This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
handy.h: Make U64, I64 available outside core
[perl5.git] / handy.h
CommitLineData
a0d0e21e 1/* handy.h
a687059c 2 *
1129b882 3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1999, 2000,
da5d8dbb 4 * 2001, 2002, 2004, 2005, 2006, 2007, 2008, 2012 by Larry Wall and others
a687059c 5 *
6e21c824
LW
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
8d063cd8 8 *
8d063cd8
LW
9 */
10
4650c663
KW
11/* IMPORTANT NOTE: Everything whose name begins with an underscore is for
12 * internal core Perl use only. */
13
6a5bc5ac
KW
14#ifndef PERL_HANDY_H_ /* Guard against nested #inclusion */
15#define PERL_HANDY_H_
9d745869 16
24792b8d
NC
17#ifndef PERL_CORE
18# define Null(type) ((type)NULL)
954c1994
GS
19
20/*
ccfc67b7 21=head1 Handy Values
954c1994 22
3609ea0d 23=for apidoc AmU||Nullch
72d33970
FC
24Null character pointer. (No longer available when C<PERL_CORE> is
25defined.)
2307c6d0 26
954c1994 27=for apidoc AmU||Nullsv
72d33970 28Null SV pointer. (No longer available when C<PERL_CORE> is defined.)
954c1994
GS
29
30=cut
31*/
32
24792b8d
NC
33# define Nullch Null(char*)
34# define Nullfp Null(PerlIO*)
35# define Nullsv Null(SV*)
36#endif
8d063cd8 37
641d3f0b
PP
38#ifdef TRUE
39#undef TRUE
40#endif
41#ifdef FALSE
42#undef FALSE
43#endif
44#define TRUE (1)
45#define FALSE (0)
46
cf3f0ffb
DM
47/* The MUTABLE_*() macros cast pointers to the types shown, in such a way
48 * (compiler permitting) that casting away const-ness will give a warning;
49 * e.g.:
50 *
51 * const SV *sv = ...;
52 * AV *av1 = (AV*)sv; <== BAD: the const has been silently cast away
53 * AV *av2 = MUTABLE_AV(sv); <== GOOD: it may warn
54 */
55
b1bc3f34 56#if defined(__GNUC__) && !defined(PERL_GCC_BRACE_GROUPS_FORBIDDEN)
f065df04 57# define MUTABLE_PTR(p) ({ void *_p = (p); _p; })
b1bc3f34
NC
58#else
59# define MUTABLE_PTR(p) ((void *) (p))
60#endif
61
a062e10d 62#define MUTABLE_AV(p) ((AV *)MUTABLE_PTR(p))
ea726b52 63#define MUTABLE_CV(p) ((CV *)MUTABLE_PTR(p))
159b6efe 64#define MUTABLE_GV(p) ((GV *)MUTABLE_PTR(p))
dbebbdb4 65#define MUTABLE_HV(p) ((HV *)MUTABLE_PTR(p))
a45c7426 66#define MUTABLE_IO(p) ((IO *)MUTABLE_PTR(p))
b1bc3f34 67#define MUTABLE_SV(p) ((SV *)MUTABLE_PTR(p))
27d4fb96 68
f789f6a4 69#if defined(I_STDBOOL) && !defined(PERL_BOOL_AS_CHAR)
bd31be4b
NC
70# include <stdbool.h>
71# ifndef HAS_BOOL
72# define HAS_BOOL 1
73# endif
74#endif
75
8e84507e 76/* bool is built-in for g++-2.6.3 and later, which might be used
c1d22f6b
GS
77 for extensions. <_G_config.h> defines _G_HAVE_BOOL, but we can't
78 be sure _G_config.h will be included before this file. _G_config.h
8e84507e 79 also defines _G_HAVE_BOOL for both gcc and g++, but only g++
c1d22f6b
GS
80 actually has bool. Hence, _G_HAVE_BOOL is pretty useless for us.
81 g++ can be identified by __GNUG__.
82 Andy Dougherty February 2000
5d94fbed 83*/
3609ea0d 84#ifdef __GNUG__ /* GNU g++ has bool built-in */
f789f6a4 85# ifndef PERL_BOOL_AS_CHAR
5d94fbed 86# ifndef HAS_BOOL
c1d22f6b 87# define HAS_BOOL 1
5d94fbed 88# endif
f789f6a4 89# endif
5d94fbed
AD
90#endif
91
92#ifndef HAS_BOOL
f789f6a4
FC
93# ifdef bool
94# undef bool
95# endif
70d5cb32 96# define bool char
c1d22f6b 97# define HAS_BOOL 1
a687059c 98#endif
0d3e774c 99
236415e2
KW
100/* cast-to-bool. A simple (bool) cast may not do the right thing: if bool is
101 * defined as char for example, then the cast from int is
102 * implementation-defined (bool)!!(cbool) in a ternary triggers a bug in xlc on
103 * AIX */
18f5643b 104#define cBOOL(cbool) ((cbool) ? (bool)1 : (bool)0)
f2338a2e 105
46c6c7e2 106/* Try to figure out __func__ or __FUNCTION__ equivalent, if any.
e352bcff
JH
107 * XXX Should really be a Configure probe, with HAS__FUNCTION__
108 * and FUNCTION__ as results.
109 * XXX Similarly, a Configure probe for __FILE__ and __LINE__ is needed. */
46c6c7e2
JH
110#if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L) || (defined(__SUNPRO_C)) /* C99 or close enough. */
111# define FUNCTION__ __func__
07798b17
AC
112#elif (defined(USING_MSVC6)) || /* MSVC6 has neither __func__ nor __FUNCTION and no good workarounds, either. */ \
113 (defined(__DECC_VER)) /* Tru64 or VMS, and strict C89 being used, but not modern enough cc (in Tur64, -c99 not known, only -std1). */
114# define FUNCTION__ ""
46c6c7e2 115#else
07798b17 116# define FUNCTION__ __FUNCTION__ /* Common extension. */
46c6c7e2
JH
117#endif
118
27d4fb96
PP
119/* XXX A note on the perl source internal type system. The
120 original intent was that I32 be *exactly* 32 bits.
121
122 Currently, we only guarantee that I32 is *at least* 32 bits.
123 Specifically, if int is 64 bits, then so is I32. (This is the case
124 for the Cray.) This has the advantage of meshing nicely with
125 standard library calls (where we pass an I32 and the library is
126 expecting an int), but the disadvantage that an I32 is not 32 bits.
127 Andy Dougherty August 1996
24fef2a7 128
dc45a647
MB
129 There is no guarantee that there is *any* integral type with
130 exactly 32 bits. It is perfectly legal for a system to have
131 sizeof(short) == sizeof(int) == sizeof(long) == 8.
693762b4 132
dc45a647
MB
133 Similarly, there is no guarantee that I16 and U16 have exactly 16
134 bits.
693762b4 135
8e84507e
NIS
136 For dealing with issues that may arise from various 32/64-bit
137 systems, we will ask Configure to check out
8175356b 138
3609ea0d
JH
139 SHORTSIZE == sizeof(short)
140 INTSIZE == sizeof(int)
141 LONGSIZE == sizeof(long)
dc45a647 142 LONGLONGSIZE == sizeof(long long) (if HAS_LONG_LONG)
3609ea0d 143 PTRSIZE == sizeof(void *)
dc45a647
MB
144 DOUBLESIZE == sizeof(double)
145 LONG_DOUBLESIZE == sizeof(long double) (if HAS_LONG_DOUBLE).
8175356b 146
27d4fb96
PP
147*/
148
69512466
JH
149#ifdef I_INTTYPES /* e.g. Linux has int64_t without <inttypes.h> */
150# include <inttypes.h>
dd0eed91
JH
151# ifdef INT32_MIN_BROKEN
152# undef INT32_MIN
153# define INT32_MIN (-2147483647-1)
154# endif
155# ifdef INT64_MIN_BROKEN
156# undef INT64_MIN
157# define INT64_MIN (-9223372036854775807LL-1)
158# endif
69512466
JH
159#endif
160
8175356b
JH
161typedef I8TYPE I8;
162typedef U8TYPE U8;
163typedef I16TYPE I16;
164typedef U16TYPE U16;
165typedef I32TYPE I32;
166typedef U32TYPE U32;
16d89be8 167
74b807c7 168#ifdef QUADKIND
8175356b
JH
169typedef I64TYPE I64;
170typedef U64TYPE U64;
16d89be8 171#endif
8175356b 172
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. */
3edba683
CB
176
177/* N.B. We would like to say HAS_QUAD here, but that doesn't actually mean what
178 * it has always been documented to mean (see RT #119753) and is explicitly turned
179 * off in perl.h outside of core with dire warnings about removing the undef.
180 */
181
182#if defined(QUADKIND)
9c39b21c
JH
183# undef PeRl_INT64_C
184# undef PeRl_UINT64_C
185/* Prefer the native integer types (int and long) over long long
186 * (which is not C89) and Win32-specific __int64. */
187# if QUADKIND == QUAD_IS_INT && INTSIZE == 8
188# define PeRl_INT64_C(c) (c)
189# define PeRl_UINT64_C(c) CAT2(c,U)
190# endif
191# if QUADKIND == QUAD_IS_LONG && LONGSIZE == 8
192# define PeRl_INT64_C(c) CAT2(c,L)
193# define PeRl_UINT64_C(c) CAT2(c,UL)
194# endif
195# if QUADKIND == QUAD_IS_LONG_LONG && defined(HAS_LONG_LONG)
196# define PeRl_INT64_C(c) CAT2(c,LL)
197# define PeRl_UINT64_C(c) CAT2(c,ULL)
198# endif
199# if QUADKIND == QUAD_IS___INT64
200# define PeRl_INT64_C(c) CAT2(c,I64)
201# define PeRl_UINT64_C(c) CAT2(c,UI64)
202# endif
203# ifndef PeRl_INT64_C
702c92eb
JH
204# define PeRl_INT64_C(c) ((I64)(c)) /* last resort */
205# define PeRl_UINT64_C(c) ((U64)(c))
9c39b21c
JH
206# endif
207/* In OS X the INT64_C/UINT64_C are defined with LL/ULL, which will
208 * not fly with C89-pedantic gcc, so let's undefine them first so that
209 * we can redefine them with our native integer preferring versions. */
8eafc6eb 210# if defined(PERL_DARWIN) && defined(PERL_GCC_PEDANTIC)
9c39b21c
JH
211# undef INT64_C
212# undef UINT64_C
213# endif
214# ifndef INT64_C
215# define INT64_C(c) PeRl_INT64_C(c)
216# endif
217# ifndef UINT64_C
218# define UINT64_C(c) PeRl_UINT64_C(c)
219# endif
e8c95190 220#endif
dc750f2e 221
d8668976 222#if defined(UINT8_MAX) && defined(INT16_MAX) && defined(INT32_MAX)
5ff3f7a4 223
5ff3f7a4
GS
224/* I8_MAX and I8_MIN constants are not defined, as I8 is an ambiguous type.
225 Please search CHAR_MAX in perl.h for further details. */
226#define U8_MAX UINT8_MAX
227#define U8_MIN UINT8_MIN
228
5ff3f7a4
GS
229#define I16_MAX INT16_MAX
230#define I16_MIN INT16_MIN
231#define U16_MAX UINT16_MAX
232#define U16_MIN UINT16_MIN
233
5ff3f7a4
GS
234#define I32_MAX INT32_MAX
235#define I32_MIN INT32_MIN
0e983133
GS
236#ifndef UINT32_MAX_BROKEN /* e.g. HP-UX with gcc messes this up */
237# define U32_MAX UINT32_MAX
238#else
239# define U32_MAX 4294967295U
240#endif
5ff3f7a4
GS
241#define U32_MIN UINT32_MIN
242
243#else
244
5c9fa16e
KA
245/* I8_MAX and I8_MIN constants are not defined, as I8 is an ambiguous type.
246 Please search CHAR_MAX in perl.h for further details. */
27d4fb96
PP
247#define U8_MAX PERL_UCHAR_MAX
248#define U8_MIN PERL_UCHAR_MIN
79072805 249
27d4fb96
PP
250#define I16_MAX PERL_SHORT_MAX
251#define I16_MIN PERL_SHORT_MIN
252#define U16_MAX PERL_USHORT_MAX
253#define U16_MIN PERL_USHORT_MIN
79072805 254
c4f23d77 255#if LONGSIZE > 4
27d4fb96
PP
256# define I32_MAX PERL_INT_MAX
257# define I32_MIN PERL_INT_MIN
258# define U32_MAX PERL_UINT_MAX
259# define U32_MIN PERL_UINT_MIN
79072805 260#else
27d4fb96
PP
261# define I32_MAX PERL_LONG_MAX
262# define I32_MIN PERL_LONG_MIN
263# define U32_MAX PERL_ULONG_MAX
264# define U32_MIN PERL_ULONG_MIN
79072805
LW
265#endif
266
5ff3f7a4
GS
267#endif
268
464decb6
KW
269/* log(2) (i.e., log base 10 of 2) is pretty close to 0.30103, just in case
270 * anyone is grepping for it */
271#define BIT_DIGITS(N) (((N)*146)/485 + 1) /* log10(2) =~ 146/485 */
fc36a67e
PP
272#define TYPE_DIGITS(T) BIT_DIGITS(sizeof(T) * 8)
273#define TYPE_CHARS(T) (TYPE_DIGITS(T) + 2) /* sign, NUL */
274
88794300 275/* Unused by core; should be deprecated */
ff68c719 276#define Ctl(ch) ((ch) & 037)
8d063cd8 277
98fce2a4
KW
278#if defined(PERL_CORE) || defined(PERL_EXT)
279# ifndef MIN
280# define MIN(a,b) ((a) < (b) ? (a) : (b))
281# endif
282# ifndef MAX
283# define MAX(a,b) ((a) > (b) ? (a) : (b))
284# endif
285#endif
286
4650c663
KW
287/* This is a helper macro to avoid preprocessor issues, replaced by nothing
288 * unless under DEBUGGING, where it expands to an assert of its argument,
289 * followed by a comma (hence the comma operator). If we just used a straight
3e94db23
JH
290 * assert(), we would get a comma with nothing before it when not DEBUGGING.
291 *
292 * We also use empty definition under Coverity since the __ASSERT__
293 * checks often check for things that Really Cannot Happen, and Coverity
294 * detects that and gets all excited. */
295
296#if defined(DEBUGGING) && !defined(__COVERITY__)
0f092d08
KW
297# define __ASSERT_(statement) assert(statement),
298#else
299# define __ASSERT_(statement)
300#endif
301
3fe05580
MHM
302/*
303=head1 SV-Body Allocation
304
1568d13a
Z
305=for apidoc Ama|SV*|newSVpvs|"literal string" s
306Like C<newSVpvn>, but takes a literal string instead of a
30a15352 307string/length pair.
3fe05580 308
1568d13a
Z
309=for apidoc Ama|SV*|newSVpvs_flags|"literal string" s|U32 flags
310Like C<newSVpvn_flags>, but takes a literal string instead of
30a15352 311a string/length pair.
84bafc02 312
1568d13a
Z
313=for apidoc Ama|SV*|newSVpvs_share|"literal string" s
314Like C<newSVpvn_share>, but takes a literal string instead of
30a15352 315a string/length pair and omits the hash parameter.
3fe05580 316
1568d13a
Z
317=for apidoc Am|void|sv_catpvs_flags|SV* sv|"literal string" s|I32 flags
318Like C<sv_catpvn_flags>, but takes a literal string instead
30a15352 319of a string/length pair.
9dcc53ea 320
1568d13a
Z
321=for apidoc Am|void|sv_catpvs_nomg|SV* sv|"literal string" s
322Like C<sv_catpvn_nomg>, but takes a literal string instead of
0c395ea5 323a string/length pair.
9dcc53ea 324
1568d13a
Z
325=for apidoc Am|void|sv_catpvs|SV* sv|"literal string" s
326Like C<sv_catpvn>, but takes a literal string instead of a
0c395ea5 327string/length pair.
3fe05580 328
1568d13a
Z
329=for apidoc Am|void|sv_catpvs_mg|SV* sv|"literal string" s
330Like C<sv_catpvn_mg>, but takes a literal string instead of a
9dcc53ea
Z
331string/length pair.
332
1568d13a
Z
333=for apidoc Am|void|sv_setpvs|SV* sv|"literal string" s
334Like C<sv_setpvn>, but takes a literal string instead of a
0c395ea5 335string/length pair.
3fe05580 336
1568d13a
Z
337=for apidoc Am|void|sv_setpvs_mg|SV* sv|"literal string" s
338Like C<sv_setpvn_mg>, but takes a literal string instead of a
9dcc53ea
Z
339string/length pair.
340
1568d13a
Z
341=for apidoc Am|SV *|sv_setref_pvs|"literal string" s
342Like C<sv_setref_pvn>, but takes a literal string instead of
0c395ea5 343a string/length pair.
9dcc53ea 344
3fe05580
MHM
345=head1 Memory Management
346
1568d13a
Z
347=for apidoc Ama|char*|savepvs|"literal string" s
348Like C<savepvn>, but takes a literal string instead of a
30a15352 349string/length pair.
3fe05580 350
1568d13a 351=for apidoc Ama|char*|savesharedpvs|"literal string" s
9dcc53ea
Z
352A version of C<savepvs()> which allocates the duplicate string in memory
353which is shared between threads.
354
3fe05580
MHM
355=head1 GV Functions
356
1568d13a
Z
357=for apidoc Am|HV*|gv_stashpvs|"literal string" name|I32 create
358Like C<gv_stashpvn>, but takes a literal string instead of a
0c395ea5 359string/length pair.
3fe05580
MHM
360
361=head1 Hash Manipulation Functions
362
1568d13a
Z
363=for apidoc Am|SV**|hv_fetchs|HV* tb|"literal string" key|I32 lval
364Like C<hv_fetch>, but takes a literal string instead of a
0c395ea5 365string/length pair.
3fe05580 366
d106cb46 367=for apidoc Am|SV**|hv_stores|HV* tb|"literal string" key|SV* val
1568d13a 368Like C<hv_store>, but takes a literal string instead of a
0c395ea5 369string/length pair
3fe05580
MHM
370and omits the hash parameter.
371
510966aa
Z
372=head1 Lexer interface
373
1568d13a 374=for apidoc Amx|void|lex_stuff_pvs|"literal string" pv|U32 flags
510966aa 375
1568d13a 376Like L</lex_stuff_pvn>, but takes a literal string instead of
0c395ea5 377a string/length pair.
510966aa 378
3fe05580
MHM
379=cut
380*/
381
926bb54c
KW
382/* concatenating with "" ensures that only literal strings are accepted as
383 * argument */
26bfd924 384#define STR_WITH_LEN(s) ("" s ""), (sizeof(s)-1)
2efa8cc7 385
926bb54c
KW
386/* note that STR_WITH_LEN() can't be used as argument to macros or functions
387 * that under some configurations might be macros, which means that it requires
388 * the full Perl_xxx(aTHX_ ...) form for any API calls where it's used.
ba3a79e7
GA
389 */
390
391/* STR_WITH_LEN() shortcuts */
392#define newSVpvs(str) Perl_newSVpvn(aTHX_ STR_WITH_LEN(str))
84bafc02
NC
393#define newSVpvs_flags(str,flags) \
394 Perl_newSVpvn_flags(aTHX_ STR_WITH_LEN(str), flags)
ba3a79e7 395#define newSVpvs_share(str) Perl_newSVpvn_share(aTHX_ STR_WITH_LEN(str), 0)
9dcc53ea
Z
396#define sv_catpvs_flags(sv, str, flags) \
397 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), flags)
398#define sv_catpvs_nomg(sv, str) \
399 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), 0)
400#define sv_catpvs(sv, str) \
401 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), SV_GMAGIC)
402#define sv_catpvs_mg(sv, str) \
403 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), SV_GMAGIC|SV_SMAGIC)
3fe05580 404#define sv_setpvs(sv, str) Perl_sv_setpvn(aTHX_ sv, STR_WITH_LEN(str))
9dcc53ea
Z
405#define sv_setpvs_mg(sv, str) Perl_sv_setpvn_mg(aTHX_ sv, STR_WITH_LEN(str))
406#define sv_setref_pvs(rv, classname, str) \
407 Perl_sv_setref_pvn(aTHX_ rv, classname, STR_WITH_LEN(str))
ba3a79e7 408#define savepvs(str) Perl_savepvn(aTHX_ STR_WITH_LEN(str))
9dcc53ea
Z
409#define savesharedpvs(str) Perl_savesharedpvn(aTHX_ STR_WITH_LEN(str))
410#define gv_stashpvs(str, create) \
411 Perl_gv_stashpvn(aTHX_ STR_WITH_LEN(str), create)
412#define gv_fetchpvs(namebeg, add, sv_type) \
413 Perl_gv_fetchpvn_flags(aTHX_ STR_WITH_LEN(namebeg), add, sv_type)
414#define gv_fetchpvn(namebeg, len, add, sv_type) \
415 Perl_gv_fetchpvn_flags(aTHX_ namebeg, len, add, sv_type)
416#define sv_catxmlpvs(dsv, str, utf8) \
417 Perl_sv_catxmlpvn(aTHX_ dsv, STR_WITH_LEN(str), utf8)
4ac46235 418
ba3a79e7 419
510966aa
Z
420#define lex_stuff_pvs(pv,flags) Perl_lex_stuff_pvn(aTHX_ STR_WITH_LEN(pv), flags)
421
b96d8cd9
NC
422#define get_cvs(str, flags) \
423 Perl_get_cvn_flags(aTHX_ STR_WITH_LEN(str), (flags))
5c1737d1 424
954c1994 425/*
ccfc67b7
JH
426=head1 Miscellaneous Functions
427
954c1994 428=for apidoc Am|bool|strNE|char* s1|char* s2
dc6b0978
KW
429Test two C<NUL>-terminated strings to see if they are different. Returns true
430or false.
954c1994
GS
431
432=for apidoc Am|bool|strEQ|char* s1|char* s2
dc6b0978
KW
433Test two C<NUL>-terminated strings to see if they are equal. Returns true or
434false.
954c1994
GS
435
436=for apidoc Am|bool|strLT|char* s1|char* s2
dc6b0978
KW
437Test two C<NUL>-terminated strings to see if the first, C<s1>, is less than the
438second, C<s2>. Returns true or false.
954c1994
GS
439
440=for apidoc Am|bool|strLE|char* s1|char* s2
dc6b0978
KW
441Test two C<NUL>-terminated strings to see if the first, C<s1>, is less than or
442equal to the second, C<s2>. Returns true or false.
954c1994
GS
443
444=for apidoc Am|bool|strGT|char* s1|char* s2
dc6b0978
KW
445Test two C<NUL>-terminated strings to see if the first, C<s1>, is greater than
446the second, C<s2>. Returns true or false.
954c1994
GS
447
448=for apidoc Am|bool|strGE|char* s1|char* s2
dc6b0978
KW
449Test two C<NUL>-terminated strings to see if the first, C<s1>, is greater than
450or equal to the second, C<s2>. Returns true or false.
954c1994
GS
451
452=for apidoc Am|bool|strnNE|char* s1|char* s2|STRLEN len
dc6b0978
KW
453Test two C<NUL>-terminated strings to see if they are different. The C<len>
454parameter indicates the number of bytes to compare. Returns true or false. (A
954c1994
GS
455wrapper for C<strncmp>).
456
457=for apidoc Am|bool|strnEQ|char* s1|char* s2|STRLEN len
dc6b0978
KW
458Test two C<NUL>-terminated strings to see if they are equal. The C<len>
459parameter indicates the number of bytes to compare. Returns true or false. (A
460wrapper for C<strncmp>).
954c1994 461
bd18bd40
KW
462=for apidoc Am|bool|memEQ|char* s1|char* s2|STRLEN len
463Test two buffers (which may contain embedded C<NUL> characters, to see if they
464are equal. The C<len> parameter indicates the number of bytes to compare.
465Returns zero if equal, or non-zero if non-equal.
466
467=for apidoc Am|bool|memNE|char* s1|char* s2|STRLEN len
468Test two buffers (which may contain embedded C<NUL> characters, to see if they
469are not equal. The C<len> parameter indicates the number of bytes to compare.
470Returns zero if non-equal, or non-zero if equal.
471
954c1994 472=cut
fc169e00
KW
473
474New macros should use the following conventions for their names (which are
475based on the underlying C library functions):
476
477 (mem | str n? ) (EQ | NE | LT | GT | GE | (( BEGIN | END ) P? )) l? s?
478
479 Each has two main parameters, string-like operands that are compared
480 against each other, as specified by the macro name. Some macros may
481 additionally have one or potentially even two length parameters. If a length
482 parameter applies to both string parameters, it will be positioned third;
483 otherwise any length parameter immediately follows the string parameter it
484 applies to.
485
486 If the prefix to the name is 'str', the string parameter is a pointer to a C
487 language string. Such a string does not contain embedded NUL bytes; its
488 length may be unknown, but can be calculated by C<strlen()>, since it is
489 terminated by a NUL, which isn't included in its length.
490
491 The optional 'n' following 'str' means that that there is a third parameter,
492 giving the maximum number of bytes to look at in each string. Even if both
493 strings are longer than the length parameter, those extra bytes will be
494 unexamined.
495
496 The 's' suffix means that the 2nd byte string parameter is a literal C
497 double-quoted string. Its length will automatically be calculated by the
498 macro, so no length parameter will ever be needed for it.
499
500 If the prefix is 'mem', the string parameters don't have to be C strings;
501 they may contain embedded NUL bytes, do not necessarily have a terminating
502 NUL, and their lengths can be known only through other means, which in
503 practice are additional parameter(s) passed to the function. All 'mem'
504 functions have at least one length parameter. Barring any 'l' or 's' suffix,
505 there is a single length parameter, in position 3, which applies to both
506 string parameters. The 's' suffix means, as described above, that the 2nd
507 string is a literal double-quoted C string (hence its length is calculated by
508 the macro, and the length parameter to the function applies just to the first
509 string parameter, and hence is positioned just after it). An 'l' suffix
510 means that the 2nd string parameter has its own length parameter, and the
511 signature will look like memFOOl(s1, l1, s2, l2).
512
513 BEGIN (and END) are for testing if the 2nd string is an initial (or final)
514 substring of the 1st string. 'P' if present indicates that the substring
515 must be a "proper" one in tha mathematical sense that the first one must be
516 strictly larger than the 2nd.
517
954c1994
GS
518*/
519
62946e08 520
75400963
KW
521#define strNE(s1,s2) (strcmp(s1,s2) != 0)
522#define strEQ(s1,s2) (strcmp(s1,s2) == 0)
8d063cd8
LW
523#define strLT(s1,s2) (strcmp(s1,s2) < 0)
524#define strLE(s1,s2) (strcmp(s1,s2) <= 0)
525#define strGT(s1,s2) (strcmp(s1,s2) > 0)
526#define strGE(s1,s2) (strcmp(s1,s2) >= 0)
62946e08 527
75400963
KW
528#define strnNE(s1,s2,l) (strncmp(s1,s2,l) != 0)
529#define strnEQ(s1,s2,l) (strncmp(s1,s2,l) == 0)
378cc40b 530
75400963
KW
531#define memNE(s1,s2,l) (memcmp(s1,s2,l) != 0)
532#define memEQ(s1,s2,l) (memcmp(s1,s2,l) == 0)
36477c24 533
085b7534 534/* memEQ and memNE where second comparand is a string constant */
568a785a 535#define memEQs(s1, l, s2) \
777fa2cb 536 (((sizeof(s2)-1) == (l)) && memEQ((s1), ("" s2 ""), (sizeof(s2)-1)))
5f50c6c9 537#define memNEs(s1, l, s2) (! memEQs(s1, l, s2))
568a785a 538
fdbb9a7c
KW
539/* Keep these private until we decide it was a good idea */
540#if defined(PERL_CORE) || defined(PERL_EXT) || defined(PERL_EXT_POSIX)
541
542#define strBEGINs(s1,s2) (strncmp(s1,"" s2 "", sizeof(s2)-1) == 0)
543
bdb7e3f0
KW
544#define memBEGINs(s1, l, s2) \
545 ( (l) >= sizeof(s2) - 1 \
546 && memEQ(s1, "" s2 "", sizeof(s2)-1))
de627158
KW
547#define memBEGINPs(s1, l, s2) \
548 ( (l) > sizeof(s2) - 1 \
549 && memEQ(s1, "" s2 "", sizeof(s2)-1))
bdb7e3f0
KW
550#define memENDs(s1, l, s2) \
551 ( (l) >= sizeof(s2) - 1 \
552 && memEQ(s1 + (l) - (sizeof(s2) - 1), "" s2 "", sizeof(s2)-1))
b80f8424
KW
553#define memENDPs(s1, l, s2) \
554 ( (l) > sizeof(s2) \
555 && memEQ(s1 + (l) - (sizeof(s2) - 1), "" s2 "", sizeof(s2)-1))
fdbb9a7c 556#endif /* End of making macros private */
bdb7e3f0 557
062b6850
KW
558#define memLT(s1,s2,l) (memcmp(s1,s2,l) < 0)
559#define memLE(s1,s2,l) (memcmp(s1,s2,l) <= 0)
560#define memGT(s1,s2,l) (memcmp(s1,s2,l) > 0)
561#define memGE(s1,s2,l) (memcmp(s1,s2,l) >= 0)
562
bbce6d69
PP
563/*
564 * Character classes.
565 *
566 * Unfortunately, the introduction of locales means that we
567 * can't trust isupper(), etc. to tell the truth. And when
568 * it comes to /\w+/ with tainting enabled, we *must* be able
569 * to trust our character classes.
570 *
571 * Therefore, the default tests in the text of Perl will be
572 * independent of locale. Any code that wants to depend on
573 * the current locale will use the tests that begin with "lc".
574 */
575
2304df62
AD
576#ifdef HAS_SETLOCALE /* XXX Is there a better test for this? */
577# ifndef CTYPE256
578# define CTYPE256
579# endif
580#endif
581
954c1994 582/*
ccfc67b7 583
dcccc8ff 584=head1 Character classification
243effed
KW
585This section is about functions (really macros) that classify characters
586into types, such as punctuation versus alphabetic, etc. Most of these are
587analogous to regular expression character classes. (See
588L<perlrecharclass/POSIX Character Classes>.) There are several variants for
589each class. (Not all macros have all variants; each item below lists the
590ones valid for it.) None are affected by C<use bytes>, and only the ones
591with C<LC> in the name are affected by the current locale.
592
593The base function, e.g., C<isALPHA()>, takes an octet (either a C<char> or a
594C<U8>) as input and returns a boolean as to whether or not the character
6aff1f14
KW
595represented by that octet is (or on non-ASCII platforms, corresponds to) an
596ASCII character in the named class based on platform, Unicode, and Perl rules.
597If the input is a number that doesn't fit in an octet, FALSE is returned.
243effed 598
c98722a4 599Variant C<isI<FOO>_A> (e.g., C<isALPHA_A()>) is identical to the base function
550da823
KW
600with no suffix C<"_A">. This variant is used to emphasize by its name that
601only ASCII-range characters can return TRUE.
4b9734bf 602
d60679e1 603Variant C<isI<FOO>_L1> imposes the Latin-1 (or EBCDIC equivalent) character set
4b9734bf
KW
604onto the platform. That is, the code points that are ASCII are unaffected,
605since ASCII is a subset of Latin-1. But the non-ASCII code points are treated
606as if they are Latin-1 characters. For example, C<isWORDCHAR_L1()> will return
607true when called with the code point 0xDF, which is a word character in both
4650c663 608ASCII and EBCDIC (though it represents different characters in each).
243effed 609
c98722a4 610Variant C<isI<FOO>_uvchr> is like the C<isI<FOO>_L1> variant, but accepts any UV code
243effed
KW
611point as input. If the code point is larger than 255, Unicode rules are used
612to determine if it is in the character class. For example,
d0da05db 613C<isWORDCHAR_uvchr(0x100)> returns TRUE, since 0x100 is LATIN CAPITAL LETTER A
6aff1f14 614WITH MACRON in Unicode, and is a word character.
243effed 615
c98722a4 616Variant C<isI<FOO>_utf8_safe> is like C<isI<FOO>_uvchr>, but is used for UTF-8
da8c1a98
KW
617encoded strings. Each call classifies one character, even if the string
618contains many. This variant takes two parameters. The first, C<p>, is a
619pointer to the first byte of the character to be classified. (Recall that it
620may take more than one byte to represent a character in UTF-8 strings.) The
621second parameter, C<e>, points to anywhere in the string beyond the first
622character, up to one byte past the end of the entire string. The suffix
623C<_safe> in the function's name indicates that it will not attempt to read
624beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is true (this
625is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the input
626character is malformed in some way, the program may croak, or the function may
627return FALSE, at the discretion of the implementation, and subject to change in
628future releases.
629
c98722a4 630Variant C<isI<FOO>_utf8> is like C<isI<FOO>_utf8_safe>, but takes just a single
da8c1a98 631parameter, C<p>, which has the same meaning as the corresponding parameter does
c98722a4 632in C<isI<FOO>_utf8_safe>. The function therefore can't check if it is reading
34aeb2e9 633beyond the end of the string. Starting in Perl v5.30, it will take a second
c98722a4 634parameter, becoming a synonym for C<isI<FOO>_utf8_safe>. At that time every
34aeb2e9 635program that uses it will have to be changed to successfully compile. In the
c98722a4 636meantime, the first runtime call to C<isI<FOO>_utf8> from each call point in the
34aeb2e9 637program will raise a deprecation warning, enabled by default. You can convert
c98722a4 638your program now to use C<isI<FOO>_utf8_safe>, and avoid the warnings, and get an
34aeb2e9
KW
639extra measure of protection, or you can wait until v5.30, when you'll be forced
640to add the C<e> parameter.
243effed 641
c98722a4 642Variant C<isI<FOO>_LC> is like the C<isI<FOO>_A> and C<isI<FOO>_L1> variants, but the
31f05a37
KW
643result is based on the current locale, which is what C<LC> in the name stands
644for. If Perl can determine that the current locale is a UTF-8 locale, it uses
645the published Unicode rules; otherwise, it uses the C library function that
646gives the named classification. For example, C<isDIGIT_LC()> when not in a
647UTF-8 locale returns the result of calling C<isdigit()>. FALSE is always
1a83413c
KW
648returned if the input won't fit into an octet. On some platforms where the C
649library function is known to be defective, Perl changes its result to follow
650the POSIX standard's rules.
243effed 651
c98722a4
KW
652Variant C<isI<FOO>_LC_uvchr> is like C<isI<FOO>_LC>, but is defined on any UV. It
653returns the same as C<isI<FOO>_LC> for input code points less than 256, and
243effed
KW
654returns the hard-coded, not-affected-by-locale, Unicode results for larger ones.
655
c98722a4 656Variant C<isI<FOO>_LC_utf8_safe> is like C<isI<FOO>_LC_uvchr>, but is used for UTF-8
da8c1a98
KW
657encoded strings. Each call classifies one character, even if the string
658contains many. This variant takes two parameters. The first, C<p>, is a
659pointer to the first byte of the character to be classified. (Recall that it
660may take more than one byte to represent a character in UTF-8 strings.) The
661second parameter, C<e>, points to anywhere in the string beyond the first
662character, up to one byte past the end of the entire string. The suffix
663C<_safe> in the function's name indicates that it will not attempt to read
664beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is true (this
665is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the input
666character is malformed in some way, the program may croak, or the function may
667return FALSE, at the discretion of the implementation, and subject to change in
668future releases.
669
c98722a4 670Variant C<isI<FOO>_LC_utf8> is like C<isI<FOO>_LC_utf8_safe>, but takes just a single
da8c1a98 671parameter, C<p>, which has the same meaning as the corresponding parameter does
c98722a4 672in C<isI<FOO>_LC_utf8_safe>. The function therefore can't check if it is reading
34aeb2e9 673beyond the end of the string. Starting in Perl v5.30, it will take a second
c98722a4 674parameter, becoming a synonym for C<isI<FOO>_LC_utf8_safe>. At that time every
34aeb2e9 675program that uses it will have to be changed to successfully compile. In the
c98722a4 676meantime, the first runtime call to C<isI<FOO>_LC_utf8> from each call point in
34aeb2e9 677the program will raise a deprecation warning, enabled by default. You can
c98722a4 678convert your program now to use C<isI<FOO>_LC_utf8_safe>, and avoid the warnings,
34aeb2e9
KW
679and get an extra measure of protection, or you can wait until v5.30, when
680you'll be forced to add the C<e> parameter.
ccfc67b7 681
954c1994 682=for apidoc Am|bool|isALPHA|char ch
2787a470 683Returns a boolean indicating whether the specified character is an
6aff1f14 684alphabetic character, analogous to C<m/[[:alpha:]]/>.
dcccc8ff
KW
685See the L<top of this section|/Character classification> for an explanation of
686variants
da8c1a98
KW
687C<isALPHA_A>, C<isALPHA_L1>, C<isALPHA_uvchr>, C<isALPHA_utf8_safe>,
688C<isALPHA_LC>, C<isALPHA_LC_uvchr>, and C<isALPHA_LC_utf8_safe>.
8a58bdcf 689
15861f94
KW
690=for apidoc Am|bool|isALPHANUMERIC|char ch
691Returns a boolean indicating whether the specified character is a either an
6aff1f14 692alphabetic character or decimal digit, analogous to C<m/[[:alnum:]]/>.
dcccc8ff
KW
693See the L<top of this section|/Character classification> for an explanation of
694variants
d0da05db 695C<isALPHANUMERIC_A>, C<isALPHANUMERIC_L1>, C<isALPHANUMERIC_uvchr>,
da8c1a98
KW
696C<isALPHANUMERIC_utf8_safe>, C<isALPHANUMERIC_LC>, C<isALPHANUMERIC_LC_uvchr>,
697and C<isALPHANUMERIC_LC_utf8_safe>.
15861f94 698
8a58bdcf
KW
699=for apidoc Am|bool|isASCII|char ch
700Returns a boolean indicating whether the specified character is one of the 128
243effed 701characters in the ASCII character set, analogous to C<m/[[:ascii:]]/>.
e5ad6aba 702On non-ASCII platforms, it returns TRUE iff this
8a58bdcf
KW
703character corresponds to an ASCII character. Variants C<isASCII_A()> and
704C<isASCII_L1()> are identical to C<isASCII()>.
dcccc8ff
KW
705See the L<top of this section|/Character classification> for an explanation of
706variants
da8c1a98
KW
707C<isASCII_uvchr>, C<isASCII_utf8_safe>, C<isASCII_LC>, C<isASCII_LC_uvchr>, and
708C<isASCII_LC_utf8_safe>. Note, however, that some platforms do not have the C
243effed
KW
709library routine C<isascii()>. In these cases, the variants whose names contain
710C<LC> are the same as the corresponding ones without.
711
d98532ea
KW
712Also note, that because all ASCII characters are UTF-8 invariant (meaning they
713have the exact same representation (always a single byte) whether encoded in
714UTF-8 or not), C<isASCII> will give the correct results when called with any
da8c1a98
KW
715byte in any string encoded or not in UTF-8. And similarly C<isASCII_utf8_safe>
716will work properly on any string encoded or not in UTF-8.
d98532ea 717
243effed
KW
718=for apidoc Am|bool|isBLANK|char ch
719Returns a boolean indicating whether the specified character is a
6aff1f14 720character considered to be a blank, analogous to C<m/[[:blank:]]/>.
dcccc8ff
KW
721See the L<top of this section|/Character classification> for an explanation of
722variants
da8c1a98
KW
723C<isBLANK_A>, C<isBLANK_L1>, C<isBLANK_uvchr>, C<isBLANK_utf8_safe>,
724C<isBLANK_LC>, C<isBLANK_LC_uvchr>, and C<isBLANK_LC_utf8_safe>. Note,
725however, that some platforms do not have the C library routine
726C<isblank()>. In these cases, the variants whose names contain C<LC> are
727the same as the corresponding ones without.
243effed
KW
728
729=for apidoc Am|bool|isCNTRL|char ch
730Returns a boolean indicating whether the specified character is a
6aff1f14 731control character, analogous to C<m/[[:cntrl:]]/>.
dcccc8ff
KW
732See the L<top of this section|/Character classification> for an explanation of
733variants
da8c1a98
KW
734C<isCNTRL_A>, C<isCNTRL_L1>, C<isCNTRL_uvchr>, C<isCNTRL_utf8_safe>,
735C<isCNTRL_LC>, C<isCNTRL_LC_uvchr>, and C<isCNTRL_LC_utf8_safe> On EBCDIC
736platforms, you almost always want to use the C<isCNTRL_L1> variant.
954c1994 737
954c1994 738=for apidoc Am|bool|isDIGIT|char ch
2787a470 739Returns a boolean indicating whether the specified character is a
6aff1f14 740digit, analogous to C<m/[[:digit:]]/>.
8a58bdcf 741Variants C<isDIGIT_A> and C<isDIGIT_L1> are identical to C<isDIGIT>.
dcccc8ff
KW
742See the L<top of this section|/Character classification> for an explanation of
743variants
da8c1a98
KW
744C<isDIGIT_uvchr>, C<isDIGIT_utf8_safe>, C<isDIGIT_LC>, C<isDIGIT_LC_uvchr>, and
745C<isDIGIT_LC_utf8_safe>.
243effed
KW
746
747=for apidoc Am|bool|isGRAPH|char ch
748Returns a boolean indicating whether the specified character is a
6aff1f14 749graphic character, analogous to C<m/[[:graph:]]/>.
dcccc8ff 750See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
751variants C<isGRAPH_A>, C<isGRAPH_L1>, C<isGRAPH_uvchr>, C<isGRAPH_utf8_safe>,
752C<isGRAPH_LC>, C<isGRAPH_LC_uvchr>, and C<isGRAPH_LC_utf8_safe>.
954c1994 753
0c82b6df 754=for apidoc Am|bool|isLOWER|char ch
2787a470 755Returns a boolean indicating whether the specified character is a
6aff1f14 756lowercase character, analogous to C<m/[[:lower:]]/>.
dcccc8ff
KW
757See the L<top of this section|/Character classification> for an explanation of
758variants
da8c1a98
KW
759C<isLOWER_A>, C<isLOWER_L1>, C<isLOWER_uvchr>, C<isLOWER_utf8_safe>,
760C<isLOWER_LC>, C<isLOWER_LC_uvchr>, and C<isLOWER_LC_utf8_safe>.
0c82b6df 761
c99e91e9 762=for apidoc Am|bool|isOCTAL|char ch
2787a470 763Returns a boolean indicating whether the specified character is an
6aff1f14 764octal digit, [0-7].
243effed
KW
765The only two variants are C<isOCTAL_A> and C<isOCTAL_L1>; each is identical to
766C<isOCTAL>.
767
768=for apidoc Am|bool|isPUNCT|char ch
769Returns a boolean indicating whether the specified character is a
6aff1f14
KW
770punctuation character, analogous to C<m/[[:punct:]]/>.
771Note that the definition of what is punctuation isn't as
243effed
KW
772straightforward as one might desire. See L<perlrecharclass/POSIX Character
773Classes> for details.
dcccc8ff 774See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
775variants C<isPUNCT_A>, C<isPUNCT_L1>, C<isPUNCT_uvchr>, C<isPUNCT_utf8_safe>,
776C<isPUNCT_LC>, C<isPUNCT_LC_uvchr>, and C<isPUNCT_LC_utf8_safe>.
c99e91e9 777
0c82b6df 778=for apidoc Am|bool|isSPACE|char ch
2787a470 779Returns a boolean indicating whether the specified character is a
6aff1f14 780whitespace character. This is analogous
398d098a 781to what C<m/\s/> matches in a regular expression. Starting in Perl 5.18
779cf272 782this also matches what C<m/[[:space:]]/> does. Prior to 5.18, only the
398d098a
KW
783locale forms of this macro (the ones with C<LC> in their names) matched
784precisely what C<m/[[:space:]]/> does. In those releases, the only difference,
785in the non-locale variants, was that C<isSPACE()> did not match a vertical tab.
786(See L</isPSXSPC> for a macro that matches a vertical tab in all releases.)
dcccc8ff
KW
787See the L<top of this section|/Character classification> for an explanation of
788variants
da8c1a98
KW
789C<isSPACE_A>, C<isSPACE_L1>, C<isSPACE_uvchr>, C<isSPACE_utf8_safe>,
790C<isSPACE_LC>, C<isSPACE_LC_uvchr>, and C<isSPACE_LC_utf8_safe>.
0c82b6df 791
398d098a
KW
792=for apidoc Am|bool|isPSXSPC|char ch
793(short for Posix Space)
779cf272
KW
794Starting in 5.18, this is identical in all its forms to the
795corresponding C<isSPACE()> macros.
398d098a
KW
796The locale forms of this macro are identical to their corresponding
797C<isSPACE()> forms in all Perl releases. In releases prior to 5.18, the
798non-locale forms differ from their C<isSPACE()> forms only in that the
799C<isSPACE()> forms don't match a Vertical Tab, and the C<isPSXSPC()> forms do.
800Otherwise they are identical. Thus this macro is analogous to what
801C<m/[[:space:]]/> matches in a regular expression.
dcccc8ff 802See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
803variants C<isPSXSPC_A>, C<isPSXSPC_L1>, C<isPSXSPC_uvchr>, C<isPSXSPC_utf8_safe>,
804C<isPSXSPC_LC>, C<isPSXSPC_LC_uvchr>, and C<isPSXSPC_LC_utf8_safe>.
398d098a 805
954c1994 806=for apidoc Am|bool|isUPPER|char ch
2787a470 807Returns a boolean indicating whether the specified character is an
6aff1f14 808uppercase character, analogous to C<m/[[:upper:]]/>.
dcccc8ff 809See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
810variants C<isUPPER_A>, C<isUPPER_L1>, C<isUPPER_uvchr>, C<isUPPER_utf8_safe>,
811C<isUPPER_LC>, C<isUPPER_LC_uvchr>, and C<isUPPER_LC_utf8_safe>.
954c1994 812
243effed 813=for apidoc Am|bool|isPRINT|char ch
8eea39dd 814Returns a boolean indicating whether the specified character is a
6aff1f14 815printable character, analogous to C<m/[[:print:]]/>.
dcccc8ff
KW
816See the L<top of this section|/Character classification> for an explanation of
817variants
da8c1a98
KW
818C<isPRINT_A>, C<isPRINT_L1>, C<isPRINT_uvchr>, C<isPRINT_utf8_safe>,
819C<isPRINT_LC>, C<isPRINT_LC_uvchr>, and C<isPRINT_LC_utf8_safe>.
243effed
KW
820
821=for apidoc Am|bool|isWORDCHAR|char ch
822Returns a boolean indicating whether the specified character is a character
823that is a word character, analogous to what C<m/\w/> and C<m/[[:word:]]/> match
824in a regular expression. A word character is an alphabetic character, a
825decimal digit, a connecting punctuation character (such as an underscore), or
826a "mark" character that attaches to one of those (like some sort of accent).
827C<isALNUM()> is a synonym provided for backward compatibility, even though a
828word character includes more than the standard C language meaning of
829alphanumeric.
dcccc8ff 830See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
831variants C<isWORDCHAR_A>, C<isWORDCHAR_L1>, C<isWORDCHAR_uvchr>, and
832C<isWORDCHAR_utf8_safe>. C<isWORDCHAR_LC>, C<isWORDCHAR_LC_uvchr>, and
833C<isWORDCHAR_LC_utf8_safe> are also as described there, but additionally
834include the platform's native underscore.
8a58bdcf
KW
835
836=for apidoc Am|bool|isXDIGIT|char ch
837Returns a boolean indicating whether the specified character is a hexadecimal
243effed
KW
838digit. In the ASCII range these are C<[0-9A-Fa-f]>. Variants C<isXDIGIT_A()>
839and C<isXDIGIT_L1()> are identical to C<isXDIGIT()>.
dcccc8ff
KW
840See the L<top of this section|/Character classification> for an explanation of
841variants
da8c1a98
KW
842C<isXDIGIT_uvchr>, C<isXDIGIT_utf8_safe>, C<isXDIGIT_LC>, C<isXDIGIT_LC_uvchr>,
843and C<isXDIGIT_LC_utf8_safe>.
243effed 844
3c3ecf18
KW
845=for apidoc Am|bool|isIDFIRST|char ch
846Returns a boolean indicating whether the specified character can be the first
847character of an identifier. This is very close to, but not quite the same as
848the official Unicode property C<XID_Start>. The difference is that this
849returns true only if the input character also matches L</isWORDCHAR>.
dcccc8ff
KW
850See the L<top of this section|/Character classification> for an explanation of
851variants
da8c1a98
KW
852C<isIDFIRST_A>, C<isIDFIRST_L1>, C<isIDFIRST_uvchr>, C<isIDFIRST_utf8_safe>,
853C<isIDFIRST_LC>, C<isIDFIRST_LC_uvchr>, and C<isIDFIRST_LC_utf8_safe>.
3c3ecf18
KW
854
855=for apidoc Am|bool|isIDCONT|char ch
856Returns a boolean indicating whether the specified character can be the
857second or succeeding character of an identifier. This is very close to, but
858not quite the same as the official Unicode property C<XID_Continue>. The
859difference is that this returns true only if the input character also matches
dcccc8ff
KW
860L</isWORDCHAR>. See the L<top of this section|/Character classification> for
861an
d0da05db 862explanation of variants C<isIDCONT_A>, C<isIDCONT_L1>, C<isIDCONT_uvchr>,
da8c1a98
KW
863C<isIDCONT_utf8_safe>, C<isIDCONT_LC>, C<isIDCONT_LC_uvchr>, and
864C<isIDCONT_LC_utf8_safe>.
3c3ecf18 865
243effed 866=head1 Miscellaneous Functions
8eea39dd 867
95a59cab 868=for apidoc Am|U8|READ_XDIGIT|char str*
243effed 869Returns the value of an ASCII-range hex digit and advances the string pointer.
95a59cab
YO
870Behaviour is only well defined when isXDIGIT(*str) is true.
871
e7c1e6c1 872=head1 Character case changing
21da7284
KW
873Perl uses "full" Unicode case mappings. This means that converting a single
874character to another case may result in a sequence of more than one character.
875For example, the uppercase of C<E<223>> (LATIN SMALL LETTER SHARP S) is the two
876character sequence C<SS>. This presents some complications The lowercase of
877all characters in the range 0..255 is a single character, and thus
878C<L</toLOWER_L1>> is furnished. But, C<toUPPER_L1> can't exist, as it couldn't
879return a valid result for all legal inputs. Instead C<L</toUPPER_uvchr>> has
880an API that does allow every possible legal result to be returned.) Likewise
881no other function that is crippled by not being able to give the correct
882results for the full range of possible inputs has been implemented here.
e7c1e6c1 883
1f607577
KW
884=for apidoc Am|U8|toUPPER|U8 ch
885Converts the specified character to uppercase. If the input is anything but an
886ASCII lowercase character, that input character itself is returned. Variant
c753c8d3 887C<toUPPER_A> is equivalent.
954c1994 888
d0da05db
KW
889=for apidoc Am|UV|toUPPER_uvchr|UV cp|U8* s|STRLEN* lenp
890Converts the code point C<cp> to its uppercase version, and
891stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
892point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
893that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
894bytes since the uppercase version may be longer than the original character.
895
896The first code point of the uppercased version is returned
21da7284
KW
897(but note, as explained at L<the top of this section|/Character case
898changing>, that there may be more.)
1f607577 899
a239b1e2
KW
900=for apidoc Am|UV|toUPPER_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
901Converts the first UTF-8 encoded character in the sequence starting at C<p> and
902extending no further than S<C<e - 1>> to its uppercase version, and
1f607577
KW
903stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
904that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
905bytes since the uppercase version may be longer than the original character.
906
907The first code point of the uppercased version is returned
21da7284
KW
908(but note, as explained at L<the top of this section|/Character case
909changing>, that there may be more).
1f607577 910
a239b1e2
KW
911The suffix C<_safe> in the function's name indicates that it will not attempt
912to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
913true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
914input character is malformed in some way, the program may croak, or the
915function may return the REPLACEMENT CHARACTER, at the discretion of the
916implementation, and subject to change in future releases.
917
918=for apidoc Am|UV|toUPPER_utf8|U8* p|U8* s|STRLEN* lenp
919This is like C<L</toUPPER_utf8_safe>>, but doesn't have the C<e>
920parameter The function therefore can't check if it is reading
607313a1
KW
921beyond the end of the string. Starting in Perl v5.30, it will take the C<e>
922parameter, becoming a synonym for C<toUPPER_utf8_safe>. At that time every
923program that uses it will have to be changed to successfully compile. In the
924meantime, the first runtime call to C<toUPPER_utf8> from each call point in the
925program will raise a deprecation warning, enabled by default. You can convert
926your program now to use C<toUPPER_utf8_safe>, and avoid the warnings, and get an
927extra measure of protection, or you can wait until v5.30, when you'll be forced
928to add the C<e> parameter.
1f607577 929
25200305
KW
930=for apidoc Am|U8|toFOLD|U8 ch
931Converts the specified character to foldcase. If the input is anything but an
932ASCII uppercase character, that input character itself is returned. Variant
933C<toFOLD_A> is equivalent. (There is no equivalent C<to_FOLD_L1> for the full
d0da05db 934Latin1 range, as the full generality of L</toFOLD_uvchr> is needed there.)
25200305 935
d0da05db
KW
936=for apidoc Am|UV|toFOLD_uvchr|UV cp|U8* s|STRLEN* lenp
937Converts the code point C<cp> to its foldcase version, and
938stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
939point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
940that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
941bytes since the foldcase version may be longer than the original character.
942
943The first code point of the foldcased version is returned
21da7284
KW
944(but note, as explained at L<the top of this section|/Character case
945changing>, that there may be more).
1f607577 946
a239b1e2
KW
947=for apidoc Am|UV|toFOLD_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
948Converts the first UTF-8 encoded character in the sequence starting at C<p> and
949extending no further than S<C<e - 1>> to its foldcase version, and
1f607577
KW
950stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
951that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
952bytes since the foldcase version may be longer than the original character.
953
954The first code point of the foldcased version is returned
21da7284
KW
955(but note, as explained at L<the top of this section|/Character case
956changing>, that there may be more).
1f607577 957
a239b1e2
KW
958The suffix C<_safe> in the function's name indicates that it will not attempt
959to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
960true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
961input character is malformed in some way, the program may croak, or the
962function may return the REPLACEMENT CHARACTER, at the discretion of the
963implementation, and subject to change in future releases.
964
965=for apidoc Am|UV|toFOLD_utf8|U8* p|U8* s|STRLEN* lenp
966This is like C<L</toFOLD_utf8_safe>>, but doesn't have the C<e>
967parameter The function therefore can't check if it is reading
607313a1
KW
968beyond the end of the string. Starting in Perl v5.30, it will take the C<e>
969parameter, becoming a synonym for C<toFOLD_utf8_safe>. At that time every
970program that uses it will have to be changed to successfully compile. In the
971meantime, the first runtime call to C<toFOLD_utf8> from each call point in the
972program will raise a deprecation warning, enabled by default. You can convert
973your program now to use C<toFOLD_utf8_safe>, and avoid the warnings, and get an
974extra measure of protection, or you can wait until v5.30, when you'll be forced
975to add the C<e> parameter.
1f607577
KW
976
977=for apidoc Am|U8|toLOWER|U8 ch
978Converts the specified character to lowercase. If the input is anything but an
979ASCII uppercase character, that input character itself is returned. Variant
c753c8d3 980C<toLOWER_A> is equivalent.
954c1994 981
1f607577 982=for apidoc Am|U8|toLOWER_L1|U8 ch
b7d90381
KW
983Converts the specified Latin1 character to lowercase. The results are
984undefined if the input doesn't fit in a byte.
1f607577
KW
985
986=for apidoc Am|U8|toLOWER_LC|U8 ch
987Converts the specified character to lowercase using the current locale's rules,
988if possible; otherwise returns the input character itself.
989
d0da05db
KW
990=for apidoc Am|UV|toLOWER_uvchr|UV cp|U8* s|STRLEN* lenp
991Converts the code point C<cp> to its lowercase version, and
992stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
993point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
994that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
995bytes since the lowercase version may be longer than the original character.
996
997The first code point of the lowercased version is returned
21da7284
KW
998(but note, as explained at L<the top of this section|/Character case
999changing>, that there may be more).
1f607577 1000
a239b1e2
KW
1001
1002=for apidoc Am|UV|toLOWER_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
1003Converts the first UTF-8 encoded character in the sequence starting at C<p> and
1004extending no further than S<C<e - 1>> to its lowercase version, and
1f607577
KW
1005stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
1006that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
1007bytes since the lowercase version may be longer than the original character.
1008
1009The first code point of the lowercased version is returned
21da7284
KW
1010(but note, as explained at L<the top of this section|/Character case
1011changing>, that there may be more).
1f607577 1012
a239b1e2
KW
1013The suffix C<_safe> in the function's name indicates that it will not attempt
1014to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
1015true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
1016input character is malformed in some way, the program may croak, or the
1017function may return the REPLACEMENT CHARACTER, at the discretion of the
1018implementation, and subject to change in future releases.
1019
1020=for apidoc Am|UV|toLOWER_utf8|U8* p|U8* s|STRLEN* lenp
1021This is like C<L</toLOWER_utf8_safe>>, but doesn't have the C<e>
1022parameter The function therefore can't check if it is reading
607313a1
KW
1023beyond the end of the string. Starting in Perl v5.30, it will take the C<e>
1024parameter, becoming a synonym for C<toLOWER_utf8_safe>. At that time every
1025program that uses it will have to be changed to successfully compile. In the
1026meantime, the first runtime call to C<toLOWER_utf8> from each call point in the
1027program will raise a deprecation warning, enabled by default. You can convert
1028your program now to use C<toLOWER_utf8_safe>, and avoid the warnings, and get an
1029extra measure of protection, or you can wait until v5.30, when you'll be forced
1030to add the C<e> parameter.
1f607577 1031
25200305
KW
1032=for apidoc Am|U8|toTITLE|U8 ch
1033Converts the specified character to titlecase. If the input is anything but an
1034ASCII lowercase character, that input character itself is returned. Variant
b7d90381 1035C<toTITLE_A> is equivalent. (There is no C<toTITLE_L1> for the full Latin1
d0da05db 1036range, as the full generality of L</toTITLE_uvchr> is needed there. Titlecase is
b7d90381 1037not a concept used in locale handling, so there is no functionality for that.)
25200305 1038
d0da05db
KW
1039=for apidoc Am|UV|toTITLE_uvchr|UV cp|U8* s|STRLEN* lenp
1040Converts the code point C<cp> to its titlecase version, and
1041stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
1042point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
1043that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
1044bytes since the titlecase version may be longer than the original character.
1045
1046The first code point of the titlecased version is returned
21da7284
KW
1047(but note, as explained at L<the top of this section|/Character case
1048changing>, that there may be more).
1f607577 1049
a239b1e2
KW
1050=for apidoc Am|UV|toTITLE_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
1051Converts the first UTF-8 encoded character in the sequence starting at C<p> and
1052extending no further than S<C<e - 1>> to its titlecase version, and
1f607577
KW
1053stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
1054that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
1055bytes since the titlecase version may be longer than the original character.
1056
1057The first code point of the titlecased version is returned
21da7284
KW
1058(but note, as explained at L<the top of this section|/Character case
1059changing>, that there may be more).
1f607577 1060
a239b1e2
KW
1061The suffix C<_safe> in the function's name indicates that it will not attempt
1062to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
1063true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
1064input character is malformed in some way, the program may croak, or the
1065function may return the REPLACEMENT CHARACTER, at the discretion of the
1066implementation, and subject to change in future releases.
1067
1068=for apidoc Am|UV|toTITLE_utf8|U8* p|U8* s|STRLEN* lenp
1069This is like C<L</toLOWER_utf8_safe>>, but doesn't have the C<e>
1070parameter The function therefore can't check if it is reading
607313a1
KW
1071beyond the end of the string. Starting in Perl v5.30, it will take the C<e>
1072parameter, becoming a synonym for C<toTITLE_utf8_safe>. At that time every
1073program that uses it will have to be changed to successfully compile. In the
1074meantime, the first runtime call to C<toTITLE_utf8> from each call point in the
1075program will raise a deprecation warning, enabled by default. You can convert
1076your program now to use C<toTITLE_utf8_safe>, and avoid the warnings, and get an
1077extra measure of protection, or you can wait until v5.30, when you'll be forced
1078to add the C<e> parameter.
1f607577 1079
954c1994 1080=cut
353c9b6f 1081
d0da05db 1082XXX Still undocumented isVERTWS_uvchr and _utf8; it's unclear what their names
1e222e4f
KW
1083really should be. Also toUPPER_LC and toFOLD_LC, which are subject to change,
1084and aren't general purpose as they don't work on U+DF, and assert against that.
243effed 1085
8a58bdcf 1086Note that these macros are repeated in Devel::PPPort, so should also be
62fa66b6
KW
1087patched there. The file as of this writing is cpan/Devel-PPPort/parts/inc/misc
1088
954c1994
GS
1089*/
1090
7c062697
KW
1091/* Specify the widest unsigned type on the platform. Use U64TYPE because U64
1092 * is known only in the perl core, and this macro can be called from outside
1093 * that */
1094#ifdef HAS_QUAD
1095# define WIDEST_UTYPE U64TYPE
1096#else
1097# define WIDEST_UTYPE U32
1098#endif
1099
3912bc88
KW
1100/* FITS_IN_8_BITS(c) returns true if c doesn't have a bit set other than in
1101 * the lower 8. It is designed to be hopefully bomb-proof, making sure that no
1102 * bits of information are lost even on a 64-bit machine, but to get the
1103 * compiler to optimize it out if possible. This is because Configure makes
1104 * sure that the machine has an 8-bit byte, so if c is stored in a byte, the
1105 * sizeof() guarantees that this evaluates to a constant true at compile time.
7e75d1a1
JH
1106 *
1107 * For Coverity, be always true, because otherwise Coverity thinks
1108 * it finds several expressions that are always true, independent
1109 * of operands. Well, they are, but that is kind of the point.
220c71bf 1110 */
7e75d1a1 1111#ifndef __COVERITY__
6c5b02ac
KW
1112 /* The '| 0' part ensures a compiler error if c is not integer (like e.g., a
1113 * pointer) */
1114#define FITS_IN_8_BITS(c) ( (sizeof(c) == 1) \
ace3ad0f 1115 || !(((WIDEST_UTYPE)((c) | 0)) & ~0xFF))
7e75d1a1
JH
1116#else
1117#define FITS_IN_8_BITS(c) (1)
1118#endif
cf301eb7 1119
41f43cc2 1120#ifdef EBCDIC
b6340bd0 1121# ifndef _ALL_SOURCE
0852beac
KW
1122 /* The native libc isascii() et.al. functions return the wrong results
1123 * on at least z/OS unless this is defined. */
b6340bd0
KW
1124# error _ALL_SOURCE should probably be defined
1125# endif
41f43cc2 1126#else
0852beac
KW
1127 /* There is a simple definition of ASCII for ASCII platforms. But the
1128 * EBCDIC one isn't so simple, so is defined using table look-up like the
9c903d59 1129 * other macros below.
3f3c579d
KW
1130 *
1131 * The cast here is used instead of '(c) >= 0', because some compilers emit
1132 * a warning that that test is always true when the parameter is an
1133 * unsigned type. khw supposes that it could be written as
1134 * && ((c) == '\0' || (c) > 0)
1135 * to avoid the message, but the cast will likely avoid extra branches even
1136 * with stupid compilers.
1137 *
1138 * The '| 0' part ensures a compiler error if c is not integer (like e.g.,
1139 * a pointer) */
9c903d59 1140# define isASCII(c) ((WIDEST_UTYPE)((c) | 0) < 128)
41f43cc2
KW
1141#endif
1142
38694112
KW
1143/* Take the eight possible bit patterns of the lower 3 bits and you get the
1144 * lower 3 bits of the 8 octal digits, in both ASCII and EBCDIC, so those bits
1145 * can be ignored. If the rest match '0', we have an octal */
1146#define isOCTAL_A(c) (((WIDEST_UTYPE)((c) | 0) & ~7) == '0')
c2da0b36 1147
9fb1bf9d 1148#ifdef H_PERL /* If have access to perl.h, lookup in its table */
f4cdb42c 1149
a500dc72
KW
1150/* Character class numbers. For internal core Perl use only. The ones less
1151 * than 32 are used in PL_charclass[] and the ones up through the one that
1152 * corresponds to <_HIGHEST_REGCOMP_DOT_H_SYNC> are used by regcomp.h and
1153 * related files. PL_charclass ones use names used in l1_char_class_tab.h but
1154 * their actual definitions are here. If that file has a name not used here,
1155 * it won't compile.
1709d539
KW
1156 *
1157 * The first group of these is ordered in what I (khw) estimate to be the
31c7f561
KW
1158 * frequency of their use. This gives a slight edge to exiting a loop earlier
1159 * (in reginclass() in regexec.c) */
1709d539
KW
1160# define _CC_WORDCHAR 0 /* \w and [:word:] */
1161# define _CC_DIGIT 1 /* \d and [:digit:] */
1162# define _CC_ALPHA 2 /* [:alpha:] */
1163# define _CC_LOWER 3 /* [:lower:] */
1164# define _CC_UPPER 4 /* [:upper:] */
1165# define _CC_PUNCT 5 /* [:punct:] */
1166# define _CC_PRINT 6 /* [:print:] */
15861f94 1167# define _CC_ALPHANUMERIC 7 /* [:alnum:] */
1709d539 1168# define _CC_GRAPH 8 /* [:graph:] */
359b005e 1169# define _CC_CASED 9 /* [:lower:] or [:upper:] under /i */
1709d539 1170
b0d691b2 1171#define _FIRST_NON_SWASH_CC 10
1709d539
KW
1172/* The character classes above are implemented with swashes. The second group
1173 * (just below) contains the ones implemented without. These are also sorted
1174 * in rough order of the frequency of their use, except that \v should be last,
1175 * as it isn't a real Posix character class, and some (small) inefficiencies in
1176 * regular expression handling would be introduced by putting it in the middle
1177 * of those that are. Also, cntrl and ascii come after the others as it may be
1178 * useful to group these which have no members that match above Latin1, (or
1179 * above ASCII in the latter case) */
1180
779cf272 1181# define _CC_SPACE 10 /* \s, [:space:] */
34aeb2e9
KW
1182# define _CC_PSXSPC _CC_SPACE /* XXX Temporary, can be removed
1183 when the deprecated isFOO_utf8()
1184 functions are removed */
b0d691b2
KW
1185# define _CC_BLANK 11 /* [:blank:] */
1186# define _CC_XDIGIT 12 /* [:xdigit:] */
779cf272
KW
1187# define _CC_CNTRL 13 /* [:cntrl:] */
1188# define _CC_ASCII 14 /* [:ascii:] */
1189# define _CC_VERTSPACE 15 /* \v */
1709d539 1190
a0947d7b
KW
1191# define _HIGHEST_REGCOMP_DOT_H_SYNC _CC_VERTSPACE
1192
1709d539 1193/* The members of the third group below do not need to be coordinated with data
3ffc8c70 1194 * structures in regcomp.[ch] and regexec.c. */
779cf272
KW
1195# define _CC_IDFIRST 16
1196# define _CC_CHARNAME_CONT 17
1197# define _CC_NONLATIN1_FOLD 18
1198# define _CC_NONLATIN1_SIMPLE_FOLD 19
1199# define _CC_QUOTEMETA 20
1200# define _CC_NON_FINAL_FOLD 21
1201# define _CC_IS_IN_SOME_FOLD 22
1202# define _CC_MNEMONIC_CNTRL 23
073c22b3 1203
34aeb2e9
KW
1204# define _CC_IDCONT 24 /* XXX Temporary, can be removed when the deprecated
1205 isFOO_utf8() functions are removed */
1206
073c22b3
KW
1207/* This next group is only used on EBCDIC platforms, so theoretically could be
1208 * shared with something entirely different that's only on ASCII platforms */
5d5376e2
KW
1209# define _CC_UTF8_START_BYTE_IS_FOR_AT_LEAST_SURROGATE 28
1210# define _CC_UTF8_IS_START 29
1211# define _CC_UTF8_IS_DOWNGRADEABLE_START 30
1212# define _CC_UTF8_IS_CONTINUATION 31
1213/* Unused: 24-27
f4cdb42c
KW
1214 * If more bits are needed, one could add a second word for non-64bit
1215 * QUAD_IS_INT systems, using some #ifdefs to distinguish between having a 2nd
37ede926
KW
1216 * word or not. The IS_IN_SOME_FOLD bit is the most easily expendable, as it
1217 * is used only for optimization (as of this writing), and differs in the
1218 * Latin1 range from the ALPHA bit only in two relatively unimportant
a500dc72 1219 * characters: the masculine and feminine ordinal indicators, so removing it
073c22b3
KW
1220 * would just cause /i regexes which match them to run less efficiently.
1221 * Similarly the EBCDIC-only bits are used just for speed, and could be
1222 * replaced by other means */
96ac0975 1223
3a371f2f
KW
1224#if defined(PERL_CORE) || defined(PERL_EXT)
1225/* An enum version of the character class numbers, to help compilers
1226 * optimize */
1227typedef enum {
3a371f2f 1228 _CC_ENUM_ALPHA = _CC_ALPHA,
e8d596e0
KW
1229 _CC_ENUM_ALPHANUMERIC = _CC_ALPHANUMERIC,
1230 _CC_ENUM_ASCII = _CC_ASCII,
1231 _CC_ENUM_BLANK = _CC_BLANK,
b0d691b2 1232 _CC_ENUM_CASED = _CC_CASED,
e8d596e0 1233 _CC_ENUM_CNTRL = _CC_CNTRL,
3a371f2f
KW
1234 _CC_ENUM_DIGIT = _CC_DIGIT,
1235 _CC_ENUM_GRAPH = _CC_GRAPH,
1236 _CC_ENUM_LOWER = _CC_LOWER,
1237 _CC_ENUM_PRINT = _CC_PRINT,
1238 _CC_ENUM_PUNCT = _CC_PUNCT,
e8d596e0 1239 _CC_ENUM_SPACE = _CC_SPACE,
3a371f2f 1240 _CC_ENUM_UPPER = _CC_UPPER,
e8d596e0 1241 _CC_ENUM_VERTSPACE = _CC_VERTSPACE,
3a371f2f 1242 _CC_ENUM_WORDCHAR = _CC_WORDCHAR,
e8d596e0 1243 _CC_ENUM_XDIGIT = _CC_XDIGIT
3a371f2f
KW
1244} _char_class_number;
1245#endif
1246
63c61c3f 1247#define POSIX_SWASH_COUNT _FIRST_NON_SWASH_CC
86f72d56 1248#define POSIX_CC_COUNT (_HIGHEST_REGCOMP_DOT_H_SYNC + 1)
63c61c3f 1249
b7d90381
KW
1250#if defined(PERL_IN_UTF8_C) \
1251 || defined(PERL_IN_REGCOMP_C) \
1252 || defined(PERL_IN_REGEXEC_C)
922e8cb4 1253# if _CC_WORDCHAR != 0 || _CC_DIGIT != 1 || _CC_ALPHA != 2 || _CC_LOWER != 3 \
b0d691b2
KW
1254 || _CC_UPPER != 4 || _CC_PUNCT != 5 || _CC_PRINT != 6 \
1255 || _CC_ALPHANUMERIC != 7 || _CC_GRAPH != 8 || _CC_CASED != 9
922e8cb4
KW
1256 #error Need to adjust order of swash_property_names[]
1257# endif
1258
b44daeeb
KW
1259/* This is declared static in each of the few files that this is #defined for
1260 * to keep them from being publicly accessible. Hence there is a small amount
1261 * of wasted space */
1262
922e8cb4
KW
1263static const char* const swash_property_names[] = {
1264 "XPosixWord",
1265 "XPosixDigit",
1266 "XPosixAlpha",
1267 "XPosixLower",
1268 "XPosixUpper",
1269 "XPosixPunct",
1270 "XPosixPrint",
1271 "XPosixAlnum",
b0d691b2
KW
1272 "XPosixGraph",
1273 "Cased"
922e8cb4
KW
1274};
1275#endif
1276
6635f04f 1277START_EXTERN_C
96ac0975
NC
1278# ifdef DOINIT
1279EXTCONST U32 PL_charclass[] = {
1280# include "l1_char_class_tab.h"
1281};
1282
1283# else /* ! DOINIT */
1284EXTCONST U32 PL_charclass[];
1285# endif
6635f04f 1286END_EXTERN_C
96ac0975 1287
265c1f46 1288 /* The 1U keeps Solaris from griping when shifting sets the uppermost bit */
430b7c70 1289# define _CC_mask(classnum) (1U << (classnum))
4650c663
KW
1290
1291 /* For internal core Perl use only: the base macro for defining macros like
1292 * isALPHA */
ff7ecfc3 1293# define _generic_isCC(c, classnum) cBOOL(FITS_IN_8_BITS(c) \
f4cd282c 1294 && (PL_charclass[(U8) (c)] & _CC_mask(classnum)))
4eeeb416 1295
f4cdb42c
KW
1296 /* The mask for the _A versions of the macros; it just adds in the bit for
1297 * ASCII. */
1298# define _CC_mask_A(classnum) (_CC_mask(classnum) | _CC_mask(_CC_ASCII))
1299
4650c663
KW
1300 /* For internal core Perl use only: the base macro for defining macros like
1301 * isALPHA_A. The foo_A version makes sure that both the desired bit and
1302 * the ASCII bit are present */
b7d90381
KW
1303# define _generic_isCC_A(c, classnum) (FITS_IN_8_BITS(c) \
1304 && ((PL_charclass[(U8) (c)] & _CC_mask_A(classnum)) \
1305 == _CC_mask_A(classnum)))
f4cdb42c 1306
f4cdb42c 1307# define isALPHA_A(c) _generic_isCC_A(c, _CC_ALPHA)
15861f94 1308# define isALPHANUMERIC_A(c) _generic_isCC_A(c, _CC_ALPHANUMERIC)
f4cdb42c
KW
1309# define isBLANK_A(c) _generic_isCC_A(c, _CC_BLANK)
1310# define isCNTRL_A(c) _generic_isCC_A(c, _CC_CNTRL)
96ca48da 1311# define isDIGIT_A(c) _generic_isCC(c, _CC_DIGIT) /* No non-ASCII digits */
f4cdb42c
KW
1312# define isGRAPH_A(c) _generic_isCC_A(c, _CC_GRAPH)
1313# define isLOWER_A(c) _generic_isCC_A(c, _CC_LOWER)
1314# define isPRINT_A(c) _generic_isCC_A(c, _CC_PRINT)
f4cdb42c
KW
1315# define isPUNCT_A(c) _generic_isCC_A(c, _CC_PUNCT)
1316# define isSPACE_A(c) _generic_isCC_A(c, _CC_SPACE)
1317# define isUPPER_A(c) _generic_isCC_A(c, _CC_UPPER)
1318# define isWORDCHAR_A(c) _generic_isCC_A(c, _CC_WORDCHAR)
b7d90381
KW
1319# define isXDIGIT_A(c) _generic_isCC(c, _CC_XDIGIT) /* No non-ASCII xdigits
1320 */
d95f8b6a 1321# define isIDFIRST_A(c) _generic_isCC_A(c, _CC_IDFIRST)
3ded5eb0
KW
1322# define isALPHA_L1(c) _generic_isCC(c, _CC_ALPHA)
1323# define isALPHANUMERIC_L1(c) _generic_isCC(c, _CC_ALPHANUMERIC)
1324# define isBLANK_L1(c) _generic_isCC(c, _CC_BLANK)
1325
1326 /* continuation character for legal NAME in \N{NAME} */
1327# define isCHARNAME_CONT(c) _generic_isCC(c, _CC_CHARNAME_CONT)
1328
1329# define isCNTRL_L1(c) _generic_isCC(c, _CC_CNTRL)
1330# define isGRAPH_L1(c) _generic_isCC(c, _CC_GRAPH)
1331# define isLOWER_L1(c) _generic_isCC(c, _CC_LOWER)
1332# define isPRINT_L1(c) _generic_isCC(c, _CC_PRINT)
b7d90381 1333# define isPSXSPC_L1(c) isSPACE_L1(c)
3ded5eb0
KW
1334# define isPUNCT_L1(c) _generic_isCC(c, _CC_PUNCT)
1335# define isSPACE_L1(c) _generic_isCC(c, _CC_SPACE)
1336# define isUPPER_L1(c) _generic_isCC(c, _CC_UPPER)
1337# define isWORDCHAR_L1(c) _generic_isCC(c, _CC_WORDCHAR)
1338# define isIDFIRST_L1(c) _generic_isCC(c, _CC_IDFIRST)
f4cdb42c 1339
0852beac
KW
1340# ifdef EBCDIC
1341# define isASCII(c) _generic_isCC(c, _CC_ASCII)
1342# endif
1343
f12c0118
KW
1344 /* Participates in a single-character fold with a character above 255 */
1345# 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)))
1346
1347 /* Like the above, but also can be part of a multi-char fold */
f4cd282c 1348# 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 1349
4eeeb416 1350# define _isQUOTEMETA(c) _generic_isCC(c, _CC_QUOTEMETA)
26faadbd 1351# define _IS_NON_FINAL_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
b7d90381 1352 _generic_isCC(c, _CC_NON_FINAL_FOLD)
37ede926 1353# define _IS_IN_SOME_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
b7d90381 1354 _generic_isCC(c, _CC_IS_IN_SOME_FOLD)
8e35b056
KW
1355# define _IS_MNEMONIC_CNTRL_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
1356 _generic_isCC(c, _CC_MNEMONIC_CNTRL)
687c8d01 1357#else /* else we don't have perl.h H_PERL */
3ded5eb0
KW
1358
1359 /* If we don't have perl.h, we are compiling a utility program. Below we
1360 * hard-code various macro definitions that wouldn't otherwise be available
fc273927 1361 * to it. Most are coded based on first principles. These are written to
74665a89 1362 * avoid EBCDIC vs. ASCII #ifdef's as much as possible. */
0852beac
KW
1363# define isDIGIT_A(c) ((c) <= '9' && (c) >= '0')
1364# define isBLANK_A(c) ((c) == ' ' || (c) == '\t')
74665a89
KW
1365# define isSPACE_A(c) (isBLANK_A(c) \
1366 || (c) == '\n' \
1367 || (c) == '\r' \
1368 || (c) == '\v' \
0852beac 1369 || (c) == '\f')
74665a89
KW
1370 /* On EBCDIC, there are gaps between 'i' and 'j'; 'r' and 's'. Same for
1371 * uppercase. The tests for those aren't necessary on ASCII, but hurt only
1372 * performance (if optimization isn't on), and allow the same code to be
1373 * used for both platform types */
1374# define isLOWER_A(c) ((c) >= 'a' && (c) <= 'z' \
1375 && ( (c) <= 'i' \
1376 || ((c) >= 'j' && (c) <= 'r') \
1377 || (c) >= 's'))
1378# define isUPPER_A(c) ((c) >= 'A' && (c) <= 'Z' \
1379 && ( (c) <= 'I' \
1380 || ((c) >= 'J' && (c) <= 'R') \
1381 || (c) >= 'S'))
a4d7a999
KW
1382# define isALPHA_A(c) (isUPPER_A(c) || isLOWER_A(c))
1383# define isALPHANUMERIC_A(c) (isALPHA_A(c) || isDIGIT_A(c))
3ded5eb0 1384# define isWORDCHAR_A(c) (isALPHANUMERIC_A(c) || (c) == '_')
0852beac 1385# define isIDFIRST_A(c) (isALPHA_A(c) || (c) == '_')
74665a89
KW
1386# define isXDIGIT_A(c) (isDIGIT_A(c) \
1387 || ((c) >= 'a' && (c) <= 'f') \
a4d7a999 1388 || ((c) <= 'F' && (c) >= 'A'))
74665a89
KW
1389# define isPUNCT_A(c) ((c) == '-' || (c) == '!' || (c) == '"' \
1390 || (c) == '#' || (c) == '$' || (c) == '%' \
1391 || (c) == '&' || (c) == '\'' || (c) == '(' \
1392 || (c) == ')' || (c) == '*' || (c) == '+' \
1393 || (c) == ',' || (c) == '.' || (c) == '/' \
1394 || (c) == ':' || (c) == ';' || (c) == '<' \
1395 || (c) == '=' || (c) == '>' || (c) == '?' \
1396 || (c) == '@' || (c) == '[' || (c) == '\\' \
1397 || (c) == ']' || (c) == '^' || (c) == '_' \
1398 || (c) == '`' || (c) == '{' || (c) == '|' \
1399 || (c) == '}' || (c) == '~')
1400# define isGRAPH_A(c) (isALPHANUMERIC_A(c) || isPUNCT_A(c))
1401# define isPRINT_A(c) (isGRAPH_A(c) || (c) == ' ')
3ded5eb0 1402
0852beac 1403# ifdef EBCDIC
74665a89
KW
1404 /* The below is accurate for the 3 EBCDIC code pages traditionally
1405 * supported by perl. The only difference between them in the controls
1406 * is the position of \n, and that is represented symbolically below */
1407# define isCNTRL_A(c) ((c) == '\0' || (c) == '\a' || (c) == '\b' \
1408 || (c) == '\f' || (c) == '\n' || (c) == '\r' \
1409 || (c) == '\t' || (c) == '\v' \
1410 || ((c) <= 3 && (c) >= 1) /* SOH, STX, ETX */ \
1411 || (c) == 7 /* U+7F DEL */ \
1412 || ((c) <= 0x13 && (c) >= 0x0E) /* SO, SI */ \
1413 /* DLE, DC[1-3] */ \
1414 || (c) == 0x18 /* U+18 CAN */ \
1415 || (c) == 0x19 /* U+19 EOM */ \
1416 || ((c) <= 0x1F && (c) >= 0x1C) /* [FGRU]S */ \
1417 || (c) == 0x26 /* U+17 ETB */ \
1418 || (c) == 0x27 /* U+1B ESC */ \
1419 || (c) == 0x2D /* U+05 ENQ */ \
1420 || (c) == 0x2E /* U+06 ACK */ \
1421 || (c) == 0x32 /* U+16 SYN */ \
1422 || (c) == 0x37 /* U+04 EOT */ \
1423 || (c) == 0x3C /* U+14 DC4 */ \
1424 || (c) == 0x3D /* U+15 NAK */ \
1425 || (c) == 0x3F)/* U+1A SUB */
0852beac 1426# define isASCII(c) (isCNTRL_A(c) || isPRINT_A(c))
74665a89
KW
1427# else /* isASCII is already defined for ASCII platforms, so can use that to
1428 define isCNTRL */
1429# define isCNTRL_A(c) (isASCII(c) && ! isPRINT_A(c))
0852beac
KW
1430# endif
1431
3ffc8c70 1432 /* The _L1 macros may be unnecessary for the utilities; I (khw) added them
caa94d35
KW
1433 * during debugging, and it seems best to keep them. We may be called
1434 * without NATIVE_TO_LATIN1 being defined. On ASCII platforms, it doesn't
1435 * do anything anyway, so make it not a problem */
1436# if ! defined(EBCDIC) && ! defined(NATIVE_TO_LATIN1)
1437# define NATIVE_TO_LATIN1(ch) (ch)
1438# endif
3ded5eb0
KW
1439# define isALPHA_L1(c) (isUPPER_L1(c) || isLOWER_L1(c))
1440# define isALPHANUMERIC_L1(c) (isALPHA_L1(c) || isDIGIT_A(c))
1441# define isBLANK_L1(c) (isBLANK_A(c) \
1442 || (FITS_IN_8_BITS(c) \
1443 && NATIVE_TO_LATIN1((U8) c) == 0xA0))
1444# define isCNTRL_L1(c) (FITS_IN_8_BITS(c) && (! isPRINT_L1(c)))
1445# define isGRAPH_L1(c) (isPRINT_L1(c) && (! isBLANK_L1(c)))
1446# define isLOWER_L1(c) (isLOWER_A(c) \
1447 || (FITS_IN_8_BITS(c) \
ae683a5f 1448 && (( NATIVE_TO_LATIN1((U8) c) >= 0xDF \
3ded5eb0
KW
1449 && NATIVE_TO_LATIN1((U8) c) != 0xF7) \
1450 || NATIVE_TO_LATIN1((U8) c) == 0xAA \
1451 || NATIVE_TO_LATIN1((U8) c) == 0xBA \
1452 || NATIVE_TO_LATIN1((U8) c) == 0xB5)))
1453# define isPRINT_L1(c) (isPRINT_A(c) \
1454 || (FITS_IN_8_BITS(c) \
1455 && NATIVE_TO_LATIN1((U8) c) >= 0xA0))
3ded5eb0
KW
1456# define isPUNCT_L1(c) (isPUNCT_A(c) \
1457 || (FITS_IN_8_BITS(c) \
ae683a5f 1458 && ( NATIVE_TO_LATIN1((U8) c) == 0xA1 \
3ded5eb0
KW
1459 || NATIVE_TO_LATIN1((U8) c) == 0xA7 \
1460 || NATIVE_TO_LATIN1((U8) c) == 0xAB \
1461 || NATIVE_TO_LATIN1((U8) c) == 0xB6 \
1462 || NATIVE_TO_LATIN1((U8) c) == 0xB7 \
1463 || NATIVE_TO_LATIN1((U8) c) == 0xBB \
1464 || NATIVE_TO_LATIN1((U8) c) == 0xBF)))
1465# define isSPACE_L1(c) (isSPACE_A(c) \
1466 || (FITS_IN_8_BITS(c) \
ae683a5f 1467 && ( NATIVE_TO_LATIN1((U8) c) == 0x85 \
3ded5eb0
KW
1468 || NATIVE_TO_LATIN1((U8) c) == 0xA0)))
1469# define isUPPER_L1(c) (isUPPER_A(c) \
1470 || (FITS_IN_8_BITS(c) \
ae683a5f 1471 && ( NATIVE_TO_LATIN1((U8) c) >= 0xC0 \
3ded5eb0
KW
1472 && NATIVE_TO_LATIN1((U8) c) <= 0xDE \
1473 && NATIVE_TO_LATIN1((U8) c) != 0xD7)))
1474# define isWORDCHAR_L1(c) (isIDFIRST_L1(c) || isDIGIT_A(c))
1475# define isIDFIRST_L1(c) (isALPHA_L1(c) || NATIVE_TO_LATIN1(c) == '_')
1476# define isCHARNAME_CONT(c) (isWORDCHAR_L1(c) \
1477 || isBLANK_L1(c) \
1478 || (c) == '-' \
1479 || (c) == '(' \
1480 || (c) == ')')
1481 /* The following are not fully accurate in the above-ASCII range. I (khw)
1482 * don't think it's necessary to be so for the purposes where this gets
1483 * compiled */
1484# define _isQUOTEMETA(c) (FITS_IN_8_BITS(c) && ! isWORDCHAR_L1(c))
1485# define _IS_IN_SOME_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) isALPHA_L1(c)
1486
1487 /* And these aren't accurate at all. They are useful only for above
1488 * Latin1, which utilities and bootstrapping don't deal with */
1489# define _IS_NON_FINAL_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) 0
6838b41e 1490# define _HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) 0
3ded5eb0
KW
1491# define _HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) 0
1492
1493 /* Many of the macros later in this file are defined in terms of these. By
1494 * implementing them with a function, which converts the class number into
1495 * a call to the desired macro, all of the later ones work. However, that
1496 * function won't be actually defined when building a utility program (no
1497 * perl.h), and so a compiler error will be generated if one is attempted
1498 * to be used. And the above-Latin1 code points require Unicode tables to
1499 * be present, something unlikely to be the case when bootstrapping */
1500# define _generic_isCC(c, classnum) \
1501 (FITS_IN_8_BITS(c) && S_bootstrap_ctype((U8) (c), (classnum), TRUE))
1502# define _generic_isCC_A(c, classnum) \
1503 (FITS_IN_8_BITS(c) && S_bootstrap_ctype((U8) (c), (classnum), FALSE))
687c8d01 1504#endif /* End of no perl.h H_PERL */
8a58bdcf 1505
e66b99e9
KW
1506#define isALPHANUMERIC(c) isALPHANUMERIC_A(c)
1507#define isALPHA(c) isALPHA_A(c)
0852beac
KW
1508#define isASCII_A(c) isASCII(c)
1509#define isASCII_L1(c) isASCII(c)
e66b99e9
KW
1510#define isBLANK(c) isBLANK_A(c)
1511#define isCNTRL(c) isCNTRL_A(c)
1512#define isDIGIT(c) isDIGIT_A(c)
1513#define isGRAPH(c) isGRAPH_A(c)
1514#define isIDFIRST(c) isIDFIRST_A(c)
1515#define isLOWER(c) isLOWER_A(c)
1516#define isPRINT(c) isPRINT_A(c)
779cf272 1517#define isPSXSPC_A(c) isSPACE_A(c)
e66b99e9 1518#define isPSXSPC(c) isPSXSPC_A(c)
779cf272 1519#define isPSXSPC_L1(c) isSPACE_L1(c)
e66b99e9
KW
1520#define isPUNCT(c) isPUNCT_A(c)
1521#define isSPACE(c) isSPACE_A(c)
1522#define isUPPER(c) isUPPER_A(c)
1523#define isWORDCHAR(c) isWORDCHAR_A(c)
1524#define isXDIGIT(c) isXDIGIT_A(c)
1525
1526/* ASCII casing. These could also be written as
1527 #define toLOWER(c) (isASCII(c) ? toLOWER_LATIN1(c) : (c))
1528 #define toUPPER(c) (isASCII(c) ? toUPPER_LATIN1_MOD(c) : (c))
1529 which uses table lookup and mask instead of subtraction. (This would
1530 work because the _MOD does not apply in the ASCII range) */
68067e4e
DM
1531#define toLOWER(c) (isUPPER(c) ? (U8)((c) + ('a' - 'A')) : (c))
1532#define toUPPER(c) (isLOWER(c) ? (U8)((c) - ('a' - 'A')) : (c))
bbce6d69 1533
25200305
KW
1534/* In the ASCII range, these are equivalent to what they're here defined to be.
1535 * But by creating these definitions, other code doesn't have to be aware of
1536 * this detail */
1537#define toFOLD(c) toLOWER(c)
25200305
KW
1538#define toTITLE(c) toUPPER(c)
1539
c753c8d3
KW
1540#define toLOWER_A(c) toLOWER(c)
1541#define toUPPER_A(c) toUPPER(c)
25200305
KW
1542#define toFOLD_A(c) toFOLD(c)
1543#define toTITLE_A(c) toTITLE(c)
1a0901db 1544
4650c663 1545/* Use table lookup for speed; returns the input itself if is out-of-range */
b2bf251f 1546#define toLOWER_LATIN1(c) ((! FITS_IN_8_BITS(c)) \
8e7c6e7d 1547 ? (c) \
f4cd282c 1548 : PL_latin1_lc[ (U8) (c) ])
c753c8d3
KW
1549#define toLOWER_L1(c) toLOWER_LATIN1(c) /* Synonym for consistency */
1550
1a0901db 1551/* Modified uc. Is correct uc except for three non-ascii chars which are
4650c663
KW
1552 * all mapped to one of them, and these need special handling; returns the
1553 * input itself if is out-of-range */
b2bf251f 1554#define toUPPER_LATIN1_MOD(c) ((! FITS_IN_8_BITS(c)) \
8e7c6e7d 1555 ? (c) \
f4cd282c 1556 : PL_mod_latin1_uc[ (U8) (c) ])
31f05a37 1557#define IN_UTF8_CTYPE_LOCALE PL_in_utf8_CTYPE_locale
84061b6a 1558
beab9ebe
KW
1559/* Use foo_LC_uvchr() instead of these for beyond the Latin1 range */
1560
1561/* For internal core Perl use only: the base macro for defining macros like
1562 * isALPHA_LC, which uses the current LC_CTYPE locale. 'c' is the code point
31f05a37
KW
1563 * (0-255) to check. In a UTF-8 locale, the result is the same as calling
1564 * isFOO_L1(); the 'utf8_locale_classnum' parameter is something like
1565 * _CC_UPPER, which gives the class number for doing this. For non-UTF-8
1566 * locales, the code to actually do the test this is passed in 'non_utf8'. If
1567 * 'c' is above 255, 0 is returned. For accessing the full range of possible
1568 * code points under locale rules, use the macros based on _generic_LC_uvchr
1569 * instead of this. */
beab9ebe
KW
1570#define _generic_LC_base(c, utf8_locale_classnum, non_utf8) \
1571 (! FITS_IN_8_BITS(c) \
1572 ? 0 \
31f05a37
KW
1573 : IN_UTF8_CTYPE_LOCALE \
1574 ? cBOOL(PL_charclass[(U8) (c)] & _CC_mask(utf8_locale_classnum)) \
beab9ebe
KW
1575 : cBOOL(non_utf8))
1576
1577/* For internal core Perl use only: a helper macro for defining macros like
1578 * isALPHA_LC. 'c' is the code point (0-255) to check. The function name to
1579 * actually do this test is passed in 'non_utf8_func', which is called on 'c',
1580 * casting 'c' to the macro _LC_CAST, which should not be parenthesized. See
1581 * _generic_LC_base for more info */
1582#define _generic_LC(c, utf8_locale_classnum, non_utf8_func) \
1583 _generic_LC_base(c,utf8_locale_classnum, \
1584 non_utf8_func( (_LC_CAST) (c)))
1585
1586/* For internal core Perl use only: like _generic_LC, but also returns TRUE if
1587 * 'c' is the platform's native underscore character */
1588#define _generic_LC_underscore(c,utf8_locale_classnum,non_utf8_func) \
1589 _generic_LC_base(c, utf8_locale_classnum, \
1590 (non_utf8_func( (_LC_CAST) (c)) \
1591 || (char)(c) == '_'))
1592
1593/* These next three are also for internal core Perl use only: case-change
1594 * helper macros */
1595#define _generic_toLOWER_LC(c, function, cast) (! FITS_IN_8_BITS(c) \
1596 ? (c) \
31f05a37
KW
1597 : (IN_UTF8_CTYPE_LOCALE) \
1598 ? PL_latin1_lc[ (U8) (c) ] \
d2a8cf7f 1599 : (cast)function((cast)(c)))
beab9ebe 1600
31f05a37
KW
1601/* Note that the result can be larger than a byte in a UTF-8 locale. It
1602 * returns a single value, so can't adequately return the upper case of LATIN
1603 * SMALL LETTER SHARP S in a UTF-8 locale (which should be a string of two
1604 * values "SS"); instead it asserts against that under DEBUGGING, and
1605 * otherwise returns its input */
beab9ebe
KW
1606#define _generic_toUPPER_LC(c, function, cast) \
1607 (! FITS_IN_8_BITS(c) \
1608 ? (c) \
31f05a37 1609 : ((! IN_UTF8_CTYPE_LOCALE) \
b7d90381 1610 ? (cast)function((cast)(c)) \
31f05a37
KW
1611 : ((((U8)(c)) == MICRO_SIGN) \
1612 ? GREEK_CAPITAL_LETTER_MU \
1613 : ((((U8)(c)) == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS) \
1614 ? LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS \
1615 : ((((U8)(c)) == LATIN_SMALL_LETTER_SHARP_S) \
1616 ? (__ASSERT_(0) (c)) \
1617 : PL_mod_latin1_uc[ (U8) (c) ])))))
1618
1619/* Note that the result can be larger than a byte in a UTF-8 locale. It
1620 * returns a single value, so can't adequately return the fold case of LATIN
1621 * SMALL LETTER SHARP S in a UTF-8 locale (which should be a string of two
1622 * values "ss"); instead it asserts against that under DEBUGGING, and
1623 * otherwise returns its input */
beab9ebe 1624#define _generic_toFOLD_LC(c, function, cast) \
4d7db1b9 1625 ((UNLIKELY((c) == MICRO_SIGN) && IN_UTF8_CTYPE_LOCALE) \
31f05a37 1626 ? GREEK_SMALL_LETTER_MU \
4d7db1b9
KW
1627 : (__ASSERT_(! IN_UTF8_CTYPE_LOCALE \
1628 || (c) != LATIN_SMALL_LETTER_SHARP_S) \
1629 _generic_toLOWER_LC(c, function, cast)))
beab9ebe 1630
84061b6a 1631/* Use the libc versions for these if available. */
f05550c0 1632#if defined(HAS_ISASCII)
84061b6a
KW
1633# define isASCII_LC(c) (FITS_IN_8_BITS(c) && isascii( (U8) (c)))
1634#else
1635# define isASCII_LC(c) isASCII(c)
1636#endif
1637
f05550c0 1638#if defined(HAS_ISBLANK)
84061b6a 1639# define isBLANK_LC(c) _generic_LC(c, _CC_BLANK, isblank)
95f34b6f 1640#else /* Unlike isASCII, varies if in a UTF-8 locale */
278b2b56 1641# define isBLANK_LC(c) ((IN_UTF8_CTYPE_LOCALE) ? isBLANK_L1(c) : isBLANK(c))
84061b6a
KW
1642#endif
1643
f05550c0 1644#define _LC_CAST U8
bbce6d69 1645
f05550c0 1646#ifdef WIN32
375f5f06
KW
1647 /* The Windows functions don't bother to follow the POSIX standard, which
1648 * for example says that something can't both be a printable and a control.
1649 * But Windows treats the \t control as a printable, and does such things
1650 * as making superscripts into both digits and punctuation. This tames
1651 * these flaws by assuming that the definitions of both controls and space
1652 * are correct, and then making sure that other definitions don't have
1653 * weirdnesses, by making sure that isalnum() isn't also ispunct(), etc.
1654 * Not all possible weirdnesses are checked for, just the ones that were
1655 * detected on actual Microsoft code pages */
1656
b7d90381
KW
1657# define isCNTRL_LC(c) _generic_LC(c, _CC_CNTRL, iscntrl)
1658# define isSPACE_LC(c) _generic_LC(c, _CC_SPACE, isspace)
1659
1660# define isALPHA_LC(c) (_generic_LC(c, _CC_ALPHA, isalpha) \
1661 && isALPHANUMERIC_LC(c))
1662# define isALPHANUMERIC_LC(c) (_generic_LC(c, _CC_ALPHANUMERIC, isalnum) && \
1663 ! isPUNCT_LC(c))
1664# define isDIGIT_LC(c) (_generic_LC(c, _CC_DIGIT, isdigit) && \
1665 isALPHANUMERIC_LC(c))
1666# define isGRAPH_LC(c) (_generic_LC(c, _CC_GRAPH, isgraph) && isPRINT_LC(c))
1667# define isIDFIRST_LC(c) (((c) == '_') \
1668 || (_generic_LC(c, _CC_IDFIRST, isalpha) && ! isPUNCT_LC(c)))
1669# define isLOWER_LC(c) (_generic_LC(c, _CC_LOWER, islower) && isALPHA_LC(c))
1670# define isPRINT_LC(c) (_generic_LC(c, _CC_PRINT, isprint) && ! isCNTRL_LC(c))
1671# define isPUNCT_LC(c) (_generic_LC(c, _CC_PUNCT, ispunct) && ! isCNTRL_LC(c))
1672# define isUPPER_LC(c) (_generic_LC(c, _CC_UPPER, isupper) && isALPHA_LC(c))
f05550c0 1673# define isWORDCHAR_LC(c) (((c) == '_') || isALPHANUMERIC_LC(c))
b7d90381
KW
1674# define isXDIGIT_LC(c) (_generic_LC(c, _CC_XDIGIT, isxdigit) \
1675 && isALPHANUMERIC_LC(c))
f05550c0
BF
1676
1677# define toLOWER_LC(c) _generic_toLOWER_LC((c), tolower, U8)
1678# define toUPPER_LC(c) _generic_toUPPER_LC((c), toupper, U8)
1679# define toFOLD_LC(c) _generic_toFOLD_LC((c), tolower, U8)
1680
1681#elif defined(CTYPE256) || (!defined(isascii) && !defined(HAS_ISASCII))
4650c663 1682 /* For most other platforms */
beab9ebe 1683
f05550c0
BF
1684# define isALPHA_LC(c) _generic_LC(c, _CC_ALPHA, isalpha)
1685# define isALPHANUMERIC_LC(c) _generic_LC(c, _CC_ALPHANUMERIC, isalnum)
1686# define isCNTRL_LC(c) _generic_LC(c, _CC_CNTRL, iscntrl)
1687# define isDIGIT_LC(c) _generic_LC(c, _CC_DIGIT, isdigit)
1688# define isGRAPH_LC(c) _generic_LC(c, _CC_GRAPH, isgraph)
1689# define isIDFIRST_LC(c) _generic_LC_underscore(c, _CC_IDFIRST, isalpha)
1690# define isLOWER_LC(c) _generic_LC(c, _CC_LOWER, islower)
1691# define isPRINT_LC(c) _generic_LC(c, _CC_PRINT, isprint)
1692# define isPUNCT_LC(c) _generic_LC(c, _CC_PUNCT, ispunct)
1693# define isSPACE_LC(c) _generic_LC(c, _CC_SPACE, isspace)
1694# define isUPPER_LC(c) _generic_LC(c, _CC_UPPER, isupper)
1695# define isWORDCHAR_LC(c) _generic_LC_underscore(c, _CC_WORDCHAR, isalnum)
1696# define isXDIGIT_LC(c) _generic_LC(c, _CC_XDIGIT, isxdigit)
1697
1698
1699# define toLOWER_LC(c) _generic_toLOWER_LC((c), tolower, U8)
1700# define toUPPER_LC(c) _generic_toUPPER_LC((c), toupper, U8)
1701# define toFOLD_LC(c) _generic_toFOLD_LC((c), tolower, U8)
1702
1703#else /* The final fallback position */
1704
b7d90381
KW
1705# define isALPHA_LC(c) (isascii(c) && isalpha(c))
1706# define isALPHANUMERIC_LC(c) (isascii(c) && isalnum(c))
1707# define isCNTRL_LC(c) (isascii(c) && iscntrl(c))
1708# define isDIGIT_LC(c) (isascii(c) && isdigit(c))
1709# define isGRAPH_LC(c) (isascii(c) && isgraph(c))
f05550c0 1710# define isIDFIRST_LC(c) (isascii(c) && (isalpha(c) || (c) == '_'))
b7d90381
KW
1711# define isLOWER_LC(c) (isascii(c) && islower(c))
1712# define isPRINT_LC(c) (isascii(c) && isprint(c))
1713# define isPUNCT_LC(c) (isascii(c) && ispunct(c))
1714# define isSPACE_LC(c) (isascii(c) && isspace(c))
1715# define isUPPER_LC(c) (isascii(c) && isupper(c))
f05550c0 1716# define isWORDCHAR_LC(c) (isascii(c) && (isalnum(c) || (c) == '_'))
b7d90381 1717# define isXDIGIT_LC(c) (isascii(c) && isxdigit(c))
f05550c0
BF
1718
1719# define toLOWER_LC(c) (isascii(c) ? tolower(c) : (c))
1720# define toUPPER_LC(c) (isascii(c) ? toupper(c) : (c))
1721# define toFOLD_LC(c) (isascii(c) ? tolower(c) : (c))
bbce6d69 1722
f05550c0 1723#endif
55204971 1724
eba68aa0
KW
1725#define isIDCONT(c) isWORDCHAR(c)
1726#define isIDCONT_A(c) isWORDCHAR_A(c)
1727#define isIDCONT_L1(c) isWORDCHAR_L1(c)
1728#define isIDCONT_LC(c) isWORDCHAR_LC(c)
13380643 1729#define isPSXSPC_LC(c) isSPACE_LC(c)
aaa51d5e 1730
4650c663 1731/* For internal core Perl use only: the base macros for defining macros like
d0da05db
KW
1732 * isALPHA_uvchr. 'c' is the code point to check. 'classnum' is the POSIX class
1733 * number defined earlier in this file. _generic_uvchr() is used for POSIX
4650c663
KW
1734 * classes where there is a macro or function 'above_latin1' that takes the
1735 * single argument 'c' and returns the desired value. These exist for those
1736 * classes which have simple definitions, avoiding the overhead of a hash
d0da05db 1737 * lookup or inversion list binary search. _generic_swash_uvchr() can be used
4650c663 1738 * for classes where that overhead is faster than a direct lookup.
d0da05db 1739 * _generic_uvchr() won't compile if 'c' isn't unsigned, as it won't match the
4650c663
KW
1740 * 'above_latin1' prototype. _generic_isCC() macro does bounds checking, so
1741 * have duplicate checks here, so could create versions of the macros that
1742 * don't, but experiments show that gcc optimizes them out anyway. */
66c17564
KW
1743
1744/* Note that all ignore 'use bytes' */
1e222e4f
KW
1745#define _generic_uvchr(classnum, above_latin1, c) ((c) < 256 \
1746 ? _generic_isCC(c, classnum) \
cd500f2f 1747 : above_latin1(c))
1e222e4f
KW
1748#define _generic_swash_uvchr(classnum, c) ((c) < 256 \
1749 ? _generic_isCC(c, classnum) \
922e8cb4 1750 : _is_uni_FOO(classnum, c))
d0da05db
KW
1751#define isALPHA_uvchr(c) _generic_swash_uvchr(_CC_ALPHA, c)
1752#define isALPHANUMERIC_uvchr(c) _generic_swash_uvchr(_CC_ALPHANUMERIC, c)
1753#define isASCII_uvchr(c) isASCII(c)
1754#define isBLANK_uvchr(c) _generic_uvchr(_CC_BLANK, is_HORIZWS_cp_high, c)
1755#define isCNTRL_uvchr(c) isCNTRL_L1(c) /* All controls are in Latin1 */
1756#define isDIGIT_uvchr(c) _generic_swash_uvchr(_CC_DIGIT, c)
1757#define isGRAPH_uvchr(c) _generic_swash_uvchr(_CC_GRAPH, c)
1e222e4f
KW
1758#define isIDCONT_uvchr(c) \
1759 _generic_uvchr(_CC_WORDCHAR, _is_uni_perl_idcont, c)
1760#define isIDFIRST_uvchr(c) \
1761 _generic_uvchr(_CC_IDFIRST, _is_uni_perl_idstart, c)
d0da05db
KW
1762#define isLOWER_uvchr(c) _generic_swash_uvchr(_CC_LOWER, c)
1763#define isPRINT_uvchr(c) _generic_swash_uvchr(_CC_PRINT, c)
1764
1765#define isPUNCT_uvchr(c) _generic_swash_uvchr(_CC_PUNCT, c)
1766#define isSPACE_uvchr(c) _generic_uvchr(_CC_SPACE, is_XPERLSPACE_cp_high, c)
1767#define isPSXSPC_uvchr(c) isSPACE_uvchr(c)
1768
1769#define isUPPER_uvchr(c) _generic_swash_uvchr(_CC_UPPER, c)
1770#define isVERTWS_uvchr(c) _generic_uvchr(_CC_VERTSPACE, is_VERTWS_cp_high, c)
1771#define isWORDCHAR_uvchr(c) _generic_swash_uvchr(_CC_WORDCHAR, c)
1772#define isXDIGIT_uvchr(c) _generic_uvchr(_CC_XDIGIT, is_XDIGIT_cp_high, c)
1773
1774#define toFOLD_uvchr(c,s,l) to_uni_fold(c,s,l)
1775#define toLOWER_uvchr(c,s,l) to_uni_lower(c,s,l)
1776#define toTITLE_uvchr(c,s,l) to_uni_title(c,s,l)
1777#define toUPPER_uvchr(c,s,l) to_uni_upper(c,s,l)
1778
1779/* For backwards compatibility, even though '_uni' should mean official Unicode
1780 * code points, in Perl it means native for those below 256 */
1781#define isALPHA_uni(c) isALPHA_uvchr(c)
1782#define isALPHANUMERIC_uni(c) isALPHANUMERIC_uvchr(c)
1783#define isASCII_uni(c) isASCII_uvchr(c)
1784#define isBLANK_uni(c) isBLANK_uvchr(c)
1785#define isCNTRL_uni(c) isCNTRL_uvchr(c)
1786#define isDIGIT_uni(c) isDIGIT_uvchr(c)
1787#define isGRAPH_uni(c) isGRAPH_uvchr(c)
1788#define isIDCONT_uni(c) isIDCONT_uvchr(c)
1789#define isIDFIRST_uni(c) isIDFIRST_uvchr(c)
1790#define isLOWER_uni(c) isLOWER_uvchr(c)
1791#define isPRINT_uni(c) isPRINT_uvchr(c)
1792#define isPUNCT_uni(c) isPUNCT_uvchr(c)
1793#define isSPACE_uni(c) isSPACE_uvchr(c)
1794#define isPSXSPC_uni(c) isPSXSPC_uvchr(c)
1795#define isUPPER_uni(c) isUPPER_uvchr(c)
1796#define isVERTWS_uni(c) isVERTWS_uvchr(c)
1797#define isWORDCHAR_uni(c) isWORDCHAR_uvchr(c)
1798#define isXDIGIT_uni(c) isXDIGIT_uvchr(c)
1799#define toFOLD_uni(c,s,l) toFOLD_uvchr(c,s,l)
1800#define toLOWER_uni(c,s,l) toLOWER_uvchr(c,s,l)
1801#define toTITLE_uni(c,s,l) toTITLE_uvchr(c,s,l)
1802#define toUPPER_uni(c,s,l) toUPPER_uvchr(c,s,l)
a0ed51b3 1803
4650c663
KW
1804/* For internal core Perl use only: the base macros for defining macros like
1805 * isALPHA_LC_uvchr. These are like isALPHA_LC, but the input can be any code
d0da05db 1806 * point, not just 0-255. Like _generic_uvchr, there are two versions, one for
4650c663 1807 * simple class definitions; the other for more complex. These are like
d0da05db 1808 * _generic_uvchr, so see it for more info. */
cd500f2f
KW
1809#define _generic_LC_uvchr(latin1, above_latin1, c) \
1810 (c < 256 ? latin1(c) : above_latin1(c))
1811#define _generic_LC_swash_uvchr(latin1, classnum, c) \
1812 (c < 256 ? latin1(c) : _is_uni_FOO(classnum, c))
1813
1814#define isALPHA_LC_uvchr(c) _generic_LC_swash_uvchr(isALPHA_LC, _CC_ALPHA, c)
1815#define isALPHANUMERIC_LC_uvchr(c) _generic_LC_swash_uvchr(isALPHANUMERIC_LC, \
1816 _CC_ALPHANUMERIC, c)
b7d90381
KW
1817#define isASCII_LC_uvchr(c) isASCII_LC(c)
1818#define isBLANK_LC_uvchr(c) _generic_LC_uvchr(isBLANK_LC, \
1819 is_HORIZWS_cp_high, c)
feeab5a9 1820#define isCNTRL_LC_uvchr(c) (c < 256 ? isCNTRL_LC(c) : 0)
cd500f2f
KW
1821#define isDIGIT_LC_uvchr(c) _generic_LC_swash_uvchr(isDIGIT_LC, _CC_DIGIT, c)
1822#define isGRAPH_LC_uvchr(c) _generic_LC_swash_uvchr(isGRAPH_LC, _CC_GRAPH, c)
b7d90381 1823#define isIDCONT_LC_uvchr(c) _generic_LC_uvchr(isIDCONT_LC, \
eba68aa0 1824 _is_uni_perl_idcont, c)
b7d90381 1825#define isIDFIRST_LC_uvchr(c) _generic_LC_uvchr(isIDFIRST_LC, \
cd500f2f
KW
1826 _is_uni_perl_idstart, c)
1827#define isLOWER_LC_uvchr(c) _generic_LC_swash_uvchr(isLOWER_LC, _CC_LOWER, c)
1828#define isPRINT_LC_uvchr(c) _generic_LC_swash_uvchr(isPRINT_LC, _CC_PRINT, c)
b7d90381 1829#define isPSXSPC_LC_uvchr(c) isSPACE_LC_uvchr(c)
cd500f2f 1830#define isPUNCT_LC_uvchr(c) _generic_LC_swash_uvchr(isPUNCT_LC, _CC_PUNCT, c)
b7d90381 1831#define isSPACE_LC_uvchr(c) _generic_LC_uvchr(isSPACE_LC, \
509fb054 1832 is_XPERLSPACE_cp_high, c)
cd500f2f 1833#define isUPPER_LC_uvchr(c) _generic_LC_swash_uvchr(isUPPER_LC, _CC_UPPER, c)
b7d90381 1834#define isWORDCHAR_LC_uvchr(c) _generic_LC_swash_uvchr(isWORDCHAR_LC, \
cd500f2f 1835 _CC_WORDCHAR, c)
b7d90381
KW
1836#define isXDIGIT_LC_uvchr(c) _generic_LC_uvchr(isXDIGIT_LC, \
1837 is_XDIGIT_cp_high, c)
e712593e 1838
b7d90381 1839#define isBLANK_LC_uni(c) isBLANK_LC_uvchr(UNI_TO_NATIVE(c))
aaa51d5e 1840
4650c663
KW
1841/* For internal core Perl use only: the base macros for defining macros like
1842 * isALPHA_utf8. These are like the earlier defined macros, but take an input
1843 * UTF-8 encoded string 'p'. If the input is in the Latin1 range, use
1844 * the Latin1 macro 'classnum' on 'p'. Otherwise use the value given by the
1845 * 'utf8' parameter. This relies on the fact that ASCII characters have the
1846 * same representation whether utf8 or not. Note that it assumes that the utf8
1847 * has been validated, and ignores 'use bytes' */
34aeb2e9
KW
1848#define _base_generic_utf8(enum_name, name, p, use_locale ) \
1849 _is_utf8_FOO(CAT2(_CC_, enum_name), \
1850 (const U8 *) p, \
1851 "is" STRINGIFY(name) "_utf8", \
1852 "is" STRINGIFY(name) "_utf8_safe", \
1853 1, use_locale, __FILE__,__LINE__)
1854
1855#define _generic_utf8(name, p) _base_generic_utf8(name, name, p, 0)
da8c1a98
KW
1856
1857/* The "_safe" macros make sure that we don't attempt to read beyond 'e', but
1858 * they don't otherwise go out of their way to look for malformed UTF-8. If
1859 * they can return accurate results without knowing if the input is otherwise
1860 * malformed, they do so. For example isASCII is accurate in spite of any
1861 * non-length malformations because it looks only at a single byte. Likewise
1862 * isDIGIT looks just at the first byte for code points 0-255, as all UTF-8
1863 * variant ones return FALSE. But, if the input has to be well-formed in order
1864 * for the results to be accurate, the macros will test and if malformed will
1865 * call a routine to die
1866 *
1867 * Except for toke.c, the macros do assume that e > p, asserting that on
1868 * DEBUGGING builds. Much code that calls these depends on this being true,
1869 * for other reasons. toke.c is treated specially as using the regular
1870 * assertion breaks it in many ways. All strings that these operate on there
1871 * are supposed to have an extra NUL character at the end, so that *e = \0. A
1872 * bunch of code in toke.c assumes that this is true, so the assertion allows
1873 * for that */
1874#ifdef PERL_IN_TOKE_C
1875# define _utf8_safe_assert(p,e) ((e) > (p) || ((e) == (p) && *(p) == '\0'))
1876#else
1877# define _utf8_safe_assert(p,e) ((e) > (p))
1878#endif
1879
1880#define _generic_utf8_safe(classnum, p, e, above_latin1) \
1881 (__ASSERT_(_utf8_safe_assert(p, e)) \
1882 (UTF8_IS_INVARIANT(*(p))) \
1883 ? _generic_isCC(*(p), classnum) \
1884 : (UTF8_IS_DOWNGRADEABLE_START(*(p)) \
1885 ? ((LIKELY((e) - (p) > 1 && UTF8_IS_CONTINUATION(*((p)+1)))) \
1886 ? _generic_isCC(EIGHT_BIT_UTF8_TO_NATIVE(*(p), *((p)+1 )), \
1887 classnum) \
1888 : (_force_out_malformed_utf8_message( \
1889 (U8 *) (p), (U8 *) (e), 0, 1), 0)) \
1890 : above_latin1))
b7d90381
KW
1891/* Like the above, but calls 'above_latin1(p)' to get the utf8 value.
1892 * 'above_latin1' can be a macro */
da8c1a98
KW
1893#define _generic_func_utf8_safe(classnum, above_latin1, p, e) \
1894 _generic_utf8_safe(classnum, p, e, above_latin1(p, e))
1895#define _generic_non_swash_utf8_safe(classnum, above_latin1, p, e) \
1896 _generic_utf8_safe(classnum, p, e, \
1897 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
1898 ? (_force_out_malformed_utf8_message( \
1899 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
1900 : above_latin1(p)))
4650c663 1901/* Like the above, but passes classnum to _isFOO_utf8(), instead of having an
cd500f2f 1902 * 'above_latin1' parameter */
da8c1a98
KW
1903#define _generic_swash_utf8_safe(classnum, p, e) \
1904_generic_utf8_safe(classnum, p, e, _is_utf8_FOO_with_len(classnum, p, e))
922e8cb4 1905
cc8ab7c0 1906/* Like the above, but should be used only when it is known that there are no
ff7ecfc3
KW
1907 * characters in the upper-Latin1 range (128-255 on ASCII platforms) which the
1908 * class is TRUE for. Hence it can skip the tests for this range.
1909 * 'above_latin1' should include its arguments */
da8c1a98
KW
1910#define _generic_utf8_safe_no_upper_latin1(classnum, p, e, above_latin1) \
1911 (__ASSERT_(_utf8_safe_assert(p, e)) \
1912 (UTF8_IS_INVARIANT(*(p))) \
1913 ? _generic_isCC(*(p), classnum) \
1914 : (UTF8_IS_DOWNGRADEABLE_START(*(p))) \
1915 ? 0 /* Note that doesn't check validity for latin1 */ \
1916 : above_latin1)
1917
84238efa
KW
1918/* NOTE that some of these macros have very similar ones in regcharclass.h.
1919 * For example, there is (at the time of this writing) an 'is_SPACE_utf8()'
1920 * there, differing in name only by an underscore from the one here
1921 * 'isSPACE_utf8(). The difference is that the ones here are probably more
1922 * efficient and smaller, using an O(1) array lookup for Latin1-range code
1923 * points; the regcharclass.h ones are implemented as a series of
1924 * "if-else-if-else ..." */
1925
34aeb2e9
KW
1926#define isALPHA_utf8(p) _generic_utf8(ALPHA, p)
1927#define isALPHANUMERIC_utf8(p) _generic_utf8(ALPHANUMERIC, p)
1928#define isASCII_utf8(p) _generic_utf8(ASCII, p)
1929#define isBLANK_utf8(p) _generic_utf8(BLANK, p)
1930#define isCNTRL_utf8(p) _generic_utf8(CNTRL, p)
1931#define isDIGIT_utf8(p) _generic_utf8(DIGIT, p)
1932#define isGRAPH_utf8(p) _generic_utf8(GRAPH, p)
1933#define isIDCONT_utf8(p) _generic_utf8(IDCONT, p)
1934#define isIDFIRST_utf8(p) _generic_utf8(IDFIRST, p)
1935#define isLOWER_utf8(p) _generic_utf8(LOWER, p)
1936#define isPRINT_utf8(p) _generic_utf8(PRINT, p)
1937#define isPSXSPC_utf8(p) _generic_utf8(PSXSPC, p)
1938#define isPUNCT_utf8(p) _generic_utf8(PUNCT, p)
1939#define isSPACE_utf8(p) _generic_utf8(SPACE, p)
1940#define isUPPER_utf8(p) _generic_utf8(UPPER, p)
1941#define isVERTWS_utf8(p) _generic_utf8(VERTSPACE, p)
1942#define isWORDCHAR_utf8(p) _generic_utf8(WORDCHAR, p)
1943#define isXDIGIT_utf8(p) _generic_utf8(XDIGIT, p)
e8fa43e2 1944
da8c1a98
KW
1945#define isALPHA_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_ALPHA, p, e)
1946#define isALPHANUMERIC_utf8_safe(p, e) \
1947 _generic_swash_utf8_safe(_CC_ALPHANUMERIC, p, e)
1948#define isASCII_utf8_safe(p, e) \
1949 /* Because ASCII is invariant under utf8, the non-utf8 macro \
1950 * works */ \
1951 (__ASSERT_(_utf8_safe_assert(p, e)) isASCII(*(p)))
1952#define isBLANK_utf8_safe(p, e) \
1953 _generic_non_swash_utf8_safe(_CC_BLANK, is_HORIZWS_high, p, e)
1954
e8fa43e2
KW
1955#ifdef EBCDIC
1956 /* Because all controls are UTF-8 invariants in EBCDIC, we can use this
1957 * more efficient macro instead of the more general one */
da8c1a98 1958# define isCNTRL_utf8_safe(p, e) \
56d02b8c 1959 (__ASSERT_(_utf8_safe_assert(p, e)) isCNTRL_L1(*(p)))
e8fa43e2 1960#else
da8c1a98 1961# define isCNTRL_utf8_safe(p, e) _generic_utf8_safe(_CC_CNTRL, p, e, 0)
e8fa43e2
KW
1962#endif
1963
da8c1a98
KW
1964#define isDIGIT_utf8_safe(p, e) \
1965 _generic_utf8_safe_no_upper_latin1(_CC_DIGIT, p, e, \
1966 _is_utf8_FOO_with_len(_CC_DIGIT, p, e))
1967#define isGRAPH_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_GRAPH, p, e)
1968#define isIDCONT_utf8_safe(p, e) _generic_func_utf8_safe(_CC_WORDCHAR, \
1969 _is_utf8_perl_idcont_with_len, p, e)
e5dcd934 1970
c11ff943
KW
1971/* To prevent S_scan_word in toke.c from hanging, we have to make sure that
1972 * IDFIRST is an alnum. See
c8362b00 1973 * http://rt.perl.org/rt3/Ticket/Display.html?id=74022 for more detail than you
f91dcd13
KW
1974 * ever wanted to know about. (In the ASCII range, there isn't a difference.)
1975 * This used to be not the XID version, but we decided to go with the more
1976 * modern Unicode definition */
da8c1a98
KW
1977#define isIDFIRST_utf8_safe(p, e) \
1978 _generic_func_utf8_safe(_CC_IDFIRST, \
1979 _is_utf8_perl_idstart_with_len, (U8 *) (p), (U8 *) (e))
1980
1981#define isLOWER_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_LOWER, p, e)
1982#define isPRINT_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_PRINT, p, e)
1983#define isPSXSPC_utf8_safe(p, e) isSPACE_utf8_safe(p, e)
1984#define isPUNCT_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_PUNCT, p, e)
1985#define isSPACE_utf8_safe(p, e) \
1986 _generic_non_swash_utf8_safe(_CC_SPACE, is_XPERLSPACE_high, p, e)
1987#define isUPPER_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_UPPER, p, e)
1988#define isVERTWS_utf8_safe(p, e) \
1989 _generic_non_swash_utf8_safe(_CC_VERTSPACE, is_VERTWS_high, p, e)
1990#define isWORDCHAR_utf8_safe(p, e) \
1991 _generic_swash_utf8_safe(_CC_WORDCHAR, p, e)
1992#define isXDIGIT_utf8_safe(p, e) \
1993 _generic_utf8_safe_no_upper_latin1(_CC_XDIGIT, p, e, \
1994 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
1995 ? (_force_out_malformed_utf8_message( \
1996 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
1997 : is_XDIGIT_high(p)))
a0ed51b3 1998
7d2b61dd 1999#define toFOLD_utf8(p,s,l) to_utf8_fold(p,s,l)
509fb054
KW
2000#define toLOWER_utf8(p,s,l) to_utf8_lower(p,s,l)
2001#define toTITLE_utf8(p,s,l) to_utf8_title(p,s,l)
2002#define toUPPER_utf8(p,s,l) to_utf8_upper(p,s,l)
2e8adce6 2003
567b353c 2004/* For internal core use only, subject to change */
607313a1
KW
2005#define _toFOLD_utf8_flags(p,e,s,l,f) _to_utf8_fold_flags (p,e,s,l,f, "", 0)
2006#define _toLOWER_utf8_flags(p,e,s,l,f) _to_utf8_lower_flags(p,e,s,l,f, "", 0)
2007#define _toTITLE_utf8_flags(p,e,s,l,f) _to_utf8_title_flags(p,e,s,l,f, "", 0)
2008#define _toUPPER_utf8_flags(p,e,s,l,f) _to_utf8_upper_flags(p,e,s,l,f, "", 0)
a1a5ec35
KW
2009
2010#define toFOLD_utf8_safe(p,e,s,l) _toFOLD_utf8_flags(p,e,s,l, FOLD_FLAGS_FULL)
2011#define toLOWER_utf8_safe(p,e,s,l) _toLOWER_utf8_flags(p,e,s,l, 0)
2012#define toTITLE_utf8_safe(p,e,s,l) _toTITLE_utf8_flags(p,e,s,l, 0)
2013#define toUPPER_utf8_safe(p,e,s,l) _toUPPER_utf8_flags(p,e,s,l, 0)
567b353c 2014
4650c663
KW
2015/* For internal core Perl use only: the base macros for defining macros like
2016 * isALPHA_LC_utf8. These are like _generic_utf8, but if the first code point
2017 * in 'p' is within the 0-255 range, it uses locale rules from the passed-in
2018 * 'macro' parameter */
34aeb2e9
KW
2019#define _generic_LC_utf8(name, p) _base_generic_utf8(name, name, p, 1)
2020
2021#define isALPHA_LC_utf8(p) _generic_LC_utf8(ALPHA, p)
2022#define isALPHANUMERIC_LC_utf8(p) _generic_LC_utf8(ALPHANUMERIC, p)
2023#define isASCII_LC_utf8(p) _generic_LC_utf8(ASCII, p)
2024#define isBLANK_LC_utf8(p) _generic_LC_utf8(BLANK, p)
2025#define isCNTRL_LC_utf8(p) _generic_LC_utf8(CNTRL, p)
2026#define isDIGIT_LC_utf8(p) _generic_LC_utf8(DIGIT, p)
2027#define isGRAPH_LC_utf8(p) _generic_LC_utf8(GRAPH, p)
2028#define isIDCONT_LC_utf8(p) _generic_LC_utf8(IDCONT, p)
2029#define isIDFIRST_LC_utf8(p) _generic_LC_utf8(IDFIRST, p)
2030#define isLOWER_LC_utf8(p) _generic_LC_utf8(LOWER, p)
2031#define isPRINT_LC_utf8(p) _generic_LC_utf8(PRINT, p)
2032#define isPSXSPC_LC_utf8(p) _generic_LC_utf8(PSXSPC, p)
2033#define isPUNCT_LC_utf8(p) _generic_LC_utf8(PUNCT, p)
2034#define isSPACE_LC_utf8(p) _generic_LC_utf8(SPACE, p)
2035#define isUPPER_LC_utf8(p) _generic_LC_utf8(UPPER, p)
2036#define isWORDCHAR_LC_utf8(p) _generic_LC_utf8(WORDCHAR, p)
2037#define isXDIGIT_LC_utf8(p) _generic_LC_utf8(XDIGIT, p)
2038
da8c1a98
KW
2039/* For internal core Perl use only: the base macros for defining macros like
2040 * isALPHA_LC_utf8_safe. These are like _generic_utf8, but if the first code
2041 * point in 'p' is within the 0-255 range, it uses locale rules from the
2042 * passed-in 'macro' parameter */
2043#define _generic_LC_utf8_safe(macro, p, e, above_latin1) \
2044 (__ASSERT_(_utf8_safe_assert(p, e)) \
2045 (UTF8_IS_INVARIANT(*(p))) \
2046 ? macro(*(p)) \
2047 : (UTF8_IS_DOWNGRADEABLE_START(*(p)) \
2048 ? ((LIKELY((e) - (p) > 1 && UTF8_IS_CONTINUATION(*((p)+1)))) \
2049 ? macro(EIGHT_BIT_UTF8_TO_NATIVE(*(p), *((p)+1))) \
2050 : (_force_out_malformed_utf8_message( \
2051 (U8 *) (p), (U8 *) (e), 0, 1), 0)) \
2052 : above_latin1))
2053
2054#define _generic_LC_swash_utf8_safe(macro, classnum, p, e) \
2055 _generic_LC_utf8_safe(macro, p, e, \
2056 _is_utf8_FOO_with_len(classnum, p, e))
2057
2058#define _generic_LC_func_utf8_safe(macro, above_latin1, p, e) \
2059 _generic_LC_utf8_safe(macro, p, e, above_latin1(p, e))
2060
2061#define _generic_LC_non_swash_utf8_safe(classnum, above_latin1, p, e) \
2062 _generic_LC_utf8_safe(classnum, p, e, \
2063 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
2064 ? (_force_out_malformed_utf8_message( \
2065 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
2066 : above_latin1(p)))
2067
2068#define isALPHANUMERIC_LC_utf8_safe(p, e) \
2069 _generic_LC_swash_utf8_safe(isALPHANUMERIC_LC, \
2070 _CC_ALPHANUMERIC, p, e)
2071#define isALPHA_LC_utf8_safe(p, e) \
2072 _generic_LC_swash_utf8_safe(isALPHA_LC, _CC_ALPHA, p, e)
2073#define isASCII_LC_utf8_safe(p, e) \
2074 (__ASSERT_(_utf8_safe_assert(p, e)) isASCII_LC(*(p)))
2075#define isBLANK_LC_utf8_safe(p, e) \
2076 _generic_LC_non_swash_utf8_safe(isBLANK_LC, is_HORIZWS_high, p, e)
2077#define isCNTRL_LC_utf8_safe(p, e) \
2078 _generic_LC_utf8_safe(isCNTRL_LC, p, e, 0)
2079#define isDIGIT_LC_utf8_safe(p, e) \
2080 _generic_LC_swash_utf8_safe(isDIGIT_LC, _CC_DIGIT, p, e)
2081#define isGRAPH_LC_utf8_safe(p, e) \
2082 _generic_LC_swash_utf8_safe(isGRAPH_LC, _CC_GRAPH, p, e)
2083#define isIDCONT_LC_utf8_safe(p, e) \
2084 _generic_LC_func_utf8_safe(isIDCONT_LC, \
2085 _is_utf8_perl_idcont_with_len, p, e)
2086#define isIDFIRST_LC_utf8_safe(p, e) \
2087 _generic_LC_func_utf8_safe(isIDFIRST_LC, \
2088 _is_utf8_perl_idstart_with_len, p, e)
2089#define isLOWER_LC_utf8_safe(p, e) \
2090 _generic_LC_swash_utf8_safe(isLOWER_LC, _CC_LOWER, p, e)
2091#define isPRINT_LC_utf8_safe(p, e) \
2092 _generic_LC_swash_utf8_safe(isPRINT_LC, _CC_PRINT, p, e)
2093#define isPSXSPC_LC_utf8_safe(p, e) isSPACE_LC_utf8_safe(p, e)
2094#define isPUNCT_LC_utf8_safe(p, e) \
2095 _generic_LC_swash_utf8_safe(isPUNCT_LC, _CC_PUNCT, p, e)
2096#define isSPACE_LC_utf8_safe(p, e) \
2097 _generic_LC_non_swash_utf8_safe(isSPACE_LC, is_XPERLSPACE_high, p, e)
2098#define isUPPER_LC_utf8_safe(p, e) \
2099 _generic_LC_swash_utf8_safe(isUPPER_LC, _CC_UPPER, p, e)
2100#define isWORDCHAR_LC_utf8_safe(p, e) \
2101 _generic_LC_swash_utf8_safe(isWORDCHAR_LC, _CC_WORDCHAR, p, e)
2102#define isXDIGIT_LC_utf8_safe(p, e) \
2103 _generic_LC_non_swash_utf8_safe(isXDIGIT_LC, is_XDIGIT_high, p, e)
aaa51d5e 2104
fbc19f27
KW
2105/* Macros for backwards compatibility and for completeness when the ASCII and
2106 * Latin1 values are identical */
b7d90381
KW
2107#define isALPHAU(c) isALPHA_L1(c)
2108#define isDIGIT_L1(c) isDIGIT_A(c)
2109#define isOCTAL(c) isOCTAL_A(c)
2110#define isOCTAL_L1(c) isOCTAL_A(c)
2111#define isXDIGIT_L1(c) isXDIGIT_A(c)
2112#define isALNUM(c) isWORDCHAR(c)
2113#define isALNUMU(c) isWORDCHAR_L1(c)
2114#define isALNUM_LC(c) isWORDCHAR_LC(c)
2115#define isALNUM_uni(c) isWORDCHAR_uni(c)
2e28f0b9 2116#define isALNUM_LC_uvchr(c) isWORDCHAR_LC_uvchr(c)
b7d90381
KW
2117#define isALNUM_utf8(p) isWORDCHAR_utf8(p)
2118#define isALNUM_LC_utf8(p) isWORDCHAR_LC_utf8(p)
2119#define isALNUMC_A(c) isALPHANUMERIC_A(c) /* Mnemonic: "C's alnum" */
2120#define isALNUMC_L1(c) isALPHANUMERIC_L1(c)
2121#define isALNUMC(c) isALPHANUMERIC(c)
2122#define isALNUMC_LC(c) isALPHANUMERIC_LC(c)
2123#define isALNUMC_uni(c) isALPHANUMERIC_uni(c)
15861f94 2124#define isALNUMC_LC_uvchr(c) isALPHANUMERIC_LC_uvchr(c)
b7d90381 2125#define isALNUMC_utf8(p) isALPHANUMERIC_utf8(p)
15861f94 2126#define isALNUMC_LC_utf8(p) isALPHANUMERIC_LC_utf8(p)
fbc19f27 2127
2bd1cbf6
KW
2128/* On EBCDIC platforms, CTRL-@ is 0, CTRL-A is 1, etc, just like on ASCII,
2129 * except that they don't necessarily mean the same characters, e.g. CTRL-D is
2130 * 4 on both systems, but that is EOT on ASCII; ST on EBCDIC.
2131 * '?' is special-cased on EBCDIC to APC, which is the control there that is
2132 * the outlier from the block that contains the other controls, just like
2133 * toCTRL('?') on ASCII yields DEL, the control that is the outlier from the C0
2134 * block. If it weren't special cased, it would yield a non-control.
88794300
KW
2135 * The conversion works both ways, so toCTRL('D') is 4, and toCTRL(4) is D,
2136 * etc. */
2bd1cbf6 2137#ifndef EBCDIC
75763b3a 2138# define toCTRL(c) (__ASSERT_(FITS_IN_8_BITS(c)) toUPPER(((U8)(c))) ^ 64)
2bd1cbf6 2139#else
75763b3a
KW
2140# define toCTRL(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
2141 ((isPRINT_A(c)) \
2142 ? (UNLIKELY((c) == '?') \
2143 ? QUESTION_MARK_CTRL \
2144 : (NATIVE_TO_LATIN1(toUPPER((U8) (c))) ^ 64)) \
2145 : (UNLIKELY((c) == QUESTION_MARK_CTRL) \
2146 ? '?' \
2147 : (LATIN1_TO_NATIVE(((U8) (c)) ^ 64)))))
2bd1cbf6 2148#endif
bbce6d69 2149
dea28490
JJ
2150/* Line numbers are unsigned, 32 bits. */
2151typedef U32 line_t;
e5dcd934 2152#define NOLINE ((line_t) 4294967295UL) /* = FFFFFFFF */
378cc40b 2153
91152fc1
DG
2154/* Helpful alias for version prescan */
2155#define is_LAX_VERSION(a,b) \
2156 (a != Perl_prescan_version(aTHX_ a, FALSE, b, NULL, NULL, NULL, NULL))
2157
2158#define is_STRICT_VERSION(a,b) \
2159 (a != Perl_prescan_version(aTHX_ a, TRUE, b, NULL, NULL, NULL, NULL))
2160
2161#define BADVERSION(a,b,c) \
2162 if (b) { \
2163 *b = c; \
2164 } \
2165 return a;
8c52afec 2166
cb27eebd
KW
2167/* Converts a character known to represent a hexadecimal digit (0-9, A-F, or
2168 * a-f) to its numeric value. READ_XDIGIT's argument is a string pointer,
2169 * which is advanced. The input is validated only by an assert() in DEBUGGING
2170 * builds. In both ASCII and EBCDIC the last 4 bits of the digits are 0-9; and
2171 * the last 4 bits of A-F and a-f are 1-6, so adding 9 yields 10-15 */
2172#define XDIGIT_VALUE(c) (__ASSERT_(isXDIGIT(c)) (0xf & (isDIGIT(c) \
2173 ? (c) \
2174 : ((c) + 9))))
902008b8
KW
2175#define READ_XDIGIT(s) (__ASSERT_(isXDIGIT(*s)) (0xf & (isDIGIT(*(s)) \
2176 ? (*(s)++) \
2177 : (*(s)++ + 9))))
95a59cab 2178
cb27eebd
KW
2179/* Converts a character known to represent an octal digit (0-7) to its numeric
2180 * value. The input is validated only by an assert() in DEBUGGING builds. In
2181 * both ASCII and EBCDIC the last 3 bits of the octal digits range from 0-7. */
2182#define OCTAL_VALUE(c) (__ASSERT_(isOCTAL(c)) (7 & (c)))
2183
305b8651
KW
2184/* Efficiently returns a boolean as to if two native characters are equivalent
2185 * case-insenstively. At least one of the characters must be one of [A-Za-z];
2186 * the ALPHA in the name is to remind you of that. This is asserted() in
2187 * DEBUGGING builds. Because [A-Za-z] are invariant under UTF-8, this macro
2188 * works (on valid input) for both non- and UTF-8-encoded bytes.
2189 *
2190 * When one of the inputs is a compile-time constant and gets folded by the
2191 * compiler, this reduces to an AND and a TEST. On both EBCDIC and ASCII
2192 * machines, 'A' and 'a' differ by a single bit; the same with the upper and
2193 * lower case of all other ASCII-range alphabetics. On ASCII platforms, they
96ca48da
KW
2194 * are 32 apart; on EBCDIC, they are 64. At compile time, this uses an
2195 * exclusive 'or' to find that bit and then inverts it to form a mask, with
2196 * just a single 0, in the bit position where the upper- and lowercase differ.
2197 * */
305b8651
KW
2198#define isALPHA_FOLD_EQ(c1, c2) \
2199 (__ASSERT_(isALPHA_A(c1) || isALPHA_A(c2)) \
2200 ((c1) & ~('A' ^ 'a')) == ((c2) & ~('A' ^ 'a')))
2201#define isALPHA_FOLD_NE(c1, c2) (! isALPHA_FOLD_EQ((c1), (c2)))
2202
8e84507e 2203/*
ccfc67b7
JH
2204=head1 Memory Management
2205
a02a5408 2206=for apidoc Am|void|Newx|void* ptr|int nitems|type
954c1994
GS
2207The XSUB-writer's interface to the C C<malloc> function.
2208
596f7718 2209Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2210
c5008215
JC
2211In 5.9.3, Newx() and friends replace the older New() API, and drops
2212the first parameter, I<x>, a debug aid which allowed callers to identify
37b8b4c9 2213themselves. This aid has been superseded by a new build option,
d10b4965 2214PERL_MEM_LOG (see L<perlhacktips/PERL_MEM_LOG>). The older API is still
c5008215
JC
2215there for use in XS modules supporting older perls.
2216
a02a5408 2217=for apidoc Am|void|Newxc|void* ptr|int nitems|type|cast
954c1994 2218The XSUB-writer's interface to the C C<malloc> function, with
fbe13c60 2219cast. See also C<L</Newx>>.
954c1994 2220
596f7718 2221Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2222
a02a5408 2223=for apidoc Am|void|Newxz|void* ptr|int nitems|type
954c1994 2224The XSUB-writer's interface to the C C<malloc> function. The allocated
fbe13c60 2225memory is zeroed with C<memzero>. See also C<L</Newx>>.
a02a5408 2226
596f7718 2227Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2228
954c1994
GS
2229=for apidoc Am|void|Renew|void* ptr|int nitems|type
2230The XSUB-writer's interface to the C C<realloc> function.
2231
596f7718 2232Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2233
954c1994
GS
2234=for apidoc Am|void|Renewc|void* ptr|int nitems|type|cast
2235The XSUB-writer's interface to the C C<realloc> function, with
2236cast.
2237
596f7718 2238Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2239
49b8b560 2240=for apidoc Am|void|Safefree|void* ptr
954c1994
GS
2241The XSUB-writer's interface to the C C<free> function.
2242
596f7718 2243This should B<ONLY> be used on memory obtained using L</"Newx"> and friends.
0d7b2759 2244
954c1994
GS
2245=for apidoc Am|void|Move|void* src|void* dest|int nitems|type
2246The XSUB-writer's interface to the C C<memmove> function. The C<src> is the
926bb54c 2247source, C<dest> is the destination, C<nitems> is the number of items, and
fbe13c60 2248C<type> is the type. Can do overlapping moves. See also C<L</Copy>>.
954c1994 2249
e90e2364 2250=for apidoc Am|void *|MoveD|void* src|void* dest|int nitems|type
796b6530 2251Like C<Move> but returns C<dest>. Useful
72d33970 2252for encouraging compilers to tail-call
e90e2364
NC
2253optimise.
2254
954c1994
GS
2255=for apidoc Am|void|Copy|void* src|void* dest|int nitems|type
2256The XSUB-writer's interface to the C C<memcpy> function. The C<src> is the
926bb54c 2257source, C<dest> is the destination, C<nitems> is the number of items, and
fbe13c60 2258C<type> is the type. May fail on overlapping copies. See also C<L</Move>>.
954c1994 2259
e90e2364
NC
2260=for apidoc Am|void *|CopyD|void* src|void* dest|int nitems|type
2261
796b6530 2262Like C<Copy> but returns C<dest>. Useful
72d33970 2263for encouraging compilers to tail-call
e90e2364
NC
2264optimise.
2265
954c1994
GS
2266=for apidoc Am|void|Zero|void* dest|int nitems|type
2267
2268The XSUB-writer's interface to the C C<memzero> function. The C<dest> is the
2269destination, C<nitems> is the number of items, and C<type> is the type.
2270
e90e2364
NC
2271=for apidoc Am|void *|ZeroD|void* dest|int nitems|type
2272
72d33970
FC
2273Like C<Zero> but returns dest. Useful
2274for encouraging compilers to tail-call
e90e2364
NC
2275optimise.
2276
da5d8dbb 2277=for apidoc Am|void|StructCopy|type *src|type *dest|type
4375e838 2278This is an architecture-independent macro to copy one structure to another.
954c1994 2279
7e337ee0
JH
2280=for apidoc Am|void|PoisonWith|void* dest|int nitems|type|U8 byte
2281
2282Fill up memory with a byte pattern (a byte repeated over and over
2283again) that hopefully catches attempts to access uninitialized memory.
2284
2285=for apidoc Am|void|PoisonNew|void* dest|int nitems|type
2286
2287PoisonWith(0xAB) for catching access to allocated but uninitialized memory.
2288
1c12ffb4 2289=for apidoc Am|void|PoisonFree|void* dest|int nitems|type
7e337ee0
JH
2290
2291PoisonWith(0xEF) for catching access to freed memory.
2292
9965345d
JH
2293=for apidoc Am|void|Poison|void* dest|int nitems|type
2294
7e337ee0 2295PoisonWith(0xEF) for catching access to freed memory.
9965345d
JH
2296
2297=cut */
954c1994 2298
561b68a9
SH
2299/* Maintained for backwards-compatibility only. Use newSV() instead. */
2300#ifndef PERL_CORE
ff06c60c 2301#define NEWSV(x,len) newSV(len)
561b68a9 2302#endif
ff06c60c 2303
b7112dce 2304#define MEM_SIZE_MAX ((MEM_SIZE)-1)
19a94d75 2305
a500027b 2306#define _PERL_STRLEN_ROUNDUP_UNCHECKED(n) (((n) - 1 + PERL_STRLEN_ROUNDUP_QUANTUM) & ~((MEM_SIZE)PERL_STRLEN_ROUNDUP_QUANTUM - 1))
e6bdf523 2307
27d5b266 2308#ifdef PERL_MALLOC_WRAP
e6bdf523
DM
2309
2310/* This expression will be constant-folded at compile time. It checks
2311 * whether or not the type of the count n is so small (e.g. U8 or U16, or
2312 * U32 on 64-bit systems) that there's no way a wrap-around could occur.
2313 * As well as avoiding the need for a run-time check in some cases, it's
2314 * designed to avoid compiler warnings like:
2315 * comparison is always false due to limited range of data type
73e8ff00
DM
2316 * It's mathematically equivalent to
2317 * max(n) * sizeof(t) > MEM_SIZE_MAX
e6bdf523
DM
2318 */
2319
2320# define _MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) \
445198b9
LM
2321 ( sizeof(MEM_SIZE) < sizeof(n) \
2322 || sizeof(t) > ((MEM_SIZE)1 << 8*(sizeof(MEM_SIZE) - sizeof(n))))
e6bdf523 2323
88f9f128 2324/* This is written in a slightly odd way to avoid various spurious
d98e5cde
DM
2325 * compiler warnings. We *want* to write the expression as
2326 * _MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) && (n > C)
2327 * (for some compile-time constant C), but even when the LHS
2328 * constant-folds to false at compile-time, g++ insists on emitting
2329 * warnings about the RHS (e.g. "comparison is always false"), so instead
2330 * we write it as
e6bdf523 2331 *
d98e5cde 2332 * (cond ? n : X) > C
88f9f128 2333 *
d98e5cde
DM
2334 * where X is a constant with X > C always false. Choosing a value for X
2335 * is tricky. If 0, some compilers will complain about 0 > C always being
2336 * false; if 1, Coverity complains when n happens to be the constant value
2337 * '1', that cond ? 1 : 1 has the same value on both branches; so use C
2338 * for X and hope that nothing else whines.
e6bdf523
DM
2339 */
2340
2341# define _MEM_WRAP_WILL_WRAP(n,t) \
88f9f128
DM
2342 ((_MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) ? (MEM_SIZE)(n) : \
2343 MEM_SIZE_MAX/sizeof(t)) > MEM_SIZE_MAX/sizeof(t))
e6bdf523
DM
2344
2345# define MEM_WRAP_CHECK(n,t) \
2346 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2347 && (croak_memory_wrap(),0))
2348
2349# define MEM_WRAP_CHECK_1(n,t,a) \
2350 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2351 && (Perl_croak_nocontext("%s",(a)),0))
2352
8b44ba4c 2353#define MEM_WRAP_CHECK_(n,t) MEM_WRAP_CHECK(n,t),
27d5b266 2354
a500027b 2355#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
2356#else
2357
410319be
NC
2358#define MEM_WRAP_CHECK(n,t)
2359#define MEM_WRAP_CHECK_1(n,t,a)
2360#define MEM_WRAP_CHECK_2(n,t,a,b)
8b44ba4c
NC
2361#define MEM_WRAP_CHECK_(n,t)
2362
a500027b 2363#define PERL_STRLEN_ROUNDUP(n) _PERL_STRLEN_ROUNDUP_UNCHECKED(n)
27d5b266 2364
1936d2a7 2365#endif
8b44ba4c 2366
fe4f188c 2367#ifdef PERL_MEM_LOG
46c6c7e2 2368/*
9f653bb5 2369 * If PERL_MEM_LOG is defined, all Newx()s, Renew()s, and Safefree()s
46c6c7e2
JH
2370 * go through functions, which are handy for debugging breakpoints, but
2371 * which more importantly get the immediate calling environment (file and
e352bcff
JH
2372 * line number, and C function name if available) passed in. This info can
2373 * then be used for logging the calls, for which one gets a sample
73d1d973 2374 * implementation unless -DPERL_MEM_LOG_NOIMPL is also defined.
3609ea0d 2375 *
46c6c7e2 2376 * Known problems:
94e892a6 2377 * - not all memory allocs get logged, only those
46c6c7e2 2378 * that go through Newx() and derivatives (while all
94e892a6 2379 * Safefrees do get logged)
46c6c7e2
JH
2380 * - __FILE__ and __LINE__ do not work everywhere
2381 * - __func__ or __FUNCTION__ even less so
2382 * - I think more goes on after the perlio frees but
2383 * the thing is that STDERR gets closed (as do all
2384 * the file descriptors)
2385 * - no deeper calling stack than the caller of the Newx()
2386 * or the kind, but do I look like a C reflection/introspection
2387 * utility to you?
2388 * - the function prototypes for the logging functions
2389 * probably should maybe be somewhere else than handy.h
2390 * - one could consider inlining (macrofying) the logging
2391 * for speed, but I am too lazy
2392 * - one could imagine recording the allocations in a hash,
2393 * (keyed by the allocation address?), and maintain that
2394 * through reallocs and frees, but how to do that without
2395 * any News() happening...?
73d1d973 2396 * - lots of -Ddefines to get useful/controllable output
b953482e 2397 * - lots of ENV reads
46c6c7e2
JH
2398 */
2399
0b0ab801 2400# ifdef PERL_CORE
73d1d973 2401# ifndef PERL_MEM_LOG_NOIMPL
0b0ab801
MHM
2402enum mem_log_type {
2403 MLT_ALLOC,
2404 MLT_REALLOC,
d7a2c63c
MHM
2405 MLT_FREE,
2406 MLT_NEW_SV,
2407 MLT_DEL_SV
0b0ab801
MHM
2408};
2409# endif
12754f92 2410# if defined(PERL_IN_SV_C) /* those are only used in sv.c */
d7a2c63c
MHM
2411void Perl_mem_log_new_sv(const SV *sv, const char *filename, const int linenumber, const char *funcname);
2412void Perl_mem_log_del_sv(const SV *sv, const char *filename, const int linenumber, const char *funcname);
12754f92 2413# endif
0b0ab801
MHM
2414# endif
2415
fe4f188c
JH
2416#endif
2417
2418#ifdef PERL_MEM_LOG
d1401ee9
MHM
2419#define MEM_LOG_ALLOC(n,t,a) Perl_mem_log_alloc(n,sizeof(t),STRINGIFY(t),a,__FILE__,__LINE__,FUNCTION__)
2420#define MEM_LOG_REALLOC(n,t,v,a) Perl_mem_log_realloc(n,sizeof(t),STRINGIFY(t),v,a,__FILE__,__LINE__,FUNCTION__)
46c6c7e2 2421#define MEM_LOG_FREE(a) Perl_mem_log_free(a,__FILE__,__LINE__,FUNCTION__)
fe4f188c
JH
2422#endif
2423
2424#ifndef MEM_LOG_ALLOC
2425#define MEM_LOG_ALLOC(n,t,a) (a)
2426#endif
2427#ifndef MEM_LOG_REALLOC
2428#define MEM_LOG_REALLOC(n,t,v,a) (a)
2429#endif
2430#ifndef MEM_LOG_FREE
2431#define MEM_LOG_FREE(a) (a)
2432#endif
2433
d1401ee9
MHM
2434#define Newx(v,n,t) (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_ALLOC(n,t,safemalloc((MEM_SIZE)((n)*sizeof(t))))))
2435#define Newxc(v,n,t,c) (v = (MEM_WRAP_CHECK_(n,t) (c*)MEM_LOG_ALLOC(n,t,safemalloc((MEM_SIZE)((n)*sizeof(t))))))
2436#define Newxz(v,n,t) (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_ALLOC(n,t,safecalloc((n),sizeof(t)))))
a6f6820f
NC
2437
2438#ifndef PERL_CORE
a02a5408
JC
2439/* pre 5.9.x compatibility */
2440#define New(x,v,n,t) Newx(v,n,t)
2441#define Newc(x,v,n,t,c) Newxc(v,n,t,c)
4541904d 2442#define Newz(x,v,n,t) Newxz(v,n,t)
a6f6820f 2443#endif
a02a5408 2444
ff68c719 2445#define Renew(v,n,t) \
d1401ee9 2446 (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_REALLOC(n,t,v,saferealloc((Malloc_t)(v),(MEM_SIZE)((n)*sizeof(t))))))
ff68c719 2447#define Renewc(v,n,t,c) \
d1401ee9 2448 (v = (MEM_WRAP_CHECK_(n,t) (c*)MEM_LOG_REALLOC(n,t,v,saferealloc((Malloc_t)(v),(MEM_SIZE)((n)*sizeof(t))))))
94010e71
NC
2449
2450#ifdef PERL_POISON
2451#define Safefree(d) \
06c0cc96 2452 ((d) ? (void)(safefree(MEM_LOG_FREE((Malloc_t)(d))), Poison(&(d), 1, Malloc_t)) : (void) 0)
94010e71 2453#else
fe4f188c 2454#define Safefree(d) safefree(MEM_LOG_FREE((Malloc_t)(d)))
94010e71 2455#endif
55497cff 2456
dbb57106
YO
2457/* assert that a valid ptr has been supplied - use this instead of assert(ptr) *
2458 * as it handles cases like constant string arguments without throwing warnings *
2459 * the cast is required, as is the inequality check, to avoid warnings */
45908e4d 2460#define perl_assert_ptr(p) assert( ((void*)(p)) != 0 )
55497cff 2461
45908e4d
YO
2462
2463#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)))
2464#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)))
2465#define Zero(d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), (void)memzero((char*)(d), (n) * sizeof(t)))
2466
2467#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)))
2468#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 2469#define ZeroD(d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), memzero((char*)(d), (n) * sizeof(t)))
e90e2364 2470
7e337ee0
JH
2471#define PoisonWith(d,n,t,b) (MEM_WRAP_CHECK_(n,t) (void)memset((char*)(d), (U8)(b), (n) * sizeof(t)))
2472#define PoisonNew(d,n,t) PoisonWith(d,n,t,0xAB)
2473#define PoisonFree(d,n,t) PoisonWith(d,n,t,0xEF)
2474#define Poison(d,n,t) PoisonFree(d,n,t)
27d5b266 2475
caa674f3
DD
2476#ifdef PERL_POISON
2477# define PERL_POISON_EXPR(x) x
2478#else
2479# define PERL_POISON_EXPR(x)
2480#endif
2481
ff68c719 2482#define StructCopy(s,d,t) (*((t*)(d)) = *((t*)(s)))
2cc61e15 2483
c3caa5c3
JH
2484/* C_ARRAY_LENGTH is the number of elements in the C array (so you
2485 * want your zero-based indices to be less than but not equal to).
2486 *
2487 * C_ARRAY_END is one past the last: half-open/half-closed range,
2488 * not last-inclusive range. */
622913ab 2489#define C_ARRAY_LENGTH(a) (sizeof(a)/sizeof((a)[0]))
c3caa5c3 2490#define C_ARRAY_END(a) ((a) + C_ARRAY_LENGTH(a))
622913ab 2491
2cc61e15
DD
2492#ifdef NEED_VA_COPY
2493# ifdef va_copy
2494# define Perl_va_copy(s, d) va_copy(d, s)
07798b17
AC
2495# elif defined(__va_copy)
2496# define Perl_va_copy(s, d) __va_copy(d, s)
2cc61e15 2497# else
07798b17 2498# define Perl_va_copy(s, d) Copy(s, d, 1, va_list)
2cc61e15
DD
2499# endif
2500#endif
2501
472d47bc
SB
2502/* convenience debug macros */
2503#ifdef USE_ITHREADS
2504#define pTHX_FORMAT "Perl interpreter: 0x%p"
2505#define pTHX__FORMAT ", Perl interpreter: 0x%p"
f54cb97a
AL
2506#define pTHX_VALUE_ (void *)my_perl,
2507#define pTHX_VALUE (void *)my_perl
2508#define pTHX__VALUE_ ,(void *)my_perl,
2509#define pTHX__VALUE ,(void *)my_perl
472d47bc 2510#else
3609ea0d 2511#define pTHX_FORMAT
472d47bc 2512#define pTHX__FORMAT
3609ea0d 2513#define pTHX_VALUE_
472d47bc 2514#define pTHX_VALUE
3609ea0d 2515#define pTHX__VALUE_
472d47bc
SB
2516#define pTHX__VALUE
2517#endif /* USE_ITHREADS */
3609ea0d 2518
2acdbac1
NC
2519/* Perl_deprecate was not part of the public API, and did not have a deprecate()
2520 shortcut macro defined without -DPERL_CORE. Neither codesearch.google.com nor
2521 CPAN::Unpack show any users outside the core. */
2522#ifdef PERL_CORE
dc6e8de0
A
2523# define deprecate(s) Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2524 "Use of " s " is deprecated")
c9680906
A
2525# define deprecate_disappears_in(when,message) \
2526 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2527 message ", and will disappear in Perl " when)
ac641426
A
2528# define deprecate_fatal_in(when,message) \
2529 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2530 message ". Its use will be fatal in Perl " when)
2acdbac1
NC
2531#endif
2532
dfff4baf
BF
2533/* Internal macros to deal with gids and uids */
2534#ifdef PERL_CORE
2535
2536# if Uid_t_size > IVSIZE
2537# define sv_setuid(sv, uid) sv_setnv((sv), (NV)(uid))
2538# define SvUID(sv) SvNV(sv)
07798b17
AC
2539# elif Uid_t_sign <= 0
2540# define sv_setuid(sv, uid) sv_setiv((sv), (IV)(uid))
2541# define SvUID(sv) SvIV(sv)
dfff4baf 2542# else
07798b17
AC
2543# define sv_setuid(sv, uid) sv_setuv((sv), (UV)(uid))
2544# define SvUID(sv) SvUV(sv)
dfff4baf
BF
2545# endif /* Uid_t_size */
2546
2547# if Gid_t_size > IVSIZE
2548# define sv_setgid(sv, gid) sv_setnv((sv), (NV)(gid))
2549# define SvGID(sv) SvNV(sv)
07798b17
AC
2550# elif Gid_t_sign <= 0
2551# define sv_setgid(sv, gid) sv_setiv((sv), (IV)(gid))
2552# define SvGID(sv) SvIV(sv)
dfff4baf 2553# else
07798b17
AC
2554# define sv_setgid(sv, gid) sv_setuv((sv), (UV)(gid))
2555# define SvGID(sv) SvUV(sv)
dfff4baf
BF
2556# endif /* Gid_t_size */
2557
2558#endif
2559
6a5bc5ac 2560#endif /* PERL_HANDY_H_ */
9d745869 2561
e9a8c099 2562/*
14d04a33 2563 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 2564 */