<memcpy> function. The C<src> is the
926bb54c 2251source, C<dest> is the destination, C<nitems> is the number of items, and
fbe13c60 2252C<type> is the type. May fail on overlapping copies. See also C<L</Move>>.
954c1994 2253
e90e2364
NC
2254=for apidoc Am|void *|CopyD|void* src|void* dest|int nitems|type
2255
796b6530 2256Like C<Copy> but returns C<dest>. Useful
72d33970 2257for encouraging compilers to tail-call
e90e2364
NC
2258optimise.
2259
954c1994
GS
2260=for apidoc Am|void|Zero|void* dest|int nitems|type
2261
2262The XSUB-writer's interface to the C C<memzero> function. The C<dest> is the
2263destination, C<nitems> is the number of items, and C<type> is the type.
2264
e90e2364
NC
2265=for apidoc Am|void *|ZeroD|void* dest|int nitems|type
2266
72d33970
FC
2267Like C<Zero> but returns dest. Useful
2268for encouraging compilers to tail-call
e90e2364
NC
2269optimise.
2270
da5d8dbb 2271=for apidoc Am|void|StructCopy|type *src|type *dest|type
4375e838 2272This is an architecture-independent macro to copy one structure to another.
954c1994 2273
7e337ee0
JH
2274=for apidoc Am|void|PoisonWith|void* dest|int nitems|type|U8 byte
2275
2276Fill up memory with a byte pattern (a byte repeated over and over
2277again) that hopefully catches attempts to access uninitialized memory.
2278
2279=for apidoc Am|void|PoisonNew|void* dest|int nitems|type
2280
2281PoisonWith(0xAB) for catching access to allocated but uninitialized memory.
2282
1c12ffb4 2283=for apidoc Am|void|PoisonFree|void* dest|int nitems|type
7e337ee0
JH
2284
2285PoisonWith(0xEF) for catching access to freed memory.
2286
9965345d
JH
2287=for apidoc Am|void|Poison|void* dest|int nitems|type
2288
7e337ee0 2289PoisonWith(0xEF) for catching access to freed memory.
9965345d
JH
2290
2291=cut */
954c1994 2292
561b68a9
SH
2293/* Maintained for backwards-compatibility only. Use newSV() instead. */
2294#ifndef PERL_CORE
ff06c60c 2295#define NEWSV(x,len) newSV(len)
561b68a9 2296#endif
ff06c60c 2297
b7112dce 2298#define MEM_SIZE_MAX ((MEM_SIZE)-1)
19a94d75 2299
a500027b 2300#define _PERL_STRLEN_ROUNDUP_UNCHECKED(n) (((n) - 1 + PERL_STRLEN_ROUNDUP_QUANTUM) & ~((MEM_SIZE)PERL_STRLEN_ROUNDUP_QUANTUM - 1))
e6bdf523 2301
27d5b266 2302#ifdef PERL_MALLOC_WRAP
e6bdf523
DM
2303
2304/* This expression will be constant-folded at compile time. It checks
2305 * whether or not the type of the count n is so small (e.g. U8 or U16, or
2306 * U32 on 64-bit systems) that there's no way a wrap-around could occur.
2307 * As well as avoiding the need for a run-time check in some cases, it's
2308 * designed to avoid compiler warnings like:
2309 * comparison is always false due to limited range of data type
73e8ff00
DM
2310 * It's mathematically equivalent to
2311 * max(n) * sizeof(t) > MEM_SIZE_MAX
e6bdf523
DM
2312 */
2313
2314# define _MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) \
445198b9
LM
2315 ( sizeof(MEM_SIZE) < sizeof(n) \
2316 || sizeof(t) > ((MEM_SIZE)1 << 8*(sizeof(MEM_SIZE) - sizeof(n))))
e6bdf523 2317
88f9f128 2318/* This is written in a slightly odd way to avoid various spurious
d98e5cde
DM
2319 * compiler warnings. We *want* to write the expression as
2320 * _MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) && (n > C)
2321 * (for some compile-time constant C), but even when the LHS
2322 * constant-folds to false at compile-time, g++ insists on emitting
2323 * warnings about the RHS (e.g. "comparison is always false"), so instead
2324 * we write it as
e6bdf523 2325 *
d98e5cde 2326 * (cond ? n : X) > C
88f9f128 2327 *
d98e5cde
DM
2328 * where X is a constant with X > C always false. Choosing a value for X
2329 * is tricky. If 0, some compilers will complain about 0 > C always being
2330 * false; if 1, Coverity complains when n happens to be the constant value
2331 * '1', that cond ? 1 : 1 has the same value on both branches; so use C
2332 * for X and hope that nothing else whines.
e6bdf523
DM
2333 */
2334
2335# define _MEM_WRAP_WILL_WRAP(n,t) \
88f9f128
DM
2336 ((_MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) ? (MEM_SIZE)(n) : \
2337 MEM_SIZE_MAX/sizeof(t)) > MEM_SIZE_MAX/sizeof(t))
e6bdf523
DM
2338
2339# define MEM_WRAP_CHECK(n,t) \
2340 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2341 && (croak_memory_wrap(),0))
2342
2343# define MEM_WRAP_CHECK_1(n,t,a) \
2344 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2345 && (Perl_croak_nocontext("%s",(a)),0))
2346
8b44ba4c 2347#define MEM_WRAP_CHECK_(n,t) MEM_WRAP_CHECK(n,t),
27d5b266 2348
a500027b 2349#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
2350#else
2351
410319be
NC
2352#define MEM_WRAP_CHECK(n,t)
2353#define MEM_WRAP_CHECK_1(n,t,a)
2354#define MEM_WRAP_CHECK_2(n,t,a,b)
8b44ba4c
NC
2355#define MEM_WRAP_CHECK_(n,t)
2356
a500027b 2357#define PERL_STRLEN_ROUNDUP(n) _PERL_STRLEN_ROUNDUP_UNCHECKED(n)
27d5b266 2358
1936d2a7 2359#endif
8b44ba4c 2360
fe4f188c 2361#ifdef PERL_MEM_LOG
46c6c7e2 2362/*
9f653bb5 2363 * If PERL_MEM_LOG is defined, all Newx()s, Renew()s, and Safefree()s
46c6c7e2
JH
2364 * go through functions, which are handy for debugging breakpoints, but
2365 * which more importantly get the immediate calling environment (file and
e352bcff
JH
2366 * line number, and C function name if available) passed in. This info can
2367 * then be used for logging the calls, for which one gets a sample
73d1d973 2368 * implementation unless -DPERL_MEM_LOG_NOIMPL is also defined.
3609ea0d 2369 *
46c6c7e2 2370 * Known problems:
94e892a6 2371 * - not all memory allocs get logged, only those
46c6c7e2 2372 * that go through Newx() and derivatives (while all
94e892a6 2373 * Safefrees do get logged)
46c6c7e2
JH
2374 * - __FILE__ and __LINE__ do not work everywhere
2375 * - __func__ or __FUNCTION__ even less so
2376 * - I think more goes on after the perlio frees but
2377 * the thing is that STDERR gets closed (as do all
2378 * the file descriptors)
2379 * - no deeper calling stack than the caller of the Newx()
2380 * or the kind, but do I look like a C reflection/introspection
2381 * utility to you?
2382 * - the function prototypes for the logging functions
2383 * probably should maybe be somewhere else than handy.h
2384 * - one could consider inlining (macrofying) the logging
2385 * for speed, but I am too lazy
2386 * - one could imagine recording the allocations in a hash,
2387 * (keyed by the allocation address?), and maintain that
2388 * through reallocs and frees, but how to do that without
2389 * any News() happening...?
73d1d973 2390 * - lots of -Ddefines to get useful/controllable output
b953482e 2391 * - lots of ENV reads
46c6c7e2
JH
2392 */
2393
0b0ab801 2394# ifdef PERL_CORE
73d1d973 2395# ifndef PERL_MEM_LOG_NOIMPL
0b0ab801
MHM
2396enum mem_log_type {
2397 MLT_ALLOC,
2398 MLT_REALLOC,
d7a2c63c
MHM
2399 MLT_FREE,
2400 MLT_NEW_SV,
2401 MLT_DEL_SV
0b0ab801
MHM
2402};
2403# endif
12754f92 2404# if defined(PERL_IN_SV_C) /* those are only used in sv.c */
d7a2c63c
MHM
2405void Perl_mem_log_new_sv(const SV *sv, const char *filename, const int linenumber, const char *funcname);
2406void Perl_mem_log_del_sv(const SV *sv, const char *filename, const int linenumber, const char *funcname);
12754f92 2407# endif
0b0ab801
MHM
2408# endif
2409
fe4f188c
JH
2410#endif
2411
2412#ifdef PERL_MEM_LOG
d1401ee9
MHM
2413#define MEM_LOG_ALLOC(n,t,a) Perl_mem_log_alloc(n,sizeof(t),STRINGIFY(t),a,__FILE__,__LINE__,FUNCTION__)
2414#define MEM_LOG_REALLOC(n,t,v,a) Perl_mem_log_realloc(n,sizeof(t),STRINGIFY(t),v,a,__FILE__,__LINE__,FUNCTION__)
46c6c7e2 2415#define MEM_LOG_FREE(a) Perl_mem_log_free(a,__FILE__,__LINE__,FUNCTION__)
fe4f188c
JH
2416#endif
2417
2418#ifndef MEM_LOG_ALLOC
2419#define MEM_LOG_ALLOC(n,t,a) (a)
2420#endif
2421#ifndef MEM_LOG_REALLOC
2422#define MEM_LOG_REALLOC(n,t,v,a) (a)
2423#endif
2424#ifndef MEM_LOG_FREE
2425#define MEM_LOG_FREE(a) (a)
2426#endif
2427
d1401ee9
MHM
2428#define Newx(v,n,t) (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_ALLOC(n,t,safemalloc((MEM_SIZE)((n)*sizeof(t))))))
2429#define Newxc(v,n,t,c) (v = (MEM_WRAP_CHECK_(n,t) (c*)MEM_LOG_ALLOC(n,t,safemalloc((MEM_SIZE)((n)*sizeof(t))))))
2430#define Newxz(v,n,t) (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_ALLOC(n,t,safecalloc((n),sizeof(t)))))
a6f6820f
NC
2431
2432#ifndef PERL_CORE
a02a5408
JC
2433/* pre 5.9.x compatibility */
2434#define New(x,v,n,t) Newx(v,n,t)
2435#define Newc(x,v,n,t,c) Newxc(v,n,t,c)
4541904d 2436#define Newz(x,v,n,t) Newxz(v,n,t)
a6f6820f 2437#endif
a02a5408 2438
ff68c719 2439#define Renew(v,n,t) \
d1401ee9 2440 (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_REALLOC(n,t,v,saferealloc((Malloc_t)(v),(MEM_SIZE)((n)*sizeof(t))))))
ff68c719 2441#define Renewc(v,n,t,c) \
d1401ee9 2442 (v = (MEM_WRAP_CHECK_(n,t) (c*)MEM_LOG_REALLOC(n,t,v,saferealloc((Malloc_t)(v),(MEM_SIZE)((n)*sizeof(t))))))
94010e71
NC
2443
2444#ifdef PERL_POISON
2445#define Safefree(d) \
06c0cc96 2446 ((d) ? (void)(safefree(MEM_LOG_FREE((Malloc_t)(d))), Poison(&(d), 1, Malloc_t)) : (void) 0)
94010e71 2447#else
fe4f188c 2448#define Safefree(d) safefree(MEM_LOG_FREE((Malloc_t)(d)))
94010e71 2449#endif
55497cff 2450
45908e4d 2451#define perl_assert_ptr(p) assert( ((void*)(p)) != 0 )
55497cff 2452
45908e4d
YO
2453
2454#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)))
2455#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)))
2456#define Zero(d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), (void)memzero((char*)(d), (n) * sizeof(t)))
2457
2458#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)))
2459#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 2460#define ZeroD(d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), memzero((char*)(d), (n) * sizeof(t)))
e90e2364 2461
7e337ee0
JH
2462#define PoisonWith(d,n,t,b) (MEM_WRAP_CHECK_(n,t) (void)memset((char*)(d), (U8)(b), (n) * sizeof(t)))
2463#define PoisonNew(d,n,t) PoisonWith(d,n,t,0xAB)
2464#define PoisonFree(d,n,t) PoisonWith(d,n,t,0xEF)
2465#define Poison(d,n,t) PoisonFree(d,n,t)
27d5b266 2466
caa674f3
DD
2467#ifdef PERL_POISON
2468# define PERL_POISON_EXPR(x) x
2469#else
2470# define PERL_POISON_EXPR(x)
2471#endif
2472
ff68c719 2473#define StructCopy(s,d,t) (*((t*)(d)) = *((t*)(s)))
2cc61e15 2474
c3caa5c3
JH
2475/* C_ARRAY_LENGTH is the number of elements in the C array (so you
2476 * want your zero-based indices to be less than but not equal to).
2477 *
2478 * C_ARRAY_END is one past the last: half-open/half-closed range,
2479 * not last-inclusive range. */
622913ab 2480#define C_ARRAY_LENGTH(a) (sizeof(a)/sizeof((a)[0]))
c3caa5c3 2481#define C_ARRAY_END(a) ((a) + C_ARRAY_LENGTH(a))
622913ab 2482
2cc61e15
DD
2483#ifdef NEED_VA_COPY
2484# ifdef va_copy
2485# define Perl_va_copy(s, d) va_copy(d, s)
07798b17
AC
2486# elif defined(__va_copy)
2487# define Perl_va_copy(s, d) __va_copy(d, s)
2cc61e15 2488# else
07798b17 2489# define Perl_va_copy(s, d) Copy(s, d, 1, va_list)
2cc61e15
DD
2490# endif
2491#endif
2492
472d47bc
SB
2493/* convenience debug macros */
2494#ifdef USE_ITHREADS
2495#define pTHX_FORMAT "Perl interpreter: 0x%p"
2496#define pTHX__FORMAT ", Perl interpreter: 0x%p"
f54cb97a
AL
2497#define pTHX_VALUE_ (void *)my_perl,
2498#define pTHX_VALUE (void *)my_perl
2499#define pTHX__VALUE_ ,(void *)my_perl,
2500#define pTHX__VALUE ,(void *)my_perl
472d47bc 2501#else
3609ea0d 2502#define pTHX_FORMAT
472d47bc 2503#define pTHX__FORMAT
3609ea0d 2504#define pTHX_VALUE_
472d47bc 2505#define pTHX_VALUE
3609ea0d 2506#define pTHX__VALUE_
472d47bc
SB
2507#define pTHX__VALUE
2508#endif /* USE_ITHREADS */
3609ea0d 2509
2acdbac1
NC
2510/* Perl_deprecate was not part of the public API, and did not have a deprecate()
2511 shortcut macro defined without -DPERL_CORE. Neither codesearch.google.com nor
2512 CPAN::Unpack show any users outside the core. */
2513#ifdef PERL_CORE
dc6e8de0
A
2514# define deprecate(s) Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2515 "Use of " s " is deprecated")
c9680906
A
2516# define deprecate_disappears_in(when,message) \
2517 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2518 message ", and will disappear in Perl " when)
ac641426
A
2519# define deprecate_fatal_in(when,message) \
2520 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2521 message ". Its use will be fatal in Perl " when)
2acdbac1
NC
2522#endif
2523
dfff4baf
BF
2524/* Internal macros to deal with gids and uids */
2525#ifdef PERL_CORE
2526
2527# if Uid_t_size > IVSIZE
2528# define sv_setuid(sv, uid) sv_setnv((sv), (NV)(uid))
2529# define SvUID(sv) SvNV(sv)
07798b17
AC
2530# elif Uid_t_sign <= 0
2531# define sv_setuid(sv, uid) sv_setiv((sv), (IV)(uid))
2532# define SvUID(sv) SvIV(sv)
dfff4baf 2533# else
07798b17
AC
2534# define sv_setuid(sv, uid) sv_setuv((sv), (UV)(uid))
2535# define SvUID(sv) SvUV(sv)
dfff4baf
BF
2536# endif /* Uid_t_size */
2537
2538# if Gid_t_size > IVSIZE
2539# define sv_setgid(sv, gid) sv_setnv((sv), (NV)(gid))
2540# define SvGID(sv) SvNV(sv)
07798b17
AC
2541# elif Gid_t_sign <= 0
2542# define sv_setgid(sv, gid) sv_setiv((sv), (IV)(gid))
2543# define SvGID(sv) SvIV(sv)
dfff4baf 2544# else
07798b17
AC
2545# define sv_setgid(sv, gid) sv_setuv((sv), (UV)(gid))
2546# define SvGID(sv) SvUV(sv)
dfff4baf
BF
2547# endif /* Gid_t_size */
2548
2549#endif
2550
6a5bc5ac 2551#endif /* PERL_HANDY_H_ */
9d745869 2552
e9a8c099 2553/*
14d04a33 2554 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 2555 */