3 perlguts - Introduction to the Perl API
7 This document attempts to describe how to use the Perl API, as well as
8 to provide some info on the basic workings of the Perl core. It is far
9 from complete and probably contains many errors. Please refer any
10 questions or comments to the author below.
16 Perl has three typedefs that handle Perl's three main data types:
22 Each typedef has specific routines that manipulate the various data types.
24 =head2 What is an "IV"?
26 Perl uses a special typedef IV which is a simple signed integer type that is
27 guaranteed to be large enough to hold a pointer (as well as an integer).
28 Additionally, there is the UV, which is simply an unsigned IV.
30 Perl also uses two special typedefs, I32 and I16, which will always be at
31 least 32-bits and 16-bits long, respectively. (Again, there are U32 and U16,
32 as well.) They will usually be exactly 32 and 16 bits long, but on Crays
33 they will both be 64 bits.
35 =head2 Working with SVs
37 An SV can be created and loaded with one command. There are five types of
38 values that can be loaded: an integer value (IV), an unsigned integer
39 value (UV), a double (NV), a string (PV), and another scalar (SV).
41 The seven routines are:
46 SV* newSVpv(const char*, STRLEN);
47 SV* newSVpvn(const char*, STRLEN);
48 SV* newSVpvf(const char*, ...);
51 C<STRLEN> is an integer type (Size_t, usually defined as size_t in
52 F<config.h>) guaranteed to be large enough to represent the size of
53 any string that perl can handle.
55 In the unlikely case of a SV requiring more complex initialisation, you
56 can create an empty SV with newSV(len). If C<len> is 0 an empty SV of
57 type NULL is returned, else an SV of type PV is returned with len + 1 (for
58 the NUL) bytes of storage allocated, accessible via SvPVX. In both cases
59 the SV has value undef.
61 SV *sv = newSV(0); /* no storage allocated */
62 SV *sv = newSV(10); /* 10 (+1) bytes of uninitialised storage allocated */
64 To change the value of an I<already-existing> SV, there are eight routines:
66 void sv_setiv(SV*, IV);
67 void sv_setuv(SV*, UV);
68 void sv_setnv(SV*, double);
69 void sv_setpv(SV*, const char*);
70 void sv_setpvn(SV*, const char*, STRLEN)
71 void sv_setpvf(SV*, const char*, ...);
72 void sv_vsetpvfn(SV*, const char*, STRLEN, va_list *, SV **, I32, bool *);
73 void sv_setsv(SV*, SV*);
75 Notice that you can choose to specify the length of the string to be
76 assigned by using C<sv_setpvn>, C<newSVpvn>, or C<newSVpv>, or you may
77 allow Perl to calculate the length by using C<sv_setpv> or by specifying
78 0 as the second argument to C<newSVpv>. Be warned, though, that Perl will
79 determine the string's length by using C<strlen>, which depends on the
80 string terminating with a NUL character.
82 The arguments of C<sv_setpvf> are processed like C<sprintf>, and the
83 formatted output becomes the value.
85 C<sv_vsetpvfn> is an analogue of C<vsprintf>, but it allows you to specify
86 either a pointer to a variable argument list or the address and length of
87 an array of SVs. The last argument points to a boolean; on return, if that
88 boolean is true, then locale-specific information has been used to format
89 the string, and the string's contents are therefore untrustworthy (see
90 L<perlsec>). This pointer may be NULL if that information is not
91 important. Note that this function requires you to specify the length of
94 The C<sv_set*()> functions are not generic enough to operate on values
95 that have "magic". See L<Magic Virtual Tables> later in this document.
97 All SVs that contain strings should be terminated with a NUL character.
98 If it is not NUL-terminated there is a risk of
99 core dumps and corruptions from code which passes the string to C
100 functions or system calls which expect a NUL-terminated string.
101 Perl's own functions typically add a trailing NUL for this reason.
102 Nevertheless, you should be very careful when you pass a string stored
103 in an SV to a C function or system call.
105 To access the actual value that an SV points to, you can use the macros:
110 SvPV(SV*, STRLEN len)
113 which will automatically coerce the actual scalar type into an IV, UV, double,
116 In the C<SvPV> macro, the length of the string returned is placed into the
117 variable C<len> (this is a macro, so you do I<not> use C<&len>). If you do
118 not care what the length of the data is, use the C<SvPV_nolen> macro.
119 Historically the C<SvPV> macro with the global variable C<PL_na> has been
120 used in this case. But that can be quite inefficient because C<PL_na> must
121 be accessed in thread-local storage in threaded Perl. In any case, remember
122 that Perl allows arbitrary strings of data that may both contain NULs and
123 might not be terminated by a NUL.
125 Also remember that C doesn't allow you to safely say C<foo(SvPV(s, len),
126 len);>. It might work with your compiler, but it won't work for everyone.
127 Break this sort of statement up into separate assignments:
135 If you want to know if the scalar value is TRUE, you can use:
139 Although Perl will automatically grow strings for you, if you need to force
140 Perl to allocate more memory for your SV, you can use the macro
142 SvGROW(SV*, STRLEN newlen)
144 which will determine if more memory needs to be allocated. If so, it will
145 call the function C<sv_grow>. Note that C<SvGROW> can only increase, not
146 decrease, the allocated memory of an SV and that it does not automatically
147 add a byte for the a trailing NUL (perl's own string functions typically do
148 C<SvGROW(sv, len + 1)>).
150 If you have an SV and want to know what kind of data Perl thinks is stored
151 in it, you can use the following macros to check the type of SV you have.
157 You can get and set the current length of the string stored in an SV with
158 the following macros:
161 SvCUR_set(SV*, I32 val)
163 You can also get a pointer to the end of the string stored in the SV
168 But note that these last three macros are valid only if C<SvPOK()> is true.
170 If you want to append something to the end of string stored in an C<SV*>,
171 you can use the following functions:
173 void sv_catpv(SV*, const char*);
174 void sv_catpvn(SV*, const char*, STRLEN);
175 void sv_catpvf(SV*, const char*, ...);
176 void sv_vcatpvfn(SV*, const char*, STRLEN, va_list *, SV **, I32, bool);
177 void sv_catsv(SV*, SV*);
179 The first function calculates the length of the string to be appended by
180 using C<strlen>. In the second, you specify the length of the string
181 yourself. The third function processes its arguments like C<sprintf> and
182 appends the formatted output. The fourth function works like C<vsprintf>.
183 You can specify the address and length of an array of SVs instead of the
184 va_list argument. The fifth function extends the string stored in the first
185 SV with the string stored in the second SV. It also forces the second SV
186 to be interpreted as a string.
188 The C<sv_cat*()> functions are not generic enough to operate on values that
189 have "magic". See L<Magic Virtual Tables> later in this document.
191 If you know the name of a scalar variable, you can get a pointer to its SV
192 by using the following:
194 SV* get_sv("package::varname", 0);
196 This returns NULL if the variable does not exist.
198 If you want to know if this variable (or any other SV) is actually C<defined>,
203 The scalar C<undef> value is stored in an SV instance called C<PL_sv_undef>.
205 Its address can be used whenever an C<SV*> is needed. Make sure that
206 you don't try to compare a random sv with C<&PL_sv_undef>. For example
207 when interfacing Perl code, it'll work correctly for:
211 But won't work when called as:
216 So to repeat always use SvOK() to check whether an sv is defined.
218 Also you have to be careful when using C<&PL_sv_undef> as a value in
219 AVs or HVs (see L<AVs, HVs and undefined values>).
221 There are also the two values C<PL_sv_yes> and C<PL_sv_no>, which contain
222 boolean TRUE and FALSE values, respectively. Like C<PL_sv_undef>, their
223 addresses can be used whenever an C<SV*> is needed.
225 Do not be fooled into thinking that C<(SV *) 0> is the same as C<&PL_sv_undef>.
229 if (I-am-to-return-a-real-value) {
230 sv = sv_2mortal(newSViv(42));
234 This code tries to return a new SV (which contains the value 42) if it should
235 return a real value, or undef otherwise. Instead it has returned a NULL
236 pointer which, somewhere down the line, will cause a segmentation violation,
237 bus error, or just weird results. Change the zero to C<&PL_sv_undef> in the
238 first line and all will be well.
240 To free an SV that you've created, call C<SvREFCNT_dec(SV*)>. Normally this
241 call is not necessary (see L<Reference Counts and Mortality>).
245 Perl provides the function C<sv_chop> to efficiently remove characters
246 from the beginning of a string; you give it an SV and a pointer to
247 somewhere inside the PV, and it discards everything before the
248 pointer. The efficiency comes by means of a little hack: instead of
249 actually removing the characters, C<sv_chop> sets the flag C<OOK>
250 (offset OK) to signal to other functions that the offset hack is in
251 effect, and it puts the number of bytes chopped off into the IV field
252 of the SV. It then moves the PV pointer (called C<SvPVX>) forward that
253 many bytes, and adjusts C<SvCUR> and C<SvLEN>.
255 Hence, at this point, the start of the buffer that we allocated lives
256 at C<SvPVX(sv) - SvIV(sv)> in memory and the PV pointer is pointing
257 into the middle of this allocated storage.
259 This is best demonstrated by example:
261 % ./perl -Ilib -MDevel::Peek -le '$a="12345"; $a=~s/.//; Dump($a)'
262 SV = PVIV(0x8128450) at 0x81340f0
264 FLAGS = (POK,OOK,pPOK)
266 PV = 0x8135781 ( "1" . ) "2345"\0
270 Here the number of bytes chopped off (1) is put into IV, and
271 C<Devel::Peek::Dump> helpfully reminds us that this is an offset. The
272 portion of the string between the "real" and the "fake" beginnings is
273 shown in parentheses, and the values of C<SvCUR> and C<SvLEN> reflect
274 the fake beginning, not the real one.
276 Something similar to the offset hack is performed on AVs to enable
277 efficient shifting and splicing off the beginning of the array; while
278 C<AvARRAY> points to the first element in the array that is visible from
279 Perl, C<AvALLOC> points to the real start of the C array. These are
280 usually the same, but a C<shift> operation can be carried out by
281 increasing C<AvARRAY> by one and decreasing C<AvFILL> and C<AvMAX>.
282 Again, the location of the real start of the C array only comes into
283 play when freeing the array. See C<av_shift> in F<av.c>.
285 =head2 What's Really Stored in an SV?
287 Recall that the usual method of determining the type of scalar you have is
288 to use C<Sv*OK> macros. Because a scalar can be both a number and a string,
289 usually these macros will always return TRUE and calling the C<Sv*V>
290 macros will do the appropriate conversion of string to integer/double or
291 integer/double to string.
293 If you I<really> need to know if you have an integer, double, or string
294 pointer in an SV, you can use the following three macros instead:
300 These will tell you if you truly have an integer, double, or string pointer
301 stored in your SV. The "p" stands for private.
303 The are various ways in which the private and public flags may differ.
304 For example, a tied SV may have a valid underlying value in the IV slot
305 (so SvIOKp is true), but the data should be accessed via the FETCH
306 routine rather than directly, so SvIOK is false. Another is when
307 numeric conversion has occurred and precision has been lost: only the
308 private flag is set on 'lossy' values. So when an NV is converted to an
309 IV with loss, SvIOKp, SvNOKp and SvNOK will be set, while SvIOK wont be.
311 In general, though, it's best to use the C<Sv*V> macros.
313 =head2 Working with AVs
315 There are two ways to create and load an AV. The first method creates an
320 The second method both creates the AV and initially populates it with SVs:
322 AV* av_make(I32 num, SV **ptr);
324 The second argument points to an array containing C<num> C<SV*>'s. Once the
325 AV has been created, the SVs can be destroyed, if so desired.
327 Once the AV has been created, the following operations are possible on AVs:
329 void av_push(AV*, SV*);
332 void av_unshift(AV*, I32 num);
334 These should be familiar operations, with the exception of C<av_unshift>.
335 This routine adds C<num> elements at the front of the array with the C<undef>
336 value. You must then use C<av_store> (described below) to assign values
337 to these new elements.
339 Here are some other functions:
342 SV** av_fetch(AV*, I32 key, I32 lval);
343 SV** av_store(AV*, I32 key, SV* val);
345 The C<av_len> function returns the highest index value in array (just
346 like $#array in Perl). If the array is empty, -1 is returned. The
347 C<av_fetch> function returns the value at index C<key>, but if C<lval>
348 is non-zero, then C<av_fetch> will store an undef value at that index.
349 The C<av_store> function stores the value C<val> at index C<key>, and does
350 not increment the reference count of C<val>. Thus the caller is responsible
351 for taking care of that, and if C<av_store> returns NULL, the caller will
352 have to decrement the reference count to avoid a memory leak. Note that
353 C<av_fetch> and C<av_store> both return C<SV**>'s, not C<SV*>'s as their
358 void av_extend(AV*, I32 key);
360 The C<av_clear> function deletes all the elements in the AV* array, but
361 does not actually delete the array itself. The C<av_undef> function will
362 delete all the elements in the array plus the array itself. The
363 C<av_extend> function extends the array so that it contains at least C<key+1>
364 elements. If C<key+1> is less than the currently allocated length of the array,
365 then nothing is done.
367 If you know the name of an array variable, you can get a pointer to its AV
368 by using the following:
370 AV* get_av("package::varname", 0);
372 This returns NULL if the variable does not exist.
374 See L<Understanding the Magic of Tied Hashes and Arrays> for more
375 information on how to use the array access functions on tied arrays.
377 =head2 Working with HVs
379 To create an HV, you use the following routine:
383 Once the HV has been created, the following operations are possible on HVs:
385 SV** hv_store(HV*, const char* key, U32 klen, SV* val, U32 hash);
386 SV** hv_fetch(HV*, const char* key, U32 klen, I32 lval);
388 The C<klen> parameter is the length of the key being passed in (Note that
389 you cannot pass 0 in as a value of C<klen> to tell Perl to measure the
390 length of the key). The C<val> argument contains the SV pointer to the
391 scalar being stored, and C<hash> is the precomputed hash value (zero if
392 you want C<hv_store> to calculate it for you). The C<lval> parameter
393 indicates whether this fetch is actually a part of a store operation, in
394 which case a new undefined value will be added to the HV with the supplied
395 key and C<hv_fetch> will return as if the value had already existed.
397 Remember that C<hv_store> and C<hv_fetch> return C<SV**>'s and not just
398 C<SV*>. To access the scalar value, you must first dereference the return
399 value. However, you should check to make sure that the return value is
400 not NULL before dereferencing it.
402 These two functions check if a hash table entry exists, and deletes it.
404 bool hv_exists(HV*, const char* key, U32 klen);
405 SV* hv_delete(HV*, const char* key, U32 klen, I32 flags);
407 If C<flags> does not include the C<G_DISCARD> flag then C<hv_delete> will
408 create and return a mortal copy of the deleted value.
410 And more miscellaneous functions:
415 Like their AV counterparts, C<hv_clear> deletes all the entries in the hash
416 table but does not actually delete the hash table. The C<hv_undef> deletes
417 both the entries and the hash table itself.
419 Perl keeps the actual data in linked list of structures with a typedef of HE.
420 These contain the actual key and value pointers (plus extra administrative
421 overhead). The key is a string pointer; the value is an C<SV*>. However,
422 once you have an C<HE*>, to get the actual key and value, use the routines
425 I32 hv_iterinit(HV*);
426 /* Prepares starting point to traverse hash table */
427 HE* hv_iternext(HV*);
428 /* Get the next entry, and return a pointer to a
429 structure that has both the key and value */
430 char* hv_iterkey(HE* entry, I32* retlen);
431 /* Get the key from an HE structure and also return
432 the length of the key string */
433 SV* hv_iterval(HV*, HE* entry);
434 /* Return an SV pointer to the value of the HE
436 SV* hv_iternextsv(HV*, char** key, I32* retlen);
437 /* This convenience routine combines hv_iternext,
438 hv_iterkey, and hv_iterval. The key and retlen
439 arguments are return values for the key and its
440 length. The value is returned in the SV* argument */
442 If you know the name of a hash variable, you can get a pointer to its HV
443 by using the following:
445 HV* get_hv("package::varname", 0);
447 This returns NULL if the variable does not exist.
449 The hash algorithm is defined in the C<PERL_HASH(hash, key, klen)> macro:
453 hash = (hash * 33) + *key++;
454 hash = hash + (hash >> 5); /* after 5.6 */
456 The last step was added in version 5.6 to improve distribution of
457 lower bits in the resulting hash value.
459 See L<Understanding the Magic of Tied Hashes and Arrays> for more
460 information on how to use the hash access functions on tied hashes.
462 =head2 Hash API Extensions
464 Beginning with version 5.004, the following functions are also supported:
466 HE* hv_fetch_ent (HV* tb, SV* key, I32 lval, U32 hash);
467 HE* hv_store_ent (HV* tb, SV* key, SV* val, U32 hash);
469 bool hv_exists_ent (HV* tb, SV* key, U32 hash);
470 SV* hv_delete_ent (HV* tb, SV* key, I32 flags, U32 hash);
472 SV* hv_iterkeysv (HE* entry);
474 Note that these functions take C<SV*> keys, which simplifies writing
475 of extension code that deals with hash structures. These functions
476 also allow passing of C<SV*> keys to C<tie> functions without forcing
477 you to stringify the keys (unlike the previous set of functions).
479 They also return and accept whole hash entries (C<HE*>), making their
480 use more efficient (since the hash number for a particular string
481 doesn't have to be recomputed every time). See L<perlapi> for detailed
484 The following macros must always be used to access the contents of hash
485 entries. Note that the arguments to these macros must be simple
486 variables, since they may get evaluated more than once. See
487 L<perlapi> for detailed descriptions of these macros.
489 HePV(HE* he, STRLEN len)
493 HeSVKEY_force(HE* he)
494 HeSVKEY_set(HE* he, SV* sv)
496 These two lower level macros are defined, but must only be used when
497 dealing with keys that are not C<SV*>s:
502 Note that both C<hv_store> and C<hv_store_ent> do not increment the
503 reference count of the stored C<val>, which is the caller's responsibility.
504 If these functions return a NULL value, the caller will usually have to
505 decrement the reference count of C<val> to avoid a memory leak.
507 =head2 AVs, HVs and undefined values
509 Sometimes you have to store undefined values in AVs or HVs. Although
510 this may be a rare case, it can be tricky. That's because you're
511 used to using C<&PL_sv_undef> if you need an undefined SV.
513 For example, intuition tells you that this XS code:
516 av_store( av, 0, &PL_sv_undef );
518 is equivalent to this Perl code:
523 Unfortunately, this isn't true. AVs use C<&PL_sv_undef> as a marker
524 for indicating that an array element has not yet been initialized.
525 Thus, C<exists $av[0]> would be true for the above Perl code, but
526 false for the array generated by the XS code.
528 Other problems can occur when storing C<&PL_sv_undef> in HVs:
530 hv_store( hv, "key", 3, &PL_sv_undef, 0 );
532 This will indeed make the value C<undef>, but if you try to modify
533 the value of C<key>, you'll get the following error:
535 Modification of non-creatable hash value attempted
537 In perl 5.8.0, C<&PL_sv_undef> was also used to mark placeholders
538 in restricted hashes. This caused such hash entries not to appear
539 when iterating over the hash or when checking for the keys
540 with the C<hv_exists> function.
542 You can run into similar problems when you store C<&PL_sv_yes> or
543 C<&PL_sv_no> into AVs or HVs. Trying to modify such elements
544 will give you the following error:
546 Modification of a read-only value attempted
548 To make a long story short, you can use the special variables
549 C<&PL_sv_undef>, C<&PL_sv_yes> and C<&PL_sv_no> with AVs and
550 HVs, but you have to make sure you know what you're doing.
552 Generally, if you want to store an undefined value in an AV
553 or HV, you should not use C<&PL_sv_undef>, but rather create a
554 new undefined value using the C<newSV> function, for example:
556 av_store( av, 42, newSV(0) );
557 hv_store( hv, "foo", 3, newSV(0), 0 );
561 References are a special type of scalar that point to other data types
562 (including references).
564 To create a reference, use either of the following functions:
566 SV* newRV_inc((SV*) thing);
567 SV* newRV_noinc((SV*) thing);
569 The C<thing> argument can be any of an C<SV*>, C<AV*>, or C<HV*>. The
570 functions are identical except that C<newRV_inc> increments the reference
571 count of the C<thing>, while C<newRV_noinc> does not. For historical
572 reasons, C<newRV> is a synonym for C<newRV_inc>.
574 Once you have a reference, you can use the following macro to dereference
579 then call the appropriate routines, casting the returned C<SV*> to either an
580 C<AV*> or C<HV*>, if required.
582 To determine if an SV is a reference, you can use the following macro:
586 To discover what type of value the reference refers to, use the following
587 macro and then check the return value.
591 The most useful types that will be returned are:
600 SVt_PVGV Glob (possibly a file handle)
601 SVt_PVMG Blessed or Magical Scalar
603 See the F<sv.h> header file for more details.
605 =head2 Blessed References and Class Objects
607 References are also used to support object-oriented programming. In perl's
608 OO lexicon, an object is simply a reference that has been blessed into a
609 package (or class). Once blessed, the programmer may now use the reference
610 to access the various methods in the class.
612 A reference can be blessed into a package with the following function:
614 SV* sv_bless(SV* sv, HV* stash);
616 The C<sv> argument must be a reference value. The C<stash> argument
617 specifies which class the reference will belong to. See
618 L<Stashes and Globs> for information on converting class names into stashes.
620 /* Still under construction */
622 The following function upgrades rv to reference if not already one.
623 Creates a new SV for rv to point to. If C<classname> is non-null, the SV
624 is blessed into the specified class. SV is returned.
626 SV* newSVrv(SV* rv, const char* classname);
628 The following three functions copy integer, unsigned integer or double
629 into an SV whose reference is C<rv>. SV is blessed if C<classname> is
632 SV* sv_setref_iv(SV* rv, const char* classname, IV iv);
633 SV* sv_setref_uv(SV* rv, const char* classname, UV uv);
634 SV* sv_setref_nv(SV* rv, const char* classname, NV iv);
636 The following function copies the pointer value (I<the address, not the
637 string!>) into an SV whose reference is rv. SV is blessed if C<classname>
640 SV* sv_setref_pv(SV* rv, const char* classname, void* pv);
642 The following function copies string into an SV whose reference is C<rv>.
643 Set length to 0 to let Perl calculate the string length. SV is blessed if
644 C<classname> is non-null.
646 SV* sv_setref_pvn(SV* rv, const char* classname, char* pv, STRLEN length);
648 The following function tests whether the SV is blessed into the specified
649 class. It does not check inheritance relationships.
651 int sv_isa(SV* sv, const char* name);
653 The following function tests whether the SV is a reference to a blessed object.
655 int sv_isobject(SV* sv);
657 The following function tests whether the SV is derived from the specified
658 class. SV can be either a reference to a blessed object or a string
659 containing a class name. This is the function implementing the
660 C<UNIVERSAL::isa> functionality.
662 bool sv_derived_from(SV* sv, const char* name);
664 To check if you've got an object derived from a specific class you have
667 if (sv_isobject(sv) && sv_derived_from(sv, class)) { ... }
669 =head2 Creating New Variables
671 To create a new Perl variable with an undef value which can be accessed from
672 your Perl script, use the following routines, depending on the variable type.
674 SV* get_sv("package::varname", GV_ADD);
675 AV* get_av("package::varname", GV_ADD);
676 HV* get_hv("package::varname", GV_ADD);
678 Notice the use of GV_ADD as the second parameter. The new variable can now
679 be set, using the routines appropriate to the data type.
681 There are additional macros whose values may be bitwise OR'ed with the
682 C<GV_ADD> argument to enable certain extra features. Those bits are:
688 Marks the variable as multiply defined, thus preventing the:
690 Name <varname> used only once: possible typo
698 Had to create <varname> unexpectedly
700 if the variable did not exist before the function was called.
704 If you do not specify a package name, the variable is created in the current
707 =head2 Reference Counts and Mortality
709 Perl uses a reference count-driven garbage collection mechanism. SVs,
710 AVs, or HVs (xV for short in the following) start their life with a
711 reference count of 1. If the reference count of an xV ever drops to 0,
712 then it will be destroyed and its memory made available for reuse.
714 This normally doesn't happen at the Perl level unless a variable is
715 undef'ed or the last variable holding a reference to it is changed or
716 overwritten. At the internal level, however, reference counts can be
717 manipulated with the following macros:
719 int SvREFCNT(SV* sv);
720 SV* SvREFCNT_inc(SV* sv);
721 void SvREFCNT_dec(SV* sv);
723 However, there is one other function which manipulates the reference
724 count of its argument. The C<newRV_inc> function, you will recall,
725 creates a reference to the specified argument. As a side effect,
726 it increments the argument's reference count. If this is not what
727 you want, use C<newRV_noinc> instead.
729 For example, imagine you want to return a reference from an XSUB function.
730 Inside the XSUB routine, you create an SV which initially has a reference
731 count of one. Then you call C<newRV_inc>, passing it the just-created SV.
732 This returns the reference as a new SV, but the reference count of the
733 SV you passed to C<newRV_inc> has been incremented to two. Now you
734 return the reference from the XSUB routine and forget about the SV.
735 But Perl hasn't! Whenever the returned reference is destroyed, the
736 reference count of the original SV is decreased to one and nothing happens.
737 The SV will hang around without any way to access it until Perl itself
738 terminates. This is a memory leak.
740 The correct procedure, then, is to use C<newRV_noinc> instead of
741 C<newRV_inc>. Then, if and when the last reference is destroyed,
742 the reference count of the SV will go to zero and it will be destroyed,
743 stopping any memory leak.
745 There are some convenience functions available that can help with the
746 destruction of xVs. These functions introduce the concept of "mortality".
747 An xV that is mortal has had its reference count marked to be decremented,
748 but not actually decremented, until "a short time later". Generally the
749 term "short time later" means a single Perl statement, such as a call to
750 an XSUB function. The actual determinant for when mortal xVs have their
751 reference count decremented depends on two macros, SAVETMPS and FREETMPS.
752 See L<perlcall> and L<perlxs> for more details on these macros.
754 "Mortalization" then is at its simplest a deferred C<SvREFCNT_dec>.
755 However, if you mortalize a variable twice, the reference count will
756 later be decremented twice.
758 "Mortal" SVs are mainly used for SVs that are placed on perl's stack.
759 For example an SV which is created just to pass a number to a called sub
760 is made mortal to have it cleaned up automatically when it's popped off
761 the stack. Similarly, results returned by XSUBs (which are pushed on the
762 stack) are often made mortal.
764 To create a mortal variable, use the functions:
768 SV* sv_mortalcopy(SV*)
770 The first call creates a mortal SV (with no value), the second converts an existing
771 SV to a mortal SV (and thus defers a call to C<SvREFCNT_dec>), and the
772 third creates a mortal copy of an existing SV.
773 Because C<sv_newmortal> gives the new SV no value,it must normally be given one
774 via C<sv_setpv>, C<sv_setiv>, etc. :
776 SV *tmp = sv_newmortal();
777 sv_setiv(tmp, an_integer);
779 As that is multiple C statements it is quite common so see this idiom instead:
781 SV *tmp = sv_2mortal(newSViv(an_integer));
784 You should be careful about creating mortal variables. Strange things
785 can happen if you make the same value mortal within multiple contexts,
786 or if you make a variable mortal multiple times. Thinking of "Mortalization"
787 as deferred C<SvREFCNT_dec> should help to minimize such problems.
788 For example if you are passing an SV which you I<know> has high enough REFCNT
789 to survive its use on the stack you need not do any mortalization.
790 If you are not sure then doing an C<SvREFCNT_inc> and C<sv_2mortal>, or
791 making a C<sv_mortalcopy> is safer.
793 The mortal routines are not just for SVs; AVs and HVs can be
794 made mortal by passing their address (type-casted to C<SV*>) to the
795 C<sv_2mortal> or C<sv_mortalcopy> routines.
797 =head2 Stashes and Globs
799 A B<stash> is a hash that contains all variables that are defined
800 within a package. Each key of the stash is a symbol
801 name (shared by all the different types of objects that have the same
802 name), and each value in the hash table is a GV (Glob Value). This GV
803 in turn contains references to the various objects of that name,
804 including (but not limited to) the following:
813 There is a single stash called C<PL_defstash> that holds the items that exist
814 in the C<main> package. To get at the items in other packages, append the
815 string "::" to the package name. The items in the C<Foo> package are in
816 the stash C<Foo::> in PL_defstash. The items in the C<Bar::Baz> package are
817 in the stash C<Baz::> in C<Bar::>'s stash.
819 To get the stash pointer for a particular package, use the function:
821 HV* gv_stashpv(const char* name, I32 flags)
822 HV* gv_stashsv(SV*, I32 flags)
824 The first function takes a literal string, the second uses the string stored
825 in the SV. Remember that a stash is just a hash table, so you get back an
826 C<HV*>. The C<flags> flag will create a new package if it is set to GV_ADD.
828 The name that C<gv_stash*v> wants is the name of the package whose symbol table
829 you want. The default package is called C<main>. If you have multiply nested
830 packages, pass their names to C<gv_stash*v>, separated by C<::> as in the Perl
833 Alternately, if you have an SV that is a blessed reference, you can find
834 out the stash pointer by using:
836 HV* SvSTASH(SvRV(SV*));
838 then use the following to get the package name itself:
840 char* HvNAME(HV* stash);
842 If you need to bless or re-bless an object you can use the following
845 SV* sv_bless(SV*, HV* stash)
847 where the first argument, an C<SV*>, must be a reference, and the second
848 argument is a stash. The returned C<SV*> can now be used in the same way
851 For more information on references and blessings, consult L<perlref>.
853 =head2 Double-Typed SVs
855 Scalar variables normally contain only one type of value, an integer,
856 double, pointer, or reference. Perl will automatically convert the
857 actual scalar data from the stored type into the requested type.
859 Some scalar variables contain more than one type of scalar data. For
860 example, the variable C<$!> contains either the numeric value of C<errno>
861 or its string equivalent from either C<strerror> or C<sys_errlist[]>.
863 To force multiple data values into an SV, you must do two things: use the
864 C<sv_set*v> routines to add the additional scalar type, then set a flag
865 so that Perl will believe it contains more than one type of data. The
866 four macros to set the flags are:
873 The particular macro you must use depends on which C<sv_set*v> routine
874 you called first. This is because every C<sv_set*v> routine turns on
875 only the bit for the particular type of data being set, and turns off
878 For example, to create a new Perl variable called "dberror" that contains
879 both the numeric and descriptive string error values, you could use the
883 extern char *dberror_list;
885 SV* sv = get_sv("dberror", GV_ADD);
886 sv_setiv(sv, (IV) dberror);
887 sv_setpv(sv, dberror_list[dberror]);
890 If the order of C<sv_setiv> and C<sv_setpv> had been reversed, then the
891 macro C<SvPOK_on> would need to be called instead of C<SvIOK_on>.
893 =head2 Magic Variables
895 [This section still under construction. Ignore everything here. Post no
896 bills. Everything not permitted is forbidden.]
898 Any SV may be magical, that is, it has special features that a normal
899 SV does not have. These features are stored in the SV structure in a
900 linked list of C<struct magic>'s, typedef'ed to C<MAGIC>.
913 Note this is current as of patchlevel 0, and could change at any time.
915 =head2 Assigning Magic
917 Perl adds magic to an SV using the sv_magic function:
919 void sv_magic(SV* sv, SV* obj, int how, const char* name, I32 namlen);
921 The C<sv> argument is a pointer to the SV that is to acquire a new magical
924 If C<sv> is not already magical, Perl uses the C<SvUPGRADE> macro to
925 convert C<sv> to type C<SVt_PVMG>. Perl then continues by adding new magic
926 to the beginning of the linked list of magical features. Any prior entry
927 of the same type of magic is deleted. Note that this can be overridden,
928 and multiple instances of the same type of magic can be associated with an
931 The C<name> and C<namlen> arguments are used to associate a string with
932 the magic, typically the name of a variable. C<namlen> is stored in the
933 C<mg_len> field and if C<name> is non-null then either a C<savepvn> copy of
934 C<name> or C<name> itself is stored in the C<mg_ptr> field, depending on
935 whether C<namlen> is greater than zero or equal to zero respectively. As a
936 special case, if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
937 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
939 The sv_magic function uses C<how> to determine which, if any, predefined
940 "Magic Virtual Table" should be assigned to the C<mg_virtual> field.
941 See the L<Magic Virtual Tables> section below. The C<how> argument is also
942 stored in the C<mg_type> field. The value of C<how> should be chosen
943 from the set of macros C<PERL_MAGIC_foo> found in F<perl.h>. Note that before
944 these macros were added, Perl internals used to directly use character
945 literals, so you may occasionally come across old code or documentation
946 referring to 'U' magic rather than C<PERL_MAGIC_uvar> for example.
948 The C<obj> argument is stored in the C<mg_obj> field of the C<MAGIC>
949 structure. If it is not the same as the C<sv> argument, the reference
950 count of the C<obj> object is incremented. If it is the same, or if
951 the C<how> argument is C<PERL_MAGIC_arylen>, or if it is a NULL pointer,
952 then C<obj> is merely stored, without the reference count being incremented.
954 See also C<sv_magicext> in L<perlapi> for a more flexible way to add magic
957 There is also a function to add magic to an C<HV>:
959 void hv_magic(HV *hv, GV *gv, int how);
961 This simply calls C<sv_magic> and coerces the C<gv> argument into an C<SV>.
963 To remove the magic from an SV, call the function sv_unmagic:
965 int sv_unmagic(SV *sv, int type);
967 The C<type> argument should be equal to the C<how> value when the C<SV>
968 was initially made magical.
970 However, note that C<sv_unmagic> removes all magic of a certain C<type> from the
971 C<SV>. If you want to remove only certain magic of a C<type> based on the magic
972 virtual table, use C<sv_unmagicext> instead:
974 int sv_unmagicext(SV *sv, int type, MGVTBL *vtbl);
976 =head2 Magic Virtual Tables
978 The C<mg_virtual> field in the C<MAGIC> structure is a pointer to an
979 C<MGVTBL>, which is a structure of function pointers and stands for
980 "Magic Virtual Table" to handle the various operations that might be
981 applied to that variable.
983 The C<MGVTBL> has five (or sometimes eight) pointers to the following
986 int (*svt_get)(SV* sv, MAGIC* mg);
987 int (*svt_set)(SV* sv, MAGIC* mg);
988 U32 (*svt_len)(SV* sv, MAGIC* mg);
989 int (*svt_clear)(SV* sv, MAGIC* mg);
990 int (*svt_free)(SV* sv, MAGIC* mg);
992 int (*svt_copy)(SV *sv, MAGIC* mg, SV *nsv, const char *name, I32 namlen);
993 int (*svt_dup)(MAGIC *mg, CLONE_PARAMS *param);
994 int (*svt_local)(SV *nsv, MAGIC *mg);
997 This MGVTBL structure is set at compile-time in F<perl.h> and there are
998 currently 32 types. These different structures contain pointers to various
999 routines that perform additional actions depending on which function is
1002 Function pointer Action taken
1003 ---------------- ------------
1004 svt_get Do something before the value of the SV is retrieved.
1005 svt_set Do something after the SV is assigned a value.
1006 svt_len Report on the SV's length.
1007 svt_clear Clear something the SV represents.
1008 svt_free Free any extra storage associated with the SV.
1010 svt_copy copy tied variable magic to a tied element
1011 svt_dup duplicate a magic structure during thread cloning
1012 svt_local copy magic to local value during 'local'
1014 For instance, the MGVTBL structure called C<vtbl_sv> (which corresponds
1015 to an C<mg_type> of C<PERL_MAGIC_sv>) contains:
1017 { magic_get, magic_set, magic_len, 0, 0 }
1019 Thus, when an SV is determined to be magical and of type C<PERL_MAGIC_sv>,
1020 if a get operation is being performed, the routine C<magic_get> is
1021 called. All the various routines for the various magical types begin
1022 with C<magic_>. NOTE: the magic routines are not considered part of
1023 the Perl API, and may not be exported by the Perl library.
1025 The last three slots are a recent addition, and for source code
1026 compatibility they are only checked for if one of the three flags
1027 MGf_COPY, MGf_DUP or MGf_LOCAL is set in mg_flags. This means that most
1028 code can continue declaring a vtable as a 5-element value. These three are
1029 currently used exclusively by the threading code, and are highly subject
1032 The current kinds of Magic Virtual Tables are:
1035 (old-style char and macro) MGVTBL Type of magic
1036 -------------------------- ------ -------------
1037 \0 PERL_MAGIC_sv vtbl_sv Special scalar variable
1038 # PERL_MAGIC_arylen vtbl_arylen Array length ($#ary)
1039 % PERL_MAGIC_rhash (none) extra data for restricted
1041 . PERL_MAGIC_pos vtbl_pos pos() lvalue
1042 : PERL_MAGIC_symtab (none) extra data for symbol tables
1043 < PERL_MAGIC_backref vtbl_backref for weak ref data
1044 @ PERL_MAGIC_arylen_p (none) to move arylen out of XPVAV
1045 A PERL_MAGIC_overload vtbl_amagic %OVERLOAD hash
1046 a PERL_MAGIC_overload_elem vtbl_amagicelem %OVERLOAD hash element
1047 B PERL_MAGIC_bm vtbl_regexp Boyer-Moore
1048 (fast string search)
1049 c PERL_MAGIC_overload_table vtbl_ovrld Holds overload table
1051 D PERL_MAGIC_regdata vtbl_regdata Regex match position data
1053 d PERL_MAGIC_regdatum vtbl_regdatum Regex match position data
1055 E PERL_MAGIC_env vtbl_env %ENV hash
1056 e PERL_MAGIC_envelem vtbl_envelem %ENV hash element
1057 f PERL_MAGIC_fm vtbl_regdata Formline ('compiled' format)
1058 G PERL_MAGIC_study vtbl_regdata study()ed string
1059 g PERL_MAGIC_regex_global vtbl_mglob m//g target
1060 H PERL_MAGIC_hints vtbl_hints %^H hash
1061 h PERL_MAGIC_hintselem vtbl_hintselem %^H hash element
1062 I PERL_MAGIC_isa vtbl_isa @ISA array
1063 i PERL_MAGIC_isaelem vtbl_isaelem @ISA array element
1064 k PERL_MAGIC_nkeys vtbl_nkeys scalar(keys()) lvalue
1065 L PERL_MAGIC_dbfile (none) Debugger %_<filename
1066 l PERL_MAGIC_dbline vtbl_dbline Debugger %_<filename element
1067 N PERL_MAGIC_shared (none) Shared between threads
1068 n PERL_MAGIC_shared_scalar (none) Shared between threads
1069 o PERL_MAGIC_collxfrm vtbl_collxfrm Locale transformation
1070 P PERL_MAGIC_tied vtbl_pack Tied array or hash
1071 p PERL_MAGIC_tiedelem vtbl_packelem Tied array or hash element
1072 q PERL_MAGIC_tiedscalar vtbl_packelem Tied scalar or handle
1073 r PERL_MAGIC_qr vtbl_regexp precompiled qr// regex
1074 S PERL_MAGIC_sig (none) %SIG hash
1075 s PERL_MAGIC_sigelem vtbl_sigelem %SIG hash element
1076 t PERL_MAGIC_taint vtbl_taint Taintedness
1077 U PERL_MAGIC_uvar vtbl_uvar Available for use by extensions
1078 u PERL_MAGIC_uvar_elem (none) Reserved for use by extensions
1079 V PERL_MAGIC_vstring (none) SV was vstring literal
1080 v PERL_MAGIC_vec vtbl_vec vec() lvalue
1081 w PERL_MAGIC_utf8 vtbl_utf8 Cached UTF-8 information
1082 x PERL_MAGIC_substr vtbl_substr substr() lvalue
1083 y PERL_MAGIC_defelem vtbl_defelem Shadow "foreach" iterator
1084 variable / smart parameter
1086 ] PERL_MAGIC_checkcall (none) inlining/mutation of call to
1088 ~ PERL_MAGIC_ext (none) Available for use by extensions
1091 When an uppercase and lowercase letter both exist in the table, then the
1092 uppercase letter is typically used to represent some kind of composite type
1093 (a list or a hash), and the lowercase letter is used to represent an element
1094 of that composite type. Some internals code makes use of this case
1095 relationship. However, 'v' and 'V' (vec and v-string) are in no way related.
1097 The C<PERL_MAGIC_ext> and C<PERL_MAGIC_uvar> magic types are defined
1098 specifically for use by extensions and will not be used by perl itself.
1099 Extensions can use C<PERL_MAGIC_ext> magic to 'attach' private information
1100 to variables (typically objects). This is especially useful because
1101 there is no way for normal perl code to corrupt this private information
1102 (unlike using extra elements of a hash object).
1104 Similarly, C<PERL_MAGIC_uvar> magic can be used much like tie() to call a
1105 C function any time a scalar's value is used or changed. The C<MAGIC>'s
1106 C<mg_ptr> field points to a C<ufuncs> structure:
1109 I32 (*uf_val)(pTHX_ IV, SV*);
1110 I32 (*uf_set)(pTHX_ IV, SV*);
1114 When the SV is read from or written to, the C<uf_val> or C<uf_set>
1115 function will be called with C<uf_index> as the first arg and a pointer to
1116 the SV as the second. A simple example of how to add C<PERL_MAGIC_uvar>
1117 magic is shown below. Note that the ufuncs structure is copied by
1118 sv_magic, so you can safely allocate it on the stack.
1126 uf.uf_val = &my_get_fn;
1127 uf.uf_set = &my_set_fn;
1129 sv_magic(sv, 0, PERL_MAGIC_uvar, (char*)&uf, sizeof(uf));
1131 Attaching C<PERL_MAGIC_uvar> to arrays is permissible but has no effect.
1133 For hashes there is a specialized hook that gives control over hash
1134 keys (but not values). This hook calls C<PERL_MAGIC_uvar> 'get' magic
1135 if the "set" function in the C<ufuncs> structure is NULL. The hook
1136 is activated whenever the hash is accessed with a key specified as
1137 an C<SV> through the functions C<hv_store_ent>, C<hv_fetch_ent>,
1138 C<hv_delete_ent>, and C<hv_exists_ent>. Accessing the key as a string
1139 through the functions without the C<..._ent> suffix circumvents the
1140 hook. See L<Hash::Util::FieldHash/GUTS> for a detailed description.
1142 Note that because multiple extensions may be using C<PERL_MAGIC_ext>
1143 or C<PERL_MAGIC_uvar> magic, it is important for extensions to take
1144 extra care to avoid conflict. Typically only using the magic on
1145 objects blessed into the same class as the extension is sufficient.
1146 For C<PERL_MAGIC_ext> magic, it is usually a good idea to define an
1147 C<MGVTBL>, even if all its fields will be C<0>, so that individual
1148 C<MAGIC> pointers can be identified as a particular kind of magic
1149 using their magic virtual table. C<mg_findext> provides an easy way
1152 STATIC MGVTBL my_vtbl = { 0, 0, 0, 0, 0, 0, 0, 0 };
1155 if ((mg = mg_findext(sv, PERL_MAGIC_ext, &my_vtbl))) {
1156 /* this is really ours, not another module's PERL_MAGIC_ext */
1157 my_priv_data_t *priv = (my_priv_data_t *)mg->mg_ptr;
1161 Also note that the C<sv_set*()> and C<sv_cat*()> functions described
1162 earlier do B<not> invoke 'set' magic on their targets. This must
1163 be done by the user either by calling the C<SvSETMAGIC()> macro after
1164 calling these functions, or by using one of the C<sv_set*_mg()> or
1165 C<sv_cat*_mg()> functions. Similarly, generic C code must call the
1166 C<SvGETMAGIC()> macro to invoke any 'get' magic if they use an SV
1167 obtained from external sources in functions that don't handle magic.
1168 See L<perlapi> for a description of these functions.
1169 For example, calls to the C<sv_cat*()> functions typically need to be
1170 followed by C<SvSETMAGIC()>, but they don't need a prior C<SvGETMAGIC()>
1171 since their implementation handles 'get' magic.
1173 =head2 Finding Magic
1175 MAGIC *mg_find(SV *sv, int type); /* Finds the magic pointer of that type */
1177 This routine returns a pointer to a C<MAGIC> structure stored in the SV.
1178 If the SV does not have that magical feature, C<NULL> is returned. If the
1179 SV has multiple instances of that magical feature, the first one will be
1180 returned. C<mg_findext> can be used to find a C<MAGIC> structure of an SV
1181 based on both it's magic type and it's magic virtual table:
1183 MAGIC *mg_findext(SV *sv, int type, MGVTBL *vtbl);
1185 Also, if the SV passed to C<mg_find> or C<mg_findext> is not of type
1186 SVt_PVMG, Perl may core dump.
1188 int mg_copy(SV* sv, SV* nsv, const char* key, STRLEN klen);
1190 This routine checks to see what types of magic C<sv> has. If the mg_type
1191 field is an uppercase letter, then the mg_obj is copied to C<nsv>, but
1192 the mg_type field is changed to be the lowercase letter.
1194 =head2 Understanding the Magic of Tied Hashes and Arrays
1196 Tied hashes and arrays are magical beasts of the C<PERL_MAGIC_tied>
1199 WARNING: As of the 5.004 release, proper usage of the array and hash
1200 access functions requires understanding a few caveats. Some
1201 of these caveats are actually considered bugs in the API, to be fixed
1202 in later releases, and are bracketed with [MAYCHANGE] below. If
1203 you find yourself actually applying such information in this section, be
1204 aware that the behavior may change in the future, umm, without warning.
1206 The perl tie function associates a variable with an object that implements
1207 the various GET, SET, etc methods. To perform the equivalent of the perl
1208 tie function from an XSUB, you must mimic this behaviour. The code below
1209 carries out the necessary steps - firstly it creates a new hash, and then
1210 creates a second hash which it blesses into the class which will implement
1211 the tie methods. Lastly it ties the two hashes together, and returns a
1212 reference to the new tied hash. Note that the code below does NOT call the
1213 TIEHASH method in the MyTie class -
1214 see L<Calling Perl Routines from within C Programs> for details on how
1225 tie = newRV_noinc((SV*)newHV());
1226 stash = gv_stashpv("MyTie", GV_ADD);
1227 sv_bless(tie, stash);
1228 hv_magic(hash, (GV*)tie, PERL_MAGIC_tied);
1229 RETVAL = newRV_noinc(hash);
1233 The C<av_store> function, when given a tied array argument, merely
1234 copies the magic of the array onto the value to be "stored", using
1235 C<mg_copy>. It may also return NULL, indicating that the value did not
1236 actually need to be stored in the array. [MAYCHANGE] After a call to
1237 C<av_store> on a tied array, the caller will usually need to call
1238 C<mg_set(val)> to actually invoke the perl level "STORE" method on the
1239 TIEARRAY object. If C<av_store> did return NULL, a call to
1240 C<SvREFCNT_dec(val)> will also be usually necessary to avoid a memory
1243 The previous paragraph is applicable verbatim to tied hash access using the
1244 C<hv_store> and C<hv_store_ent> functions as well.
1246 C<av_fetch> and the corresponding hash functions C<hv_fetch> and
1247 C<hv_fetch_ent> actually return an undefined mortal value whose magic
1248 has been initialized using C<mg_copy>. Note the value so returned does not
1249 need to be deallocated, as it is already mortal. [MAYCHANGE] But you will
1250 need to call C<mg_get()> on the returned value in order to actually invoke
1251 the perl level "FETCH" method on the underlying TIE object. Similarly,
1252 you may also call C<mg_set()> on the return value after possibly assigning
1253 a suitable value to it using C<sv_setsv>, which will invoke the "STORE"
1254 method on the TIE object. [/MAYCHANGE]
1257 In other words, the array or hash fetch/store functions don't really
1258 fetch and store actual values in the case of tied arrays and hashes. They
1259 merely call C<mg_copy> to attach magic to the values that were meant to be
1260 "stored" or "fetched". Later calls to C<mg_get> and C<mg_set> actually
1261 do the job of invoking the TIE methods on the underlying objects. Thus
1262 the magic mechanism currently implements a kind of lazy access to arrays
1265 Currently (as of perl version 5.004), use of the hash and array access
1266 functions requires the user to be aware of whether they are operating on
1267 "normal" hashes and arrays, or on their tied variants. The API may be
1268 changed to provide more transparent access to both tied and normal data
1269 types in future versions.
1272 You would do well to understand that the TIEARRAY and TIEHASH interfaces
1273 are mere sugar to invoke some perl method calls while using the uniform hash
1274 and array syntax. The use of this sugar imposes some overhead (typically
1275 about two to four extra opcodes per FETCH/STORE operation, in addition to
1276 the creation of all the mortal variables required to invoke the methods).
1277 This overhead will be comparatively small if the TIE methods are themselves
1278 substantial, but if they are only a few statements long, the overhead
1279 will not be insignificant.
1281 =head2 Localizing changes
1283 Perl has a very handy construction
1290 This construction is I<approximately> equivalent to
1299 The biggest difference is that the first construction would
1300 reinstate the initial value of $var, irrespective of how control exits
1301 the block: C<goto>, C<return>, C<die>/C<eval>, etc. It is a little bit
1302 more efficient as well.
1304 There is a way to achieve a similar task from C via Perl API: create a
1305 I<pseudo-block>, and arrange for some changes to be automatically
1306 undone at the end of it, either explicit, or via a non-local exit (via
1307 die()). A I<block>-like construct is created by a pair of
1308 C<ENTER>/C<LEAVE> macros (see L<perlcall/"Returning a Scalar">).
1309 Such a construct may be created specially for some important localized
1310 task, or an existing one (like boundaries of enclosing Perl
1311 subroutine/block, or an existing pair for freeing TMPs) may be
1312 used. (In the second case the overhead of additional localization must
1313 be almost negligible.) Note that any XSUB is automatically enclosed in
1314 an C<ENTER>/C<LEAVE> pair.
1316 Inside such a I<pseudo-block> the following service is available:
1320 =item C<SAVEINT(int i)>
1322 =item C<SAVEIV(IV i)>
1324 =item C<SAVEI32(I32 i)>
1326 =item C<SAVELONG(long i)>
1328 These macros arrange things to restore the value of integer variable
1329 C<i> at the end of enclosing I<pseudo-block>.
1331 =item C<SAVESPTR(s)>
1333 =item C<SAVEPPTR(p)>
1335 These macros arrange things to restore the value of pointers C<s> and
1336 C<p>. C<s> must be a pointer of a type which survives conversion to
1337 C<SV*> and back, C<p> should be able to survive conversion to C<char*>
1340 =item C<SAVEFREESV(SV *sv)>
1342 The refcount of C<sv> would be decremented at the end of
1343 I<pseudo-block>. This is similar to C<sv_2mortal> in that it is also a
1344 mechanism for doing a delayed C<SvREFCNT_dec>. However, while C<sv_2mortal>
1345 extends the lifetime of C<sv> until the beginning of the next statement,
1346 C<SAVEFREESV> extends it until the end of the enclosing scope. These
1347 lifetimes can be wildly different.
1349 Also compare C<SAVEMORTALIZESV>.
1351 =item C<SAVEMORTALIZESV(SV *sv)>
1353 Just like C<SAVEFREESV>, but mortalizes C<sv> at the end of the current
1354 scope instead of decrementing its reference count. This usually has the
1355 effect of keeping C<sv> alive until the statement that called the currently
1356 live scope has finished executing.
1358 =item C<SAVEFREEOP(OP *op)>
1360 The C<OP *> is op_free()ed at the end of I<pseudo-block>.
1362 =item C<SAVEFREEPV(p)>
1364 The chunk of memory which is pointed to by C<p> is Safefree()ed at the
1365 end of I<pseudo-block>.
1367 =item C<SAVECLEARSV(SV *sv)>
1369 Clears a slot in the current scratchpad which corresponds to C<sv> at
1370 the end of I<pseudo-block>.
1372 =item C<SAVEDELETE(HV *hv, char *key, I32 length)>
1374 The key C<key> of C<hv> is deleted at the end of I<pseudo-block>. The
1375 string pointed to by C<key> is Safefree()ed. If one has a I<key> in
1376 short-lived storage, the corresponding string may be reallocated like
1379 SAVEDELETE(PL_defstash, savepv(tmpbuf), strlen(tmpbuf));
1381 =item C<SAVEDESTRUCTOR(DESTRUCTORFUNC_NOCONTEXT_t f, void *p)>
1383 At the end of I<pseudo-block> the function C<f> is called with the
1386 =item C<SAVEDESTRUCTOR_X(DESTRUCTORFUNC_t f, void *p)>
1388 At the end of I<pseudo-block> the function C<f> is called with the
1389 implicit context argument (if any), and C<p>.
1391 =item C<SAVESTACK_POS()>
1393 The current offset on the Perl internal stack (cf. C<SP>) is restored
1394 at the end of I<pseudo-block>.
1398 The following API list contains functions, thus one needs to
1399 provide pointers to the modifiable data explicitly (either C pointers,
1400 or Perlish C<GV *>s). Where the above macros take C<int>, a similar
1401 function takes C<int *>.
1405 =item C<SV* save_scalar(GV *gv)>
1407 Equivalent to Perl code C<local $gv>.
1409 =item C<AV* save_ary(GV *gv)>
1411 =item C<HV* save_hash(GV *gv)>
1413 Similar to C<save_scalar>, but localize C<@gv> and C<%gv>.
1415 =item C<void save_item(SV *item)>
1417 Duplicates the current value of C<SV>, on the exit from the current
1418 C<ENTER>/C<LEAVE> I<pseudo-block> will restore the value of C<SV>
1419 using the stored value. It doesn't handle magic. Use C<save_scalar> if
1422 =item C<void save_list(SV **sarg, I32 maxsarg)>
1424 A variant of C<save_item> which takes multiple arguments via an array
1425 C<sarg> of C<SV*> of length C<maxsarg>.
1427 =item C<SV* save_svref(SV **sptr)>
1429 Similar to C<save_scalar>, but will reinstate an C<SV *>.
1431 =item C<void save_aptr(AV **aptr)>
1433 =item C<void save_hptr(HV **hptr)>
1435 Similar to C<save_svref>, but localize C<AV *> and C<HV *>.
1439 The C<Alias> module implements localization of the basic types within the
1440 I<caller's scope>. People who are interested in how to localize things in
1441 the containing scope should take a look there too.
1445 =head2 XSUBs and the Argument Stack
1447 The XSUB mechanism is a simple way for Perl programs to access C subroutines.
1448 An XSUB routine will have a stack that contains the arguments from the Perl
1449 program, and a way to map from the Perl data structures to a C equivalent.
1451 The stack arguments are accessible through the C<ST(n)> macro, which returns
1452 the C<n>'th stack argument. Argument 0 is the first argument passed in the
1453 Perl subroutine call. These arguments are C<SV*>, and can be used anywhere
1456 Most of the time, output from the C routine can be handled through use of
1457 the RETVAL and OUTPUT directives. However, there are some cases where the
1458 argument stack is not already long enough to handle all the return values.
1459 An example is the POSIX tzname() call, which takes no arguments, but returns
1460 two, the local time zone's standard and summer time abbreviations.
1462 To handle this situation, the PPCODE directive is used and the stack is
1463 extended using the macro:
1467 where C<SP> is the macro that represents the local copy of the stack pointer,
1468 and C<num> is the number of elements the stack should be extended by.
1470 Now that there is room on the stack, values can be pushed on it using C<PUSHs>
1471 macro. The pushed values will often need to be "mortal" (See
1472 L</Reference Counts and Mortality>):
1474 PUSHs(sv_2mortal(newSViv(an_integer)))
1475 PUSHs(sv_2mortal(newSVuv(an_unsigned_integer)))
1476 PUSHs(sv_2mortal(newSVnv(a_double)))
1477 PUSHs(sv_2mortal(newSVpv("Some String",0)))
1478 /* Although the last example is better written as the more efficient: */
1479 PUSHs(newSVpvs_flags("Some String", SVs_TEMP))
1481 And now the Perl program calling C<tzname>, the two values will be assigned
1484 ($standard_abbrev, $summer_abbrev) = POSIX::tzname;
1486 An alternate (and possibly simpler) method to pushing values on the stack is
1491 This macro automatically adjust the stack for you, if needed. Thus, you
1492 do not need to call C<EXTEND> to extend the stack.
1494 Despite their suggestions in earlier versions of this document the macros
1495 C<(X)PUSH[iunp]> are I<not> suited to XSUBs which return multiple results.
1496 For that, either stick to the C<(X)PUSHs> macros shown above, or use the new
1497 C<m(X)PUSH[iunp]> macros instead; see L</Putting a C value on Perl stack>.
1499 For more information, consult L<perlxs> and L<perlxstut>.
1501 =head2 Autoloading with XSUBs
1503 If an AUTOLOAD routine is an XSUB, as with Perl subroutines, Perl puts the
1504 fully-qualified name of the autoloaded subroutine in the $AUTOLOAD variable
1505 of the XSUB's package.
1507 But it also puts the same information in certain fields of the XSUB itself:
1509 HV *stash = CvSTASH(cv);
1510 const char *subname = SvPVX(cv);
1511 STRLEN name_length = SvCUR(cv); /* in bytes */
1512 U32 is_utf8 = SvUTF8(cv);
1514 C<SvPVX(cv)> contains just the sub name itself, not including the package.
1515 For an AUTOLOAD routine in UNIVERSAL or one of its superclasses,
1516 C<CvSTASH(cv)> returns NULL during a method call on a nonexistent package.
1518 B<Note>: Setting $AUTOLOAD stopped working in 5.6.1, which did not support
1519 XS AUTOLOAD subs at all. Perl 5.8.0 introduced the use of fields in the
1520 XSUB itself. Perl 5.16.0 restored the setting of $AUTOLOAD. If you need
1521 to support 5.8-5.14, use the XSUB's fields.
1523 =head2 Calling Perl Routines from within C Programs
1525 There are four routines that can be used to call a Perl subroutine from
1526 within a C program. These four are:
1528 I32 call_sv(SV*, I32);
1529 I32 call_pv(const char*, I32);
1530 I32 call_method(const char*, I32);
1531 I32 call_argv(const char*, I32, register char**);
1533 The routine most often used is C<call_sv>. The C<SV*> argument
1534 contains either the name of the Perl subroutine to be called, or a
1535 reference to the subroutine. The second argument consists of flags
1536 that control the context in which the subroutine is called, whether
1537 or not the subroutine is being passed arguments, how errors should be
1538 trapped, and how to treat return values.
1540 All four routines return the number of arguments that the subroutine returned
1543 These routines used to be called C<perl_call_sv>, etc., before Perl v5.6.0,
1544 but those names are now deprecated; macros of the same name are provided for
1547 When using any of these routines (except C<call_argv>), the programmer
1548 must manipulate the Perl stack. These include the following macros and
1563 For a detailed description of calling conventions from C to Perl,
1564 consult L<perlcall>.
1566 =head2 Memory Allocation
1570 All memory meant to be used with the Perl API functions should be manipulated
1571 using the macros described in this section. The macros provide the necessary
1572 transparency between differences in the actual malloc implementation that is
1575 It is suggested that you enable the version of malloc that is distributed
1576 with Perl. It keeps pools of various sizes of unallocated memory in
1577 order to satisfy allocation requests more quickly. However, on some
1578 platforms, it may cause spurious malloc or free errors.
1580 The following three macros are used to initially allocate memory :
1582 Newx(pointer, number, type);
1583 Newxc(pointer, number, type, cast);
1584 Newxz(pointer, number, type);
1586 The first argument C<pointer> should be the name of a variable that will
1587 point to the newly allocated memory.
1589 The second and third arguments C<number> and C<type> specify how many of
1590 the specified type of data structure should be allocated. The argument
1591 C<type> is passed to C<sizeof>. The final argument to C<Newxc>, C<cast>,
1592 should be used if the C<pointer> argument is different from the C<type>
1595 Unlike the C<Newx> and C<Newxc> macros, the C<Newxz> macro calls C<memzero>
1596 to zero out all the newly allocated memory.
1600 Renew(pointer, number, type);
1601 Renewc(pointer, number, type, cast);
1604 These three macros are used to change a memory buffer size or to free a
1605 piece of memory no longer needed. The arguments to C<Renew> and C<Renewc>
1606 match those of C<New> and C<Newc> with the exception of not needing the
1607 "magic cookie" argument.
1611 Move(source, dest, number, type);
1612 Copy(source, dest, number, type);
1613 Zero(dest, number, type);
1615 These three macros are used to move, copy, or zero out previously allocated
1616 memory. The C<source> and C<dest> arguments point to the source and
1617 destination starting points. Perl will move, copy, or zero out C<number>
1618 instances of the size of the C<type> data structure (using the C<sizeof>
1623 The most recent development releases of Perl has been experimenting with
1624 removing Perl's dependency on the "normal" standard I/O suite and allowing
1625 other stdio implementations to be used. This involves creating a new
1626 abstraction layer that then calls whichever implementation of stdio Perl
1627 was compiled with. All XSUBs should now use the functions in the PerlIO
1628 abstraction layer and not make any assumptions about what kind of stdio
1631 For a complete description of the PerlIO abstraction, consult L<perlapio>.
1633 =head2 Putting a C value on Perl stack
1635 A lot of opcodes (this is an elementary operation in the internal perl
1636 stack machine) put an SV* on the stack. However, as an optimization
1637 the corresponding SV is (usually) not recreated each time. The opcodes
1638 reuse specially assigned SVs (I<target>s) which are (as a corollary)
1639 not constantly freed/created.
1641 Each of the targets is created only once (but see
1642 L<Scratchpads and recursion> below), and when an opcode needs to put
1643 an integer, a double, or a string on stack, it just sets the
1644 corresponding parts of its I<target> and puts the I<target> on stack.
1646 The macro to put this target on stack is C<PUSHTARG>, and it is
1647 directly used in some opcodes, as well as indirectly in zillions of
1648 others, which use it via C<(X)PUSH[iunp]>.
1650 Because the target is reused, you must be careful when pushing multiple
1651 values on the stack. The following code will not do what you think:
1656 This translates as "set C<TARG> to 10, push a pointer to C<TARG> onto
1657 the stack; set C<TARG> to 20, push a pointer to C<TARG> onto the stack".
1658 At the end of the operation, the stack does not contain the values 10
1659 and 20, but actually contains two pointers to C<TARG>, which we have set
1662 If you need to push multiple different values then you should either use
1663 the C<(X)PUSHs> macros, or else use the new C<m(X)PUSH[iunp]> macros,
1664 none of which make use of C<TARG>. The C<(X)PUSHs> macros simply push an
1665 SV* on the stack, which, as noted under L</XSUBs and the Argument Stack>,
1666 will often need to be "mortal". The new C<m(X)PUSH[iunp]> macros make
1667 this a little easier to achieve by creating a new mortal for you (via
1668 C<(X)PUSHmortal>), pushing that onto the stack (extending it if necessary
1669 in the case of the C<mXPUSH[iunp]> macros), and then setting its value.
1670 Thus, instead of writing this to "fix" the example above:
1672 XPUSHs(sv_2mortal(newSViv(10)))
1673 XPUSHs(sv_2mortal(newSViv(20)))
1675 you can simply write:
1680 On a related note, if you do use C<(X)PUSH[iunp]>, then you're going to
1681 need a C<dTARG> in your variable declarations so that the C<*PUSH*>
1682 macros can make use of the local variable C<TARG>. See also C<dTARGET>
1687 The question remains on when the SVs which are I<target>s for opcodes
1688 are created. The answer is that they are created when the current
1689 unit--a subroutine or a file (for opcodes for statements outside of
1690 subroutines)--is compiled. During this time a special anonymous Perl
1691 array is created, which is called a scratchpad for the current unit.
1693 A scratchpad keeps SVs which are lexicals for the current unit and are
1694 targets for opcodes. One can deduce that an SV lives on a scratchpad
1695 by looking on its flags: lexicals have C<SVs_PADMY> set, and
1696 I<target>s have C<SVs_PADTMP> set.
1698 The correspondence between OPs and I<target>s is not 1-to-1. Different
1699 OPs in the compile tree of the unit can use the same target, if this
1700 would not conflict with the expected life of the temporary.
1702 =head2 Scratchpads and recursion
1704 In fact it is not 100% true that a compiled unit contains a pointer to
1705 the scratchpad AV. In fact it contains a pointer to an AV of
1706 (initially) one element, and this element is the scratchpad AV. Why do
1707 we need an extra level of indirection?
1709 The answer is B<recursion>, and maybe B<threads>. Both
1710 these can create several execution pointers going into the same
1711 subroutine. For the subroutine-child not write over the temporaries
1712 for the subroutine-parent (lifespan of which covers the call to the
1713 child), the parent and the child should have different
1714 scratchpads. (I<And> the lexicals should be separate anyway!)
1716 So each subroutine is born with an array of scratchpads (of length 1).
1717 On each entry to the subroutine it is checked that the current
1718 depth of the recursion is not more than the length of this array, and
1719 if it is, new scratchpad is created and pushed into the array.
1721 The I<target>s on this scratchpad are C<undef>s, but they are already
1722 marked with correct flags.
1724 =head1 Compiled code
1728 Here we describe the internal form your code is converted to by
1729 Perl. Start with a simple example:
1733 This is converted to a tree similar to this one:
1741 (but slightly more complicated). This tree reflects the way Perl
1742 parsed your code, but has nothing to do with the execution order.
1743 There is an additional "thread" going through the nodes of the tree
1744 which shows the order of execution of the nodes. In our simplified
1745 example above it looks like:
1747 $b ---> $c ---> + ---> $a ---> assign-to
1749 But with the actual compile tree for C<$a = $b + $c> it is different:
1750 some nodes I<optimized away>. As a corollary, though the actual tree
1751 contains more nodes than our simplified example, the execution order
1752 is the same as in our example.
1754 =head2 Examining the tree
1756 If you have your perl compiled for debugging (usually done with
1757 C<-DDEBUGGING> on the C<Configure> command line), you may examine the
1758 compiled tree by specifying C<-Dx> on the Perl command line. The
1759 output takes several lines per node, and for C<$b+$c> it looks like
1764 FLAGS = (SCALAR,KIDS)
1766 TYPE = null ===> (4)
1768 FLAGS = (SCALAR,KIDS)
1770 3 TYPE = gvsv ===> 4
1776 TYPE = null ===> (5)
1778 FLAGS = (SCALAR,KIDS)
1780 4 TYPE = gvsv ===> 5
1786 This tree has 5 nodes (one per C<TYPE> specifier), only 3 of them are
1787 not optimized away (one per number in the left column). The immediate
1788 children of the given node correspond to C<{}> pairs on the same level
1789 of indentation, thus this listing corresponds to the tree:
1797 The execution order is indicated by C<===E<gt>> marks, thus it is C<3
1798 4 5 6> (node C<6> is not included into above listing), i.e.,
1799 C<gvsv gvsv add whatever>.
1801 Each of these nodes represents an op, a fundamental operation inside the
1802 Perl core. The code which implements each operation can be found in the
1803 F<pp*.c> files; the function which implements the op with type C<gvsv>
1804 is C<pp_gvsv>, and so on. As the tree above shows, different ops have
1805 different numbers of children: C<add> is a binary operator, as one would
1806 expect, and so has two children. To accommodate the various different
1807 numbers of children, there are various types of op data structure, and
1808 they link together in different ways.
1810 The simplest type of op structure is C<OP>: this has no children. Unary
1811 operators, C<UNOP>s, have one child, and this is pointed to by the
1812 C<op_first> field. Binary operators (C<BINOP>s) have not only an
1813 C<op_first> field but also an C<op_last> field. The most complex type of
1814 op is a C<LISTOP>, which has any number of children. In this case, the
1815 first child is pointed to by C<op_first> and the last child by
1816 C<op_last>. The children in between can be found by iteratively
1817 following the C<op_sibling> pointer from the first child to the last.
1819 There are also two other op types: a C<PMOP> holds a regular expression,
1820 and has no children, and a C<LOOP> may or may not have children. If the
1821 C<op_children> field is non-zero, it behaves like a C<LISTOP>. To
1822 complicate matters, if a C<UNOP> is actually a C<null> op after
1823 optimization (see L</Compile pass 2: context propagation>) it will still
1824 have children in accordance with its former type.
1826 Another way to examine the tree is to use a compiler back-end module, such
1829 =head2 Compile pass 1: check routines
1831 The tree is created by the compiler while I<yacc> code feeds it
1832 the constructions it recognizes. Since I<yacc> works bottom-up, so does
1833 the first pass of perl compilation.
1835 What makes this pass interesting for perl developers is that some
1836 optimization may be performed on this pass. This is optimization by
1837 so-called "check routines". The correspondence between node names
1838 and corresponding check routines is described in F<opcode.pl> (do not
1839 forget to run C<make regen_headers> if you modify this file).
1841 A check routine is called when the node is fully constructed except
1842 for the execution-order thread. Since at this time there are no
1843 back-links to the currently constructed node, one can do most any
1844 operation to the top-level node, including freeing it and/or creating
1845 new nodes above/below it.
1847 The check routine returns the node which should be inserted into the
1848 tree (if the top-level node was not modified, check routine returns
1851 By convention, check routines have names C<ck_*>. They are usually
1852 called from C<new*OP> subroutines (or C<convert>) (which in turn are
1853 called from F<perly.y>).
1855 =head2 Compile pass 1a: constant folding
1857 Immediately after the check routine is called the returned node is
1858 checked for being compile-time executable. If it is (the value is
1859 judged to be constant) it is immediately executed, and a I<constant>
1860 node with the "return value" of the corresponding subtree is
1861 substituted instead. The subtree is deleted.
1863 If constant folding was not performed, the execution-order thread is
1866 =head2 Compile pass 2: context propagation
1868 When a context for a part of compile tree is known, it is propagated
1869 down through the tree. At this time the context can have 5 values
1870 (instead of 2 for runtime context): void, boolean, scalar, list, and
1871 lvalue. In contrast with the pass 1 this pass is processed from top
1872 to bottom: a node's context determines the context for its children.
1874 Additional context-dependent optimizations are performed at this time.
1875 Since at this moment the compile tree contains back-references (via
1876 "thread" pointers), nodes cannot be free()d now. To allow
1877 optimized-away nodes at this stage, such nodes are null()ified instead
1878 of free()ing (i.e. their type is changed to OP_NULL).
1880 =head2 Compile pass 3: peephole optimization
1882 After the compile tree for a subroutine (or for an C<eval> or a file)
1883 is created, an additional pass over the code is performed. This pass
1884 is neither top-down or bottom-up, but in the execution order (with
1885 additional complications for conditionals). Optimizations performed
1886 at this stage are subject to the same restrictions as in the pass 2.
1888 Peephole optimizations are done by calling the function pointed to
1889 by the global variable C<PL_peepp>. By default, C<PL_peepp> just
1890 calls the function pointed to by the global variable C<PL_rpeepp>.
1891 By default, that performs some basic op fixups and optimisations along
1892 the execution-order op chain, and recursively calls C<PL_rpeepp> for
1893 each side chain of ops (resulting from conditionals). Extensions may
1894 provide additional optimisations or fixups, hooking into either the
1895 per-subroutine or recursive stage, like this:
1897 static peep_t prev_peepp;
1898 static void my_peep(pTHX_ OP *o)
1900 /* custom per-subroutine optimisation goes here */
1902 /* custom per-subroutine optimisation may also go here */
1905 prev_peepp = PL_peepp;
1908 static peep_t prev_rpeepp;
1909 static void my_rpeep(pTHX_ OP *o)
1912 for(; o; o = o->op_next) {
1913 /* custom per-op optimisation goes here */
1915 prev_rpeepp(orig_o);
1918 prev_rpeepp = PL_rpeepp;
1919 PL_rpeepp = my_rpeep;
1921 =head2 Pluggable runops
1923 The compile tree is executed in a runops function. There are two runops
1924 functions, in F<run.c> and in F<dump.c>. C<Perl_runops_debug> is used
1925 with DEBUGGING and C<Perl_runops_standard> is used otherwise. For fine
1926 control over the execution of the compile tree it is possible to provide
1927 your own runops function.
1929 It's probably best to copy one of the existing runops functions and
1930 change it to suit your needs. Then, in the BOOT section of your XS
1933 PL_runops = my_runops;
1935 This function should be as efficient as possible to keep your programs
1936 running as fast as possible.
1938 =head2 Compile-time scope hooks
1940 As of perl 5.14 it is possible to hook into the compile-time lexical
1941 scope mechanism using C<Perl_blockhook_register>. This is used like
1944 STATIC void my_start_hook(pTHX_ int full);
1945 STATIC BHK my_hooks;
1948 BhkENTRY_set(&my_hooks, bhk_start, my_start_hook);
1949 Perl_blockhook_register(aTHX_ &my_hooks);
1951 This will arrange to have C<my_start_hook> called at the start of
1952 compiling every lexical scope. The available hooks are:
1956 =item C<void bhk_start(pTHX_ int full)>
1958 This is called just after starting a new lexical scope. Note that Perl
1963 creates two scopes: the first starts at the C<(> and has C<full == 1>,
1964 the second starts at the C<{> and has C<full == 0>. Both end at the
1965 C<}>, so calls to C<start> and C<pre/post_end> will match. Anything
1966 pushed onto the save stack by this hook will be popped just before the
1967 scope ends (between the C<pre_> and C<post_end> hooks, in fact).
1969 =item C<void bhk_pre_end(pTHX_ OP **o)>
1971 This is called at the end of a lexical scope, just before unwinding the
1972 stack. I<o> is the root of the optree representing the scope; it is a
1973 double pointer so you can replace the OP if you need to.
1975 =item C<void bhk_post_end(pTHX_ OP **o)>
1977 This is called at the end of a lexical scope, just after unwinding the
1978 stack. I<o> is as above. Note that it is possible for calls to C<pre_>
1979 and C<post_end> to nest, if there is something on the save stack that
1982 =item C<void bhk_eval(pTHX_ OP *const o)>
1984 This is called just before starting to compile an C<eval STRING>, C<do
1985 FILE>, C<require> or C<use>, after the eval has been set up. I<o> is the
1986 OP that requested the eval, and will normally be an C<OP_ENTEREVAL>,
1987 C<OP_DOFILE> or C<OP_REQUIRE>.
1991 Once you have your hook functions, you need a C<BHK> structure to put
1992 them in. It's best to allocate it statically, since there is no way to
1993 free it once it's registered. The function pointers should be inserted
1994 into this structure using the C<BhkENTRY_set> macro, which will also set
1995 flags indicating which entries are valid. If you do need to allocate
1996 your C<BHK> dynamically for some reason, be sure to zero it before you
1999 Once registered, there is no mechanism to switch these hooks off, so if
2000 that is necessary you will need to do this yourself. An entry in C<%^H>
2001 is probably the best way, so the effect is lexically scoped; however it
2002 is also possible to use the C<BhkDISABLE> and C<BhkENABLE> macros to
2003 temporarily switch entries on and off. You should also be aware that
2004 generally speaking at least one scope will have opened before your
2005 extension is loaded, so you will see some C<pre/post_end> pairs that
2006 didn't have a matching C<start>.
2008 =head1 Examining internal data structures with the C<dump> functions
2010 To aid debugging, the source file F<dump.c> contains a number of
2011 functions which produce formatted output of internal data structures.
2013 The most commonly used of these functions is C<Perl_sv_dump>; it's used
2014 for dumping SVs, AVs, HVs, and CVs. The C<Devel::Peek> module calls
2015 C<sv_dump> to produce debugging output from Perl-space, so users of that
2016 module should already be familiar with its format.
2018 C<Perl_op_dump> can be used to dump an C<OP> structure or any of its
2019 derivatives, and produces output similar to C<perl -Dx>; in fact,
2020 C<Perl_dump_eval> will dump the main root of the code being evaluated,
2021 exactly like C<-Dx>.
2023 Other useful functions are C<Perl_dump_sub>, which turns a C<GV> into an
2024 op tree, C<Perl_dump_packsubs> which calls C<Perl_dump_sub> on all the
2025 subroutines in a package like so: (Thankfully, these are all xsubs, so
2026 there is no op tree)
2028 (gdb) print Perl_dump_packsubs(PL_defstash)
2030 SUB attributes::bootstrap = (xsub 0x811fedc 0)
2032 SUB UNIVERSAL::can = (xsub 0x811f50c 0)
2034 SUB UNIVERSAL::isa = (xsub 0x811f304 0)
2036 SUB UNIVERSAL::VERSION = (xsub 0x811f7ac 0)
2038 SUB DynaLoader::boot_DynaLoader = (xsub 0x805b188 0)
2040 and C<Perl_dump_all>, which dumps all the subroutines in the stash and
2041 the op tree of the main root.
2043 =head1 How multiple interpreters and concurrency are supported
2045 =head2 Background and PERL_IMPLICIT_CONTEXT
2047 The Perl interpreter can be regarded as a closed box: it has an API
2048 for feeding it code or otherwise making it do things, but it also has
2049 functions for its own use. This smells a lot like an object, and
2050 there are ways for you to build Perl so that you can have multiple
2051 interpreters, with one interpreter represented either as a C structure,
2052 or inside a thread-specific structure. These structures contain all
2053 the context, the state of that interpreter.
2055 One macro controls the major Perl build flavor: MULTIPLICITY. The
2056 MULTIPLICITY build has a C structure that packages all the interpreter
2057 state. With multiplicity-enabled perls, PERL_IMPLICIT_CONTEXT is also
2058 normally defined, and enables the support for passing in a "hidden" first
2059 argument that represents all three data structures. MULTIPLICITY makes
2060 multi-threaded perls possible (with the ithreads threading model, related
2061 to the macro USE_ITHREADS.)
2063 Two other "encapsulation" macros are the PERL_GLOBAL_STRUCT and
2064 PERL_GLOBAL_STRUCT_PRIVATE (the latter turns on the former, and the
2065 former turns on MULTIPLICITY.) The PERL_GLOBAL_STRUCT causes all the
2066 internal variables of Perl to be wrapped inside a single global struct,
2067 struct perl_vars, accessible as (globals) &PL_Vars or PL_VarsPtr or
2068 the function Perl_GetVars(). The PERL_GLOBAL_STRUCT_PRIVATE goes
2069 one step further, there is still a single struct (allocated in main()
2070 either from heap or from stack) but there are no global data symbols
2071 pointing to it. In either case the global struct should be initialised
2072 as the very first thing in main() using Perl_init_global_struct() and
2073 correspondingly tear it down after perl_free() using Perl_free_global_struct(),
2074 please see F<miniperlmain.c> for usage details. You may also need
2075 to use C<dVAR> in your coding to "declare the global variables"
2076 when you are using them. dTHX does this for you automatically.
2078 To see whether you have non-const data you can use a BSD-compatible C<nm>:
2080 nm libperl.a | grep -v ' [TURtr] '
2082 If this displays any C<D> or C<d> symbols, you have non-const data.
2084 For backward compatibility reasons defining just PERL_GLOBAL_STRUCT
2085 doesn't actually hide all symbols inside a big global struct: some
2086 PerlIO_xxx vtables are left visible. The PERL_GLOBAL_STRUCT_PRIVATE
2087 then hides everything (see how the PERLIO_FUNCS_DECL is used).
2089 All this obviously requires a way for the Perl internal functions to be
2090 either subroutines taking some kind of structure as the first
2091 argument, or subroutines taking nothing as the first argument. To
2092 enable these two very different ways of building the interpreter,
2093 the Perl source (as it does in so many other situations) makes heavy
2094 use of macros and subroutine naming conventions.
2096 First problem: deciding which functions will be public API functions and
2097 which will be private. All functions whose names begin C<S_> are private
2098 (think "S" for "secret" or "static"). All other functions begin with
2099 "Perl_", but just because a function begins with "Perl_" does not mean it is
2100 part of the API. (See L</Internal Functions>.) The easiest way to be B<sure> a
2101 function is part of the API is to find its entry in L<perlapi>.
2102 If it exists in L<perlapi>, it's part of the API. If it doesn't, and you
2103 think it should be (i.e., you need it for your extension), send mail via
2104 L<perlbug> explaining why you think it should be.
2106 Second problem: there must be a syntax so that the same subroutine
2107 declarations and calls can pass a structure as their first argument,
2108 or pass nothing. To solve this, the subroutines are named and
2109 declared in a particular way. Here's a typical start of a static
2110 function used within the Perl guts:
2113 S_incline(pTHX_ char *s)
2115 STATIC becomes "static" in C, and may be #define'd to nothing in some
2116 configurations in future.
2118 A public function (i.e. part of the internal API, but not necessarily
2119 sanctioned for use in extensions) begins like this:
2122 Perl_sv_setiv(pTHX_ SV* dsv, IV num)
2124 C<pTHX_> is one of a number of macros (in F<perl.h>) that hide the
2125 details of the interpreter's context. THX stands for "thread", "this",
2126 or "thingy", as the case may be. (And no, George Lucas is not involved. :-)
2127 The first character could be 'p' for a B<p>rototype, 'a' for B<a>rgument,
2128 or 'd' for B<d>eclaration, so we have C<pTHX>, C<aTHX> and C<dTHX>, and
2131 When Perl is built without options that set PERL_IMPLICIT_CONTEXT, there is no
2132 first argument containing the interpreter's context. The trailing underscore
2133 in the pTHX_ macro indicates that the macro expansion needs a comma
2134 after the context argument because other arguments follow it. If
2135 PERL_IMPLICIT_CONTEXT is not defined, pTHX_ will be ignored, and the
2136 subroutine is not prototyped to take the extra argument. The form of the
2137 macro without the trailing underscore is used when there are no additional
2140 When a core function calls another, it must pass the context. This
2141 is normally hidden via macros. Consider C<sv_setiv>. It expands into
2142 something like this:
2144 #ifdef PERL_IMPLICIT_CONTEXT
2145 #define sv_setiv(a,b) Perl_sv_setiv(aTHX_ a, b)
2146 /* can't do this for vararg functions, see below */
2148 #define sv_setiv Perl_sv_setiv
2151 This works well, and means that XS authors can gleefully write:
2155 and still have it work under all the modes Perl could have been
2158 This doesn't work so cleanly for varargs functions, though, as macros
2159 imply that the number of arguments is known in advance. Instead we
2160 either need to spell them out fully, passing C<aTHX_> as the first
2161 argument (the Perl core tends to do this with functions like
2162 Perl_warner), or use a context-free version.
2164 The context-free version of Perl_warner is called
2165 Perl_warner_nocontext, and does not take the extra argument. Instead
2166 it does dTHX; to get the context from thread-local storage. We
2167 C<#define warner Perl_warner_nocontext> so that extensions get source
2168 compatibility at the expense of performance. (Passing an arg is
2169 cheaper than grabbing it from thread-local storage.)
2171 You can ignore [pad]THXx when browsing the Perl headers/sources.
2172 Those are strictly for use within the core. Extensions and embedders
2173 need only be aware of [pad]THX.
2175 =head2 So what happened to dTHR?
2177 C<dTHR> was introduced in perl 5.005 to support the older thread model.
2178 The older thread model now uses the C<THX> mechanism to pass context
2179 pointers around, so C<dTHR> is not useful any more. Perl 5.6.0 and
2180 later still have it for backward source compatibility, but it is defined
2183 =head2 How do I use all this in extensions?
2185 When Perl is built with PERL_IMPLICIT_CONTEXT, extensions that call
2186 any functions in the Perl API will need to pass the initial context
2187 argument somehow. The kicker is that you will need to write it in
2188 such a way that the extension still compiles when Perl hasn't been
2189 built with PERL_IMPLICIT_CONTEXT enabled.
2191 There are three ways to do this. First, the easy but inefficient way,
2192 which is also the default, in order to maintain source compatibility
2193 with extensions: whenever F<XSUB.h> is #included, it redefines the aTHX
2194 and aTHX_ macros to call a function that will return the context.
2195 Thus, something like:
2199 in your extension will translate to this when PERL_IMPLICIT_CONTEXT is
2202 Perl_sv_setiv(Perl_get_context(), sv, num);
2204 or to this otherwise:
2206 Perl_sv_setiv(sv, num);
2208 You have to do nothing new in your extension to get this; since
2209 the Perl library provides Perl_get_context(), it will all just
2212 The second, more efficient way is to use the following template for
2215 #define PERL_NO_GET_CONTEXT /* we want efficiency */
2220 STATIC void my_private_function(int arg1, int arg2);
2223 my_private_function(int arg1, int arg2)
2225 dTHX; /* fetch context */
2226 ... call many Perl API functions ...
2231 MODULE = Foo PACKAGE = Foo
2239 my_private_function(arg, 10);
2241 Note that the only two changes from the normal way of writing an
2242 extension is the addition of a C<#define PERL_NO_GET_CONTEXT> before
2243 including the Perl headers, followed by a C<dTHX;> declaration at
2244 the start of every function that will call the Perl API. (You'll
2245 know which functions need this, because the C compiler will complain
2246 that there's an undeclared identifier in those functions.) No changes
2247 are needed for the XSUBs themselves, because the XS() macro is
2248 correctly defined to pass in the implicit context if needed.
2250 The third, even more efficient way is to ape how it is done within
2254 #define PERL_NO_GET_CONTEXT /* we want efficiency */
2259 /* pTHX_ only needed for functions that call Perl API */
2260 STATIC void my_private_function(pTHX_ int arg1, int arg2);
2263 my_private_function(pTHX_ int arg1, int arg2)
2265 /* dTHX; not needed here, because THX is an argument */
2266 ... call Perl API functions ...
2271 MODULE = Foo PACKAGE = Foo
2279 my_private_function(aTHX_ arg, 10);
2281 This implementation never has to fetch the context using a function
2282 call, since it is always passed as an extra argument. Depending on
2283 your needs for simplicity or efficiency, you may mix the previous
2284 two approaches freely.
2286 Never add a comma after C<pTHX> yourself--always use the form of the
2287 macro with the underscore for functions that take explicit arguments,
2288 or the form without the argument for functions with no explicit arguments.
2290 If one is compiling Perl with the C<-DPERL_GLOBAL_STRUCT> the C<dVAR>
2291 definition is needed if the Perl global variables (see F<perlvars.h>
2292 or F<globvar.sym>) are accessed in the function and C<dTHX> is not
2293 used (the C<dTHX> includes the C<dVAR> if necessary). One notices
2294 the need for C<dVAR> only with the said compile-time define, because
2295 otherwise the Perl global variables are visible as-is.
2297 =head2 Should I do anything special if I call perl from multiple threads?
2299 If you create interpreters in one thread and then proceed to call them in
2300 another, you need to make sure perl's own Thread Local Storage (TLS) slot is
2301 initialized correctly in each of those threads.
2303 The C<perl_alloc> and C<perl_clone> API functions will automatically set
2304 the TLS slot to the interpreter they created, so that there is no need to do
2305 anything special if the interpreter is always accessed in the same thread that
2306 created it, and that thread did not create or call any other interpreters
2307 afterwards. If that is not the case, you have to set the TLS slot of the
2308 thread before calling any functions in the Perl API on that particular
2309 interpreter. This is done by calling the C<PERL_SET_CONTEXT> macro in that
2310 thread as the first thing you do:
2312 /* do this before doing anything else with some_perl */
2313 PERL_SET_CONTEXT(some_perl);
2315 ... other Perl API calls on some_perl go here ...
2317 =head2 Future Plans and PERL_IMPLICIT_SYS
2319 Just as PERL_IMPLICIT_CONTEXT provides a way to bundle up everything
2320 that the interpreter knows about itself and pass it around, so too are
2321 there plans to allow the interpreter to bundle up everything it knows
2322 about the environment it's running on. This is enabled with the
2323 PERL_IMPLICIT_SYS macro. Currently it only works with USE_ITHREADS on
2326 This allows the ability to provide an extra pointer (called the "host"
2327 environment) for all the system calls. This makes it possible for
2328 all the system stuff to maintain their own state, broken down into
2329 seven C structures. These are thin wrappers around the usual system
2330 calls (see F<win32/perllib.c>) for the default perl executable, but for a
2331 more ambitious host (like the one that would do fork() emulation) all
2332 the extra work needed to pretend that different interpreters are
2333 actually different "processes", would be done here.
2335 The Perl engine/interpreter and the host are orthogonal entities.
2336 There could be one or more interpreters in a process, and one or
2337 more "hosts", with free association between them.
2339 =head1 Internal Functions
2341 All of Perl's internal functions which will be exposed to the outside
2342 world are prefixed by C<Perl_> so that they will not conflict with XS
2343 functions or functions used in a program in which Perl is embedded.
2344 Similarly, all global variables begin with C<PL_>. (By convention,
2345 static functions start with C<S_>.)
2347 Inside the Perl core (C<PERL_CORE> defined), you can get at the functions
2348 either with or without the C<Perl_> prefix, thanks to a bunch of defines
2349 that live in F<embed.h>. Note that extension code should I<not> set
2350 C<PERL_CORE>; this exposes the full perl internals, and is likely to cause
2351 breakage of the XS in each new perl release.
2353 The file F<embed.h> is generated automatically from
2354 F<embed.pl> and F<embed.fnc>. F<embed.pl> also creates the prototyping
2355 header files for the internal functions, generates the documentation
2356 and a lot of other bits and pieces. It's important that when you add
2357 a new function to the core or change an existing one, you change the
2358 data in the table in F<embed.fnc> as well. Here's a sample entry from
2361 Apd |SV** |av_fetch |AV* ar|I32 key|I32 lval
2363 The second column is the return type, the third column the name. Columns
2364 after that are the arguments. The first column is a set of flags:
2370 This function is a part of the public API. All such functions should also
2371 have 'd', very few do not.
2375 This function has a C<Perl_> prefix; i.e. it is defined as
2380 This function has documentation using the C<apidoc> feature which we'll
2381 look at in a second. Some functions have 'd' but not 'A'; docs are good.
2385 Other available flags are:
2391 This is a static function and is defined as C<STATIC S_whatever>, and
2392 usually called within the sources as C<whatever(...)>.
2396 This does not need a interpreter context, so the definition has no
2397 C<pTHX>, and it follows that callers don't use C<aTHX>. (See
2398 L</Background and PERL_IMPLICIT_CONTEXT>.)
2402 This function never returns; C<croak>, C<exit> and friends.
2406 This function takes a variable number of arguments, C<printf> style.
2407 The argument list should end with C<...>, like this:
2409 Afprd |void |croak |const char* pat|...
2413 This function is part of the experimental development API, and may change
2414 or disappear without notice.
2418 This function should not have a compatibility macro to define, say,
2419 C<Perl_parse> to C<parse>. It must be called as C<Perl_parse>.
2423 This function isn't exported out of the Perl core.
2427 This is implemented as a macro.
2431 This function is explicitly exported.
2435 This function is visible to extensions included in the Perl core.
2439 Binary backward compatibility; this function is a macro but also has
2440 a C<Perl_> implementation (which is exported).
2444 See the comments at the top of C<embed.fnc> for others.
2448 If you edit F<embed.pl> or F<embed.fnc>, you will need to run
2449 C<make regen_headers> to force a rebuild of F<embed.h> and other
2450 auto-generated files.
2452 =head2 Formatted Printing of IVs, UVs, and NVs
2454 If you are printing IVs, UVs, or NVS instead of the stdio(3) style
2455 formatting codes like C<%d>, C<%ld>, C<%f>, you should use the
2456 following macros for portability
2461 UVxf UV in hexadecimal
2466 These will take care of 64-bit integers and long doubles.
2469 printf("IV is %"IVdf"\n", iv);
2471 The IVdf will expand to whatever is the correct format for the IVs.
2473 If you are printing addresses of pointers, use UVxf combined
2474 with PTR2UV(), do not use %lx or %p.
2476 =head2 Pointer-To-Integer and Integer-To-Pointer
2478 Because pointer size does not necessarily equal integer size,
2479 use the follow macros to do it right.
2484 INT2PTR(pointertotype, integer)
2489 SV *sv = INT2PTR(SV*, iv);
2496 =head2 Exception Handling
2498 There are a couple of macros to do very basic exception handling in XS
2499 modules. You have to define C<NO_XSLOCKS> before including F<XSUB.h> to
2500 be able to use these macros:
2505 You can use these macros if you call code that may croak, but you need
2506 to do some cleanup before giving control back to Perl. For example:
2508 dXCPT; /* set up necessary variables */
2511 code_that_may_croak();
2516 /* do cleanup here */
2520 Note that you always have to rethrow an exception that has been
2521 caught. Using these macros, it is not possible to just catch the
2522 exception and ignore it. If you have to ignore the exception, you
2523 have to use the C<call_*> function.
2525 The advantage of using the above macros is that you don't have
2526 to setup an extra function for C<call_*>, and that using these
2527 macros is faster than using C<call_*>.
2529 =head2 Source Documentation
2531 There's an effort going on to document the internal functions and
2532 automatically produce reference manuals from them - L<perlapi> is one
2533 such manual which details all the functions which are available to XS
2534 writers. L<perlintern> is the autogenerated manual for the functions
2535 which are not part of the API and are supposedly for internal use only.
2537 Source documentation is created by putting POD comments into the C
2541 =for apidoc sv_setiv
2543 Copies an integer into the given SV. Does not handle 'set' magic. See
2549 Please try and supply some documentation if you add functions to the
2552 =head2 Backwards compatibility
2554 The Perl API changes over time. New functions are added or the interfaces
2555 of existing functions are changed. The C<Devel::PPPort> module tries to
2556 provide compatibility code for some of these changes, so XS writers don't
2557 have to code it themselves when supporting multiple versions of Perl.
2559 C<Devel::PPPort> generates a C header file F<ppport.h> that can also
2560 be run as a Perl script. To generate F<ppport.h>, run:
2562 perl -MDevel::PPPort -eDevel::PPPort::WriteFile
2564 Besides checking existing XS code, the script can also be used to retrieve
2565 compatibility information for various API calls using the C<--api-info>
2566 command line switch. For example:
2568 % perl ppport.h --api-info=sv_magicext
2570 For details, see C<perldoc ppport.h>.
2572 =head1 Unicode Support
2574 Perl 5.6.0 introduced Unicode support. It's important for porters and XS
2575 writers to understand this support and make sure that the code they
2576 write does not corrupt Unicode data.
2578 =head2 What B<is> Unicode, anyway?
2580 In the olden, less enlightened times, we all used to use ASCII. Most of
2581 us did, anyway. The big problem with ASCII is that it's American. Well,
2582 no, that's not actually the problem; the problem is that it's not
2583 particularly useful for people who don't use the Roman alphabet. What
2584 used to happen was that particular languages would stick their own
2585 alphabet in the upper range of the sequence, between 128 and 255. Of
2586 course, we then ended up with plenty of variants that weren't quite
2587 ASCII, and the whole point of it being a standard was lost.
2589 Worse still, if you've got a language like Chinese or
2590 Japanese that has hundreds or thousands of characters, then you really
2591 can't fit them into a mere 256, so they had to forget about ASCII
2592 altogether, and build their own systems using pairs of numbers to refer
2595 To fix this, some people formed Unicode, Inc. and
2596 produced a new character set containing all the characters you can
2597 possibly think of and more. There are several ways of representing these
2598 characters, and the one Perl uses is called UTF-8. UTF-8 uses
2599 a variable number of bytes to represent a character. You can learn more
2600 about Unicode and Perl's Unicode model in L<perlunicode>.
2602 =head2 How can I recognise a UTF-8 string?
2604 You can't. This is because UTF-8 data is stored in bytes just like
2605 non-UTF-8 data. The Unicode character 200, (C<0xC8> for you hex types)
2606 capital E with a grave accent, is represented by the two bytes
2607 C<v196.172>. Unfortunately, the non-Unicode string C<chr(196).chr(172)>
2608 has that byte sequence as well. So you can't tell just by looking - this
2609 is what makes Unicode input an interesting problem.
2611 In general, you either have to know what you're dealing with, or you
2612 have to guess. The API function C<is_utf8_string> can help; it'll tell
2613 you if a string contains only valid UTF-8 characters. However, it can't
2614 do the work for you. On a character-by-character basis, C<is_utf8_char>
2615 will tell you whether the current character in a string is valid UTF-8.
2617 =head2 How does UTF-8 represent Unicode characters?
2619 As mentioned above, UTF-8 uses a variable number of bytes to store a
2620 character. Characters with values 0...127 are stored in one byte, just
2621 like good ol' ASCII. Character 128 is stored as C<v194.128>; this
2622 continues up to character 191, which is C<v194.191>. Now we've run out of
2623 bits (191 is binary C<10111111>) so we move on; 192 is C<v195.128>. And
2624 so it goes on, moving to three bytes at character 2048.
2626 Assuming you know you're dealing with a UTF-8 string, you can find out
2627 how long the first character in it is with the C<UTF8SKIP> macro:
2629 char *utf = "\305\233\340\240\201";
2632 len = UTF8SKIP(utf); /* len is 2 here */
2634 len = UTF8SKIP(utf); /* len is 3 here */
2636 Another way to skip over characters in a UTF-8 string is to use
2637 C<utf8_hop>, which takes a string and a number of characters to skip
2638 over. You're on your own about bounds checking, though, so don't use it
2641 All bytes in a multi-byte UTF-8 character will have the high bit set,
2642 so you can test if you need to do something special with this
2643 character like this (the UTF8_IS_INVARIANT() is a macro that tests
2644 whether the byte can be encoded as a single byte even in UTF-8):
2647 UV uv; /* Note: a UV, not a U8, not a char */
2649 if (!UTF8_IS_INVARIANT(*utf))
2650 /* Must treat this as UTF-8 */
2651 uv = utf8_to_uv(utf);
2653 /* OK to treat this character as a byte */
2656 You can also see in that example that we use C<utf8_to_uv> to get the
2657 value of the character; the inverse function C<uv_to_utf8> is available
2658 for putting a UV into UTF-8:
2660 if (!UTF8_IS_INVARIANT(uv))
2661 /* Must treat this as UTF8 */
2662 utf8 = uv_to_utf8(utf8, uv);
2664 /* OK to treat this character as a byte */
2667 You B<must> convert characters to UVs using the above functions if
2668 you're ever in a situation where you have to match UTF-8 and non-UTF-8
2669 characters. You may not skip over UTF-8 characters in this case. If you
2670 do this, you'll lose the ability to match hi-bit non-UTF-8 characters;
2671 for instance, if your UTF-8 string contains C<v196.172>, and you skip
2672 that character, you can never match a C<chr(200)> in a non-UTF-8 string.
2675 =head2 How does Perl store UTF-8 strings?
2677 Currently, Perl deals with Unicode strings and non-Unicode strings
2678 slightly differently. A flag in the SV, C<SVf_UTF8>, indicates that the
2679 string is internally encoded as UTF-8. Without it, the byte value is the
2680 codepoint number and vice versa (in other words, the string is encoded
2681 as iso-8859-1, but C<use feature 'unicode_strings'> is needed to get iso-8859-1
2682 semantics). You can check and manipulate this flag with the
2689 This flag has an important effect on Perl's treatment of the string: if
2690 Unicode data is not properly distinguished, regular expressions,
2691 C<length>, C<substr> and other string handling operations will have
2692 undesirable results.
2694 The problem comes when you have, for instance, a string that isn't
2695 flagged as UTF-8, and contains a byte sequence that could be UTF-8 -
2696 especially when combining non-UTF-8 and UTF-8 strings.
2698 Never forget that the C<SVf_UTF8> flag is separate to the PV value; you
2699 need be sure you don't accidentally knock it off while you're
2700 manipulating SVs. More specifically, you cannot expect to do this:
2709 nsv = newSVpvn(p, len);
2711 The C<char*> string does not tell you the whole story, and you can't
2712 copy or reconstruct an SV just by copying the string value. Check if the
2713 old SV has the UTF8 flag set, and act accordingly:
2717 nsv = newSVpvn(p, len);
2721 In fact, your C<frobnicate> function should be made aware of whether or
2722 not it's dealing with UTF-8 data, so that it can handle the string
2725 Since just passing an SV to an XS function and copying the data of
2726 the SV is not enough to copy the UTF8 flags, even less right is just
2727 passing a C<char *> to an XS function.
2729 =head2 How do I convert a string to UTF-8?
2731 If you're mixing UTF-8 and non-UTF-8 strings, it is necessary to upgrade
2732 one of the strings to UTF-8. If you've got an SV, the easiest way to do
2735 sv_utf8_upgrade(sv);
2737 However, you must not do this, for example:
2740 sv_utf8_upgrade(left);
2742 If you do this in a binary operator, you will actually change one of the
2743 strings that came into the operator, and, while it shouldn't be noticeable
2744 by the end user, it can cause problems in deficient code.
2746 Instead, C<bytes_to_utf8> will give you a UTF-8-encoded B<copy> of its
2747 string argument. This is useful for having the data available for
2748 comparisons and so on, without harming the original SV. There's also
2749 C<utf8_to_bytes> to go the other way, but naturally, this will fail if
2750 the string contains any characters above 255 that can't be represented
2753 =head2 Is there anything else I need to know?
2755 Not really. Just remember these things:
2761 There's no way to tell if a string is UTF-8 or not. You can tell if an SV
2762 is UTF-8 by looking at is C<SvUTF8> flag. Don't forget to set the flag if
2763 something should be UTF-8. Treat the flag as part of the PV, even though
2764 it's not - if you pass on the PV to somewhere, pass on the flag too.
2768 If a string is UTF-8, B<always> use C<utf8_to_uv> to get at the value,
2769 unless C<UTF8_IS_INVARIANT(*s)> in which case you can use C<*s>.
2773 When writing a character C<uv> to a UTF-8 string, B<always> use
2774 C<uv_to_utf8>, unless C<UTF8_IS_INVARIANT(uv))> in which case
2775 you can use C<*s = uv>.
2779 Mixing UTF-8 and non-UTF-8 strings is tricky. Use C<bytes_to_utf8> to get
2780 a new string which is UTF-8 encoded, and then combine them.
2784 =head1 Custom Operators
2786 Custom operator support is a new experimental feature that allows you to
2787 define your own ops. This is primarily to allow the building of
2788 interpreters for other languages in the Perl core, but it also allows
2789 optimizations through the creation of "macro-ops" (ops which perform the
2790 functions of multiple ops which are usually executed together, such as
2791 C<gvsv, gvsv, add>.)
2793 This feature is implemented as a new op type, C<OP_CUSTOM>. The Perl
2794 core does not "know" anything special about this op type, and so it will
2795 not be involved in any optimizations. This also means that you can
2796 define your custom ops to be any op structure - unary, binary, list and
2799 It's important to know what custom operators won't do for you. They
2800 won't let you add new syntax to Perl, directly. They won't even let you
2801 add new keywords, directly. In fact, they won't change the way Perl
2802 compiles a program at all. You have to do those changes yourself, after
2803 Perl has compiled the program. You do this either by manipulating the op
2804 tree using a C<CHECK> block and the C<B::Generate> module, or by adding
2805 a custom peephole optimizer with the C<optimize> module.
2807 When you do this, you replace ordinary Perl ops with custom ops by
2808 creating ops with the type C<OP_CUSTOM> and the C<pp_addr> of your own
2809 PP function. This should be defined in XS code, and should look like
2810 the PP ops in C<pp_*.c>. You are responsible for ensuring that your op
2811 takes the appropriate number of values from the stack, and you are
2812 responsible for adding stack marks if necessary.
2814 You should also "register" your op with the Perl interpreter so that it
2815 can produce sensible error and warning messages. Since it is possible to
2816 have multiple custom ops within the one "logical" op type C<OP_CUSTOM>,
2817 Perl uses the value of C<< o->op_ppaddr >> to determine which custom op
2818 it is dealing with. You should create an C<XOP> structure for each
2819 ppaddr you use, set the properties of the custom op with
2820 C<XopENTRY_set>, and register the structure against the ppaddr using
2821 C<Perl_custom_op_register>. A trivial example might look like:
2824 static OP *my_pp(pTHX);
2827 XopENTRY_set(&my_xop, xop_name, "myxop");
2828 XopENTRY_set(&my_xop, xop_desc, "Useless custom op");
2829 Perl_custom_op_register(aTHX_ my_pp, &my_xop);
2831 The available fields in the structure are:
2837 A short name for your op. This will be included in some error messages,
2838 and will also be returned as C<< $op->name >> by the L<B|B> module, so
2839 it will appear in the output of module like L<B::Concise|B::Concise>.
2843 A short description of the function of the op.
2847 Which of the various C<*OP> structures this op uses. This should be one of
2848 the C<OA_*> constants from F<op.h>, namely
2868 =item OA_PVOP_OR_SVOP
2870 This should be interpreted as 'C<PVOP>' only. The C<_OR_SVOP> is because
2871 the only core C<PVOP>, C<OP_TRANS>, can sometimes be a C<SVOP> instead.
2879 The other C<OA_*> constants should not be used.
2883 This member is of type C<Perl_cpeep_t>, which expands to C<void
2884 (*Perl_cpeep_t)(aTHX_ OP *o, OP *oldop)>. If it is set, this function
2885 will be called from C<Perl_rpeep> when ops of this type are encountered
2886 by the peephole optimizer. I<o> is the OP that needs optimizing;
2887 I<oldop> is the previous OP optimized, whose C<op_next> points to I<o>.
2891 C<B::Generate> directly supports the creation of custom ops by name.
2895 Until May 1997, this document was maintained by Jeff Okamoto
2896 E<lt>okamoto@corp.hp.comE<gt>. It is now maintained as part of Perl
2897 itself by the Perl 5 Porters E<lt>perl5-porters@perl.orgE<gt>.
2899 With lots of help and suggestions from Dean Roehrich, Malcolm Beattie,
2900 Andreas Koenig, Paul Hudson, Ilya Zakharevich, Paul Marquess, Neil
2901 Bowers, Matthew Green, Tim Bunce, Spider Boardman, Ulrich Pfeifer,
2902 Stephen McCamant, and Gurusamy Sarathy.
2906 L<perlapi>, L<perlintern>, L<perlxs>, L<perlembed>