This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Update perlguts for NULL in AVs
[perl5.git] / pod / perlguts.pod
CommitLineData
a0d0e21e
LW
1=head1 NAME
2
954c1994 3perlguts - Introduction to the Perl API
a0d0e21e
LW
4
5=head1 DESCRIPTION
6
b3b6085d 7This document attempts to describe how to use the Perl API, as well as
06f6df17 8to provide some info on the basic workings of the Perl core. It is far
b3b6085d
PP
9from complete and probably contains many errors. Please refer any
10questions or comments to the author below.
a0d0e21e 11
0a753a76
PP
12=head1 Variables
13
5f05dabc 14=head2 Datatypes
a0d0e21e
LW
15
16Perl has three typedefs that handle Perl's three main data types:
17
18 SV Scalar Value
19 AV Array Value
20 HV Hash Value
21
d1b91892 22Each typedef has specific routines that manipulate the various data types.
a0d0e21e
LW
23
24=head2 What is an "IV"?
25
954c1994 26Perl uses a special typedef IV which is a simple signed integer type that is
5f05dabc 27guaranteed to be large enough to hold a pointer (as well as an integer).
954c1994 28Additionally, there is the UV, which is simply an unsigned IV.
a0d0e21e 29
d1b91892 30Perl also uses two special typedefs, I32 and I16, which will always be at
954c1994 31least 32-bits and 16-bits long, respectively. (Again, there are U32 and U16,
20dbd849
NC
32as well.) They will usually be exactly 32 and 16 bits long, but on Crays
33they will both be 64 bits.
a0d0e21e 34
54310121 35=head2 Working with SVs
a0d0e21e 36
20dbd849
NC
37An SV can be created and loaded with one command. There are five types of
38values that can be loaded: an integer value (IV), an unsigned integer
39value (UV), a double (NV), a string (PV), and another scalar (SV).
61984ee1
KW
40("PV" stands for "Pointer Value". You might think that it is misnamed
41because it is described as pointing only to strings. However, it is
d6605d24
KW
42possible to have it point to other things For example, it could point
43to an array of UVs. But,
61984ee1
KW
44using it for non-strings requires care, as the underlying assumption of
45much of the internals is that PVs are just for strings. Often, for
46example, a trailing NUL is tacked on automatically. The non-string use
47is documented only in this paragraph.)
a0d0e21e 48
20dbd849 49The seven routines are:
a0d0e21e
LW
50
51 SV* newSViv(IV);
20dbd849 52 SV* newSVuv(UV);
a0d0e21e 53 SV* newSVnv(double);
06f6df17
RGS
54 SV* newSVpv(const char*, STRLEN);
55 SV* newSVpvn(const char*, STRLEN);
46fc3d4c 56 SV* newSVpvf(const char*, ...);
a0d0e21e
LW
57 SV* newSVsv(SV*);
58
06f6df17
RGS
59C<STRLEN> is an integer type (Size_t, usually defined as size_t in
60F<config.h>) guaranteed to be large enough to represent the size of
61any string that perl can handle.
62
3bf17896 63In the unlikely case of a SV requiring more complex initialization, you
06f6df17
RGS
64can create an empty SV with newSV(len). If C<len> is 0 an empty SV of
65type NULL is returned, else an SV of type PV is returned with len + 1 (for
66the NUL) bytes of storage allocated, accessible via SvPVX. In both cases
da8c5729 67the SV has the undef value.
20dbd849 68
06f6df17 69 SV *sv = newSV(0); /* no storage allocated */
a9b0660e
KW
70 SV *sv = newSV(10); /* 10 (+1) bytes of uninitialised storage
71 * allocated */
20dbd849 72
06f6df17 73To change the value of an I<already-existing> SV, there are eight routines:
a0d0e21e
LW
74
75 void sv_setiv(SV*, IV);
deb3007b 76 void sv_setuv(SV*, UV);
a0d0e21e 77 void sv_setnv(SV*, double);
08105a92 78 void sv_setpv(SV*, const char*);
06f6df17 79 void sv_setpvn(SV*, const char*, STRLEN)
46fc3d4c 80 void sv_setpvf(SV*, const char*, ...);
a9b0660e
KW
81 void sv_vsetpvfn(SV*, const char*, STRLEN, va_list *,
82 SV **, I32, bool *);
a0d0e21e
LW
83 void sv_setsv(SV*, SV*);
84
85Notice that you can choose to specify the length of the string to be
9da1e3b5
MUN
86assigned by using C<sv_setpvn>, C<newSVpvn>, or C<newSVpv>, or you may
87allow Perl to calculate the length by using C<sv_setpv> or by specifying
880 as the second argument to C<newSVpv>. Be warned, though, that Perl will
89determine the string's length by using C<strlen>, which depends on the
a9b0660e
KW
90string terminating with a NUL character, and not otherwise containing
91NULs.
9abd00ed
GS
92
93The arguments of C<sv_setpvf> are processed like C<sprintf>, and the
94formatted output becomes the value.
95
328bf373 96C<sv_vsetpvfn> is an analogue of C<vsprintf>, but it allows you to specify
9abd00ed
GS
97either a pointer to a variable argument list or the address and length of
98an array of SVs. The last argument points to a boolean; on return, if that
99boolean is true, then locale-specific information has been used to format
c2611fb3 100the string, and the string's contents are therefore untrustworthy (see
9abd00ed
GS
101L<perlsec>). This pointer may be NULL if that information is not
102important. Note that this function requires you to specify the length of
103the format.
104
9da1e3b5
MUN
105The C<sv_set*()> functions are not generic enough to operate on values
106that have "magic". See L<Magic Virtual Tables> later in this document.
a0d0e21e 107
a3cb178b
GS
108All SVs that contain strings should be terminated with a NUL character.
109If it is not NUL-terminated there is a risk of
5f05dabc
PP
110core dumps and corruptions from code which passes the string to C
111functions or system calls which expect a NUL-terminated string.
112Perl's own functions typically add a trailing NUL for this reason.
113Nevertheless, you should be very careful when you pass a string stored
114in an SV to a C function or system call.
115
a0d0e21e
LW
116To access the actual value that an SV points to, you can use the macros:
117
118 SvIV(SV*)
954c1994 119 SvUV(SV*)
a0d0e21e
LW
120 SvNV(SV*)
121 SvPV(SV*, STRLEN len)
1fa8b10d 122 SvPV_nolen(SV*)
a0d0e21e 123
954c1994 124which will automatically coerce the actual scalar type into an IV, UV, double,
a0d0e21e
LW
125or string.
126
127In the C<SvPV> macro, the length of the string returned is placed into the
1fa8b10d
JD
128variable C<len> (this is a macro, so you do I<not> use C<&len>). If you do
129not care what the length of the data is, use the C<SvPV_nolen> macro.
130Historically the C<SvPV> macro with the global variable C<PL_na> has been
131used in this case. But that can be quite inefficient because C<PL_na> must
132be accessed in thread-local storage in threaded Perl. In any case, remember
133that Perl allows arbitrary strings of data that may both contain NULs and
134might not be terminated by a NUL.
a0d0e21e 135
ce2f5d8f
KA
136Also remember that C doesn't allow you to safely say C<foo(SvPV(s, len),
137len);>. It might work with your compiler, but it won't work for everyone.
138Break this sort of statement up into separate assignments:
139
1aa6ea50
JC
140 SV *s;
141 STRLEN len;
61955433 142 char *ptr;
1aa6ea50
JC
143 ptr = SvPV(s, len);
144 foo(ptr, len);
ce2f5d8f 145
07fa94a1 146If you want to know if the scalar value is TRUE, you can use:
a0d0e21e
LW
147
148 SvTRUE(SV*)
149
150Although Perl will automatically grow strings for you, if you need to force
151Perl to allocate more memory for your SV, you can use the macro
152
153 SvGROW(SV*, STRLEN newlen)
154
155which will determine if more memory needs to be allocated. If so, it will
156call the function C<sv_grow>. Note that C<SvGROW> can only increase, not
5f05dabc 157decrease, the allocated memory of an SV and that it does not automatically
a9b0660e 158add space for the trailing NUL byte (perl's own string functions typically do
8ebc5c01 159C<SvGROW(sv, len + 1)>).
a0d0e21e
LW
160
161If you have an SV and want to know what kind of data Perl thinks is stored
162in it, you can use the following macros to check the type of SV you have.
163
164 SvIOK(SV*)
165 SvNOK(SV*)
166 SvPOK(SV*)
167
168You can get and set the current length of the string stored in an SV with
169the following macros:
170
171 SvCUR(SV*)
172 SvCUR_set(SV*, I32 val)
173
cb1a09d0
AD
174You can also get a pointer to the end of the string stored in the SV
175with the macro:
176
177 SvEND(SV*)
178
179But note that these last three macros are valid only if C<SvPOK()> is true.
a0d0e21e 180
d1b91892
AD
181If you want to append something to the end of string stored in an C<SV*>,
182you can use the following functions:
183
08105a92 184 void sv_catpv(SV*, const char*);
e65f3abd 185 void sv_catpvn(SV*, const char*, STRLEN);
46fc3d4c 186 void sv_catpvf(SV*, const char*, ...);
a9b0660e
KW
187 void sv_vcatpvfn(SV*, const char*, STRLEN, va_list *, SV **,
188 I32, bool);
d1b91892
AD
189 void sv_catsv(SV*, SV*);
190
191The first function calculates the length of the string to be appended by
192using C<strlen>. In the second, you specify the length of the string
46fc3d4c 193yourself. The third function processes its arguments like C<sprintf> and
9abd00ed
GS
194appends the formatted output. The fourth function works like C<vsprintf>.
195You can specify the address and length of an array of SVs instead of the
196va_list argument. The fifth function extends the string stored in the first
197SV with the string stored in the second SV. It also forces the second SV
198to be interpreted as a string.
199
200The C<sv_cat*()> functions are not generic enough to operate on values that
201have "magic". See L<Magic Virtual Tables> later in this document.
d1b91892 202
a0d0e21e
LW
203If you know the name of a scalar variable, you can get a pointer to its SV
204by using the following:
205
64ace3f8 206 SV* get_sv("package::varname", 0);
a0d0e21e
LW
207
208This returns NULL if the variable does not exist.
209
d1b91892 210If you want to know if this variable (or any other SV) is actually C<defined>,
a0d0e21e
LW
211you can call:
212
213 SvOK(SV*)
214
06f6df17 215The scalar C<undef> value is stored in an SV instance called C<PL_sv_undef>.
9adebda4
SB
216
217Its address can be used whenever an C<SV*> is needed. Make sure that
218you don't try to compare a random sv with C<&PL_sv_undef>. For example
219when interfacing Perl code, it'll work correctly for:
220
221 foo(undef);
222
223But won't work when called as:
224
225 $x = undef;
226 foo($x);
227
228So to repeat always use SvOK() to check whether an sv is defined.
229
230Also you have to be careful when using C<&PL_sv_undef> as a value in
231AVs or HVs (see L<AVs, HVs and undefined values>).
a0d0e21e 232
06f6df17
RGS
233There are also the two values C<PL_sv_yes> and C<PL_sv_no>, which contain
234boolean TRUE and FALSE values, respectively. Like C<PL_sv_undef>, their
235addresses can be used whenever an C<SV*> is needed.
a0d0e21e 236
9cde0e7f 237Do not be fooled into thinking that C<(SV *) 0> is the same as C<&PL_sv_undef>.
a0d0e21e
LW
238Take this code:
239
240 SV* sv = (SV*) 0;
241 if (I-am-to-return-a-real-value) {
242 sv = sv_2mortal(newSViv(42));
243 }
244 sv_setsv(ST(0), sv);
245
246This code tries to return a new SV (which contains the value 42) if it should
04343c6d 247return a real value, or undef otherwise. Instead it has returned a NULL
a0d0e21e 248pointer which, somewhere down the line, will cause a segmentation violation,
06f6df17
RGS
249bus error, or just weird results. Change the zero to C<&PL_sv_undef> in the
250first line and all will be well.
a0d0e21e
LW
251
252To free an SV that you've created, call C<SvREFCNT_dec(SV*)>. Normally this
3fe9a6f1 253call is not necessary (see L<Reference Counts and Mortality>).
a0d0e21e 254
94dde4fb
SC
255=head2 Offsets
256
257Perl provides the function C<sv_chop> to efficiently remove characters
258from the beginning of a string; you give it an SV and a pointer to
da75cd15 259somewhere inside the PV, and it discards everything before the
94dde4fb
SC
260pointer. The efficiency comes by means of a little hack: instead of
261actually removing the characters, C<sv_chop> sets the flag C<OOK>
262(offset OK) to signal to other functions that the offset hack is in
883bb8c0
KW
263effect, and it moves the PV pointer (called C<SvPVX>) forward
264by the number of bytes chopped off, and adjusts C<SvCUR> and C<SvLEN>
265accordingly. (A portion of the space between the old and new PV
266pointers is used to store the count of chopped bytes.)
94dde4fb
SC
267
268Hence, at this point, the start of the buffer that we allocated lives
269at C<SvPVX(sv) - SvIV(sv)> in memory and the PV pointer is pointing
270into the middle of this allocated storage.
271
272This is best demonstrated by example:
273
274 % ./perl -Ilib -MDevel::Peek -le '$a="12345"; $a=~s/.//; Dump($a)'
275 SV = PVIV(0x8128450) at 0x81340f0
276 REFCNT = 1
277 FLAGS = (POK,OOK,pPOK)
278 IV = 1 (OFFSET)
279 PV = 0x8135781 ( "1" . ) "2345"\0
280 CUR = 4
281 LEN = 5
282
283Here the number of bytes chopped off (1) is put into IV, and
284C<Devel::Peek::Dump> helpfully reminds us that this is an offset. The
285portion of the string between the "real" and the "fake" beginnings is
286shown in parentheses, and the values of C<SvCUR> and C<SvLEN> reflect
287the fake beginning, not the real one.
288
fe854a6f 289Something similar to the offset hack is performed on AVs to enable
319cef53
SC
290efficient shifting and splicing off the beginning of the array; while
291C<AvARRAY> points to the first element in the array that is visible from
292Perl, C<AvALLOC> points to the real start of the C array. These are
293usually the same, but a C<shift> operation can be carried out by
6de131f0 294increasing C<AvARRAY> by one and decreasing C<AvFILL> and C<AvMAX>.
319cef53
SC
295Again, the location of the real start of the C array only comes into
296play when freeing the array. See C<av_shift> in F<av.c>.
297
d1b91892 298=head2 What's Really Stored in an SV?
a0d0e21e
LW
299
300Recall that the usual method of determining the type of scalar you have is
5f05dabc 301to use C<Sv*OK> macros. Because a scalar can be both a number and a string,
d1b91892 302usually these macros will always return TRUE and calling the C<Sv*V>
a0d0e21e
LW
303macros will do the appropriate conversion of string to integer/double or
304integer/double to string.
305
306If you I<really> need to know if you have an integer, double, or string
307pointer in an SV, you can use the following three macros instead:
308
309 SvIOKp(SV*)
310 SvNOKp(SV*)
311 SvPOKp(SV*)
312
313These will tell you if you truly have an integer, double, or string pointer
d1b91892 314stored in your SV. The "p" stands for private.
a0d0e21e 315
da8c5729 316There are various ways in which the private and public flags may differ.
9e9796d6
JH
317For example, a tied SV may have a valid underlying value in the IV slot
318(so SvIOKp is true), but the data should be accessed via the FETCH
319routine rather than directly, so SvIOK is false. Another is when
d7f8936a 320numeric conversion has occurred and precision has been lost: only the
9e9796d6
JH
321private flag is set on 'lossy' values. So when an NV is converted to an
322IV with loss, SvIOKp, SvNOKp and SvNOK will be set, while SvIOK wont be.
323
07fa94a1 324In general, though, it's best to use the C<Sv*V> macros.
a0d0e21e 325
54310121 326=head2 Working with AVs
a0d0e21e 327
07fa94a1
JO
328There are two ways to create and load an AV. The first method creates an
329empty AV:
a0d0e21e
LW
330
331 AV* newAV();
332
54310121 333The second method both creates the AV and initially populates it with SVs:
a0d0e21e 334
c70927a6 335 AV* av_make(SSize_t num, SV **ptr);
a0d0e21e 336
5f05dabc 337The second argument points to an array containing C<num> C<SV*>'s. Once the
54310121 338AV has been created, the SVs can be destroyed, if so desired.
a0d0e21e 339
da8c5729 340Once the AV has been created, the following operations are possible on it:
a0d0e21e
LW
341
342 void av_push(AV*, SV*);
343 SV* av_pop(AV*);
344 SV* av_shift(AV*);
c70927a6 345 void av_unshift(AV*, SSize_t num);
a0d0e21e
LW
346
347These should be familiar operations, with the exception of C<av_unshift>.
348This routine adds C<num> elements at the front of the array with the C<undef>
349value. You must then use C<av_store> (described below) to assign values
350to these new elements.
351
352Here are some other functions:
353
c70927a6
FC
354 SSize_t av_top_index(AV*);
355 SV** av_fetch(AV*, SSize_t key, I32 lval);
356 SV** av_store(AV*, SSize_t key, SV* val);
a0d0e21e 357
dab460cd 358The C<av_top_index> function returns the highest index value in an array (just
5f05dabc
PP
359like $#array in Perl). If the array is empty, -1 is returned. The
360C<av_fetch> function returns the value at index C<key>, but if C<lval>
361is non-zero, then C<av_fetch> will store an undef value at that index.
04343c6d
GS
362The C<av_store> function stores the value C<val> at index C<key>, and does
363not increment the reference count of C<val>. Thus the caller is responsible
364for taking care of that, and if C<av_store> returns NULL, the caller will
365have to decrement the reference count to avoid a memory leak. Note that
366C<av_fetch> and C<av_store> both return C<SV**>'s, not C<SV*>'s as their
367return value.
d1b91892 368
da8c5729
MH
369A few more:
370
a0d0e21e 371 void av_clear(AV*);
a0d0e21e 372 void av_undef(AV*);
c70927a6 373 void av_extend(AV*, SSize_t key);
5f05dabc
PP
374
375The C<av_clear> function deletes all the elements in the AV* array, but
376does not actually delete the array itself. The C<av_undef> function will
377delete all the elements in the array plus the array itself. The
adc882cf
GS
378C<av_extend> function extends the array so that it contains at least C<key+1>
379elements. If C<key+1> is less than the currently allocated length of the array,
380then nothing is done.
a0d0e21e
LW
381
382If you know the name of an array variable, you can get a pointer to its AV
383by using the following:
384
cbfd0a87 385 AV* get_av("package::varname", 0);
a0d0e21e
LW
386
387This returns NULL if the variable does not exist.
388
04343c6d
GS
389See L<Understanding the Magic of Tied Hashes and Arrays> for more
390information on how to use the array access functions on tied arrays.
391
54310121 392=head2 Working with HVs
a0d0e21e
LW
393
394To create an HV, you use the following routine:
395
396 HV* newHV();
397
da8c5729 398Once the HV has been created, the following operations are possible on it:
a0d0e21e 399
08105a92
GS
400 SV** hv_store(HV*, const char* key, U32 klen, SV* val, U32 hash);
401 SV** hv_fetch(HV*, const char* key, U32 klen, I32 lval);
a0d0e21e 402
5f05dabc
PP
403The C<klen> parameter is the length of the key being passed in (Note that
404you cannot pass 0 in as a value of C<klen> to tell Perl to measure the
405length of the key). The C<val> argument contains the SV pointer to the
54310121 406scalar being stored, and C<hash> is the precomputed hash value (zero if
5f05dabc
PP
407you want C<hv_store> to calculate it for you). The C<lval> parameter
408indicates whether this fetch is actually a part of a store operation, in
409which case a new undefined value will be added to the HV with the supplied
410key and C<hv_fetch> will return as if the value had already existed.
a0d0e21e 411
5f05dabc
PP
412Remember that C<hv_store> and C<hv_fetch> return C<SV**>'s and not just
413C<SV*>. To access the scalar value, you must first dereference the return
414value. However, you should check to make sure that the return value is
415not NULL before dereferencing it.
a0d0e21e 416
da8c5729
MH
417The first of these two functions checks if a hash table entry exists, and the
418second deletes it.
a0d0e21e 419
08105a92
GS
420 bool hv_exists(HV*, const char* key, U32 klen);
421 SV* hv_delete(HV*, const char* key, U32 klen, I32 flags);
a0d0e21e 422
5f05dabc
PP
423If C<flags> does not include the C<G_DISCARD> flag then C<hv_delete> will
424create and return a mortal copy of the deleted value.
425
a0d0e21e
LW
426And more miscellaneous functions:
427
428 void hv_clear(HV*);
a0d0e21e 429 void hv_undef(HV*);
5f05dabc
PP
430
431Like their AV counterparts, C<hv_clear> deletes all the entries in the hash
432table but does not actually delete the hash table. The C<hv_undef> deletes
433both the entries and the hash table itself.
a0d0e21e 434
a9b0660e 435Perl keeps the actual data in a linked list of structures with a typedef of HE.
d1b91892
AD
436These contain the actual key and value pointers (plus extra administrative
437overhead). The key is a string pointer; the value is an C<SV*>. However,
438once you have an C<HE*>, to get the actual key and value, use the routines
439specified below.
440
a0d0e21e
LW
441 I32 hv_iterinit(HV*);
442 /* Prepares starting point to traverse hash table */
443 HE* hv_iternext(HV*);
444 /* Get the next entry, and return a pointer to a
445 structure that has both the key and value */
446 char* hv_iterkey(HE* entry, I32* retlen);
447 /* Get the key from an HE structure and also return
448 the length of the key string */
cb1a09d0 449 SV* hv_iterval(HV*, HE* entry);
d1be9408 450 /* Return an SV pointer to the value of the HE
a0d0e21e 451 structure */
cb1a09d0 452 SV* hv_iternextsv(HV*, char** key, I32* retlen);
d1b91892
AD
453 /* This convenience routine combines hv_iternext,
454 hv_iterkey, and hv_iterval. The key and retlen
455 arguments are return values for the key and its
456 length. The value is returned in the SV* argument */
a0d0e21e
LW
457
458If you know the name of a hash variable, you can get a pointer to its HV
459by using the following:
460
6673a63c 461 HV* get_hv("package::varname", 0);
a0d0e21e
LW
462
463This returns NULL if the variable does not exist.
464
a43e7901 465The hash algorithm is defined in the C<PERL_HASH> macro:
a0d0e21e 466
a43e7901 467 PERL_HASH(hash, key, klen)
ab192400 468
a43e7901
YO
469The exact implementation of this macro varies by architecture and version
470of perl, and the return value may change per invocation, so the value
471is only valid for the duration of a single perl process.
a0d0e21e 472
04343c6d
GS
473See L<Understanding the Magic of Tied Hashes and Arrays> for more
474information on how to use the hash access functions on tied hashes.
475
1e422769
PP
476=head2 Hash API Extensions
477
478Beginning with version 5.004, the following functions are also supported:
479
480 HE* hv_fetch_ent (HV* tb, SV* key, I32 lval, U32 hash);
481 HE* hv_store_ent (HV* tb, SV* key, SV* val, U32 hash);
c47ff5f1 482
1e422769
PP
483 bool hv_exists_ent (HV* tb, SV* key, U32 hash);
484 SV* hv_delete_ent (HV* tb, SV* key, I32 flags, U32 hash);
c47ff5f1 485
1e422769
PP
486 SV* hv_iterkeysv (HE* entry);
487
488Note that these functions take C<SV*> keys, which simplifies writing
489of extension code that deals with hash structures. These functions
490also allow passing of C<SV*> keys to C<tie> functions without forcing
491you to stringify the keys (unlike the previous set of functions).
492
493They also return and accept whole hash entries (C<HE*>), making their
494use more efficient (since the hash number for a particular string
4a4eefd0
GS
495doesn't have to be recomputed every time). See L<perlapi> for detailed
496descriptions.
1e422769
PP
497
498The following macros must always be used to access the contents of hash
499entries. Note that the arguments to these macros must be simple
500variables, since they may get evaluated more than once. See
4a4eefd0 501L<perlapi> for detailed descriptions of these macros.
1e422769
PP
502
503 HePV(HE* he, STRLEN len)
504 HeVAL(HE* he)
505 HeHASH(HE* he)
506 HeSVKEY(HE* he)
507 HeSVKEY_force(HE* he)
508 HeSVKEY_set(HE* he, SV* sv)
509
510These two lower level macros are defined, but must only be used when
511dealing with keys that are not C<SV*>s:
512
513 HeKEY(HE* he)
514 HeKLEN(HE* he)
515
04343c6d
GS
516Note that both C<hv_store> and C<hv_store_ent> do not increment the
517reference count of the stored C<val>, which is the caller's responsibility.
518If these functions return a NULL value, the caller will usually have to
519decrement the reference count of C<val> to avoid a memory leak.
1e422769 520
a9381218
MHM
521=head2 AVs, HVs and undefined values
522
523Sometimes you have to store undefined values in AVs or HVs. Although
524this may be a rare case, it can be tricky. That's because you're
525used to using C<&PL_sv_undef> if you need an undefined SV.
526
527For example, intuition tells you that this XS code:
528
529 AV *av = newAV();
530 av_store( av, 0, &PL_sv_undef );
531
532is equivalent to this Perl code:
533
534 my @av;
535 $av[0] = undef;
536
f3c4ec28 537Unfortunately, this isn't true. In perl 5.18 and earlier, AVs use C<&PL_sv_undef> as a marker
a9381218
MHM
538for indicating that an array element has not yet been initialized.
539Thus, C<exists $av[0]> would be true for the above Perl code, but
f3c4ec28
FC
540false for the array generated by the XS code. In perl 5.20, storing
541&PL_sv_undef will create a read-only element, because the scalar
542&PL_sv_undef itself is stored, not a copy.
a9381218 543
f3c4ec28 544Similar problems can occur when storing C<&PL_sv_undef> in HVs:
a9381218
MHM
545
546 hv_store( hv, "key", 3, &PL_sv_undef, 0 );
547
548This will indeed make the value C<undef>, but if you try to modify
549the value of C<key>, you'll get the following error:
550
551 Modification of non-creatable hash value attempted
552
553In perl 5.8.0, C<&PL_sv_undef> was also used to mark placeholders
554in restricted hashes. This caused such hash entries not to appear
555when iterating over the hash or when checking for the keys
556with the C<hv_exists> function.
557
8abccac8
AR
558You can run into similar problems when you store C<&PL_sv_yes> or
559C<&PL_sv_no> into AVs or HVs. Trying to modify such elements
a9381218
MHM
560will give you the following error:
561
562 Modification of a read-only value attempted
563
564To make a long story short, you can use the special variables
8abccac8 565C<&PL_sv_undef>, C<&PL_sv_yes> and C<&PL_sv_no> with AVs and
a9381218
MHM
566HVs, but you have to make sure you know what you're doing.
567
568Generally, if you want to store an undefined value in an AV
569or HV, you should not use C<&PL_sv_undef>, but rather create a
570new undefined value using the C<newSV> function, for example:
571
572 av_store( av, 42, newSV(0) );
573 hv_store( hv, "foo", 3, newSV(0), 0 );
574
a0d0e21e
LW
575=head2 References
576
d1b91892 577References are a special type of scalar that point to other data types
a9b0660e 578(including other references).
a0d0e21e 579
07fa94a1 580To create a reference, use either of the following functions:
a0d0e21e 581
5f05dabc
PP
582 SV* newRV_inc((SV*) thing);
583 SV* newRV_noinc((SV*) thing);
a0d0e21e 584
5f05dabc 585The C<thing> argument can be any of an C<SV*>, C<AV*>, or C<HV*>. The
07fa94a1
JO
586functions are identical except that C<newRV_inc> increments the reference
587count of the C<thing>, while C<newRV_noinc> does not. For historical
588reasons, C<newRV> is a synonym for C<newRV_inc>.
589
590Once you have a reference, you can use the following macro to dereference
591the reference:
a0d0e21e
LW
592
593 SvRV(SV*)
594
595then call the appropriate routines, casting the returned C<SV*> to either an
d1b91892 596C<AV*> or C<HV*>, if required.
a0d0e21e 597
d1b91892 598To determine if an SV is a reference, you can use the following macro:
a0d0e21e
LW
599
600 SvROK(SV*)
601
07fa94a1
JO
602To discover what type of value the reference refers to, use the following
603macro and then check the return value.
d1b91892
AD
604
605 SvTYPE(SvRV(SV*))
606
607The most useful types that will be returned are:
608
a5e62da0
FC
609 < SVt_PVAV Scalar
610 SVt_PVAV Array
611 SVt_PVHV Hash
612 SVt_PVCV Code
613 SVt_PVGV Glob (possibly a file handle)
614
615See L<perlapi/svtype> for more details.
d1b91892 616
cb1a09d0
AD
617=head2 Blessed References and Class Objects
618
06f6df17 619References are also used to support object-oriented programming. In perl's
cb1a09d0
AD
620OO lexicon, an object is simply a reference that has been blessed into a
621package (or class). Once blessed, the programmer may now use the reference
622to access the various methods in the class.
623
624A reference can be blessed into a package with the following function:
625
626 SV* sv_bless(SV* sv, HV* stash);
627
06f6df17
RGS
628The C<sv> argument must be a reference value. The C<stash> argument
629specifies which class the reference will belong to. See
2ae324a7 630L<Stashes and Globs> for information on converting class names into stashes.
cb1a09d0
AD
631
632/* Still under construction */
633
ddd2cc91
DM
634The following function upgrades rv to reference if not already one.
635Creates a new SV for rv to point to. If C<classname> is non-null, the SV
636is blessed into the specified class. SV is returned.
cb1a09d0 637
08105a92 638 SV* newSVrv(SV* rv, const char* classname);
cb1a09d0 639
ddd2cc91
DM
640The following three functions copy integer, unsigned integer or double
641into an SV whose reference is C<rv>. SV is blessed if C<classname> is
642non-null.
cb1a09d0 643
08105a92 644 SV* sv_setref_iv(SV* rv, const char* classname, IV iv);
e1c57cef 645 SV* sv_setref_uv(SV* rv, const char* classname, UV uv);
08105a92 646 SV* sv_setref_nv(SV* rv, const char* classname, NV iv);
cb1a09d0 647
ddd2cc91
DM
648The following function copies the pointer value (I<the address, not the
649string!>) into an SV whose reference is rv. SV is blessed if C<classname>
650is non-null.
cb1a09d0 651
ddd2cc91 652 SV* sv_setref_pv(SV* rv, const char* classname, void* pv);
cb1a09d0 653
a9b0660e 654The following function copies a string into an SV whose reference is C<rv>.
ddd2cc91
DM
655Set length to 0 to let Perl calculate the string length. SV is blessed if
656C<classname> is non-null.
cb1a09d0 657
a9b0660e
KW
658 SV* sv_setref_pvn(SV* rv, const char* classname, char* pv,
659 STRLEN length);
cb1a09d0 660
ddd2cc91
DM
661The following function tests whether the SV is blessed into the specified
662class. It does not check inheritance relationships.
9abd00ed 663
08105a92 664 int sv_isa(SV* sv, const char* name);
9abd00ed 665
ddd2cc91 666The following function tests whether the SV is a reference to a blessed object.
9abd00ed
GS
667
668 int sv_isobject(SV* sv);
669
ddd2cc91
DM
670The following function tests whether the SV is derived from the specified
671class. SV can be either a reference to a blessed object or a string
672containing a class name. This is the function implementing the
673C<UNIVERSAL::isa> functionality.
9abd00ed 674
08105a92 675 bool sv_derived_from(SV* sv, const char* name);
9abd00ed 676
00aadd71 677To check if you've got an object derived from a specific class you have
9abd00ed
GS
678to write:
679
680 if (sv_isobject(sv) && sv_derived_from(sv, class)) { ... }
cb1a09d0 681
5f05dabc 682=head2 Creating New Variables
cb1a09d0 683
5f05dabc
PP
684To create a new Perl variable with an undef value which can be accessed from
685your Perl script, use the following routines, depending on the variable type.
cb1a09d0 686
64ace3f8 687 SV* get_sv("package::varname", GV_ADD);
cbfd0a87 688 AV* get_av("package::varname", GV_ADD);
6673a63c 689 HV* get_hv("package::varname", GV_ADD);
cb1a09d0 690
058a5f6c 691Notice the use of GV_ADD as the second parameter. The new variable can now
cb1a09d0
AD
692be set, using the routines appropriate to the data type.
693
5f05dabc 694There are additional macros whose values may be bitwise OR'ed with the
058a5f6c 695C<GV_ADD> argument to enable certain extra features. Those bits are:
cb1a09d0 696
9a68f1db
SB
697=over
698
699=item GV_ADDMULTI
700
701Marks the variable as multiply defined, thus preventing the:
702
703 Name <varname> used only once: possible typo
704
705warning.
706
9a68f1db
SB
707=item GV_ADDWARN
708
709Issues the warning:
710
711 Had to create <varname> unexpectedly
712
713if the variable did not exist before the function was called.
714
715=back
cb1a09d0 716
07fa94a1
JO
717If you do not specify a package name, the variable is created in the current
718package.
cb1a09d0 719
5f05dabc 720=head2 Reference Counts and Mortality
a0d0e21e 721
d1be9408 722Perl uses a reference count-driven garbage collection mechanism. SVs,
54310121 723AVs, or HVs (xV for short in the following) start their life with a
55497cff 724reference count of 1. If the reference count of an xV ever drops to 0,
07fa94a1 725then it will be destroyed and its memory made available for reuse.
55497cff
PP
726
727This normally doesn't happen at the Perl level unless a variable is
5f05dabc
PP
728undef'ed or the last variable holding a reference to it is changed or
729overwritten. At the internal level, however, reference counts can be
55497cff
PP
730manipulated with the following macros:
731
732 int SvREFCNT(SV* sv);
5f05dabc 733 SV* SvREFCNT_inc(SV* sv);
55497cff
PP
734 void SvREFCNT_dec(SV* sv);
735
736However, there is one other function which manipulates the reference
07fa94a1
JO
737count of its argument. The C<newRV_inc> function, you will recall,
738creates a reference to the specified argument. As a side effect,
739it increments the argument's reference count. If this is not what
740you want, use C<newRV_noinc> instead.
741
742For example, imagine you want to return a reference from an XSUB function.
743Inside the XSUB routine, you create an SV which initially has a reference
744count of one. Then you call C<newRV_inc>, passing it the just-created SV.
5f05dabc
PP
745This returns the reference as a new SV, but the reference count of the
746SV you passed to C<newRV_inc> has been incremented to two. Now you
07fa94a1
JO
747return the reference from the XSUB routine and forget about the SV.
748But Perl hasn't! Whenever the returned reference is destroyed, the
749reference count of the original SV is decreased to one and nothing happens.
750The SV will hang around without any way to access it until Perl itself
751terminates. This is a memory leak.
5f05dabc
PP
752
753The correct procedure, then, is to use C<newRV_noinc> instead of
faed5253
JO
754C<newRV_inc>. Then, if and when the last reference is destroyed,
755the reference count of the SV will go to zero and it will be destroyed,
07fa94a1 756stopping any memory leak.
55497cff 757
5f05dabc 758There are some convenience functions available that can help with the
54310121 759destruction of xVs. These functions introduce the concept of "mortality".
07fa94a1
JO
760An xV that is mortal has had its reference count marked to be decremented,
761but not actually decremented, until "a short time later". Generally the
762term "short time later" means a single Perl statement, such as a call to
54310121 763an XSUB function. The actual determinant for when mortal xVs have their
07fa94a1
JO
764reference count decremented depends on two macros, SAVETMPS and FREETMPS.
765See L<perlcall> and L<perlxs> for more details on these macros.
55497cff
PP
766
767"Mortalization" then is at its simplest a deferred C<SvREFCNT_dec>.
768However, if you mortalize a variable twice, the reference count will
769later be decremented twice.
770
00aadd71
NIS
771"Mortal" SVs are mainly used for SVs that are placed on perl's stack.
772For example an SV which is created just to pass a number to a called sub
06f6df17
RGS
773is made mortal to have it cleaned up automatically when it's popped off
774the stack. Similarly, results returned by XSUBs (which are pushed on the
775stack) are often made mortal.
a0d0e21e
LW
776
777To create a mortal variable, use the functions:
778
779 SV* sv_newmortal()
780 SV* sv_2mortal(SV*)
781 SV* sv_mortalcopy(SV*)
782
00aadd71 783The first call creates a mortal SV (with no value), the second converts an existing
5f05dabc
PP
784SV to a mortal SV (and thus defers a call to C<SvREFCNT_dec>), and the
785third creates a mortal copy of an existing SV.
da8c5729 786Because C<sv_newmortal> gives the new SV no value, it must normally be given one
9a68f1db 787via C<sv_setpv>, C<sv_setiv>, etc. :
00aadd71
NIS
788
789 SV *tmp = sv_newmortal();
790 sv_setiv(tmp, an_integer);
791
792As that is multiple C statements it is quite common so see this idiom instead:
793
794 SV *tmp = sv_2mortal(newSViv(an_integer));
795
796
797You should be careful about creating mortal variables. Strange things
798can happen if you make the same value mortal within multiple contexts,
799or if you make a variable mortal multiple times. Thinking of "Mortalization"
800as deferred C<SvREFCNT_dec> should help to minimize such problems.
da8c5729 801For example if you are passing an SV which you I<know> has a high enough REFCNT
00aadd71
NIS
802to survive its use on the stack you need not do any mortalization.
803If you are not sure then doing an C<SvREFCNT_inc> and C<sv_2mortal>, or
804making a C<sv_mortalcopy> is safer.
a0d0e21e 805
ac036724 806The mortal routines are not just for SVs; AVs and HVs can be
faed5253 807made mortal by passing their address (type-casted to C<SV*>) to the
07fa94a1 808C<sv_2mortal> or C<sv_mortalcopy> routines.
a0d0e21e 809
5f05dabc 810=head2 Stashes and Globs
a0d0e21e 811
06f6df17
RGS
812A B<stash> is a hash that contains all variables that are defined
813within a package. Each key of the stash is a symbol
aa689395
PP
814name (shared by all the different types of objects that have the same
815name), and each value in the hash table is a GV (Glob Value). This GV
816in turn contains references to the various objects of that name,
817including (but not limited to) the following:
cb1a09d0 818
a0d0e21e
LW
819 Scalar Value
820 Array Value
821 Hash Value
a3cb178b 822 I/O Handle
a0d0e21e
LW
823 Format
824 Subroutine
825
06f6df17
RGS
826There is a single stash called C<PL_defstash> that holds the items that exist
827in the C<main> package. To get at the items in other packages, append the
828string "::" to the package name. The items in the C<Foo> package are in
829the stash C<Foo::> in PL_defstash. The items in the C<Bar::Baz> package are
830in the stash C<Baz::> in C<Bar::>'s stash.
a0d0e21e 831
d1b91892 832To get the stash pointer for a particular package, use the function:
a0d0e21e 833
da51bb9b
NC
834 HV* gv_stashpv(const char* name, I32 flags)
835 HV* gv_stashsv(SV*, I32 flags)
a0d0e21e
LW
836
837The first function takes a literal string, the second uses the string stored
d1b91892 838in the SV. Remember that a stash is just a hash table, so you get back an
da51bb9b 839C<HV*>. The C<flags> flag will create a new package if it is set to GV_ADD.
a0d0e21e
LW
840
841The name that C<gv_stash*v> wants is the name of the package whose symbol table
842you want. The default package is called C<main>. If you have multiply nested
d1b91892
AD
843packages, pass their names to C<gv_stash*v>, separated by C<::> as in the Perl
844language itself.
a0d0e21e
LW
845
846Alternately, if you have an SV that is a blessed reference, you can find
847out the stash pointer by using:
848
849 HV* SvSTASH(SvRV(SV*));
850
851then use the following to get the package name itself:
852
853 char* HvNAME(HV* stash);
854
5f05dabc
PP
855If you need to bless or re-bless an object you can use the following
856function:
a0d0e21e
LW
857
858 SV* sv_bless(SV*, HV* stash)
859
860where the first argument, an C<SV*>, must be a reference, and the second
861argument is a stash. The returned C<SV*> can now be used in the same way
862as any other SV.
863
d1b91892
AD
864For more information on references and blessings, consult L<perlref>.
865
54310121 866=head2 Double-Typed SVs
0a753a76
PP
867
868Scalar variables normally contain only one type of value, an integer,
869double, pointer, or reference. Perl will automatically convert the
870actual scalar data from the stored type into the requested type.
871
872Some scalar variables contain more than one type of scalar data. For
873example, the variable C<$!> contains either the numeric value of C<errno>
874or its string equivalent from either C<strerror> or C<sys_errlist[]>.
875
876To force multiple data values into an SV, you must do two things: use the
877C<sv_set*v> routines to add the additional scalar type, then set a flag
878so that Perl will believe it contains more than one type of data. The
879four macros to set the flags are:
880
881 SvIOK_on
882 SvNOK_on
883 SvPOK_on
884 SvROK_on
885
886The particular macro you must use depends on which C<sv_set*v> routine
887you called first. This is because every C<sv_set*v> routine turns on
888only the bit for the particular type of data being set, and turns off
889all the rest.
890
891For example, to create a new Perl variable called "dberror" that contains
892both the numeric and descriptive string error values, you could use the
893following code:
894
895 extern int dberror;
896 extern char *dberror_list;
897
64ace3f8 898 SV* sv = get_sv("dberror", GV_ADD);
0a753a76
PP
899 sv_setiv(sv, (IV) dberror);
900 sv_setpv(sv, dberror_list[dberror]);
901 SvIOK_on(sv);
902
903If the order of C<sv_setiv> and C<sv_setpv> had been reversed, then the
904macro C<SvPOK_on> would need to be called instead of C<SvIOK_on>.
905
906=head2 Magic Variables
a0d0e21e 907
d1b91892
AD
908[This section still under construction. Ignore everything here. Post no
909bills. Everything not permitted is forbidden.]
910
d1b91892
AD
911Any SV may be magical, that is, it has special features that a normal
912SV does not have. These features are stored in the SV structure in a
5f05dabc 913linked list of C<struct magic>'s, typedef'ed to C<MAGIC>.
d1b91892
AD
914
915 struct magic {
916 MAGIC* mg_moremagic;
917 MGVTBL* mg_virtual;
918 U16 mg_private;
919 char mg_type;
920 U8 mg_flags;
b205eb13 921 I32 mg_len;
d1b91892
AD
922 SV* mg_obj;
923 char* mg_ptr;
d1b91892
AD
924 };
925
926Note this is current as of patchlevel 0, and could change at any time.
927
928=head2 Assigning Magic
929
930Perl adds magic to an SV using the sv_magic function:
931
a9b0660e 932 void sv_magic(SV* sv, SV* obj, int how, const char* name, I32 namlen);
d1b91892
AD
933
934The C<sv> argument is a pointer to the SV that is to acquire a new magical
935feature.
936
937If C<sv> is not already magical, Perl uses the C<SvUPGRADE> macro to
645c22ef
DM
938convert C<sv> to type C<SVt_PVMG>. Perl then continues by adding new magic
939to the beginning of the linked list of magical features. Any prior entry
940of the same type of magic is deleted. Note that this can be overridden,
941and multiple instances of the same type of magic can be associated with an
942SV.
d1b91892 943
54310121
PP
944The C<name> and C<namlen> arguments are used to associate a string with
945the magic, typically the name of a variable. C<namlen> is stored in the
2d8d5d5a
SH
946C<mg_len> field and if C<name> is non-null then either a C<savepvn> copy of
947C<name> or C<name> itself is stored in the C<mg_ptr> field, depending on
948whether C<namlen> is greater than zero or equal to zero respectively. As a
949special case, if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
950to contain an C<SV*> and is stored as-is with its REFCNT incremented.
d1b91892
AD
951
952The sv_magic function uses C<how> to determine which, if any, predefined
953"Magic Virtual Table" should be assigned to the C<mg_virtual> field.
06f6df17 954See the L<Magic Virtual Tables> section below. The C<how> argument is also
14befaf4 955stored in the C<mg_type> field. The value of C<how> should be chosen
06f6df17 956from the set of macros C<PERL_MAGIC_foo> found in F<perl.h>. Note that before
645c22ef 957these macros were added, Perl internals used to directly use character
14befaf4 958literals, so you may occasionally come across old code or documentation
75d0f26d 959referring to 'U' magic rather than C<PERL_MAGIC_uvar> for example.
d1b91892
AD
960
961The C<obj> argument is stored in the C<mg_obj> field of the C<MAGIC>
962structure. If it is not the same as the C<sv> argument, the reference
963count of the C<obj> object is incremented. If it is the same, or if
645c22ef 964the C<how> argument is C<PERL_MAGIC_arylen>, or if it is a NULL pointer,
14befaf4 965then C<obj> is merely stored, without the reference count being incremented.
d1b91892 966
2d8d5d5a
SH
967See also C<sv_magicext> in L<perlapi> for a more flexible way to add magic
968to an SV.
969
cb1a09d0
AD
970There is also a function to add magic to an C<HV>:
971
972 void hv_magic(HV *hv, GV *gv, int how);
973
974This simply calls C<sv_magic> and coerces the C<gv> argument into an C<SV>.
975
976To remove the magic from an SV, call the function sv_unmagic:
977
70a53b35 978 int sv_unmagic(SV *sv, int type);
cb1a09d0
AD
979
980The C<type> argument should be equal to the C<how> value when the C<SV>
981was initially made magical.
982
f6ee7b17
FR
983However, note that C<sv_unmagic> removes all magic of a certain C<type> from the
984C<SV>. If you want to remove only certain magic of a C<type> based on the magic
985virtual table, use C<sv_unmagicext> instead:
986
987 int sv_unmagicext(SV *sv, int type, MGVTBL *vtbl);
988
d1b91892
AD
989=head2 Magic Virtual Tables
990
d1be9408 991The C<mg_virtual> field in the C<MAGIC> structure is a pointer to an
d1b91892
AD
992C<MGVTBL>, which is a structure of function pointers and stands for
993"Magic Virtual Table" to handle the various operations that might be
994applied to that variable.
995
301cb7e8
DM
996The C<MGVTBL> has five (or sometimes eight) pointers to the following
997routine types:
d1b91892
AD
998
999 int (*svt_get)(SV* sv, MAGIC* mg);
1000 int (*svt_set)(SV* sv, MAGIC* mg);
1001 U32 (*svt_len)(SV* sv, MAGIC* mg);
1002 int (*svt_clear)(SV* sv, MAGIC* mg);
1003 int (*svt_free)(SV* sv, MAGIC* mg);
1004
a9b0660e
KW
1005 int (*svt_copy)(SV *sv, MAGIC* mg, SV *nsv,
1006 const char *name, I32 namlen);
301cb7e8
DM
1007 int (*svt_dup)(MAGIC *mg, CLONE_PARAMS *param);
1008 int (*svt_local)(SV *nsv, MAGIC *mg);
1009
1010
06f6df17 1011This MGVTBL structure is set at compile-time in F<perl.h> and there are
b7a0f54c
SM
1012currently 32 types. These different structures contain pointers to various
1013routines that perform additional actions depending on which function is
1014being called.
d1b91892 1015
a9b0660e
KW
1016 Function pointer Action taken
1017 ---------------- ------------
1018 svt_get Do something before the value of the SV is
1019 retrieved.
1020 svt_set Do something after the SV is assigned a value.
1021 svt_len Report on the SV's length.
1022 svt_clear Clear something the SV represents.
1023 svt_free Free any extra storage associated with the SV.
d1b91892 1024
a9b0660e
KW
1025 svt_copy copy tied variable magic to a tied element
1026 svt_dup duplicate a magic structure during thread cloning
1027 svt_local copy magic to local value during 'local'
301cb7e8 1028
d1b91892 1029For instance, the MGVTBL structure called C<vtbl_sv> (which corresponds
14befaf4 1030to an C<mg_type> of C<PERL_MAGIC_sv>) contains:
d1b91892
AD
1031
1032 { magic_get, magic_set, magic_len, 0, 0 }
1033
14befaf4
DM
1034Thus, when an SV is determined to be magical and of type C<PERL_MAGIC_sv>,
1035if a get operation is being performed, the routine C<magic_get> is
1036called. All the various routines for the various magical types begin
1037with C<magic_>. NOTE: the magic routines are not considered part of
1038the Perl API, and may not be exported by the Perl library.
d1b91892 1039
301cb7e8
DM
1040The last three slots are a recent addition, and for source code
1041compatibility they are only checked for if one of the three flags
1042MGf_COPY, MGf_DUP or MGf_LOCAL is set in mg_flags. This means that most
1043code can continue declaring a vtable as a 5-element value. These three are
1044currently used exclusively by the threading code, and are highly subject
1045to change.
1046
d1b91892
AD
1047The current kinds of Magic Virtual Tables are:
1048
f1f5ddd7
FC
1049=for comment
1050This table is generated by regen/mg_vtable.pl. Any changes made here
1051will be lost.
1052
1053=for mg_vtable.pl begin
1054
a9b0660e 1055 mg_type
bd6e6c12
FC
1056 (old-style char and macro) MGVTBL Type of magic
1057 -------------------------- ------ -------------
1058 \0 PERL_MAGIC_sv vtbl_sv Special scalar variable
1059 # PERL_MAGIC_arylen vtbl_arylen Array length ($#ary)
1060 % PERL_MAGIC_rhash (none) extra data for restricted
1061 hashes
2a388207 1062 & PERL_MAGIC_proto (none) my sub prototype CV
bd6e6c12
FC
1063 . PERL_MAGIC_pos vtbl_pos pos() lvalue
1064 : PERL_MAGIC_symtab (none) extra data for symbol
1065 tables
1066 < PERL_MAGIC_backref vtbl_backref for weak ref data
1067 @ PERL_MAGIC_arylen_p (none) to move arylen out of XPVAV
1068 B PERL_MAGIC_bm vtbl_regexp Boyer-Moore
1069 (fast string search)
1070 c PERL_MAGIC_overload_table vtbl_ovrld Holds overload table
1071 (AMT) on stash
1072 D PERL_MAGIC_regdata vtbl_regdata Regex match position data
1073 (@+ and @- vars)
1074 d PERL_MAGIC_regdatum vtbl_regdatum Regex match position data
1075 element
1076 E PERL_MAGIC_env vtbl_env %ENV hash
1077 e PERL_MAGIC_envelem vtbl_envelem %ENV hash element
eccba044 1078 f PERL_MAGIC_fm vtbl_regexp Formline
bd6e6c12 1079 ('compiled' format)
bd6e6c12
FC
1080 g PERL_MAGIC_regex_global vtbl_mglob m//g target
1081 H PERL_MAGIC_hints vtbl_hints %^H hash
1082 h PERL_MAGIC_hintselem vtbl_hintselem %^H hash element
1083 I PERL_MAGIC_isa vtbl_isa @ISA array
1084 i PERL_MAGIC_isaelem vtbl_isaelem @ISA array element
1085 k PERL_MAGIC_nkeys vtbl_nkeys scalar(keys()) lvalue
1086 L PERL_MAGIC_dbfile (none) Debugger %_<filename
1087 l PERL_MAGIC_dbline vtbl_dbline Debugger %_<filename
1088 element
1089 N PERL_MAGIC_shared (none) Shared between threads
1090 n PERL_MAGIC_shared_scalar (none) Shared between threads
1091 o PERL_MAGIC_collxfrm vtbl_collxfrm Locale transformation
1092 P PERL_MAGIC_tied vtbl_pack Tied array or hash
1093 p PERL_MAGIC_tiedelem vtbl_packelem Tied array or hash element
1094 q PERL_MAGIC_tiedscalar vtbl_packelem Tied scalar or handle
1095 r PERL_MAGIC_qr vtbl_regexp precompiled qr// regex
1096 S PERL_MAGIC_sig (none) %SIG hash
1097 s PERL_MAGIC_sigelem vtbl_sigelem %SIG hash element
1098 t PERL_MAGIC_taint vtbl_taint Taintedness
1099 U PERL_MAGIC_uvar vtbl_uvar Available for use by
1100 extensions
1101 u PERL_MAGIC_uvar_elem (none) Reserved for use by
1102 extensions
4499db73 1103 V PERL_MAGIC_vstring (none) SV was vstring literal
bd6e6c12
FC
1104 v PERL_MAGIC_vec vtbl_vec vec() lvalue
1105 w PERL_MAGIC_utf8 vtbl_utf8 Cached UTF-8 information
1106 x PERL_MAGIC_substr vtbl_substr substr() lvalue
1107 y PERL_MAGIC_defelem vtbl_defelem Shadow "foreach" iterator
1108 variable / smart parameter
1109 vivification
1110 ] PERL_MAGIC_checkcall vtbl_checkcall inlining/mutation of call
1111 to this CV
1112 ~ PERL_MAGIC_ext (none) Available for use by
1113 extensions
0cbee0a4 1114
f1f5ddd7 1115=for mg_vtable.pl end
d1b91892 1116
68dc0745 1117When an uppercase and lowercase letter both exist in the table, then the
92f0c265
JP
1118uppercase letter is typically used to represent some kind of composite type
1119(a list or a hash), and the lowercase letter is used to represent an element
1120of that composite type. Some internals code makes use of this case
1121relationship. However, 'v' and 'V' (vec and v-string) are in no way related.
14befaf4
DM
1122
1123The C<PERL_MAGIC_ext> and C<PERL_MAGIC_uvar> magic types are defined
1124specifically for use by extensions and will not be used by perl itself.
1125Extensions can use C<PERL_MAGIC_ext> magic to 'attach' private information
1126to variables (typically objects). This is especially useful because
1127there is no way for normal perl code to corrupt this private information
1128(unlike using extra elements of a hash object).
1129
1130Similarly, C<PERL_MAGIC_uvar> magic can be used much like tie() to call a
1131C function any time a scalar's value is used or changed. The C<MAGIC>'s
bdbeb323
SM
1132C<mg_ptr> field points to a C<ufuncs> structure:
1133
1134 struct ufuncs {
a9402793
AB
1135 I32 (*uf_val)(pTHX_ IV, SV*);
1136 I32 (*uf_set)(pTHX_ IV, SV*);
bdbeb323
SM
1137 IV uf_index;
1138 };
1139
1140When the SV is read from or written to, the C<uf_val> or C<uf_set>
14befaf4
DM
1141function will be called with C<uf_index> as the first arg and a pointer to
1142the SV as the second. A simple example of how to add C<PERL_MAGIC_uvar>
1526ead6
AB
1143magic is shown below. Note that the ufuncs structure is copied by
1144sv_magic, so you can safely allocate it on the stack.
1145
1146 void
1147 Umagic(sv)
1148 SV *sv;
1149 PREINIT:
1150 struct ufuncs uf;
1151 CODE:
1152 uf.uf_val = &my_get_fn;
1153 uf.uf_set = &my_set_fn;
1154 uf.uf_index = 0;
14befaf4 1155 sv_magic(sv, 0, PERL_MAGIC_uvar, (char*)&uf, sizeof(uf));
5f05dabc 1156
1e73acc8
AS
1157Attaching C<PERL_MAGIC_uvar> to arrays is permissible but has no effect.
1158
1159For hashes there is a specialized hook that gives control over hash
1160keys (but not values). This hook calls C<PERL_MAGIC_uvar> 'get' magic
1161if the "set" function in the C<ufuncs> structure is NULL. The hook
1162is activated whenever the hash is accessed with a key specified as
1163an C<SV> through the functions C<hv_store_ent>, C<hv_fetch_ent>,
1164C<hv_delete_ent>, and C<hv_exists_ent>. Accessing the key as a string
1165through the functions without the C<..._ent> suffix circumvents the
4509d391 1166hook. See L<Hash::Util::FieldHash/GUTS> for a detailed description.
1e73acc8 1167
14befaf4
DM
1168Note that because multiple extensions may be using C<PERL_MAGIC_ext>
1169or C<PERL_MAGIC_uvar> magic, it is important for extensions to take
1170extra care to avoid conflict. Typically only using the magic on
1171objects blessed into the same class as the extension is sufficient.
2f07f21a
FR
1172For C<PERL_MAGIC_ext> magic, it is usually a good idea to define an
1173C<MGVTBL>, even if all its fields will be C<0>, so that individual
1174C<MAGIC> pointers can be identified as a particular kind of magic
f6ee7b17
FR
1175using their magic virtual table. C<mg_findext> provides an easy way
1176to do that:
2f07f21a
FR
1177
1178 STATIC MGVTBL my_vtbl = { 0, 0, 0, 0, 0, 0, 0, 0 };
1179
1180 MAGIC *mg;
f6ee7b17
FR
1181 if ((mg = mg_findext(sv, PERL_MAGIC_ext, &my_vtbl))) {
1182 /* this is really ours, not another module's PERL_MAGIC_ext */
1183 my_priv_data_t *priv = (my_priv_data_t *)mg->mg_ptr;
1184 ...
2f07f21a 1185 }
5f05dabc 1186
ef50df4b
GS
1187Also note that the C<sv_set*()> and C<sv_cat*()> functions described
1188earlier do B<not> invoke 'set' magic on their targets. This must
1189be done by the user either by calling the C<SvSETMAGIC()> macro after
1190calling these functions, or by using one of the C<sv_set*_mg()> or
1191C<sv_cat*_mg()> functions. Similarly, generic C code must call the
1192C<SvGETMAGIC()> macro to invoke any 'get' magic if they use an SV
1193obtained from external sources in functions that don't handle magic.
4a4eefd0 1194See L<perlapi> for a description of these functions.
189b2af5
GS
1195For example, calls to the C<sv_cat*()> functions typically need to be
1196followed by C<SvSETMAGIC()>, but they don't need a prior C<SvGETMAGIC()>
1197since their implementation handles 'get' magic.
1198
d1b91892
AD
1199=head2 Finding Magic
1200
a9b0660e
KW
1201 MAGIC *mg_find(SV *sv, int type); /* Finds the magic pointer of that
1202 * type */
f6ee7b17
FR
1203
1204This routine returns a pointer to a C<MAGIC> structure stored in the SV.
1205If the SV does not have that magical feature, C<NULL> is returned. If the
1206SV has multiple instances of that magical feature, the first one will be
1207returned. C<mg_findext> can be used to find a C<MAGIC> structure of an SV
da8c5729 1208based on both its magic type and its magic virtual table:
f6ee7b17
FR
1209
1210 MAGIC *mg_findext(SV *sv, int type, MGVTBL *vtbl);
d1b91892 1211
f6ee7b17
FR
1212Also, if the SV passed to C<mg_find> or C<mg_findext> is not of type
1213SVt_PVMG, Perl may core dump.
d1b91892 1214
08105a92 1215 int mg_copy(SV* sv, SV* nsv, const char* key, STRLEN klen);
d1b91892
AD
1216
1217This routine checks to see what types of magic C<sv> has. If the mg_type
68dc0745
PP
1218field is an uppercase letter, then the mg_obj is copied to C<nsv>, but
1219the mg_type field is changed to be the lowercase letter.
a0d0e21e 1220
04343c6d
GS
1221=head2 Understanding the Magic of Tied Hashes and Arrays
1222
14befaf4
DM
1223Tied hashes and arrays are magical beasts of the C<PERL_MAGIC_tied>
1224magic type.
9edb2b46
GS
1225
1226WARNING: As of the 5.004 release, proper usage of the array and hash
1227access functions requires understanding a few caveats. Some
1228of these caveats are actually considered bugs in the API, to be fixed
1229in later releases, and are bracketed with [MAYCHANGE] below. If
1230you find yourself actually applying such information in this section, be
1231aware that the behavior may change in the future, umm, without warning.
04343c6d 1232
1526ead6 1233The perl tie function associates a variable with an object that implements
9a68f1db 1234the various GET, SET, etc methods. To perform the equivalent of the perl
1526ead6
AB
1235tie function from an XSUB, you must mimic this behaviour. The code below
1236carries out the necessary steps - firstly it creates a new hash, and then
1237creates a second hash which it blesses into the class which will implement
1238the tie methods. Lastly it ties the two hashes together, and returns a
1239reference to the new tied hash. Note that the code below does NOT call the
1240TIEHASH method in the MyTie class -
1241see L<Calling Perl Routines from within C Programs> for details on how
1242to do this.
1243
1244 SV*
1245 mytie()
1246 PREINIT:
1247 HV *hash;
1248 HV *stash;
1249 SV *tie;
1250 CODE:
1251 hash = newHV();
1252 tie = newRV_noinc((SV*)newHV());
da51bb9b 1253 stash = gv_stashpv("MyTie", GV_ADD);
1526ead6 1254 sv_bless(tie, stash);
899e16d0 1255 hv_magic(hash, (GV*)tie, PERL_MAGIC_tied);
1526ead6
AB
1256 RETVAL = newRV_noinc(hash);
1257 OUTPUT:
1258 RETVAL
1259
04343c6d
GS
1260The C<av_store> function, when given a tied array argument, merely
1261copies the magic of the array onto the value to be "stored", using
1262C<mg_copy>. It may also return NULL, indicating that the value did not
9edb2b46
GS
1263actually need to be stored in the array. [MAYCHANGE] After a call to
1264C<av_store> on a tied array, the caller will usually need to call
1265C<mg_set(val)> to actually invoke the perl level "STORE" method on the
1266TIEARRAY object. If C<av_store> did return NULL, a call to
1267C<SvREFCNT_dec(val)> will also be usually necessary to avoid a memory
1268leak. [/MAYCHANGE]
04343c6d
GS
1269
1270The previous paragraph is applicable verbatim to tied hash access using the
1271C<hv_store> and C<hv_store_ent> functions as well.
1272
1273C<av_fetch> and the corresponding hash functions C<hv_fetch> and
1274C<hv_fetch_ent> actually return an undefined mortal value whose magic
1275has been initialized using C<mg_copy>. Note the value so returned does not
9edb2b46
GS
1276need to be deallocated, as it is already mortal. [MAYCHANGE] But you will
1277need to call C<mg_get()> on the returned value in order to actually invoke
1278the perl level "FETCH" method on the underlying TIE object. Similarly,
04343c6d
GS
1279you may also call C<mg_set()> on the return value after possibly assigning
1280a suitable value to it using C<sv_setsv>, which will invoke the "STORE"
9edb2b46 1281method on the TIE object. [/MAYCHANGE]
04343c6d 1282
9edb2b46 1283[MAYCHANGE]
04343c6d
GS
1284In other words, the array or hash fetch/store functions don't really
1285fetch and store actual values in the case of tied arrays and hashes. They
1286merely call C<mg_copy> to attach magic to the values that were meant to be
1287"stored" or "fetched". Later calls to C<mg_get> and C<mg_set> actually
1288do the job of invoking the TIE methods on the underlying objects. Thus
9edb2b46 1289the magic mechanism currently implements a kind of lazy access to arrays
04343c6d
GS
1290and hashes.
1291
1292Currently (as of perl version 5.004), use of the hash and array access
1293functions requires the user to be aware of whether they are operating on
9edb2b46
GS
1294"normal" hashes and arrays, or on their tied variants. The API may be
1295changed to provide more transparent access to both tied and normal data
1296types in future versions.
1297[/MAYCHANGE]
04343c6d
GS
1298
1299You would do well to understand that the TIEARRAY and TIEHASH interfaces
1300are mere sugar to invoke some perl method calls while using the uniform hash
1301and array syntax. The use of this sugar imposes some overhead (typically
1302about two to four extra opcodes per FETCH/STORE operation, in addition to
1303the creation of all the mortal variables required to invoke the methods).
1304This overhead will be comparatively small if the TIE methods are themselves
1305substantial, but if they are only a few statements long, the overhead
1306will not be insignificant.
1307
d1c897a1
IZ
1308=head2 Localizing changes
1309
1310Perl has a very handy construction
1311
1312 {
1313 local $var = 2;
1314 ...
1315 }
1316
1317This construction is I<approximately> equivalent to
1318
1319 {
1320 my $oldvar = $var;
1321 $var = 2;
1322 ...
1323 $var = $oldvar;
1324 }
1325
1326The biggest difference is that the first construction would
1327reinstate the initial value of $var, irrespective of how control exits
9a68f1db 1328the block: C<goto>, C<return>, C<die>/C<eval>, etc. It is a little bit
d1c897a1
IZ
1329more efficient as well.
1330
1331There is a way to achieve a similar task from C via Perl API: create a
1332I<pseudo-block>, and arrange for some changes to be automatically
1333undone at the end of it, either explicit, or via a non-local exit (via
1334die()). A I<block>-like construct is created by a pair of
b687b08b
TC
1335C<ENTER>/C<LEAVE> macros (see L<perlcall/"Returning a Scalar">).
1336Such a construct may be created specially for some important localized
1337task, or an existing one (like boundaries of enclosing Perl
1338subroutine/block, or an existing pair for freeing TMPs) may be
1339used. (In the second case the overhead of additional localization must
1340be almost negligible.) Note that any XSUB is automatically enclosed in
1341an C<ENTER>/C<LEAVE> pair.
d1c897a1
IZ
1342
1343Inside such a I<pseudo-block> the following service is available:
1344
13a2d996 1345=over 4
d1c897a1
IZ
1346
1347=item C<SAVEINT(int i)>
1348
1349=item C<SAVEIV(IV i)>
1350
1351=item C<SAVEI32(I32 i)>
1352
1353=item C<SAVELONG(long i)>
1354
1355These macros arrange things to restore the value of integer variable
1356C<i> at the end of enclosing I<pseudo-block>.
1357
1358=item C<SAVESPTR(s)>
1359
1360=item C<SAVEPPTR(p)>
1361
1362These macros arrange things to restore the value of pointers C<s> and
1363C<p>. C<s> must be a pointer of a type which survives conversion to
1364C<SV*> and back, C<p> should be able to survive conversion to C<char*>
1365and back.
1366
1367=item C<SAVEFREESV(SV *sv)>
1368
1369The refcount of C<sv> would be decremented at the end of
26d9b02f
JH
1370I<pseudo-block>. This is similar to C<sv_2mortal> in that it is also a
1371mechanism for doing a delayed C<SvREFCNT_dec>. However, while C<sv_2mortal>
1372extends the lifetime of C<sv> until the beginning of the next statement,
1373C<SAVEFREESV> extends it until the end of the enclosing scope. These
1374lifetimes can be wildly different.
1375
1376Also compare C<SAVEMORTALIZESV>.
1377
1378=item C<SAVEMORTALIZESV(SV *sv)>
1379
1380Just like C<SAVEFREESV>, but mortalizes C<sv> at the end of the current
1381scope instead of decrementing its reference count. This usually has the
1382effect of keeping C<sv> alive until the statement that called the currently
1383live scope has finished executing.
d1c897a1
IZ
1384
1385=item C<SAVEFREEOP(OP *op)>
1386
1387The C<OP *> is op_free()ed at the end of I<pseudo-block>.
1388
1389=item C<SAVEFREEPV(p)>
1390
1391The chunk of memory which is pointed to by C<p> is Safefree()ed at the
1392end of I<pseudo-block>.
1393
1394=item C<SAVECLEARSV(SV *sv)>
1395
1396Clears a slot in the current scratchpad which corresponds to C<sv> at
1397the end of I<pseudo-block>.
1398
1399=item C<SAVEDELETE(HV *hv, char *key, I32 length)>
1400
1401The key C<key> of C<hv> is deleted at the end of I<pseudo-block>. The
1402string pointed to by C<key> is Safefree()ed. If one has a I<key> in
1403short-lived storage, the corresponding string may be reallocated like
1404this:
1405
9cde0e7f 1406 SAVEDELETE(PL_defstash, savepv(tmpbuf), strlen(tmpbuf));
d1c897a1 1407
c76ac1ee 1408=item C<SAVEDESTRUCTOR(DESTRUCTORFUNC_NOCONTEXT_t f, void *p)>
d1c897a1
IZ
1409
1410At the end of I<pseudo-block> the function C<f> is called with the
c76ac1ee
GS
1411only argument C<p>.
1412
1413=item C<SAVEDESTRUCTOR_X(DESTRUCTORFUNC_t f, void *p)>
1414
1415At the end of I<pseudo-block> the function C<f> is called with the
1416implicit context argument (if any), and C<p>.
d1c897a1
IZ
1417
1418=item C<SAVESTACK_POS()>
1419
1420The current offset on the Perl internal stack (cf. C<SP>) is restored
1421at the end of I<pseudo-block>.
1422
1423=back
1424
1425The following API list contains functions, thus one needs to
1426provide pointers to the modifiable data explicitly (either C pointers,
00aadd71 1427or Perlish C<GV *>s). Where the above macros take C<int>, a similar
d1c897a1
IZ
1428function takes C<int *>.
1429
13a2d996 1430=over 4
d1c897a1
IZ
1431
1432=item C<SV* save_scalar(GV *gv)>
1433
1434Equivalent to Perl code C<local $gv>.
1435
1436=item C<AV* save_ary(GV *gv)>
1437
1438=item C<HV* save_hash(GV *gv)>
1439
1440Similar to C<save_scalar>, but localize C<@gv> and C<%gv>.
1441
1442=item C<void save_item(SV *item)>
1443
1444Duplicates the current value of C<SV>, on the exit from the current
1445C<ENTER>/C<LEAVE> I<pseudo-block> will restore the value of C<SV>
038fcae3
SB
1446using the stored value. It doesn't handle magic. Use C<save_scalar> if
1447magic is affected.
d1c897a1
IZ
1448
1449=item C<void save_list(SV **sarg, I32 maxsarg)>
1450
1451A variant of C<save_item> which takes multiple arguments via an array
1452C<sarg> of C<SV*> of length C<maxsarg>.
1453
1454=item C<SV* save_svref(SV **sptr)>
1455
d1be9408 1456Similar to C<save_scalar>, but will reinstate an C<SV *>.
d1c897a1
IZ
1457
1458=item C<void save_aptr(AV **aptr)>
1459
1460=item C<void save_hptr(HV **hptr)>
1461
1462Similar to C<save_svref>, but localize C<AV *> and C<HV *>.
1463
1464=back
1465
1466The C<Alias> module implements localization of the basic types within the
1467I<caller's scope>. People who are interested in how to localize things in
1468the containing scope should take a look there too.
1469
0a753a76 1470=head1 Subroutines
a0d0e21e 1471
68dc0745 1472=head2 XSUBs and the Argument Stack
5f05dabc
PP
1473
1474The XSUB mechanism is a simple way for Perl programs to access C subroutines.
1475An XSUB routine will have a stack that contains the arguments from the Perl
1476program, and a way to map from the Perl data structures to a C equivalent.
1477
1478The stack arguments are accessible through the C<ST(n)> macro, which returns
1479the C<n>'th stack argument. Argument 0 is the first argument passed in the
1480Perl subroutine call. These arguments are C<SV*>, and can be used anywhere
1481an C<SV*> is used.
1482
1483Most of the time, output from the C routine can be handled through use of
1484the RETVAL and OUTPUT directives. However, there are some cases where the
1485argument stack is not already long enough to handle all the return values.
1486An example is the POSIX tzname() call, which takes no arguments, but returns
1487two, the local time zone's standard and summer time abbreviations.
1488
1489To handle this situation, the PPCODE directive is used and the stack is
1490extended using the macro:
1491
924508f0 1492 EXTEND(SP, num);
5f05dabc 1493
924508f0
GS
1494where C<SP> is the macro that represents the local copy of the stack pointer,
1495and C<num> is the number of elements the stack should be extended by.
5f05dabc 1496
00aadd71 1497Now that there is room on the stack, values can be pushed on it using C<PUSHs>
06f6df17 1498macro. The pushed values will often need to be "mortal" (See
d82b684c 1499L</Reference Counts and Mortality>):
5f05dabc 1500
00aadd71 1501 PUSHs(sv_2mortal(newSViv(an_integer)))
d82b684c
SH
1502 PUSHs(sv_2mortal(newSVuv(an_unsigned_integer)))
1503 PUSHs(sv_2mortal(newSVnv(a_double)))
00aadd71 1504 PUSHs(sv_2mortal(newSVpv("Some String",0)))
a9b0660e
KW
1505 /* Although the last example is better written as the more
1506 * efficient: */
a3179684 1507 PUSHs(newSVpvs_flags("Some String", SVs_TEMP))
5f05dabc
PP
1508
1509And now the Perl program calling C<tzname>, the two values will be assigned
1510as in:
1511
1512 ($standard_abbrev, $summer_abbrev) = POSIX::tzname;
1513
1514An alternate (and possibly simpler) method to pushing values on the stack is
00aadd71 1515to use the macro:
5f05dabc 1516
5f05dabc
PP
1517 XPUSHs(SV*)
1518
da8c5729 1519This macro automatically adjusts the stack for you, if needed. Thus, you
5f05dabc 1520do not need to call C<EXTEND> to extend the stack.
00aadd71
NIS
1521
1522Despite their suggestions in earlier versions of this document the macros
d82b684c
SH
1523C<(X)PUSH[iunp]> are I<not> suited to XSUBs which return multiple results.
1524For that, either stick to the C<(X)PUSHs> macros shown above, or use the new
1525C<m(X)PUSH[iunp]> macros instead; see L</Putting a C value on Perl stack>.
5f05dabc
PP
1526
1527For more information, consult L<perlxs> and L<perlxstut>.
1528
5b36e945
FC
1529=head2 Autoloading with XSUBs
1530
1531If an AUTOLOAD routine is an XSUB, as with Perl subroutines, Perl puts the
1532fully-qualified name of the autoloaded subroutine in the $AUTOLOAD variable
1533of the XSUB's package.
1534
1535But it also puts the same information in certain fields of the XSUB itself:
1536
1537 HV *stash = CvSTASH(cv);
1538 const char *subname = SvPVX(cv);
1539 STRLEN name_length = SvCUR(cv); /* in bytes */
1540 U32 is_utf8 = SvUTF8(cv);
f703fc96 1541
5b36e945 1542C<SvPVX(cv)> contains just the sub name itself, not including the package.
d8893903
FC
1543For an AUTOLOAD routine in UNIVERSAL or one of its superclasses,
1544C<CvSTASH(cv)> returns NULL during a method call on a nonexistent package.
5b36e945
FC
1545
1546B<Note>: Setting $AUTOLOAD stopped working in 5.6.1, which did not support
1547XS AUTOLOAD subs at all. Perl 5.8.0 introduced the use of fields in the
1548XSUB itself. Perl 5.16.0 restored the setting of $AUTOLOAD. If you need
1549to support 5.8-5.14, use the XSUB's fields.
1550
5f05dabc 1551=head2 Calling Perl Routines from within C Programs
a0d0e21e
LW
1552
1553There are four routines that can be used to call a Perl subroutine from
1554within a C program. These four are:
1555
954c1994
GS
1556 I32 call_sv(SV*, I32);
1557 I32 call_pv(const char*, I32);
1558 I32 call_method(const char*, I32);
5aaab254 1559 I32 call_argv(const char*, I32, char**);
a0d0e21e 1560
954c1994 1561The routine most often used is C<call_sv>. The C<SV*> argument
d1b91892
AD
1562contains either the name of the Perl subroutine to be called, or a
1563reference to the subroutine. The second argument consists of flags
1564that control the context in which the subroutine is called, whether
1565or not the subroutine is being passed arguments, how errors should be
1566trapped, and how to treat return values.
a0d0e21e
LW
1567
1568All four routines return the number of arguments that the subroutine returned
1569on the Perl stack.
1570
9a68f1db 1571These routines used to be called C<perl_call_sv>, etc., before Perl v5.6.0,
954c1994
GS
1572but those names are now deprecated; macros of the same name are provided for
1573compatibility.
1574
1575When using any of these routines (except C<call_argv>), the programmer
d1b91892
AD
1576must manipulate the Perl stack. These include the following macros and
1577functions:
a0d0e21e
LW
1578
1579 dSP
924508f0 1580 SP
a0d0e21e
LW
1581 PUSHMARK()
1582 PUTBACK
1583 SPAGAIN
1584 ENTER
1585 SAVETMPS
1586 FREETMPS
1587 LEAVE
1588 XPUSH*()
cb1a09d0 1589 POP*()
a0d0e21e 1590
5f05dabc
PP
1591For a detailed description of calling conventions from C to Perl,
1592consult L<perlcall>.
a0d0e21e 1593
5f05dabc 1594=head2 Memory Allocation
a0d0e21e 1595
06f6df17
RGS
1596=head3 Allocation
1597
86058a2d
GS
1598All memory meant to be used with the Perl API functions should be manipulated
1599using the macros described in this section. The macros provide the necessary
1600transparency between differences in the actual malloc implementation that is
1601used within perl.
1602
1603It is suggested that you enable the version of malloc that is distributed
5f05dabc 1604with Perl. It keeps pools of various sizes of unallocated memory in
07fa94a1
JO
1605order to satisfy allocation requests more quickly. However, on some
1606platforms, it may cause spurious malloc or free errors.
d1b91892 1607
06f6df17
RGS
1608The following three macros are used to initially allocate memory :
1609
9f653bb5
SH
1610 Newx(pointer, number, type);
1611 Newxc(pointer, number, type, cast);
1612 Newxz(pointer, number, type);
d1b91892 1613
9f653bb5 1614The first argument C<pointer> should be the name of a variable that will
5f05dabc 1615point to the newly allocated memory.
d1b91892 1616
9f653bb5 1617The second and third arguments C<number> and C<type> specify how many of
d1b91892 1618the specified type of data structure should be allocated. The argument
9f653bb5 1619C<type> is passed to C<sizeof>. The final argument to C<Newxc>, C<cast>,
d1b91892
AD
1620should be used if the C<pointer> argument is different from the C<type>
1621argument.
1622
9f653bb5 1623Unlike the C<Newx> and C<Newxc> macros, the C<Newxz> macro calls C<memzero>
d1b91892
AD
1624to zero out all the newly allocated memory.
1625
06f6df17
RGS
1626=head3 Reallocation
1627
d1b91892
AD
1628 Renew(pointer, number, type);
1629 Renewc(pointer, number, type, cast);
1630 Safefree(pointer)
1631
1632These three macros are used to change a memory buffer size or to free a
1633piece of memory no longer needed. The arguments to C<Renew> and C<Renewc>
1634match those of C<New> and C<Newc> with the exception of not needing the
1635"magic cookie" argument.
1636
06f6df17
RGS
1637=head3 Moving
1638
d1b91892
AD
1639 Move(source, dest, number, type);
1640 Copy(source, dest, number, type);
1641 Zero(dest, number, type);
1642
1643These three macros are used to move, copy, or zero out previously allocated
1644memory. The C<source> and C<dest> arguments point to the source and
1645destination starting points. Perl will move, copy, or zero out C<number>
1646instances of the size of the C<type> data structure (using the C<sizeof>
1647function).
a0d0e21e 1648
5f05dabc 1649=head2 PerlIO
ce3d39e2 1650
da8c5729 1651The most recent development releases of Perl have been experimenting with
5f05dabc
PP
1652removing Perl's dependency on the "normal" standard I/O suite and allowing
1653other stdio implementations to be used. This involves creating a new
1654abstraction layer that then calls whichever implementation of stdio Perl
68dc0745 1655was compiled with. All XSUBs should now use the functions in the PerlIO
5f05dabc
PP
1656abstraction layer and not make any assumptions about what kind of stdio
1657is being used.
1658
1659For a complete description of the PerlIO abstraction, consult L<perlapio>.
1660
8ebc5c01 1661=head2 Putting a C value on Perl stack
ce3d39e2
IZ
1662
1663A lot of opcodes (this is an elementary operation in the internal perl
1664stack machine) put an SV* on the stack. However, as an optimization
1665the corresponding SV is (usually) not recreated each time. The opcodes
1666reuse specially assigned SVs (I<target>s) which are (as a corollary)
1667not constantly freed/created.
1668
0a753a76 1669Each of the targets is created only once (but see
ce3d39e2
IZ
1670L<Scratchpads and recursion> below), and when an opcode needs to put
1671an integer, a double, or a string on stack, it just sets the
1672corresponding parts of its I<target> and puts the I<target> on stack.
1673
1674The macro to put this target on stack is C<PUSHTARG>, and it is
1675directly used in some opcodes, as well as indirectly in zillions of
d82b684c 1676others, which use it via C<(X)PUSH[iunp]>.
ce3d39e2 1677
1bd1c0d5
SC
1678Because the target is reused, you must be careful when pushing multiple
1679values on the stack. The following code will not do what you think:
1680
1681 XPUSHi(10);
1682 XPUSHi(20);
1683
1684This translates as "set C<TARG> to 10, push a pointer to C<TARG> onto
1685the stack; set C<TARG> to 20, push a pointer to C<TARG> onto the stack".
1686At the end of the operation, the stack does not contain the values 10
1687and 20, but actually contains two pointers to C<TARG>, which we have set
d82b684c 1688to 20.
1bd1c0d5 1689
d82b684c
SH
1690If you need to push multiple different values then you should either use
1691the C<(X)PUSHs> macros, or else use the new C<m(X)PUSH[iunp]> macros,
1692none of which make use of C<TARG>. The C<(X)PUSHs> macros simply push an
1693SV* on the stack, which, as noted under L</XSUBs and the Argument Stack>,
1694will often need to be "mortal". The new C<m(X)PUSH[iunp]> macros make
1695this a little easier to achieve by creating a new mortal for you (via
1696C<(X)PUSHmortal>), pushing that onto the stack (extending it if necessary
1697in the case of the C<mXPUSH[iunp]> macros), and then setting its value.
1698Thus, instead of writing this to "fix" the example above:
1699
1700 XPUSHs(sv_2mortal(newSViv(10)))
1701 XPUSHs(sv_2mortal(newSViv(20)))
1702
1703you can simply write:
1704
1705 mXPUSHi(10)
1706 mXPUSHi(20)
1707
1708On a related note, if you do use C<(X)PUSH[iunp]>, then you're going to
1bd1c0d5 1709need a C<dTARG> in your variable declarations so that the C<*PUSH*>
d82b684c
SH
1710macros can make use of the local variable C<TARG>. See also C<dTARGET>
1711and C<dXSTARG>.
1bd1c0d5 1712
8ebc5c01 1713=head2 Scratchpads
ce3d39e2 1714
54310121 1715The question remains on when the SVs which are I<target>s for opcodes
ac036724 1716are created. The answer is that they are created when the current
1717unit--a subroutine or a file (for opcodes for statements outside of
1718subroutines)--is compiled. During this time a special anonymous Perl
1719array is created, which is called a scratchpad for the current unit.
ce3d39e2 1720
54310121 1721A scratchpad keeps SVs which are lexicals for the current unit and are
d777b41a
FC
1722targets for opcodes. A previous version of this document
1723stated that one can deduce that an SV lives on a scratchpad
ce3d39e2 1724by looking on its flags: lexicals have C<SVs_PADMY> set, and
d777b41a
FC
1725I<target>s have C<SVs_PADTMP> set. But this have never been fully true.
1726C<SVs_PADMY> could be set on a variable that no longer resides in any pad.
1727While I<target>s do have C<SVs_PADTMP> set, it can also be set on variables
1728that have never resided in a pad, but nonetheless act like I<target>s.
ce3d39e2 1729
54310121
PP
1730The correspondence between OPs and I<target>s is not 1-to-1. Different
1731OPs in the compile tree of the unit can use the same target, if this
ce3d39e2
IZ
1732would not conflict with the expected life of the temporary.
1733
2ae324a7 1734=head2 Scratchpads and recursion
ce3d39e2
IZ
1735
1736In fact it is not 100% true that a compiled unit contains a pointer to
1737the scratchpad AV. In fact it contains a pointer to an AV of
1738(initially) one element, and this element is the scratchpad AV. Why do
1739we need an extra level of indirection?
1740
9a68f1db 1741The answer is B<recursion>, and maybe B<threads>. Both
ce3d39e2
IZ
1742these can create several execution pointers going into the same
1743subroutine. For the subroutine-child not write over the temporaries
1744for the subroutine-parent (lifespan of which covers the call to the
1745child), the parent and the child should have different
1746scratchpads. (I<And> the lexicals should be separate anyway!)
1747
5f05dabc
PP
1748So each subroutine is born with an array of scratchpads (of length 1).
1749On each entry to the subroutine it is checked that the current
ce3d39e2
IZ
1750depth of the recursion is not more than the length of this array, and
1751if it is, new scratchpad is created and pushed into the array.
1752
1753The I<target>s on this scratchpad are C<undef>s, but they are already
1754marked with correct flags.
1755
0a753a76
PP
1756=head1 Compiled code
1757
1758=head2 Code tree
1759
1760Here we describe the internal form your code is converted to by
1761Perl. Start with a simple example:
1762
1763 $a = $b + $c;
1764
1765This is converted to a tree similar to this one:
1766
1767 assign-to
1768 / \
1769 + $a
1770 / \
1771 $b $c
1772
7b8d334a 1773(but slightly more complicated). This tree reflects the way Perl
0a753a76
PP
1774parsed your code, but has nothing to do with the execution order.
1775There is an additional "thread" going through the nodes of the tree
1776which shows the order of execution of the nodes. In our simplified
1777example above it looks like:
1778
1779 $b ---> $c ---> + ---> $a ---> assign-to
1780
1781But with the actual compile tree for C<$a = $b + $c> it is different:
1782some nodes I<optimized away>. As a corollary, though the actual tree
1783contains more nodes than our simplified example, the execution order
1784is the same as in our example.
1785
1786=head2 Examining the tree
1787
06f6df17
RGS
1788If you have your perl compiled for debugging (usually done with
1789C<-DDEBUGGING> on the C<Configure> command line), you may examine the
0a753a76
PP
1790compiled tree by specifying C<-Dx> on the Perl command line. The
1791output takes several lines per node, and for C<$b+$c> it looks like
1792this:
1793
1794 5 TYPE = add ===> 6
1795 TARG = 1
1796 FLAGS = (SCALAR,KIDS)
1797 {
1798 TYPE = null ===> (4)
1799 (was rv2sv)
1800 FLAGS = (SCALAR,KIDS)
1801 {
1802 3 TYPE = gvsv ===> 4
1803 FLAGS = (SCALAR)
1804 GV = main::b
1805 }
1806 }
1807 {
1808 TYPE = null ===> (5)
1809 (was rv2sv)
1810 FLAGS = (SCALAR,KIDS)
1811 {
1812 4 TYPE = gvsv ===> 5
1813 FLAGS = (SCALAR)
1814 GV = main::c
1815 }
1816 }
1817
1818This tree has 5 nodes (one per C<TYPE> specifier), only 3 of them are
1819not optimized away (one per number in the left column). The immediate
1820children of the given node correspond to C<{}> pairs on the same level
1821of indentation, thus this listing corresponds to the tree:
1822
1823 add
1824 / \
1825 null null
1826 | |
1827 gvsv gvsv
1828
1829The execution order is indicated by C<===E<gt>> marks, thus it is C<3
18304 5 6> (node C<6> is not included into above listing), i.e.,
1831C<gvsv gvsv add whatever>.
1832
9afa14e3
SC
1833Each of these nodes represents an op, a fundamental operation inside the
1834Perl core. The code which implements each operation can be found in the
1835F<pp*.c> files; the function which implements the op with type C<gvsv>
1836is C<pp_gvsv>, and so on. As the tree above shows, different ops have
1837different numbers of children: C<add> is a binary operator, as one would
1838expect, and so has two children. To accommodate the various different
1839numbers of children, there are various types of op data structure, and
1840they link together in different ways.
1841
1842The simplest type of op structure is C<OP>: this has no children. Unary
1843operators, C<UNOP>s, have one child, and this is pointed to by the
1844C<op_first> field. Binary operators (C<BINOP>s) have not only an
1845C<op_first> field but also an C<op_last> field. The most complex type of
1846op is a C<LISTOP>, which has any number of children. In this case, the
1847first child is pointed to by C<op_first> and the last child by
1848C<op_last>. The children in between can be found by iteratively
1849following the C<op_sibling> pointer from the first child to the last.
1850
1851There are also two other op types: a C<PMOP> holds a regular expression,
1852and has no children, and a C<LOOP> may or may not have children. If the
1853C<op_children> field is non-zero, it behaves like a C<LISTOP>. To
1854complicate matters, if a C<UNOP> is actually a C<null> op after
1855optimization (see L</Compile pass 2: context propagation>) it will still
1856have children in accordance with its former type.
1857
06f6df17
RGS
1858Another way to examine the tree is to use a compiler back-end module, such
1859as L<B::Concise>.
1860
0a753a76
PP
1861=head2 Compile pass 1: check routines
1862
8870b5c7
GS
1863The tree is created by the compiler while I<yacc> code feeds it
1864the constructions it recognizes. Since I<yacc> works bottom-up, so does
0a753a76
PP
1865the first pass of perl compilation.
1866
1867What makes this pass interesting for perl developers is that some
1868optimization may be performed on this pass. This is optimization by
8870b5c7 1869so-called "check routines". The correspondence between node names
0a753a76
PP
1870and corresponding check routines is described in F<opcode.pl> (do not
1871forget to run C<make regen_headers> if you modify this file).
1872
1873A check routine is called when the node is fully constructed except
7b8d334a 1874for the execution-order thread. Since at this time there are no
0a753a76
PP
1875back-links to the currently constructed node, one can do most any
1876operation to the top-level node, including freeing it and/or creating
1877new nodes above/below it.
1878
1879The check routine returns the node which should be inserted into the
1880tree (if the top-level node was not modified, check routine returns
1881its argument).
1882
1883By convention, check routines have names C<ck_*>. They are usually
1884called from C<new*OP> subroutines (or C<convert>) (which in turn are
1885called from F<perly.y>).
1886
1887=head2 Compile pass 1a: constant folding
1888
1889Immediately after the check routine is called the returned node is
1890checked for being compile-time executable. If it is (the value is
1891judged to be constant) it is immediately executed, and a I<constant>
1892node with the "return value" of the corresponding subtree is
1893substituted instead. The subtree is deleted.
1894
1895If constant folding was not performed, the execution-order thread is
1896created.
1897
1898=head2 Compile pass 2: context propagation
1899
1900When a context for a part of compile tree is known, it is propagated
a3cb178b 1901down through the tree. At this time the context can have 5 values
0a753a76
PP
1902(instead of 2 for runtime context): void, boolean, scalar, list, and
1903lvalue. In contrast with the pass 1 this pass is processed from top
1904to bottom: a node's context determines the context for its children.
1905
1906Additional context-dependent optimizations are performed at this time.
1907Since at this moment the compile tree contains back-references (via
1908"thread" pointers), nodes cannot be free()d now. To allow
1909optimized-away nodes at this stage, such nodes are null()ified instead
1910of free()ing (i.e. their type is changed to OP_NULL).
1911
1912=head2 Compile pass 3: peephole optimization
1913
1914After the compile tree for a subroutine (or for an C<eval> or a file)
1915is created, an additional pass over the code is performed. This pass
1916is neither top-down or bottom-up, but in the execution order (with
9ea12537
Z
1917additional complications for conditionals). Optimizations performed
1918at this stage are subject to the same restrictions as in the pass 2.
1919
1920Peephole optimizations are done by calling the function pointed to
1921by the global variable C<PL_peepp>. By default, C<PL_peepp> just
1922calls the function pointed to by the global variable C<PL_rpeepp>.
1923By default, that performs some basic op fixups and optimisations along
1924the execution-order op chain, and recursively calls C<PL_rpeepp> for
1925each side chain of ops (resulting from conditionals). Extensions may
1926provide additional optimisations or fixups, hooking into either the
1927per-subroutine or recursive stage, like this:
1928
1929 static peep_t prev_peepp;
1930 static void my_peep(pTHX_ OP *o)
1931 {
1932 /* custom per-subroutine optimisation goes here */
f0358462 1933 prev_peepp(aTHX_ o);
9ea12537
Z
1934 /* custom per-subroutine optimisation may also go here */
1935 }
1936 BOOT:
1937 prev_peepp = PL_peepp;
1938 PL_peepp = my_peep;
1939
1940 static peep_t prev_rpeepp;
1941 static void my_rpeep(pTHX_ OP *o)
1942 {
1943 OP *orig_o = o;
1944 for(; o; o = o->op_next) {
1945 /* custom per-op optimisation goes here */
1946 }
f0358462 1947 prev_rpeepp(aTHX_ orig_o);
9ea12537
Z
1948 }
1949 BOOT:
1950 prev_rpeepp = PL_rpeepp;
1951 PL_rpeepp = my_rpeep;
0a753a76 1952
1ba7f851
PJ
1953=head2 Pluggable runops
1954
1955The compile tree is executed in a runops function. There are two runops
1388f78e
RGS
1956functions, in F<run.c> and in F<dump.c>. C<Perl_runops_debug> is used
1957with DEBUGGING and C<Perl_runops_standard> is used otherwise. For fine
1958control over the execution of the compile tree it is possible to provide
1959your own runops function.
1ba7f851
PJ
1960
1961It's probably best to copy one of the existing runops functions and
1962change it to suit your needs. Then, in the BOOT section of your XS
1963file, add the line:
1964
1965 PL_runops = my_runops;
1966
1967This function should be as efficient as possible to keep your programs
1968running as fast as possible.
1969
fd85fad2
BM
1970=head2 Compile-time scope hooks
1971
1972As of perl 5.14 it is possible to hook into the compile-time lexical
1973scope mechanism using C<Perl_blockhook_register>. This is used like
1974this:
1975
1976 STATIC void my_start_hook(pTHX_ int full);
1977 STATIC BHK my_hooks;
1978
1979 BOOT:
a88d97bf 1980 BhkENTRY_set(&my_hooks, bhk_start, my_start_hook);
fd85fad2
BM
1981 Perl_blockhook_register(aTHX_ &my_hooks);
1982
1983This will arrange to have C<my_start_hook> called at the start of
1984compiling every lexical scope. The available hooks are:
1985
1986=over 4
1987
a88d97bf 1988=item C<void bhk_start(pTHX_ int full)>
fd85fad2
BM
1989
1990This is called just after starting a new lexical scope. Note that Perl
1991code like
1992
1993 if ($x) { ... }
1994
1995creates two scopes: the first starts at the C<(> and has C<full == 1>,
1996the second starts at the C<{> and has C<full == 0>. Both end at the
1997C<}>, so calls to C<start> and C<pre/post_end> will match. Anything
1998pushed onto the save stack by this hook will be popped just before the
1999scope ends (between the C<pre_> and C<post_end> hooks, in fact).
2000
a88d97bf 2001=item C<void bhk_pre_end(pTHX_ OP **o)>
fd85fad2
BM
2002
2003This is called at the end of a lexical scope, just before unwinding the
2004stack. I<o> is the root of the optree representing the scope; it is a
2005double pointer so you can replace the OP if you need to.
2006
a88d97bf 2007=item C<void bhk_post_end(pTHX_ OP **o)>
fd85fad2
BM
2008
2009This is called at the end of a lexical scope, just after unwinding the
2010stack. I<o> is as above. Note that it is possible for calls to C<pre_>
2011and C<post_end> to nest, if there is something on the save stack that
2012calls string eval.
2013
a88d97bf 2014=item C<void bhk_eval(pTHX_ OP *const o)>
fd85fad2
BM
2015
2016This is called just before starting to compile an C<eval STRING>, C<do
2017FILE>, C<require> or C<use>, after the eval has been set up. I<o> is the
2018OP that requested the eval, and will normally be an C<OP_ENTEREVAL>,
2019C<OP_DOFILE> or C<OP_REQUIRE>.
2020
2021=back
2022
2023Once you have your hook functions, you need a C<BHK> structure to put
2024them in. It's best to allocate it statically, since there is no way to
2025free it once it's registered. The function pointers should be inserted
2026into this structure using the C<BhkENTRY_set> macro, which will also set
2027flags indicating which entries are valid. If you do need to allocate
2028your C<BHK> dynamically for some reason, be sure to zero it before you
2029start.
2030
2031Once registered, there is no mechanism to switch these hooks off, so if
2032that is necessary you will need to do this yourself. An entry in C<%^H>
a3e07c87
BM
2033is probably the best way, so the effect is lexically scoped; however it
2034is also possible to use the C<BhkDISABLE> and C<BhkENABLE> macros to
2035temporarily switch entries on and off. You should also be aware that
2036generally speaking at least one scope will have opened before your
2037extension is loaded, so you will see some C<pre/post_end> pairs that
2038didn't have a matching C<start>.
fd85fad2 2039
9afa14e3
SC
2040=head1 Examining internal data structures with the C<dump> functions
2041
2042To aid debugging, the source file F<dump.c> contains a number of
2043functions which produce formatted output of internal data structures.
2044
2045The most commonly used of these functions is C<Perl_sv_dump>; it's used
2046for dumping SVs, AVs, HVs, and CVs. The C<Devel::Peek> module calls
2047C<sv_dump> to produce debugging output from Perl-space, so users of that
00aadd71 2048module should already be familiar with its format.
9afa14e3
SC
2049
2050C<Perl_op_dump> can be used to dump an C<OP> structure or any of its
210b36aa 2051derivatives, and produces output similar to C<perl -Dx>; in fact,
9afa14e3
SC
2052C<Perl_dump_eval> will dump the main root of the code being evaluated,
2053exactly like C<-Dx>.
2054
2055Other useful functions are C<Perl_dump_sub>, which turns a C<GV> into an
2056op tree, C<Perl_dump_packsubs> which calls C<Perl_dump_sub> on all the
2057subroutines in a package like so: (Thankfully, these are all xsubs, so
2058there is no op tree)
2059
2060 (gdb) print Perl_dump_packsubs(PL_defstash)
2061
2062 SUB attributes::bootstrap = (xsub 0x811fedc 0)
2063
2064 SUB UNIVERSAL::can = (xsub 0x811f50c 0)
2065
2066 SUB UNIVERSAL::isa = (xsub 0x811f304 0)
2067
2068 SUB UNIVERSAL::VERSION = (xsub 0x811f7ac 0)
2069
2070 SUB DynaLoader::boot_DynaLoader = (xsub 0x805b188 0)
2071
2072and C<Perl_dump_all>, which dumps all the subroutines in the stash and
2073the op tree of the main root.
2074
954c1994 2075=head1 How multiple interpreters and concurrency are supported
ee072b34 2076
ee072b34
GS
2077=head2 Background and PERL_IMPLICIT_CONTEXT
2078
2079The Perl interpreter can be regarded as a closed box: it has an API
2080for feeding it code or otherwise making it do things, but it also has
2081functions for its own use. This smells a lot like an object, and
2082there are ways for you to build Perl so that you can have multiple
acfe0abc
GS
2083interpreters, with one interpreter represented either as a C structure,
2084or inside a thread-specific structure. These structures contain all
2085the context, the state of that interpreter.
2086
7b52221d
RGS
2087One macro controls the major Perl build flavor: MULTIPLICITY. The
2088MULTIPLICITY build has a C structure that packages all the interpreter
2089state. With multiplicity-enabled perls, PERL_IMPLICIT_CONTEXT is also
2090normally defined, and enables the support for passing in a "hidden" first
2091argument that represents all three data structures. MULTIPLICITY makes
1a64a5e6 2092multi-threaded perls possible (with the ithreads threading model, related
7b52221d 2093to the macro USE_ITHREADS.)
54aff467 2094
27da23d5
JH
2095Two other "encapsulation" macros are the PERL_GLOBAL_STRUCT and
2096PERL_GLOBAL_STRUCT_PRIVATE (the latter turns on the former, and the
2097former turns on MULTIPLICITY.) The PERL_GLOBAL_STRUCT causes all the
2098internal variables of Perl to be wrapped inside a single global struct,
2099struct perl_vars, accessible as (globals) &PL_Vars or PL_VarsPtr or
2100the function Perl_GetVars(). The PERL_GLOBAL_STRUCT_PRIVATE goes
2101one step further, there is still a single struct (allocated in main()
2102either from heap or from stack) but there are no global data symbols
3bf17896 2103pointing to it. In either case the global struct should be initialized
27da23d5
JH
2104as the very first thing in main() using Perl_init_global_struct() and
2105correspondingly tear it down after perl_free() using Perl_free_global_struct(),
2106please see F<miniperlmain.c> for usage details. You may also need
2107to use C<dVAR> in your coding to "declare the global variables"
2108when you are using them. dTHX does this for you automatically.
2109
bc028b6b
JH
2110To see whether you have non-const data you can use a BSD-compatible C<nm>:
2111
2112 nm libperl.a | grep -v ' [TURtr] '
2113
2114If this displays any C<D> or C<d> symbols, you have non-const data.
2115
27da23d5
JH
2116For backward compatibility reasons defining just PERL_GLOBAL_STRUCT
2117doesn't actually hide all symbols inside a big global struct: some
2118PerlIO_xxx vtables are left visible. The PERL_GLOBAL_STRUCT_PRIVATE
2119then hides everything (see how the PERLIO_FUNCS_DECL is used).
2120
54aff467 2121All this obviously requires a way for the Perl internal functions to be
acfe0abc 2122either subroutines taking some kind of structure as the first
ee072b34 2123argument, or subroutines taking nothing as the first argument. To
acfe0abc 2124enable these two very different ways of building the interpreter,
ee072b34
GS
2125the Perl source (as it does in so many other situations) makes heavy
2126use of macros and subroutine naming conventions.
2127
54aff467 2128First problem: deciding which functions will be public API functions and
00aadd71 2129which will be private. All functions whose names begin C<S_> are private
954c1994
GS
2130(think "S" for "secret" or "static"). All other functions begin with
2131"Perl_", but just because a function begins with "Perl_" does not mean it is
00aadd71
NIS
2132part of the API. (See L</Internal Functions>.) The easiest way to be B<sure> a
2133function is part of the API is to find its entry in L<perlapi>.
2134If it exists in L<perlapi>, it's part of the API. If it doesn't, and you
2135think it should be (i.e., you need it for your extension), send mail via
a422fd2d 2136L<perlbug> explaining why you think it should be.
ee072b34
GS
2137
2138Second problem: there must be a syntax so that the same subroutine
2139declarations and calls can pass a structure as their first argument,
2140or pass nothing. To solve this, the subroutines are named and
2141declared in a particular way. Here's a typical start of a static
2142function used within the Perl guts:
2143
2144 STATIC void
2145 S_incline(pTHX_ char *s)
2146
acfe0abc 2147STATIC becomes "static" in C, and may be #define'd to nothing in some
da8c5729 2148configurations in the future.
ee072b34 2149
651a3225
GS
2150A public function (i.e. part of the internal API, but not necessarily
2151sanctioned for use in extensions) begins like this:
ee072b34
GS
2152
2153 void
2307c6d0 2154 Perl_sv_setiv(pTHX_ SV* dsv, IV num)
ee072b34 2155
0147cd53 2156C<pTHX_> is one of a number of macros (in F<perl.h>) that hide the
ee072b34
GS
2157details of the interpreter's context. THX stands for "thread", "this",
2158or "thingy", as the case may be. (And no, George Lucas is not involved. :-)
2159The first character could be 'p' for a B<p>rototype, 'a' for B<a>rgument,
a7486cbb
JH
2160or 'd' for B<d>eclaration, so we have C<pTHX>, C<aTHX> and C<dTHX>, and
2161their variants.
ee072b34 2162
a7486cbb
JH
2163When Perl is built without options that set PERL_IMPLICIT_CONTEXT, there is no
2164first argument containing the interpreter's context. The trailing underscore
ee072b34
GS
2165in the pTHX_ macro indicates that the macro expansion needs a comma
2166after the context argument because other arguments follow it. If
2167PERL_IMPLICIT_CONTEXT is not defined, pTHX_ will be ignored, and the
54aff467
GS
2168subroutine is not prototyped to take the extra argument. The form of the
2169macro without the trailing underscore is used when there are no additional
ee072b34
GS
2170explicit arguments.
2171
54aff467 2172When a core function calls another, it must pass the context. This
2307c6d0 2173is normally hidden via macros. Consider C<sv_setiv>. It expands into
ee072b34
GS
2174something like this:
2175
2307c6d0
SB
2176 #ifdef PERL_IMPLICIT_CONTEXT
2177 #define sv_setiv(a,b) Perl_sv_setiv(aTHX_ a, b)
ee072b34 2178 /* can't do this for vararg functions, see below */
2307c6d0
SB
2179 #else
2180 #define sv_setiv Perl_sv_setiv
2181 #endif
ee072b34
GS
2182
2183This works well, and means that XS authors can gleefully write:
2184
2307c6d0 2185 sv_setiv(foo, bar);
ee072b34
GS
2186
2187and still have it work under all the modes Perl could have been
2188compiled with.
2189
ee072b34
GS
2190This doesn't work so cleanly for varargs functions, though, as macros
2191imply that the number of arguments is known in advance. Instead we
2192either need to spell them out fully, passing C<aTHX_> as the first
2193argument (the Perl core tends to do this with functions like
2194Perl_warner), or use a context-free version.
2195
2196The context-free version of Perl_warner is called
2197Perl_warner_nocontext, and does not take the extra argument. Instead
2198it does dTHX; to get the context from thread-local storage. We
2199C<#define warner Perl_warner_nocontext> so that extensions get source
2200compatibility at the expense of performance. (Passing an arg is
2201cheaper than grabbing it from thread-local storage.)
2202
acfe0abc 2203You can ignore [pad]THXx when browsing the Perl headers/sources.
ee072b34
GS
2204Those are strictly for use within the core. Extensions and embedders
2205need only be aware of [pad]THX.
2206
a7486cbb
JH
2207=head2 So what happened to dTHR?
2208
2209C<dTHR> was introduced in perl 5.005 to support the older thread model.
2210The older thread model now uses the C<THX> mechanism to pass context
2211pointers around, so C<dTHR> is not useful any more. Perl 5.6.0 and
2212later still have it for backward source compatibility, but it is defined
2213to be a no-op.
2214
ee072b34
GS
2215=head2 How do I use all this in extensions?
2216
2217When Perl is built with PERL_IMPLICIT_CONTEXT, extensions that call
2218any functions in the Perl API will need to pass the initial context
2219argument somehow. The kicker is that you will need to write it in
2220such a way that the extension still compiles when Perl hasn't been
2221built with PERL_IMPLICIT_CONTEXT enabled.
2222
2223There are three ways to do this. First, the easy but inefficient way,
2224which is also the default, in order to maintain source compatibility
0147cd53 2225with extensions: whenever F<XSUB.h> is #included, it redefines the aTHX
ee072b34
GS
2226and aTHX_ macros to call a function that will return the context.
2227Thus, something like:
2228
2307c6d0 2229 sv_setiv(sv, num);
ee072b34 2230
4375e838 2231in your extension will translate to this when PERL_IMPLICIT_CONTEXT is
54aff467 2232in effect:
ee072b34 2233
2307c6d0 2234 Perl_sv_setiv(Perl_get_context(), sv, num);
ee072b34 2235
54aff467 2236or to this otherwise:
ee072b34 2237
2307c6d0 2238 Perl_sv_setiv(sv, num);
ee072b34 2239
da8c5729 2240You don't have to do anything new in your extension to get this; since
2fa86c13 2241the Perl library provides Perl_get_context(), it will all just
ee072b34
GS
2242work.
2243
2244The second, more efficient way is to use the following template for
2245your Foo.xs:
2246
c52f9dcd
JH
2247 #define PERL_NO_GET_CONTEXT /* we want efficiency */
2248 #include "EXTERN.h"
2249 #include "perl.h"
2250 #include "XSUB.h"
ee072b34 2251
fd061412 2252 STATIC void my_private_function(int arg1, int arg2);
ee072b34 2253
fd061412 2254 STATIC void
c52f9dcd
JH
2255 my_private_function(int arg1, int arg2)
2256 {
2257 dTHX; /* fetch context */
2258 ... call many Perl API functions ...
2259 }
ee072b34
GS
2260
2261 [... etc ...]
2262
c52f9dcd 2263 MODULE = Foo PACKAGE = Foo
ee072b34 2264
c52f9dcd 2265 /* typical XSUB */
ee072b34 2266
c52f9dcd
JH
2267 void
2268 my_xsub(arg)
2269 int arg
2270 CODE:
2271 my_private_function(arg, 10);
ee072b34
GS
2272
2273Note that the only two changes from the normal way of writing an
2274extension is the addition of a C<#define PERL_NO_GET_CONTEXT> before
2275including the Perl headers, followed by a C<dTHX;> declaration at
2276the start of every function that will call the Perl API. (You'll
2277know which functions need this, because the C compiler will complain
2278that there's an undeclared identifier in those functions.) No changes
2279are needed for the XSUBs themselves, because the XS() macro is
2280correctly defined to pass in the implicit context if needed.
2281
2282The third, even more efficient way is to ape how it is done within
2283the Perl guts:
2284
2285
c52f9dcd
JH
2286 #define PERL_NO_GET_CONTEXT /* we want efficiency */
2287 #include "EXTERN.h"
2288 #include "perl.h"
2289 #include "XSUB.h"
ee072b34
GS
2290
2291 /* pTHX_ only needed for functions that call Perl API */
fd061412 2292 STATIC void my_private_function(pTHX_ int arg1, int arg2);
ee072b34 2293
fd061412 2294 STATIC void
c52f9dcd
JH
2295 my_private_function(pTHX_ int arg1, int arg2)
2296 {
2297 /* dTHX; not needed here, because THX is an argument */
2298 ... call Perl API functions ...
2299 }
ee072b34
GS
2300
2301 [... etc ...]
2302
c52f9dcd 2303 MODULE = Foo PACKAGE = Foo
ee072b34 2304
c52f9dcd 2305 /* typical XSUB */
ee072b34 2306
c52f9dcd
JH
2307 void
2308 my_xsub(arg)
2309 int arg
2310 CODE:
2311 my_private_function(aTHX_ arg, 10);
ee072b34
GS
2312
2313This implementation never has to fetch the context using a function
2314call, since it is always passed as an extra argument. Depending on
2315your needs for simplicity or efficiency, you may mix the previous
2316two approaches freely.
2317
651a3225
GS
2318Never add a comma after C<pTHX> yourself--always use the form of the
2319macro with the underscore for functions that take explicit arguments,
2320or the form without the argument for functions with no explicit arguments.
ee072b34 2321
27da23d5
JH
2322If one is compiling Perl with the C<-DPERL_GLOBAL_STRUCT> the C<dVAR>
2323definition is needed if the Perl global variables (see F<perlvars.h>
2324or F<globvar.sym>) are accessed in the function and C<dTHX> is not
2325used (the C<dTHX> includes the C<dVAR> if necessary). One notices
2326the need for C<dVAR> only with the said compile-time define, because
2327otherwise the Perl global variables are visible as-is.
2328
a7486cbb
JH
2329=head2 Should I do anything special if I call perl from multiple threads?
2330
2331If you create interpreters in one thread and then proceed to call them in
2332another, you need to make sure perl's own Thread Local Storage (TLS) slot is
2333initialized correctly in each of those threads.
2334
2335The C<perl_alloc> and C<perl_clone> API functions will automatically set
2336the TLS slot to the interpreter they created, so that there is no need to do
2337anything special if the interpreter is always accessed in the same thread that
2338created it, and that thread did not create or call any other interpreters
2339afterwards. If that is not the case, you have to set the TLS slot of the
2340thread before calling any functions in the Perl API on that particular
2341interpreter. This is done by calling the C<PERL_SET_CONTEXT> macro in that
2342thread as the first thing you do:
2343
2344 /* do this before doing anything else with some_perl */
2345 PERL_SET_CONTEXT(some_perl);
2346
2347 ... other Perl API calls on some_perl go here ...
2348
ee072b34
GS
2349=head2 Future Plans and PERL_IMPLICIT_SYS
2350
2351Just as PERL_IMPLICIT_CONTEXT provides a way to bundle up everything
2352that the interpreter knows about itself and pass it around, so too are
2353there plans to allow the interpreter to bundle up everything it knows
2354about the environment it's running on. This is enabled with the
7b52221d
RGS
2355PERL_IMPLICIT_SYS macro. Currently it only works with USE_ITHREADS on
2356Windows.
ee072b34
GS
2357
2358This allows the ability to provide an extra pointer (called the "host"
2359environment) for all the system calls. This makes it possible for
2360all the system stuff to maintain their own state, broken down into
2361seven C structures. These are thin wrappers around the usual system
0147cd53 2362calls (see F<win32/perllib.c>) for the default perl executable, but for a
ee072b34
GS
2363more ambitious host (like the one that would do fork() emulation) all
2364the extra work needed to pretend that different interpreters are
2365actually different "processes", would be done here.
2366
2367The Perl engine/interpreter and the host are orthogonal entities.
2368There could be one or more interpreters in a process, and one or
2369more "hosts", with free association between them.
2370
a422fd2d
SC
2371=head1 Internal Functions
2372
2373All of Perl's internal functions which will be exposed to the outside
06f6df17 2374world are prefixed by C<Perl_> so that they will not conflict with XS
a422fd2d
SC
2375functions or functions used in a program in which Perl is embedded.
2376Similarly, all global variables begin with C<PL_>. (By convention,
06f6df17 2377static functions start with C<S_>.)
a422fd2d 2378
0972ecdf
DM
2379Inside the Perl core (C<PERL_CORE> defined), you can get at the functions
2380either with or without the C<Perl_> prefix, thanks to a bunch of defines
2381that live in F<embed.h>. Note that extension code should I<not> set
2382C<PERL_CORE>; this exposes the full perl internals, and is likely to cause
2383breakage of the XS in each new perl release.
2384
2385The file F<embed.h> is generated automatically from
dc9b1d22
MHM
2386F<embed.pl> and F<embed.fnc>. F<embed.pl> also creates the prototyping
2387header files for the internal functions, generates the documentation
2388and a lot of other bits and pieces. It's important that when you add
2389a new function to the core or change an existing one, you change the
2390data in the table in F<embed.fnc> as well. Here's a sample entry from
2391that table:
a422fd2d
SC
2392
2393 Apd |SV** |av_fetch |AV* ar|I32 key|I32 lval
2394
2395The second column is the return type, the third column the name. Columns
2396after that are the arguments. The first column is a set of flags:
2397
2398=over 3
2399
2400=item A
2401
1aa6ea50
JC
2402This function is a part of the public API. All such functions should also
2403have 'd', very few do not.
a422fd2d
SC
2404
2405=item p
2406
1aa6ea50
JC
2407This function has a C<Perl_> prefix; i.e. it is defined as
2408C<Perl_av_fetch>.
a422fd2d
SC
2409
2410=item d
2411
2412This function has documentation using the C<apidoc> feature which we'll
1aa6ea50 2413look at in a second. Some functions have 'd' but not 'A'; docs are good.
a422fd2d
SC
2414
2415=back
2416
2417Other available flags are:
2418
2419=over 3
2420
2421=item s
2422
1aa6ea50
JC
2423This is a static function and is defined as C<STATIC S_whatever>, and
2424usually called within the sources as C<whatever(...)>.
a422fd2d
SC
2425
2426=item n
2427
da8c5729 2428This does not need an interpreter context, so the definition has no
1aa6ea50 2429C<pTHX>, and it follows that callers don't use C<aTHX>. (See
d3a43cd8 2430L</Background and PERL_IMPLICIT_CONTEXT>.)
a422fd2d
SC
2431
2432=item r
2433
2434This function never returns; C<croak>, C<exit> and friends.
2435
2436=item f
2437
2438This function takes a variable number of arguments, C<printf> style.
2439The argument list should end with C<...>, like this:
2440
2441 Afprd |void |croak |const char* pat|...
2442
a7486cbb 2443=item M
a422fd2d 2444
00aadd71 2445This function is part of the experimental development API, and may change
a422fd2d
SC
2446or disappear without notice.
2447
2448=item o
2449
2450This function should not have a compatibility macro to define, say,
2451C<Perl_parse> to C<parse>. It must be called as C<Perl_parse>.
2452
a422fd2d
SC
2453=item x
2454
2455This function isn't exported out of the Perl core.
2456
dc9b1d22
MHM
2457=item m
2458
2459This is implemented as a macro.
2460
2461=item X
2462
2463This function is explicitly exported.
2464
2465=item E
2466
2467This function is visible to extensions included in the Perl core.
2468
2469=item b
2470
2471Binary backward compatibility; this function is a macro but also has
2472a C<Perl_> implementation (which is exported).
2473
1aa6ea50
JC
2474=item others
2475
2476See the comments at the top of C<embed.fnc> for others.
2477
a422fd2d
SC
2478=back
2479
dc9b1d22
MHM
2480If you edit F<embed.pl> or F<embed.fnc>, you will need to run
2481C<make regen_headers> to force a rebuild of F<embed.h> and other
2482auto-generated files.
a422fd2d 2483
6b4667fc 2484=head2 Formatted Printing of IVs, UVs, and NVs
9dd9db0b 2485
6b4667fc
A
2486If you are printing IVs, UVs, or NVS instead of the stdio(3) style
2487formatting codes like C<%d>, C<%ld>, C<%f>, you should use the
2488following macros for portability
9dd9db0b 2489
c52f9dcd
JH
2490 IVdf IV in decimal
2491 UVuf UV in decimal
2492 UVof UV in octal
2493 UVxf UV in hexadecimal
2494 NVef NV %e-like
2495 NVff NV %f-like
2496 NVgf NV %g-like
9dd9db0b 2497
6b4667fc
A
2498These will take care of 64-bit integers and long doubles.
2499For example:
2500
c52f9dcd 2501 printf("IV is %"IVdf"\n", iv);
6b4667fc
A
2502
2503The IVdf will expand to whatever is the correct format for the IVs.
9dd9db0b 2504
8908e76d
JH
2505If you are printing addresses of pointers, use UVxf combined
2506with PTR2UV(), do not use %lx or %p.
2507
2508=head2 Pointer-To-Integer and Integer-To-Pointer
2509
2510Because pointer size does not necessarily equal integer size,
2511use the follow macros to do it right.
2512
c52f9dcd
JH
2513 PTR2UV(pointer)
2514 PTR2IV(pointer)
2515 PTR2NV(pointer)
2516 INT2PTR(pointertotype, integer)
8908e76d
JH
2517
2518For example:
2519
c52f9dcd
JH
2520 IV iv = ...;
2521 SV *sv = INT2PTR(SV*, iv);
8908e76d
JH
2522
2523and
2524
c52f9dcd
JH
2525 AV *av = ...;
2526 UV uv = PTR2UV(av);
8908e76d 2527
0ca3a874
MHM
2528=head2 Exception Handling
2529
9b5c3821
MHM
2530There are a couple of macros to do very basic exception handling in XS
2531modules. You have to define C<NO_XSLOCKS> before including F<XSUB.h> to
2532be able to use these macros:
2533
2534 #define NO_XSLOCKS
2535 #include "XSUB.h"
2536
2537You can use these macros if you call code that may croak, but you need
2538to do some cleanup before giving control back to Perl. For example:
0ca3a874 2539
d7f8936a 2540 dXCPT; /* set up necessary variables */
0ca3a874
MHM
2541
2542 XCPT_TRY_START {
2543 code_that_may_croak();
2544 } XCPT_TRY_END
2545
2546 XCPT_CATCH
2547 {
2548 /* do cleanup here */
2549 XCPT_RETHROW;
2550 }
2551
2552Note that you always have to rethrow an exception that has been
2553caught. Using these macros, it is not possible to just catch the
2554exception and ignore it. If you have to ignore the exception, you
2555have to use the C<call_*> function.
2556
2557The advantage of using the above macros is that you don't have
2558to setup an extra function for C<call_*>, and that using these
2559macros is faster than using C<call_*>.
2560
a422fd2d
SC
2561=head2 Source Documentation
2562
2563There's an effort going on to document the internal functions and
2564automatically produce reference manuals from them - L<perlapi> is one
2565such manual which details all the functions which are available to XS
2566writers. L<perlintern> is the autogenerated manual for the functions
2567which are not part of the API and are supposedly for internal use only.
2568
2569Source documentation is created by putting POD comments into the C
2570source, like this:
2571
2572 /*
2573 =for apidoc sv_setiv
2574
2575 Copies an integer into the given SV. Does not handle 'set' magic. See
2576 C<sv_setiv_mg>.
2577
2578 =cut
2579 */
2580
2581Please try and supply some documentation if you add functions to the
2582Perl core.
2583
0d098d33
MHM
2584=head2 Backwards compatibility
2585
2586The Perl API changes over time. New functions are added or the interfaces
2587of existing functions are changed. The C<Devel::PPPort> module tries to
2588provide compatibility code for some of these changes, so XS writers don't
2589have to code it themselves when supporting multiple versions of Perl.
2590
2591C<Devel::PPPort> generates a C header file F<ppport.h> that can also
2592be run as a Perl script. To generate F<ppport.h>, run:
2593
2594 perl -MDevel::PPPort -eDevel::PPPort::WriteFile
2595
2596Besides checking existing XS code, the script can also be used to retrieve
2597compatibility information for various API calls using the C<--api-info>
2598command line switch. For example:
2599
2600 % perl ppport.h --api-info=sv_magicext
2601
2602For details, see C<perldoc ppport.h>.
2603
a422fd2d
SC
2604=head1 Unicode Support
2605
2606Perl 5.6.0 introduced Unicode support. It's important for porters and XS
2607writers to understand this support and make sure that the code they
2608write does not corrupt Unicode data.
2609
2610=head2 What B<is> Unicode, anyway?
2611
2612In the olden, less enlightened times, we all used to use ASCII. Most of
2613us did, anyway. The big problem with ASCII is that it's American. Well,
2614no, that's not actually the problem; the problem is that it's not
2615particularly useful for people who don't use the Roman alphabet. What
2616used to happen was that particular languages would stick their own
2617alphabet in the upper range of the sequence, between 128 and 255. Of
2618course, we then ended up with plenty of variants that weren't quite
2619ASCII, and the whole point of it being a standard was lost.
2620
2621Worse still, if you've got a language like Chinese or
2622Japanese that has hundreds or thousands of characters, then you really
2623can't fit them into a mere 256, so they had to forget about ASCII
2624altogether, and build their own systems using pairs of numbers to refer
2625to one character.
2626
2627To fix this, some people formed Unicode, Inc. and
2628produced a new character set containing all the characters you can
2629possibly think of and more. There are several ways of representing these
1e54db1a 2630characters, and the one Perl uses is called UTF-8. UTF-8 uses
2575c402
JW
2631a variable number of bytes to represent a character. You can learn more
2632about Unicode and Perl's Unicode model in L<perlunicode>.
a422fd2d 2633
1e54db1a 2634=head2 How can I recognise a UTF-8 string?
a422fd2d 2635
1e54db1a
JH
2636You can't. This is because UTF-8 data is stored in bytes just like
2637non-UTF-8 data. The Unicode character 200, (C<0xC8> for you hex types)
a422fd2d
SC
2638capital E with a grave accent, is represented by the two bytes
2639C<v196.172>. Unfortunately, the non-Unicode string C<chr(196).chr(172)>
2640has that byte sequence as well. So you can't tell just by looking - this
2641is what makes Unicode input an interesting problem.
2642
2575c402
JW
2643In general, you either have to know what you're dealing with, or you
2644have to guess. The API function C<is_utf8_string> can help; it'll tell
2645you if a string contains only valid UTF-8 characters. However, it can't
49f4c4e4
KW
2646do the work for you. On a character-by-character basis,
2647C<is_utf8_char_buf>
2575c402 2648will tell you whether the current character in a string is valid UTF-8.
a422fd2d 2649
1e54db1a 2650=head2 How does UTF-8 represent Unicode characters?
a422fd2d 2651
1e54db1a 2652As mentioned above, UTF-8 uses a variable number of bytes to store a
2575c402
JW
2653character. Characters with values 0...127 are stored in one byte, just
2654like good ol' ASCII. Character 128 is stored as C<v194.128>; this
a31a806a 2655continues up to character 191, which is C<v194.191>. Now we've run out of
a422fd2d
SC
2656bits (191 is binary C<10111111>) so we move on; 192 is C<v195.128>. And
2657so it goes on, moving to three bytes at character 2048.
2658
1e54db1a 2659Assuming you know you're dealing with a UTF-8 string, you can find out
a422fd2d
SC
2660how long the first character in it is with the C<UTF8SKIP> macro:
2661
2662 char *utf = "\305\233\340\240\201";
2663 I32 len;
2664
2665 len = UTF8SKIP(utf); /* len is 2 here */
2666 utf += len;
2667 len = UTF8SKIP(utf); /* len is 3 here */
2668
1e54db1a 2669Another way to skip over characters in a UTF-8 string is to use
a422fd2d
SC
2670C<utf8_hop>, which takes a string and a number of characters to skip
2671over. You're on your own about bounds checking, though, so don't use it
2672lightly.
2673
1e54db1a 2674All bytes in a multi-byte UTF-8 character will have the high bit set,
3a2263fe
RGS
2675so you can test if you need to do something special with this
2676character like this (the UTF8_IS_INVARIANT() is a macro that tests
9f98c7fe 2677whether the byte is encoded as a single byte even in UTF-8):
a422fd2d 2678
3a2263fe 2679 U8 *utf;
4b88fb76 2680 U8 *utf_end; /* 1 beyond buffer pointed to by utf */
3a2263fe 2681 UV uv; /* Note: a UV, not a U8, not a char */
95701e00 2682 STRLEN len; /* length of character in bytes */
a422fd2d 2683
3a2263fe 2684 if (!UTF8_IS_INVARIANT(*utf))
1e54db1a 2685 /* Must treat this as UTF-8 */
4b88fb76 2686 uv = utf8_to_uvchr_buf(utf, utf_end, &len);
a422fd2d
SC
2687 else
2688 /* OK to treat this character as a byte */
2689 uv = *utf;
2690
4b88fb76 2691You can also see in that example that we use C<utf8_to_uvchr_buf> to get the
95701e00 2692value of the character; the inverse function C<uvchr_to_utf8> is available
1e54db1a 2693for putting a UV into UTF-8:
a422fd2d 2694
3a2263fe 2695 if (!UTF8_IS_INVARIANT(uv))
a422fd2d 2696 /* Must treat this as UTF8 */
95701e00 2697 utf8 = uvchr_to_utf8(utf8, uv);
a422fd2d
SC
2698 else
2699 /* OK to treat this character as a byte */
2700 *utf8++ = uv;
2701
2702You B<must> convert characters to UVs using the above functions if
1e54db1a
JH
2703you're ever in a situation where you have to match UTF-8 and non-UTF-8
2704characters. You may not skip over UTF-8 characters in this case. If you
2705do this, you'll lose the ability to match hi-bit non-UTF-8 characters;
2706for instance, if your UTF-8 string contains C<v196.172>, and you skip
2707that character, you can never match a C<chr(200)> in a non-UTF-8 string.
a422fd2d
SC
2708So don't do that!
2709
1e54db1a 2710=head2 How does Perl store UTF-8 strings?
a422fd2d
SC
2711
2712Currently, Perl deals with Unicode strings and non-Unicode strings
2575c402
JW
2713slightly differently. A flag in the SV, C<SVf_UTF8>, indicates that the
2714string is internally encoded as UTF-8. Without it, the byte value is the
2715codepoint number and vice versa (in other words, the string is encoded
e1b711da
KW
2716as iso-8859-1, but C<use feature 'unicode_strings'> is needed to get iso-8859-1
2717semantics). You can check and manipulate this flag with the
2575c402 2718following macros:
a422fd2d
SC
2719
2720 SvUTF8(sv)
2721 SvUTF8_on(sv)
2722 SvUTF8_off(sv)
2723
2724This flag has an important effect on Perl's treatment of the string: if
2725Unicode data is not properly distinguished, regular expressions,
2726C<length>, C<substr> and other string handling operations will have
2727undesirable results.
2728
2729The problem comes when you have, for instance, a string that isn't
2575c402 2730flagged as UTF-8, and contains a byte sequence that could be UTF-8 -
1e54db1a 2731especially when combining non-UTF-8 and UTF-8 strings.
a422fd2d
SC
2732
2733Never forget that the C<SVf_UTF8> flag is separate to the PV value; you
2734need be sure you don't accidentally knock it off while you're
2735manipulating SVs. More specifically, you cannot expect to do this:
2736
2737 SV *sv;
2738 SV *nsv;
2739 STRLEN len;
2740 char *p;
2741
2742 p = SvPV(sv, len);
2743 frobnicate(p);
2744 nsv = newSVpvn(p, len);
2745
2746The C<char*> string does not tell you the whole story, and you can't
2747copy or reconstruct an SV just by copying the string value. Check if the
2575c402 2748old SV has the UTF8 flag set, and act accordingly:
a422fd2d
SC
2749
2750 p = SvPV(sv, len);
2751 frobnicate(p);
2752 nsv = newSVpvn(p, len);
2753 if (SvUTF8(sv))
2754 SvUTF8_on(nsv);
2755
2756In fact, your C<frobnicate> function should be made aware of whether or
1e54db1a 2757not it's dealing with UTF-8 data, so that it can handle the string
a422fd2d
SC
2758appropriately.
2759
3a2263fe 2760Since just passing an SV to an XS function and copying the data of
2575c402 2761the SV is not enough to copy the UTF8 flags, even less right is just
3a2263fe
RGS
2762passing a C<char *> to an XS function.
2763
1e54db1a 2764=head2 How do I convert a string to UTF-8?
a422fd2d 2765
2575c402
JW
2766If you're mixing UTF-8 and non-UTF-8 strings, it is necessary to upgrade
2767one of the strings to UTF-8. If you've got an SV, the easiest way to do
2768this is:
a422fd2d
SC
2769
2770 sv_utf8_upgrade(sv);
2771
2772However, you must not do this, for example:
2773
2774 if (!SvUTF8(left))
2775 sv_utf8_upgrade(left);
2776
2777If you do this in a binary operator, you will actually change one of the
b1866b2d 2778strings that came into the operator, and, while it shouldn't be noticeable
2575c402 2779by the end user, it can cause problems in deficient code.
a422fd2d 2780
1e54db1a 2781Instead, C<bytes_to_utf8> will give you a UTF-8-encoded B<copy> of its
a422fd2d 2782string argument. This is useful for having the data available for
b1866b2d 2783comparisons and so on, without harming the original SV. There's also
a422fd2d
SC
2784C<utf8_to_bytes> to go the other way, but naturally, this will fail if
2785the string contains any characters above 255 that can't be represented
2786in a single byte.
2787
2788=head2 Is there anything else I need to know?
2789
2790Not really. Just remember these things:
2791
2792=over 3
2793
2794=item *
2795
1e54db1a 2796There's no way to tell if a string is UTF-8 or not. You can tell if an SV
da8c5729 2797is UTF-8 by looking at its C<SvUTF8> flag. Don't forget to set the flag if
1e54db1a 2798something should be UTF-8. Treat the flag as part of the PV, even though
a422fd2d
SC
2799it's not - if you pass on the PV to somewhere, pass on the flag too.
2800
2801=item *
2802
4b88fb76 2803If a string is UTF-8, B<always> use C<utf8_to_uvchr_buf> to get at the value,
3a2263fe 2804unless C<UTF8_IS_INVARIANT(*s)> in which case you can use C<*s>.
a422fd2d
SC
2805
2806=item *
2807
1e54db1a 2808When writing a character C<uv> to a UTF-8 string, B<always> use
95701e00 2809C<uvchr_to_utf8>, unless C<UTF8_IS_INVARIANT(uv))> in which case
3a2263fe 2810you can use C<*s = uv>.
a422fd2d
SC
2811
2812=item *
2813
1e54db1a 2814Mixing UTF-8 and non-UTF-8 strings is tricky. Use C<bytes_to_utf8> to get
2bbc8d55 2815a new string which is UTF-8 encoded, and then combine them.
a422fd2d
SC
2816
2817=back
2818
53e06cf0
SC
2819=head1 Custom Operators
2820
2a0fd0f1 2821Custom operator support is an experimental feature that allows you to
53e06cf0
SC
2822define your own ops. This is primarily to allow the building of
2823interpreters for other languages in the Perl core, but it also allows
2824optimizations through the creation of "macro-ops" (ops which perform the
2825functions of multiple ops which are usually executed together, such as
1aa6ea50 2826C<gvsv, gvsv, add>.)
53e06cf0 2827
b455bf3f 2828This feature is implemented as a new op type, C<OP_CUSTOM>. The Perl
53e06cf0
SC
2829core does not "know" anything special about this op type, and so it will
2830not be involved in any optimizations. This also means that you can
2831define your custom ops to be any op structure - unary, binary, list and
2832so on - you like.
2833
2834It's important to know what custom operators won't do for you. They
2835won't let you add new syntax to Perl, directly. They won't even let you
2836add new keywords, directly. In fact, they won't change the way Perl
2837compiles a program at all. You have to do those changes yourself, after
2838Perl has compiled the program. You do this either by manipulating the op
2839tree using a C<CHECK> block and the C<B::Generate> module, or by adding
2840a custom peephole optimizer with the C<optimize> module.
2841
2842When you do this, you replace ordinary Perl ops with custom ops by
407f86e1 2843creating ops with the type C<OP_CUSTOM> and the C<op_ppaddr> of your own
53e06cf0
SC
2844PP function. This should be defined in XS code, and should look like
2845the PP ops in C<pp_*.c>. You are responsible for ensuring that your op
2846takes the appropriate number of values from the stack, and you are
2847responsible for adding stack marks if necessary.
2848
2849You should also "register" your op with the Perl interpreter so that it
2850can produce sensible error and warning messages. Since it is possible to
2851have multiple custom ops within the one "logical" op type C<OP_CUSTOM>,
9733086d
BM
2852Perl uses the value of C<< o->op_ppaddr >> to determine which custom op
2853it is dealing with. You should create an C<XOP> structure for each
2854ppaddr you use, set the properties of the custom op with
2855C<XopENTRY_set>, and register the structure against the ppaddr using
2856C<Perl_custom_op_register>. A trivial example might look like:
2857
2858 static XOP my_xop;
2859 static OP *my_pp(pTHX);
2860
2861 BOOT:
2862 XopENTRY_set(&my_xop, xop_name, "myxop");
2863 XopENTRY_set(&my_xop, xop_desc, "Useless custom op");
2864 Perl_custom_op_register(aTHX_ my_pp, &my_xop);
2865
2866The available fields in the structure are:
2867
2868=over 4
2869
2870=item xop_name
2871
2872A short name for your op. This will be included in some error messages,
2873and will also be returned as C<< $op->name >> by the L<B|B> module, so
2874it will appear in the output of module like L<B::Concise|B::Concise>.
2875
2876=item xop_desc
2877
2878A short description of the function of the op.
2879
2880=item xop_class
2881
2882Which of the various C<*OP> structures this op uses. This should be one of
2883the C<OA_*> constants from F<op.h>, namely
2884
2885=over 4
2886
2887=item OA_BASEOP
2888
2889=item OA_UNOP
2890
2891=item OA_BINOP
2892
2893=item OA_LOGOP
2894
2895=item OA_LISTOP
2896
2897=item OA_PMOP
2898
2899=item OA_SVOP
2900
2901=item OA_PADOP
2902
2903=item OA_PVOP_OR_SVOP
2904
2905This should be interpreted as 'C<PVOP>' only. The C<_OR_SVOP> is because
2906the only core C<PVOP>, C<OP_TRANS>, can sometimes be a C<SVOP> instead.
2907
2908=item OA_LOOP
2909
2910=item OA_COP
2911
2912=back
2913
2914The other C<OA_*> constants should not be used.
2915
2916=item xop_peep
2917
2918This member is of type C<Perl_cpeep_t>, which expands to C<void
2919(*Perl_cpeep_t)(aTHX_ OP *o, OP *oldop)>. If it is set, this function
2920will be called from C<Perl_rpeep> when ops of this type are encountered
2921by the peephole optimizer. I<o> is the OP that needs optimizing;
2922I<oldop> is the previous OP optimized, whose C<op_next> points to I<o>.
2923
2924=back
53e06cf0 2925
e7d4c058 2926C<B::Generate> directly supports the creation of custom ops by name.
53e06cf0 2927
954c1994 2928=head1 AUTHORS
e89caa19 2929
954c1994 2930Until May 1997, this document was maintained by Jeff Okamoto
9b5bb84f
SB
2931E<lt>okamoto@corp.hp.comE<gt>. It is now maintained as part of Perl
2932itself by the Perl 5 Porters E<lt>perl5-porters@perl.orgE<gt>.
cb1a09d0 2933
954c1994
GS
2934With lots of help and suggestions from Dean Roehrich, Malcolm Beattie,
2935Andreas Koenig, Paul Hudson, Ilya Zakharevich, Paul Marquess, Neil
2936Bowers, Matthew Green, Tim Bunce, Spider Boardman, Ulrich Pfeifer,
2937Stephen McCamant, and Gurusamy Sarathy.
cb1a09d0 2938
954c1994 2939=head1 SEE ALSO
cb1a09d0 2940
ba555bf5 2941L<perlapi>, L<perlintern>, L<perlxs>, L<perlembed>