3 perlguts - Introduction to the Perl API
7 This document attempts to describe how to use the Perl API, as well as
8 containing 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*, int);
47 SV* newSVpvn(const char*, int);
48 SV* newSVpvf(const char*, ...);
51 If you require more complex initialisation you can create an empty SV with
52 newSV(len). If C<len> is 0 an empty SV of type NULL is returned, else an
53 SV of type PV is returned with len + 1 (for the NUL) bytes of storage
54 allocated, accessible via SvPVX. In both cases the SV has value undef.
56 SV* newSV(0); /* no storage allocated */
57 SV* newSV(10); /* 10 (+1) bytes of uninitialised storage allocated */
59 To change the value of an *already-existing* SV, there are eight routines:
61 void sv_setiv(SV*, IV);
62 void sv_setuv(SV*, UV);
63 void sv_setnv(SV*, double);
64 void sv_setpv(SV*, const char*);
65 void sv_setpvn(SV*, const char*, int)
66 void sv_setpvf(SV*, const char*, ...);
67 void sv_vsetpvfn(SV*, const char*, STRLEN, va_list *, SV **, I32, bool *);
68 void sv_setsv(SV*, SV*);
70 Notice that you can choose to specify the length of the string to be
71 assigned by using C<sv_setpvn>, C<newSVpvn>, or C<newSVpv>, or you may
72 allow Perl to calculate the length by using C<sv_setpv> or by specifying
73 0 as the second argument to C<newSVpv>. Be warned, though, that Perl will
74 determine the string's length by using C<strlen>, which depends on the
75 string terminating with a NUL character.
77 The arguments of C<sv_setpvf> are processed like C<sprintf>, and the
78 formatted output becomes the value.
80 C<sv_vsetpvfn> is an analogue of C<vsprintf>, but it allows you to specify
81 either a pointer to a variable argument list or the address and length of
82 an array of SVs. The last argument points to a boolean; on return, if that
83 boolean is true, then locale-specific information has been used to format
84 the string, and the string's contents are therefore untrustworthy (see
85 L<perlsec>). This pointer may be NULL if that information is not
86 important. Note that this function requires you to specify the length of
89 STRLEN is an integer type (Size_t, usually defined as size_t in
90 config.h) guaranteed to be large enough to represent the size of
91 any string that perl can handle.
93 The C<sv_set*()> functions are not generic enough to operate on values
94 that have "magic". See L<Magic Virtual Tables> later in this document.
96 All SVs that contain strings should be terminated with a NUL character.
97 If it is not NUL-terminated there is a risk of
98 core dumps and corruptions from code which passes the string to C
99 functions or system calls which expect a NUL-terminated string.
100 Perl's own functions typically add a trailing NUL for this reason.
101 Nevertheless, you should be very careful when you pass a string stored
102 in an SV to a C function or system call.
104 To access the actual value that an SV points to, you can use the macros:
109 SvPV(SV*, STRLEN len)
112 which will automatically coerce the actual scalar type into an IV, UV, double,
115 In the C<SvPV> macro, the length of the string returned is placed into the
116 variable C<len> (this is a macro, so you do I<not> use C<&len>). If you do
117 not care what the length of the data is, use the C<SvPV_nolen> macro.
118 Historically the C<SvPV> macro with the global variable C<PL_na> has been
119 used in this case. But that can be quite inefficient because C<PL_na> must
120 be accessed in thread-local storage in threaded Perl. In any case, remember
121 that Perl allows arbitrary strings of data that may both contain NULs and
122 might not be terminated by a NUL.
124 Also remember that C doesn't allow you to safely say C<foo(SvPV(s, len),
125 len);>. It might work with your compiler, but it won't work for everyone.
126 Break this sort of statement up into separate assignments:
134 If you want to know if the scalar value is TRUE, you can use:
138 Although Perl will automatically grow strings for you, if you need to force
139 Perl to allocate more memory for your SV, you can use the macro
141 SvGROW(SV*, STRLEN newlen)
143 which will determine if more memory needs to be allocated. If so, it will
144 call the function C<sv_grow>. Note that C<SvGROW> can only increase, not
145 decrease, the allocated memory of an SV and that it does not automatically
146 add a byte for the a trailing NUL (perl's own string functions typically do
147 C<SvGROW(sv, len + 1)>).
149 If you have an SV and want to know what kind of data Perl thinks is stored
150 in it, you can use the following macros to check the type of SV you have.
156 You can get and set the current length of the string stored in an SV with
157 the following macros:
160 SvCUR_set(SV*, I32 val)
162 You can also get a pointer to the end of the string stored in the SV
167 But note that these last three macros are valid only if C<SvPOK()> is true.
169 If you want to append something to the end of string stored in an C<SV*>,
170 you can use the following functions:
172 void sv_catpv(SV*, const char*);
173 void sv_catpvn(SV*, const char*, STRLEN);
174 void sv_catpvf(SV*, const char*, ...);
175 void sv_vcatpvfn(SV*, const char*, STRLEN, va_list *, SV **, I32, bool);
176 void sv_catsv(SV*, SV*);
178 The first function calculates the length of the string to be appended by
179 using C<strlen>. In the second, you specify the length of the string
180 yourself. The third function processes its arguments like C<sprintf> and
181 appends the formatted output. The fourth function works like C<vsprintf>.
182 You can specify the address and length of an array of SVs instead of the
183 va_list argument. The fifth function extends the string stored in the first
184 SV with the string stored in the second SV. It also forces the second SV
185 to be interpreted as a string.
187 The C<sv_cat*()> functions are not generic enough to operate on values that
188 have "magic". See L<Magic Virtual Tables> later in this document.
190 If you know the name of a scalar variable, you can get a pointer to its SV
191 by using the following:
193 SV* get_sv("package::varname", FALSE);
195 This returns NULL if the variable does not exist.
197 If you want to know if this variable (or any other SV) is actually C<defined>,
202 The scalar C<undef> value is stored in an SV instance called C<PL_sv_undef>. Its
203 address can be used whenever an C<SV*> is needed.
205 There are also the two values C<PL_sv_yes> and C<PL_sv_no>, which contain Boolean
206 TRUE and FALSE values, respectively. Like C<PL_sv_undef>, their addresses can
207 be used whenever an C<SV*> is needed.
209 Do not be fooled into thinking that C<(SV *) 0> is the same as C<&PL_sv_undef>.
213 if (I-am-to-return-a-real-value) {
214 sv = sv_2mortal(newSViv(42));
218 This code tries to return a new SV (which contains the value 42) if it should
219 return a real value, or undef otherwise. Instead it has returned a NULL
220 pointer which, somewhere down the line, will cause a segmentation violation,
221 bus error, or just weird results. Change the zero to C<&PL_sv_undef> in the first
222 line and all will be well.
224 To free an SV that you've created, call C<SvREFCNT_dec(SV*)>. Normally this
225 call is not necessary (see L<Reference Counts and Mortality>).
229 Perl provides the function C<sv_chop> to efficiently remove characters
230 from the beginning of a string; you give it an SV and a pointer to
231 somewhere inside the PV, and it discards everything before the
232 pointer. The efficiency comes by means of a little hack: instead of
233 actually removing the characters, C<sv_chop> sets the flag C<OOK>
234 (offset OK) to signal to other functions that the offset hack is in
235 effect, and it puts the number of bytes chopped off into the IV field
236 of the SV. It then moves the PV pointer (called C<SvPVX>) forward that
237 many bytes, and adjusts C<SvCUR> and C<SvLEN>.
239 Hence, at this point, the start of the buffer that we allocated lives
240 at C<SvPVX(sv) - SvIV(sv)> in memory and the PV pointer is pointing
241 into the middle of this allocated storage.
243 This is best demonstrated by example:
245 % ./perl -Ilib -MDevel::Peek -le '$a="12345"; $a=~s/.//; Dump($a)'
246 SV = PVIV(0x8128450) at 0x81340f0
248 FLAGS = (POK,OOK,pPOK)
250 PV = 0x8135781 ( "1" . ) "2345"\0
254 Here the number of bytes chopped off (1) is put into IV, and
255 C<Devel::Peek::Dump> helpfully reminds us that this is an offset. The
256 portion of the string between the "real" and the "fake" beginnings is
257 shown in parentheses, and the values of C<SvCUR> and C<SvLEN> reflect
258 the fake beginning, not the real one.
260 Something similar to the offset hack is performed on AVs to enable
261 efficient shifting and splicing off the beginning of the array; while
262 C<AvARRAY> points to the first element in the array that is visible from
263 Perl, C<AvALLOC> points to the real start of the C array. These are
264 usually the same, but a C<shift> operation can be carried out by
265 increasing C<AvARRAY> by one and decreasing C<AvFILL> and C<AvLEN>.
266 Again, the location of the real start of the C array only comes into
267 play when freeing the array. See C<av_shift> in F<av.c>.
269 =head2 What's Really Stored in an SV?
271 Recall that the usual method of determining the type of scalar you have is
272 to use C<Sv*OK> macros. Because a scalar can be both a number and a string,
273 usually these macros will always return TRUE and calling the C<Sv*V>
274 macros will do the appropriate conversion of string to integer/double or
275 integer/double to string.
277 If you I<really> need to know if you have an integer, double, or string
278 pointer in an SV, you can use the following three macros instead:
284 These will tell you if you truly have an integer, double, or string pointer
285 stored in your SV. The "p" stands for private.
287 The are various ways in which the private and public flags may differ.
288 For example, a tied SV may have a valid underlying value in the IV slot
289 (so SvIOKp is true), but the data should be accessed via the FETCH
290 routine rather than directly, so SvIOK is false. Another is when
291 numeric conversion has occured and precision has been lost: only the
292 private flag is set on 'lossy' values. So when an NV is converted to an
293 IV with loss, SvIOKp, SvNOKp and SvNOK will be set, while SvIOK wont be.
295 In general, though, it's best to use the C<Sv*V> macros.
297 =head2 Working with AVs
299 There are two ways to create and load an AV. The first method creates an
304 The second method both creates the AV and initially populates it with SVs:
306 AV* av_make(I32 num, SV **ptr);
308 The second argument points to an array containing C<num> C<SV*>'s. Once the
309 AV has been created, the SVs can be destroyed, if so desired.
311 Once the AV has been created, the following operations are possible on AVs:
313 void av_push(AV*, SV*);
316 void av_unshift(AV*, I32 num);
318 These should be familiar operations, with the exception of C<av_unshift>.
319 This routine adds C<num> elements at the front of the array with the C<undef>
320 value. You must then use C<av_store> (described below) to assign values
321 to these new elements.
323 Here are some other functions:
326 SV** av_fetch(AV*, I32 key, I32 lval);
327 SV** av_store(AV*, I32 key, SV* val);
329 The C<av_len> function returns the highest index value in array (just
330 like $#array in Perl). If the array is empty, -1 is returned. The
331 C<av_fetch> function returns the value at index C<key>, but if C<lval>
332 is non-zero, then C<av_fetch> will store an undef value at that index.
333 The C<av_store> function stores the value C<val> at index C<key>, and does
334 not increment the reference count of C<val>. Thus the caller is responsible
335 for taking care of that, and if C<av_store> returns NULL, the caller will
336 have to decrement the reference count to avoid a memory leak. Note that
337 C<av_fetch> and C<av_store> both return C<SV**>'s, not C<SV*>'s as their
342 void av_extend(AV*, I32 key);
344 The C<av_clear> function deletes all the elements in the AV* array, but
345 does not actually delete the array itself. The C<av_undef> function will
346 delete all the elements in the array plus the array itself. The
347 C<av_extend> function extends the array so that it contains at least C<key+1>
348 elements. If C<key+1> is less than the currently allocated length of the array,
349 then nothing is done.
351 If you know the name of an array variable, you can get a pointer to its AV
352 by using the following:
354 AV* get_av("package::varname", FALSE);
356 This returns NULL if the variable does not exist.
358 See L<Understanding the Magic of Tied Hashes and Arrays> for more
359 information on how to use the array access functions on tied arrays.
361 =head2 Working with HVs
363 To create an HV, you use the following routine:
367 Once the HV has been created, the following operations are possible on HVs:
369 SV** hv_store(HV*, const char* key, U32 klen, SV* val, U32 hash);
370 SV** hv_fetch(HV*, const char* key, U32 klen, I32 lval);
372 The C<klen> parameter is the length of the key being passed in (Note that
373 you cannot pass 0 in as a value of C<klen> to tell Perl to measure the
374 length of the key). The C<val> argument contains the SV pointer to the
375 scalar being stored, and C<hash> is the precomputed hash value (zero if
376 you want C<hv_store> to calculate it for you). The C<lval> parameter
377 indicates whether this fetch is actually a part of a store operation, in
378 which case a new undefined value will be added to the HV with the supplied
379 key and C<hv_fetch> will return as if the value had already existed.
381 Remember that C<hv_store> and C<hv_fetch> return C<SV**>'s and not just
382 C<SV*>. To access the scalar value, you must first dereference the return
383 value. However, you should check to make sure that the return value is
384 not NULL before dereferencing it.
386 These two functions check if a hash table entry exists, and deletes it.
388 bool hv_exists(HV*, const char* key, U32 klen);
389 SV* hv_delete(HV*, const char* key, U32 klen, I32 flags);
391 If C<flags> does not include the C<G_DISCARD> flag then C<hv_delete> will
392 create and return a mortal copy of the deleted value.
394 And more miscellaneous functions:
399 Like their AV counterparts, C<hv_clear> deletes all the entries in the hash
400 table but does not actually delete the hash table. The C<hv_undef> deletes
401 both the entries and the hash table itself.
403 Perl keeps the actual data in linked list of structures with a typedef of HE.
404 These contain the actual key and value pointers (plus extra administrative
405 overhead). The key is a string pointer; the value is an C<SV*>. However,
406 once you have an C<HE*>, to get the actual key and value, use the routines
409 I32 hv_iterinit(HV*);
410 /* Prepares starting point to traverse hash table */
411 HE* hv_iternext(HV*);
412 /* Get the next entry, and return a pointer to a
413 structure that has both the key and value */
414 char* hv_iterkey(HE* entry, I32* retlen);
415 /* Get the key from an HE structure and also return
416 the length of the key string */
417 SV* hv_iterval(HV*, HE* entry);
418 /* Return an SV pointer to the value of the HE
420 SV* hv_iternextsv(HV*, char** key, I32* retlen);
421 /* This convenience routine combines hv_iternext,
422 hv_iterkey, and hv_iterval. The key and retlen
423 arguments are return values for the key and its
424 length. The value is returned in the SV* argument */
426 If you know the name of a hash variable, you can get a pointer to its HV
427 by using the following:
429 HV* get_hv("package::varname", FALSE);
431 This returns NULL if the variable does not exist.
433 The hash algorithm is defined in the C<PERL_HASH(hash, key, klen)> macro:
437 hash = (hash * 33) + *key++;
438 hash = hash + (hash >> 5); /* after 5.6 */
440 The last step was added in version 5.6 to improve distribution of
441 lower bits in the resulting hash value.
443 See L<Understanding the Magic of Tied Hashes and Arrays> for more
444 information on how to use the hash access functions on tied hashes.
446 =head2 Hash API Extensions
448 Beginning with version 5.004, the following functions are also supported:
450 HE* hv_fetch_ent (HV* tb, SV* key, I32 lval, U32 hash);
451 HE* hv_store_ent (HV* tb, SV* key, SV* val, U32 hash);
453 bool hv_exists_ent (HV* tb, SV* key, U32 hash);
454 SV* hv_delete_ent (HV* tb, SV* key, I32 flags, U32 hash);
456 SV* hv_iterkeysv (HE* entry);
458 Note that these functions take C<SV*> keys, which simplifies writing
459 of extension code that deals with hash structures. These functions
460 also allow passing of C<SV*> keys to C<tie> functions without forcing
461 you to stringify the keys (unlike the previous set of functions).
463 They also return and accept whole hash entries (C<HE*>), making their
464 use more efficient (since the hash number for a particular string
465 doesn't have to be recomputed every time). See L<perlapi> for detailed
468 The following macros must always be used to access the contents of hash
469 entries. Note that the arguments to these macros must be simple
470 variables, since they may get evaluated more than once. See
471 L<perlapi> for detailed descriptions of these macros.
473 HePV(HE* he, STRLEN len)
477 HeSVKEY_force(HE* he)
478 HeSVKEY_set(HE* he, SV* sv)
480 These two lower level macros are defined, but must only be used when
481 dealing with keys that are not C<SV*>s:
486 Note that both C<hv_store> and C<hv_store_ent> do not increment the
487 reference count of the stored C<val>, which is the caller's responsibility.
488 If these functions return a NULL value, the caller will usually have to
489 decrement the reference count of C<val> to avoid a memory leak.
493 References are a special type of scalar that point to other data types
494 (including references).
496 To create a reference, use either of the following functions:
498 SV* newRV_inc((SV*) thing);
499 SV* newRV_noinc((SV*) thing);
501 The C<thing> argument can be any of an C<SV*>, C<AV*>, or C<HV*>. The
502 functions are identical except that C<newRV_inc> increments the reference
503 count of the C<thing>, while C<newRV_noinc> does not. For historical
504 reasons, C<newRV> is a synonym for C<newRV_inc>.
506 Once you have a reference, you can use the following macro to dereference
511 then call the appropriate routines, casting the returned C<SV*> to either an
512 C<AV*> or C<HV*>, if required.
514 To determine if an SV is a reference, you can use the following macro:
518 To discover what type of value the reference refers to, use the following
519 macro and then check the return value.
523 The most useful types that will be returned are:
532 SVt_PVGV Glob (possible a file handle)
533 SVt_PVMG Blessed or Magical Scalar
535 See the sv.h header file for more details.
537 =head2 Blessed References and Class Objects
539 References are also used to support object-oriented programming. In the
540 OO lexicon, an object is simply a reference that has been blessed into a
541 package (or class). Once blessed, the programmer may now use the reference
542 to access the various methods in the class.
544 A reference can be blessed into a package with the following function:
546 SV* sv_bless(SV* sv, HV* stash);
548 The C<sv> argument must be a reference. The C<stash> argument specifies
549 which class the reference will belong to. See
550 L<Stashes and Globs> for information on converting class names into stashes.
552 /* Still under construction */
554 Upgrades rv to reference if not already one. Creates new SV for rv to
555 point to. If C<classname> is non-null, the SV is blessed into the specified
556 class. SV is returned.
558 SV* newSVrv(SV* rv, const char* classname);
560 Copies integer, unsigned integer or double into an SV whose reference is C<rv>. SV is blessed
561 if C<classname> is non-null.
563 SV* sv_setref_iv(SV* rv, const char* classname, IV iv);
564 SV* sv_setref_uv(SV* rv, const char* classname, UV uv);
565 SV* sv_setref_nv(SV* rv, const char* classname, NV iv);
567 Copies the pointer value (I<the address, not the string!>) into an SV whose
568 reference is rv. SV is blessed if C<classname> is non-null.
570 SV* sv_setref_pv(SV* rv, const char* classname, PV iv);
572 Copies string into an SV whose reference is C<rv>. Set length to 0 to let
573 Perl calculate the string length. SV is blessed if C<classname> is non-null.
575 SV* sv_setref_pvn(SV* rv, const char* classname, PV iv, STRLEN length);
577 Tests whether the SV is blessed into the specified class. It does not
578 check inheritance relationships.
580 int sv_isa(SV* sv, const char* name);
582 Tests whether the SV is a reference to a blessed object.
584 int sv_isobject(SV* sv);
586 Tests whether the SV is derived from the specified class. SV can be either
587 a reference to a blessed object or a string containing a class name. This
588 is the function implementing the C<UNIVERSAL::isa> functionality.
590 bool sv_derived_from(SV* sv, const char* name);
592 To check if you've got an object derived from a specific class you have
595 if (sv_isobject(sv) && sv_derived_from(sv, class)) { ... }
597 =head2 Creating New Variables
599 To create a new Perl variable with an undef value which can be accessed from
600 your Perl script, use the following routines, depending on the variable type.
602 SV* get_sv("package::varname", TRUE);
603 AV* get_av("package::varname", TRUE);
604 HV* get_hv("package::varname", TRUE);
606 Notice the use of TRUE as the second parameter. The new variable can now
607 be set, using the routines appropriate to the data type.
609 There are additional macros whose values may be bitwise OR'ed with the
610 C<TRUE> argument to enable certain extra features. Those bits are:
616 Marks the variable as multiply defined, thus preventing the:
618 Name <varname> used only once: possible typo
626 Had to create <varname> unexpectedly
628 if the variable did not exist before the function was called.
632 If you do not specify a package name, the variable is created in the current
635 =head2 Reference Counts and Mortality
637 Perl uses a reference count-driven garbage collection mechanism. SVs,
638 AVs, or HVs (xV for short in the following) start their life with a
639 reference count of 1. If the reference count of an xV ever drops to 0,
640 then it will be destroyed and its memory made available for reuse.
642 This normally doesn't happen at the Perl level unless a variable is
643 undef'ed or the last variable holding a reference to it is changed or
644 overwritten. At the internal level, however, reference counts can be
645 manipulated with the following macros:
647 int SvREFCNT(SV* sv);
648 SV* SvREFCNT_inc(SV* sv);
649 void SvREFCNT_dec(SV* sv);
651 However, there is one other function which manipulates the reference
652 count of its argument. The C<newRV_inc> function, you will recall,
653 creates a reference to the specified argument. As a side effect,
654 it increments the argument's reference count. If this is not what
655 you want, use C<newRV_noinc> instead.
657 For example, imagine you want to return a reference from an XSUB function.
658 Inside the XSUB routine, you create an SV which initially has a reference
659 count of one. Then you call C<newRV_inc>, passing it the just-created SV.
660 This returns the reference as a new SV, but the reference count of the
661 SV you passed to C<newRV_inc> has been incremented to two. Now you
662 return the reference from the XSUB routine and forget about the SV.
663 But Perl hasn't! Whenever the returned reference is destroyed, the
664 reference count of the original SV is decreased to one and nothing happens.
665 The SV will hang around without any way to access it until Perl itself
666 terminates. This is a memory leak.
668 The correct procedure, then, is to use C<newRV_noinc> instead of
669 C<newRV_inc>. Then, if and when the last reference is destroyed,
670 the reference count of the SV will go to zero and it will be destroyed,
671 stopping any memory leak.
673 There are some convenience functions available that can help with the
674 destruction of xVs. These functions introduce the concept of "mortality".
675 An xV that is mortal has had its reference count marked to be decremented,
676 but not actually decremented, until "a short time later". Generally the
677 term "short time later" means a single Perl statement, such as a call to
678 an XSUB function. The actual determinant for when mortal xVs have their
679 reference count decremented depends on two macros, SAVETMPS and FREETMPS.
680 See L<perlcall> and L<perlxs> for more details on these macros.
682 "Mortalization" then is at its simplest a deferred C<SvREFCNT_dec>.
683 However, if you mortalize a variable twice, the reference count will
684 later be decremented twice.
686 "Mortal" SVs are mainly used for SVs that are placed on perl's stack.
687 For example an SV which is created just to pass a number to a called sub
688 is made mortal to have it cleaned up automatically when stack is popped.
689 Similarly results returned by XSUBs (which go in the stack) are often
692 To create a mortal variable, use the functions:
696 SV* sv_mortalcopy(SV*)
698 The first call creates a mortal SV (with no value), the second converts an existing
699 SV to a mortal SV (and thus defers a call to C<SvREFCNT_dec>), and the
700 third creates a mortal copy of an existing SV.
701 Because C<sv_newmortal> gives the new SV no value,it must normally be given one
702 via C<sv_setpv>, C<sv_setiv>, etc. :
704 SV *tmp = sv_newmortal();
705 sv_setiv(tmp, an_integer);
707 As that is multiple C statements it is quite common so see this idiom instead:
709 SV *tmp = sv_2mortal(newSViv(an_integer));
712 You should be careful about creating mortal variables. Strange things
713 can happen if you make the same value mortal within multiple contexts,
714 or if you make a variable mortal multiple times. Thinking of "Mortalization"
715 as deferred C<SvREFCNT_dec> should help to minimize such problems.
716 For example if you are passing an SV which you I<know> has high enough REFCNT
717 to survive its use on the stack you need not do any mortalization.
718 If you are not sure then doing an C<SvREFCNT_inc> and C<sv_2mortal>, or
719 making a C<sv_mortalcopy> is safer.
721 The mortal routines are not just for SVs -- AVs and HVs can be
722 made mortal by passing their address (type-casted to C<SV*>) to the
723 C<sv_2mortal> or C<sv_mortalcopy> routines.
725 =head2 Stashes and Globs
727 A "stash" is a hash that contains all of the different objects that
728 are contained within a package. Each key of the stash is a symbol
729 name (shared by all the different types of objects that have the same
730 name), and each value in the hash table is a GV (Glob Value). This GV
731 in turn contains references to the various objects of that name,
732 including (but not limited to) the following:
741 There is a single stash called "PL_defstash" that holds the items that exist
742 in the "main" package. To get at the items in other packages, append the
743 string "::" to the package name. The items in the "Foo" package are in
744 the stash "Foo::" in PL_defstash. The items in the "Bar::Baz" package are
745 in the stash "Baz::" in "Bar::"'s stash.
747 To get the stash pointer for a particular package, use the function:
749 HV* gv_stashpv(const char* name, I32 create)
750 HV* gv_stashsv(SV*, I32 create)
752 The first function takes a literal string, the second uses the string stored
753 in the SV. Remember that a stash is just a hash table, so you get back an
754 C<HV*>. The C<create> flag will create a new package if it is set.
756 The name that C<gv_stash*v> wants is the name of the package whose symbol table
757 you want. The default package is called C<main>. If you have multiply nested
758 packages, pass their names to C<gv_stash*v>, separated by C<::> as in the Perl
761 Alternately, if you have an SV that is a blessed reference, you can find
762 out the stash pointer by using:
764 HV* SvSTASH(SvRV(SV*));
766 then use the following to get the package name itself:
768 char* HvNAME(HV* stash);
770 If you need to bless or re-bless an object you can use the following
773 SV* sv_bless(SV*, HV* stash)
775 where the first argument, an C<SV*>, must be a reference, and the second
776 argument is a stash. The returned C<SV*> can now be used in the same way
779 For more information on references and blessings, consult L<perlref>.
781 =head2 Double-Typed SVs
783 Scalar variables normally contain only one type of value, an integer,
784 double, pointer, or reference. Perl will automatically convert the
785 actual scalar data from the stored type into the requested type.
787 Some scalar variables contain more than one type of scalar data. For
788 example, the variable C<$!> contains either the numeric value of C<errno>
789 or its string equivalent from either C<strerror> or C<sys_errlist[]>.
791 To force multiple data values into an SV, you must do two things: use the
792 C<sv_set*v> routines to add the additional scalar type, then set a flag
793 so that Perl will believe it contains more than one type of data. The
794 four macros to set the flags are:
801 The particular macro you must use depends on which C<sv_set*v> routine
802 you called first. This is because every C<sv_set*v> routine turns on
803 only the bit for the particular type of data being set, and turns off
806 For example, to create a new Perl variable called "dberror" that contains
807 both the numeric and descriptive string error values, you could use the
811 extern char *dberror_list;
813 SV* sv = get_sv("dberror", TRUE);
814 sv_setiv(sv, (IV) dberror);
815 sv_setpv(sv, dberror_list[dberror]);
818 If the order of C<sv_setiv> and C<sv_setpv> had been reversed, then the
819 macro C<SvPOK_on> would need to be called instead of C<SvIOK_on>.
821 =head2 Magic Variables
823 [This section still under construction. Ignore everything here. Post no
824 bills. Everything not permitted is forbidden.]
826 Any SV may be magical, that is, it has special features that a normal
827 SV does not have. These features are stored in the SV structure in a
828 linked list of C<struct magic>'s, typedef'ed to C<MAGIC>.
841 Note this is current as of patchlevel 0, and could change at any time.
843 =head2 Assigning Magic
845 Perl adds magic to an SV using the sv_magic function:
847 void sv_magic(SV* sv, SV* obj, int how, const char* name, I32 namlen);
849 The C<sv> argument is a pointer to the SV that is to acquire a new magical
852 If C<sv> is not already magical, Perl uses the C<SvUPGRADE> macro to
853 convert C<sv> to type C<SVt_PVMG>. Perl then continues by adding new magic
854 to the beginning of the linked list of magical features. Any prior entry
855 of the same type of magic is deleted. Note that this can be overridden,
856 and multiple instances of the same type of magic can be associated with an
859 The C<name> and C<namlen> arguments are used to associate a string with
860 the magic, typically the name of a variable. C<namlen> is stored in the
861 C<mg_len> field and if C<name> is non-null and C<namlen> E<gt>= 0 a malloc'd
862 copy of the name is stored in C<mg_ptr> field.
864 The sv_magic function uses C<how> to determine which, if any, predefined
865 "Magic Virtual Table" should be assigned to the C<mg_virtual> field.
866 See the "Magic Virtual Table" section below. The C<how> argument is also
867 stored in the C<mg_type> field. The value of C<how> should be chosen
868 from the set of macros C<PERL_MAGIC_foo> found perl.h. Note that before
869 these macros were added, Perl internals used to directly use character
870 literals, so you may occasionally come across old code or documentation
871 referring to 'U' magic rather than C<PERL_MAGIC_uvar> for example.
873 The C<obj> argument is stored in the C<mg_obj> field of the C<MAGIC>
874 structure. If it is not the same as the C<sv> argument, the reference
875 count of the C<obj> object is incremented. If it is the same, or if
876 the C<how> argument is C<PERL_MAGIC_arylen>, or if it is a NULL pointer,
877 then C<obj> is merely stored, without the reference count being incremented.
879 There is also a function to add magic to an C<HV>:
881 void hv_magic(HV *hv, GV *gv, int how);
883 This simply calls C<sv_magic> and coerces the C<gv> argument into an C<SV>.
885 To remove the magic from an SV, call the function sv_unmagic:
887 void sv_unmagic(SV *sv, int type);
889 The C<type> argument should be equal to the C<how> value when the C<SV>
890 was initially made magical.
892 =head2 Magic Virtual Tables
894 The C<mg_virtual> field in the C<MAGIC> structure is a pointer to an
895 C<MGVTBL>, which is a structure of function pointers and stands for
896 "Magic Virtual Table" to handle the various operations that might be
897 applied to that variable.
899 The C<MGVTBL> has five pointers to the following routine types:
901 int (*svt_get)(SV* sv, MAGIC* mg);
902 int (*svt_set)(SV* sv, MAGIC* mg);
903 U32 (*svt_len)(SV* sv, MAGIC* mg);
904 int (*svt_clear)(SV* sv, MAGIC* mg);
905 int (*svt_free)(SV* sv, MAGIC* mg);
907 This MGVTBL structure is set at compile-time in C<perl.h> and there are
908 currently 19 types (or 21 with overloading turned on). These different
909 structures contain pointers to various routines that perform additional
910 actions depending on which function is being called.
912 Function pointer Action taken
913 ---------------- ------------
914 svt_get Do something before the value of the SV is retrieved.
915 svt_set Do something after the SV is assigned a value.
916 svt_len Report on the SV's length.
917 svt_clear Clear something the SV represents.
918 svt_free Free any extra storage associated with the SV.
920 For instance, the MGVTBL structure called C<vtbl_sv> (which corresponds
921 to an C<mg_type> of C<PERL_MAGIC_sv>) contains:
923 { magic_get, magic_set, magic_len, 0, 0 }
925 Thus, when an SV is determined to be magical and of type C<PERL_MAGIC_sv>,
926 if a get operation is being performed, the routine C<magic_get> is
927 called. All the various routines for the various magical types begin
928 with C<magic_>. NOTE: the magic routines are not considered part of
929 the Perl API, and may not be exported by the Perl library.
931 The current kinds of Magic Virtual Tables are:
934 (old-style char and macro) MGVTBL Type of magic
935 -------------------------- ------ ----------------------------
936 \0 PERL_MAGIC_sv vtbl_sv Special scalar variable
937 A PERL_MAGIC_overload vtbl_amagic %OVERLOAD hash
938 a PERL_MAGIC_overload_elem vtbl_amagicelem %OVERLOAD hash element
939 c PERL_MAGIC_overload_table (none) Holds overload table (AMT)
941 B PERL_MAGIC_bm vtbl_bm Boyer-Moore (fast string search)
942 D PERL_MAGIC_regdata vtbl_regdata Regex match position data
944 d PERL_MAGIC_regdatum vtbl_regdatum Regex match position data
946 E PERL_MAGIC_env vtbl_env %ENV hash
947 e PERL_MAGIC_envelem vtbl_envelem %ENV hash element
948 f PERL_MAGIC_fm vtbl_fm Formline ('compiled' format)
949 g PERL_MAGIC_regex_global vtbl_mglob m//g target / study()ed string
950 I PERL_MAGIC_isa vtbl_isa @ISA array
951 i PERL_MAGIC_isaelem vtbl_isaelem @ISA array element
952 k PERL_MAGIC_nkeys vtbl_nkeys scalar(keys()) lvalue
953 L PERL_MAGIC_dbfile (none) Debugger %_<filename
954 l PERL_MAGIC_dbline vtbl_dbline Debugger %_<filename element
955 m PERL_MAGIC_mutex vtbl_mutex ???
956 o PERL_MAGIC_collxfrm vtbl_collxfrm Locale collate transformation
957 P PERL_MAGIC_tied vtbl_pack Tied array or hash
958 p PERL_MAGIC_tiedelem vtbl_packelem Tied array or hash element
959 q PERL_MAGIC_tiedscalar vtbl_packelem Tied scalar or handle
960 r PERL_MAGIC_qr vtbl_qr precompiled qr// regex
961 S PERL_MAGIC_sig vtbl_sig %SIG hash
962 s PERL_MAGIC_sigelem vtbl_sigelem %SIG hash element
963 t PERL_MAGIC_taint vtbl_taint Taintedness
964 U PERL_MAGIC_uvar vtbl_uvar Available for use by extensions
965 v PERL_MAGIC_vec vtbl_vec vec() lvalue
966 V PERL_MAGIC_vstring (none) v-string scalars
967 x PERL_MAGIC_substr vtbl_substr substr() lvalue
968 y PERL_MAGIC_defelem vtbl_defelem Shadow "foreach" iterator
969 variable / smart parameter
971 * PERL_MAGIC_glob vtbl_glob GV (typeglob)
972 # PERL_MAGIC_arylen vtbl_arylen Array length ($#ary)
973 . PERL_MAGIC_pos vtbl_pos pos() lvalue
974 < PERL_MAGIC_backref vtbl_backref ???
975 ~ PERL_MAGIC_ext (none) Available for use by extensions
977 When an uppercase and lowercase letter both exist in the table, then the
978 uppercase letter is typically used to represent some kind of composite type
979 (a list or a hash), and the lowercase letter is used to represent an element
980 of that composite type. Some internals code makes use of this case
981 relationship. However, 'v' and 'V' (vec and v-string) are in no way related.
983 The C<PERL_MAGIC_ext> and C<PERL_MAGIC_uvar> magic types are defined
984 specifically for use by extensions and will not be used by perl itself.
985 Extensions can use C<PERL_MAGIC_ext> magic to 'attach' private information
986 to variables (typically objects). This is especially useful because
987 there is no way for normal perl code to corrupt this private information
988 (unlike using extra elements of a hash object).
990 Similarly, C<PERL_MAGIC_uvar> magic can be used much like tie() to call a
991 C function any time a scalar's value is used or changed. The C<MAGIC>'s
992 C<mg_ptr> field points to a C<ufuncs> structure:
995 I32 (*uf_val)(pTHX_ IV, SV*);
996 I32 (*uf_set)(pTHX_ IV, SV*);
1000 When the SV is read from or written to, the C<uf_val> or C<uf_set>
1001 function will be called with C<uf_index> as the first arg and a pointer to
1002 the SV as the second. A simple example of how to add C<PERL_MAGIC_uvar>
1003 magic is shown below. Note that the ufuncs structure is copied by
1004 sv_magic, so you can safely allocate it on the stack.
1012 uf.uf_val = &my_get_fn;
1013 uf.uf_set = &my_set_fn;
1015 sv_magic(sv, 0, PERL_MAGIC_uvar, (char*)&uf, sizeof(uf));
1017 Note that because multiple extensions may be using C<PERL_MAGIC_ext>
1018 or C<PERL_MAGIC_uvar> magic, it is important for extensions to take
1019 extra care to avoid conflict. Typically only using the magic on
1020 objects blessed into the same class as the extension is sufficient.
1021 For C<PERL_MAGIC_ext> magic, it may also be appropriate to add an I32
1022 'signature' at the top of the private data area and check that.
1024 Also note that the C<sv_set*()> and C<sv_cat*()> functions described
1025 earlier do B<not> invoke 'set' magic on their targets. This must
1026 be done by the user either by calling the C<SvSETMAGIC()> macro after
1027 calling these functions, or by using one of the C<sv_set*_mg()> or
1028 C<sv_cat*_mg()> functions. Similarly, generic C code must call the
1029 C<SvGETMAGIC()> macro to invoke any 'get' magic if they use an SV
1030 obtained from external sources in functions that don't handle magic.
1031 See L<perlapi> for a description of these functions.
1032 For example, calls to the C<sv_cat*()> functions typically need to be
1033 followed by C<SvSETMAGIC()>, but they don't need a prior C<SvGETMAGIC()>
1034 since their implementation handles 'get' magic.
1036 =head2 Finding Magic
1038 MAGIC* mg_find(SV*, int type); /* Finds the magic pointer of that type */
1040 This routine returns a pointer to the C<MAGIC> structure stored in the SV.
1041 If the SV does not have that magical feature, C<NULL> is returned. Also,
1042 if the SV is not of type SVt_PVMG, Perl may core dump.
1044 int mg_copy(SV* sv, SV* nsv, const char* key, STRLEN klen);
1046 This routine checks to see what types of magic C<sv> has. If the mg_type
1047 field is an uppercase letter, then the mg_obj is copied to C<nsv>, but
1048 the mg_type field is changed to be the lowercase letter.
1050 =head2 Understanding the Magic of Tied Hashes and Arrays
1052 Tied hashes and arrays are magical beasts of the C<PERL_MAGIC_tied>
1055 WARNING: As of the 5.004 release, proper usage of the array and hash
1056 access functions requires understanding a few caveats. Some
1057 of these caveats are actually considered bugs in the API, to be fixed
1058 in later releases, and are bracketed with [MAYCHANGE] below. If
1059 you find yourself actually applying such information in this section, be
1060 aware that the behavior may change in the future, umm, without warning.
1062 The perl tie function associates a variable with an object that implements
1063 the various GET, SET, etc methods. To perform the equivalent of the perl
1064 tie function from an XSUB, you must mimic this behaviour. The code below
1065 carries out the necessary steps - firstly it creates a new hash, and then
1066 creates a second hash which it blesses into the class which will implement
1067 the tie methods. Lastly it ties the two hashes together, and returns a
1068 reference to the new tied hash. Note that the code below does NOT call the
1069 TIEHASH method in the MyTie class -
1070 see L<Calling Perl Routines from within C Programs> for details on how
1081 tie = newRV_noinc((SV*)newHV());
1082 stash = gv_stashpv("MyTie", TRUE);
1083 sv_bless(tie, stash);
1084 hv_magic(hash, (GV*)tie, PERL_MAGIC_tied);
1085 RETVAL = newRV_noinc(hash);
1089 The C<av_store> function, when given a tied array argument, merely
1090 copies the magic of the array onto the value to be "stored", using
1091 C<mg_copy>. It may also return NULL, indicating that the value did not
1092 actually need to be stored in the array. [MAYCHANGE] After a call to
1093 C<av_store> on a tied array, the caller will usually need to call
1094 C<mg_set(val)> to actually invoke the perl level "STORE" method on the
1095 TIEARRAY object. If C<av_store> did return NULL, a call to
1096 C<SvREFCNT_dec(val)> will also be usually necessary to avoid a memory
1099 The previous paragraph is applicable verbatim to tied hash access using the
1100 C<hv_store> and C<hv_store_ent> functions as well.
1102 C<av_fetch> and the corresponding hash functions C<hv_fetch> and
1103 C<hv_fetch_ent> actually return an undefined mortal value whose magic
1104 has been initialized using C<mg_copy>. Note the value so returned does not
1105 need to be deallocated, as it is already mortal. [MAYCHANGE] But you will
1106 need to call C<mg_get()> on the returned value in order to actually invoke
1107 the perl level "FETCH" method on the underlying TIE object. Similarly,
1108 you may also call C<mg_set()> on the return value after possibly assigning
1109 a suitable value to it using C<sv_setsv>, which will invoke the "STORE"
1110 method on the TIE object. [/MAYCHANGE]
1113 In other words, the array or hash fetch/store functions don't really
1114 fetch and store actual values in the case of tied arrays and hashes. They
1115 merely call C<mg_copy> to attach magic to the values that were meant to be
1116 "stored" or "fetched". Later calls to C<mg_get> and C<mg_set> actually
1117 do the job of invoking the TIE methods on the underlying objects. Thus
1118 the magic mechanism currently implements a kind of lazy access to arrays
1121 Currently (as of perl version 5.004), use of the hash and array access
1122 functions requires the user to be aware of whether they are operating on
1123 "normal" hashes and arrays, or on their tied variants. The API may be
1124 changed to provide more transparent access to both tied and normal data
1125 types in future versions.
1128 You would do well to understand that the TIEARRAY and TIEHASH interfaces
1129 are mere sugar to invoke some perl method calls while using the uniform hash
1130 and array syntax. The use of this sugar imposes some overhead (typically
1131 about two to four extra opcodes per FETCH/STORE operation, in addition to
1132 the creation of all the mortal variables required to invoke the methods).
1133 This overhead will be comparatively small if the TIE methods are themselves
1134 substantial, but if they are only a few statements long, the overhead
1135 will not be insignificant.
1137 =head2 Localizing changes
1139 Perl has a very handy construction
1146 This construction is I<approximately> equivalent to
1155 The biggest difference is that the first construction would
1156 reinstate the initial value of $var, irrespective of how control exits
1157 the block: C<goto>, C<return>, C<die>/C<eval>, etc. It is a little bit
1158 more efficient as well.
1160 There is a way to achieve a similar task from C via Perl API: create a
1161 I<pseudo-block>, and arrange for some changes to be automatically
1162 undone at the end of it, either explicit, or via a non-local exit (via
1163 die()). A I<block>-like construct is created by a pair of
1164 C<ENTER>/C<LEAVE> macros (see L<perlcall/"Returning a Scalar">).
1165 Such a construct may be created specially for some important localized
1166 task, or an existing one (like boundaries of enclosing Perl
1167 subroutine/block, or an existing pair for freeing TMPs) may be
1168 used. (In the second case the overhead of additional localization must
1169 be almost negligible.) Note that any XSUB is automatically enclosed in
1170 an C<ENTER>/C<LEAVE> pair.
1172 Inside such a I<pseudo-block> the following service is available:
1176 =item C<SAVEINT(int i)>
1178 =item C<SAVEIV(IV i)>
1180 =item C<SAVEI32(I32 i)>
1182 =item C<SAVELONG(long i)>
1184 These macros arrange things to restore the value of integer variable
1185 C<i> at the end of enclosing I<pseudo-block>.
1187 =item C<SAVESPTR(s)>
1189 =item C<SAVEPPTR(p)>
1191 These macros arrange things to restore the value of pointers C<s> and
1192 C<p>. C<s> must be a pointer of a type which survives conversion to
1193 C<SV*> and back, C<p> should be able to survive conversion to C<char*>
1196 =item C<SAVEFREESV(SV *sv)>
1198 The refcount of C<sv> would be decremented at the end of
1199 I<pseudo-block>. This is similar to C<sv_2mortal> in that it is also a
1200 mechanism for doing a delayed C<SvREFCNT_dec>. However, while C<sv_2mortal>
1201 extends the lifetime of C<sv> until the beginning of the next statement,
1202 C<SAVEFREESV> extends it until the end of the enclosing scope. These
1203 lifetimes can be wildly different.
1205 Also compare C<SAVEMORTALIZESV>.
1207 =item C<SAVEMORTALIZESV(SV *sv)>
1209 Just like C<SAVEFREESV>, but mortalizes C<sv> at the end of the current
1210 scope instead of decrementing its reference count. This usually has the
1211 effect of keeping C<sv> alive until the statement that called the currently
1212 live scope has finished executing.
1214 =item C<SAVEFREEOP(OP *op)>
1216 The C<OP *> is op_free()ed at the end of I<pseudo-block>.
1218 =item C<SAVEFREEPV(p)>
1220 The chunk of memory which is pointed to by C<p> is Safefree()ed at the
1221 end of I<pseudo-block>.
1223 =item C<SAVECLEARSV(SV *sv)>
1225 Clears a slot in the current scratchpad which corresponds to C<sv> at
1226 the end of I<pseudo-block>.
1228 =item C<SAVEDELETE(HV *hv, char *key, I32 length)>
1230 The key C<key> of C<hv> is deleted at the end of I<pseudo-block>. The
1231 string pointed to by C<key> is Safefree()ed. If one has a I<key> in
1232 short-lived storage, the corresponding string may be reallocated like
1235 SAVEDELETE(PL_defstash, savepv(tmpbuf), strlen(tmpbuf));
1237 =item C<SAVEDESTRUCTOR(DESTRUCTORFUNC_NOCONTEXT_t f, void *p)>
1239 At the end of I<pseudo-block> the function C<f> is called with the
1242 =item C<SAVEDESTRUCTOR_X(DESTRUCTORFUNC_t f, void *p)>
1244 At the end of I<pseudo-block> the function C<f> is called with the
1245 implicit context argument (if any), and C<p>.
1247 =item C<SAVESTACK_POS()>
1249 The current offset on the Perl internal stack (cf. C<SP>) is restored
1250 at the end of I<pseudo-block>.
1254 The following API list contains functions, thus one needs to
1255 provide pointers to the modifiable data explicitly (either C pointers,
1256 or Perlish C<GV *>s). Where the above macros take C<int>, a similar
1257 function takes C<int *>.
1261 =item C<SV* save_scalar(GV *gv)>
1263 Equivalent to Perl code C<local $gv>.
1265 =item C<AV* save_ary(GV *gv)>
1267 =item C<HV* save_hash(GV *gv)>
1269 Similar to C<save_scalar>, but localize C<@gv> and C<%gv>.
1271 =item C<void save_item(SV *item)>
1273 Duplicates the current value of C<SV>, on the exit from the current
1274 C<ENTER>/C<LEAVE> I<pseudo-block> will restore the value of C<SV>
1275 using the stored value.
1277 =item C<void save_list(SV **sarg, I32 maxsarg)>
1279 A variant of C<save_item> which takes multiple arguments via an array
1280 C<sarg> of C<SV*> of length C<maxsarg>.
1282 =item C<SV* save_svref(SV **sptr)>
1284 Similar to C<save_scalar>, but will reinstate an C<SV *>.
1286 =item C<void save_aptr(AV **aptr)>
1288 =item C<void save_hptr(HV **hptr)>
1290 Similar to C<save_svref>, but localize C<AV *> and C<HV *>.
1294 The C<Alias> module implements localization of the basic types within the
1295 I<caller's scope>. People who are interested in how to localize things in
1296 the containing scope should take a look there too.
1300 =head2 XSUBs and the Argument Stack
1302 The XSUB mechanism is a simple way for Perl programs to access C subroutines.
1303 An XSUB routine will have a stack that contains the arguments from the Perl
1304 program, and a way to map from the Perl data structures to a C equivalent.
1306 The stack arguments are accessible through the C<ST(n)> macro, which returns
1307 the C<n>'th stack argument. Argument 0 is the first argument passed in the
1308 Perl subroutine call. These arguments are C<SV*>, and can be used anywhere
1311 Most of the time, output from the C routine can be handled through use of
1312 the RETVAL and OUTPUT directives. However, there are some cases where the
1313 argument stack is not already long enough to handle all the return values.
1314 An example is the POSIX tzname() call, which takes no arguments, but returns
1315 two, the local time zone's standard and summer time abbreviations.
1317 To handle this situation, the PPCODE directive is used and the stack is
1318 extended using the macro:
1322 where C<SP> is the macro that represents the local copy of the stack pointer,
1323 and C<num> is the number of elements the stack should be extended by.
1325 Now that there is room on the stack, values can be pushed on it using C<PUSHs>
1326 macro. The values pushed will often need to be "mortal" (See L</Reference Counts and Mortality>).
1328 PUSHs(sv_2mortal(newSViv(an_integer)))
1329 PUSHs(sv_2mortal(newSVpv("Some String",0)))
1330 PUSHs(sv_2mortal(newSVnv(3.141592)))
1332 And now the Perl program calling C<tzname>, the two values will be assigned
1335 ($standard_abbrev, $summer_abbrev) = POSIX::tzname;
1337 An alternate (and possibly simpler) method to pushing values on the stack is
1342 This macro automatically adjust the stack for you, if needed. Thus, you
1343 do not need to call C<EXTEND> to extend the stack.
1345 Despite their suggestions in earlier versions of this document the macros
1346 C<PUSHi>, C<PUSHn> and C<PUSHp> are I<not> suited to XSUBs which return
1347 multiple results, see L</Putting a C value on Perl stack>.
1349 For more information, consult L<perlxs> and L<perlxstut>.
1351 =head2 Calling Perl Routines from within C Programs
1353 There are four routines that can be used to call a Perl subroutine from
1354 within a C program. These four are:
1356 I32 call_sv(SV*, I32);
1357 I32 call_pv(const char*, I32);
1358 I32 call_method(const char*, I32);
1359 I32 call_argv(const char*, I32, register char**);
1361 The routine most often used is C<call_sv>. The C<SV*> argument
1362 contains either the name of the Perl subroutine to be called, or a
1363 reference to the subroutine. The second argument consists of flags
1364 that control the context in which the subroutine is called, whether
1365 or not the subroutine is being passed arguments, how errors should be
1366 trapped, and how to treat return values.
1368 All four routines return the number of arguments that the subroutine returned
1371 These routines used to be called C<perl_call_sv>, etc., before Perl v5.6.0,
1372 but those names are now deprecated; macros of the same name are provided for
1375 When using any of these routines (except C<call_argv>), the programmer
1376 must manipulate the Perl stack. These include the following macros and
1391 For a detailed description of calling conventions from C to Perl,
1392 consult L<perlcall>.
1394 =head2 Memory Allocation
1396 All memory meant to be used with the Perl API functions should be manipulated
1397 using the macros described in this section. The macros provide the necessary
1398 transparency between differences in the actual malloc implementation that is
1401 It is suggested that you enable the version of malloc that is distributed
1402 with Perl. It keeps pools of various sizes of unallocated memory in
1403 order to satisfy allocation requests more quickly. However, on some
1404 platforms, it may cause spurious malloc or free errors.
1406 New(x, pointer, number, type);
1407 Newc(x, pointer, number, type, cast);
1408 Newz(x, pointer, number, type);
1410 These three macros are used to initially allocate memory.
1412 The first argument C<x> was a "magic cookie" that was used to keep track
1413 of who called the macro, to help when debugging memory problems. However,
1414 the current code makes no use of this feature (most Perl developers now
1415 use run-time memory checkers), so this argument can be any number.
1417 The second argument C<pointer> should be the name of a variable that will
1418 point to the newly allocated memory.
1420 The third and fourth arguments C<number> and C<type> specify how many of
1421 the specified type of data structure should be allocated. The argument
1422 C<type> is passed to C<sizeof>. The final argument to C<Newc>, C<cast>,
1423 should be used if the C<pointer> argument is different from the C<type>
1426 Unlike the C<New> and C<Newc> macros, the C<Newz> macro calls C<memzero>
1427 to zero out all the newly allocated memory.
1429 Renew(pointer, number, type);
1430 Renewc(pointer, number, type, cast);
1433 These three macros are used to change a memory buffer size or to free a
1434 piece of memory no longer needed. The arguments to C<Renew> and C<Renewc>
1435 match those of C<New> and C<Newc> with the exception of not needing the
1436 "magic cookie" argument.
1438 Move(source, dest, number, type);
1439 Copy(source, dest, number, type);
1440 Zero(dest, number, type);
1442 These three macros are used to move, copy, or zero out previously allocated
1443 memory. The C<source> and C<dest> arguments point to the source and
1444 destination starting points. Perl will move, copy, or zero out C<number>
1445 instances of the size of the C<type> data structure (using the C<sizeof>
1450 The most recent development releases of Perl has been experimenting with
1451 removing Perl's dependency on the "normal" standard I/O suite and allowing
1452 other stdio implementations to be used. This involves creating a new
1453 abstraction layer that then calls whichever implementation of stdio Perl
1454 was compiled with. All XSUBs should now use the functions in the PerlIO
1455 abstraction layer and not make any assumptions about what kind of stdio
1458 For a complete description of the PerlIO abstraction, consult L<perlapio>.
1460 =head2 Putting a C value on Perl stack
1462 A lot of opcodes (this is an elementary operation in the internal perl
1463 stack machine) put an SV* on the stack. However, as an optimization
1464 the corresponding SV is (usually) not recreated each time. The opcodes
1465 reuse specially assigned SVs (I<target>s) which are (as a corollary)
1466 not constantly freed/created.
1468 Each of the targets is created only once (but see
1469 L<Scratchpads and recursion> below), and when an opcode needs to put
1470 an integer, a double, or a string on stack, it just sets the
1471 corresponding parts of its I<target> and puts the I<target> on stack.
1473 The macro to put this target on stack is C<PUSHTARG>, and it is
1474 directly used in some opcodes, as well as indirectly in zillions of
1475 others, which use it via C<(X)PUSH[pni]>.
1477 Because the target is reused, you must be careful when pushing multiple
1478 values on the stack. The following code will not do what you think:
1483 This translates as "set C<TARG> to 10, push a pointer to C<TARG> onto
1484 the stack; set C<TARG> to 20, push a pointer to C<TARG> onto the stack".
1485 At the end of the operation, the stack does not contain the values 10
1486 and 20, but actually contains two pointers to C<TARG>, which we have set
1487 to 20. If you need to push multiple different values, use C<XPUSHs>,
1488 which bypasses C<TARG>.
1490 On a related note, if you do use C<(X)PUSH[npi]>, then you're going to
1491 need a C<dTARG> in your variable declarations so that the C<*PUSH*>
1492 macros can make use of the local variable C<TARG>.
1496 The question remains on when the SVs which are I<target>s for opcodes
1497 are created. The answer is that they are created when the current unit --
1498 a subroutine or a file (for opcodes for statements outside of
1499 subroutines) -- is compiled. During this time a special anonymous Perl
1500 array is created, which is called a scratchpad for the current
1503 A scratchpad keeps SVs which are lexicals for the current unit and are
1504 targets for opcodes. One can deduce that an SV lives on a scratchpad
1505 by looking on its flags: lexicals have C<SVs_PADMY> set, and
1506 I<target>s have C<SVs_PADTMP> set.
1508 The correspondence between OPs and I<target>s is not 1-to-1. Different
1509 OPs in the compile tree of the unit can use the same target, if this
1510 would not conflict with the expected life of the temporary.
1512 =head2 Scratchpads and recursion
1514 In fact it is not 100% true that a compiled unit contains a pointer to
1515 the scratchpad AV. In fact it contains a pointer to an AV of
1516 (initially) one element, and this element is the scratchpad AV. Why do
1517 we need an extra level of indirection?
1519 The answer is B<recursion>, and maybe B<threads>. Both
1520 these can create several execution pointers going into the same
1521 subroutine. For the subroutine-child not write over the temporaries
1522 for the subroutine-parent (lifespan of which covers the call to the
1523 child), the parent and the child should have different
1524 scratchpads. (I<And> the lexicals should be separate anyway!)
1526 So each subroutine is born with an array of scratchpads (of length 1).
1527 On each entry to the subroutine it is checked that the current
1528 depth of the recursion is not more than the length of this array, and
1529 if it is, new scratchpad is created and pushed into the array.
1531 The I<target>s on this scratchpad are C<undef>s, but they are already
1532 marked with correct flags.
1534 =head1 Compiled code
1538 Here we describe the internal form your code is converted to by
1539 Perl. Start with a simple example:
1543 This is converted to a tree similar to this one:
1551 (but slightly more complicated). This tree reflects the way Perl
1552 parsed your code, but has nothing to do with the execution order.
1553 There is an additional "thread" going through the nodes of the tree
1554 which shows the order of execution of the nodes. In our simplified
1555 example above it looks like:
1557 $b ---> $c ---> + ---> $a ---> assign-to
1559 But with the actual compile tree for C<$a = $b + $c> it is different:
1560 some nodes I<optimized away>. As a corollary, though the actual tree
1561 contains more nodes than our simplified example, the execution order
1562 is the same as in our example.
1564 =head2 Examining the tree
1566 If you have your perl compiled for debugging (usually done with C<-D
1567 optimize=-g> on C<Configure> command line), you may examine the
1568 compiled tree by specifying C<-Dx> on the Perl command line. The
1569 output takes several lines per node, and for C<$b+$c> it looks like
1574 FLAGS = (SCALAR,KIDS)
1576 TYPE = null ===> (4)
1578 FLAGS = (SCALAR,KIDS)
1580 3 TYPE = gvsv ===> 4
1586 TYPE = null ===> (5)
1588 FLAGS = (SCALAR,KIDS)
1590 4 TYPE = gvsv ===> 5
1596 This tree has 5 nodes (one per C<TYPE> specifier), only 3 of them are
1597 not optimized away (one per number in the left column). The immediate
1598 children of the given node correspond to C<{}> pairs on the same level
1599 of indentation, thus this listing corresponds to the tree:
1607 The execution order is indicated by C<===E<gt>> marks, thus it is C<3
1608 4 5 6> (node C<6> is not included into above listing), i.e.,
1609 C<gvsv gvsv add whatever>.
1611 Each of these nodes represents an op, a fundamental operation inside the
1612 Perl core. The code which implements each operation can be found in the
1613 F<pp*.c> files; the function which implements the op with type C<gvsv>
1614 is C<pp_gvsv>, and so on. As the tree above shows, different ops have
1615 different numbers of children: C<add> is a binary operator, as one would
1616 expect, and so has two children. To accommodate the various different
1617 numbers of children, there are various types of op data structure, and
1618 they link together in different ways.
1620 The simplest type of op structure is C<OP>: this has no children. Unary
1621 operators, C<UNOP>s, have one child, and this is pointed to by the
1622 C<op_first> field. Binary operators (C<BINOP>s) have not only an
1623 C<op_first> field but also an C<op_last> field. The most complex type of
1624 op is a C<LISTOP>, which has any number of children. In this case, the
1625 first child is pointed to by C<op_first> and the last child by
1626 C<op_last>. The children in between can be found by iteratively
1627 following the C<op_sibling> pointer from the first child to the last.
1629 There are also two other op types: a C<PMOP> holds a regular expression,
1630 and has no children, and a C<LOOP> may or may not have children. If the
1631 C<op_children> field is non-zero, it behaves like a C<LISTOP>. To
1632 complicate matters, if a C<UNOP> is actually a C<null> op after
1633 optimization (see L</Compile pass 2: context propagation>) it will still
1634 have children in accordance with its former type.
1636 =head2 Compile pass 1: check routines
1638 The tree is created by the compiler while I<yacc> code feeds it
1639 the constructions it recognizes. Since I<yacc> works bottom-up, so does
1640 the first pass of perl compilation.
1642 What makes this pass interesting for perl developers is that some
1643 optimization may be performed on this pass. This is optimization by
1644 so-called "check routines". The correspondence between node names
1645 and corresponding check routines is described in F<opcode.pl> (do not
1646 forget to run C<make regen_headers> if you modify this file).
1648 A check routine is called when the node is fully constructed except
1649 for the execution-order thread. Since at this time there are no
1650 back-links to the currently constructed node, one can do most any
1651 operation to the top-level node, including freeing it and/or creating
1652 new nodes above/below it.
1654 The check routine returns the node which should be inserted into the
1655 tree (if the top-level node was not modified, check routine returns
1658 By convention, check routines have names C<ck_*>. They are usually
1659 called from C<new*OP> subroutines (or C<convert>) (which in turn are
1660 called from F<perly.y>).
1662 =head2 Compile pass 1a: constant folding
1664 Immediately after the check routine is called the returned node is
1665 checked for being compile-time executable. If it is (the value is
1666 judged to be constant) it is immediately executed, and a I<constant>
1667 node with the "return value" of the corresponding subtree is
1668 substituted instead. The subtree is deleted.
1670 If constant folding was not performed, the execution-order thread is
1673 =head2 Compile pass 2: context propagation
1675 When a context for a part of compile tree is known, it is propagated
1676 down through the tree. At this time the context can have 5 values
1677 (instead of 2 for runtime context): void, boolean, scalar, list, and
1678 lvalue. In contrast with the pass 1 this pass is processed from top
1679 to bottom: a node's context determines the context for its children.
1681 Additional context-dependent optimizations are performed at this time.
1682 Since at this moment the compile tree contains back-references (via
1683 "thread" pointers), nodes cannot be free()d now. To allow
1684 optimized-away nodes at this stage, such nodes are null()ified instead
1685 of free()ing (i.e. their type is changed to OP_NULL).
1687 =head2 Compile pass 3: peephole optimization
1689 After the compile tree for a subroutine (or for an C<eval> or a file)
1690 is created, an additional pass over the code is performed. This pass
1691 is neither top-down or bottom-up, but in the execution order (with
1692 additional complications for conditionals). These optimizations are
1693 done in the subroutine peep(). Optimizations performed at this stage
1694 are subject to the same restrictions as in the pass 2.
1696 =head2 Pluggable runops
1698 The compile tree is executed in a runops function. There are two runops
1699 functions in F<run.c>. C<Perl_runops_debug> is used with DEBUGGING and
1700 C<Perl_runops_standard> is used otherwise. For fine control over the
1701 execution of the compile tree it is possible to provide your own runops
1704 It's probably best to copy one of the existing runops functions and
1705 change it to suit your needs. Then, in the BOOT section of your XS
1708 PL_runops = my_runops;
1710 This function should be as efficient as possible to keep your programs
1711 running as fast as possible.
1713 =head1 Examining internal data structures with the C<dump> functions
1715 To aid debugging, the source file F<dump.c> contains a number of
1716 functions which produce formatted output of internal data structures.
1718 The most commonly used of these functions is C<Perl_sv_dump>; it's used
1719 for dumping SVs, AVs, HVs, and CVs. The C<Devel::Peek> module calls
1720 C<sv_dump> to produce debugging output from Perl-space, so users of that
1721 module should already be familiar with its format.
1723 C<Perl_op_dump> can be used to dump an C<OP> structure or any of its
1724 derivatives, and produces output similar to C<perl -Dx>; in fact,
1725 C<Perl_dump_eval> will dump the main root of the code being evaluated,
1726 exactly like C<-Dx>.
1728 Other useful functions are C<Perl_dump_sub>, which turns a C<GV> into an
1729 op tree, C<Perl_dump_packsubs> which calls C<Perl_dump_sub> on all the
1730 subroutines in a package like so: (Thankfully, these are all xsubs, so
1731 there is no op tree)
1733 (gdb) print Perl_dump_packsubs(PL_defstash)
1735 SUB attributes::bootstrap = (xsub 0x811fedc 0)
1737 SUB UNIVERSAL::can = (xsub 0x811f50c 0)
1739 SUB UNIVERSAL::isa = (xsub 0x811f304 0)
1741 SUB UNIVERSAL::VERSION = (xsub 0x811f7ac 0)
1743 SUB DynaLoader::boot_DynaLoader = (xsub 0x805b188 0)
1745 and C<Perl_dump_all>, which dumps all the subroutines in the stash and
1746 the op tree of the main root.
1748 =head1 How multiple interpreters and concurrency are supported
1750 =head2 Background and PERL_IMPLICIT_CONTEXT
1752 The Perl interpreter can be regarded as a closed box: it has an API
1753 for feeding it code or otherwise making it do things, but it also has
1754 functions for its own use. This smells a lot like an object, and
1755 there are ways for you to build Perl so that you can have multiple
1756 interpreters, with one interpreter represented either as a C structure,
1757 or inside a thread-specific structure. These structures contain all
1758 the context, the state of that interpreter.
1760 Two macros control the major Perl build flavors: MULTIPLICITY and
1761 USE_5005THREADS. The MULTIPLICITY build has a C structure
1762 that packages all the interpreter state, and there is a similar thread-specific
1763 data structure under USE_5005THREADS. In both cases,
1764 PERL_IMPLICIT_CONTEXT is also normally defined, and enables the
1765 support for passing in a "hidden" first argument that represents all three
1768 All this obviously requires a way for the Perl internal functions to be
1769 either subroutines taking some kind of structure as the first
1770 argument, or subroutines taking nothing as the first argument. To
1771 enable these two very different ways of building the interpreter,
1772 the Perl source (as it does in so many other situations) makes heavy
1773 use of macros and subroutine naming conventions.
1775 First problem: deciding which functions will be public API functions and
1776 which will be private. All functions whose names begin C<S_> are private
1777 (think "S" for "secret" or "static"). All other functions begin with
1778 "Perl_", but just because a function begins with "Perl_" does not mean it is
1779 part of the API. (See L</Internal Functions>.) The easiest way to be B<sure> a
1780 function is part of the API is to find its entry in L<perlapi>.
1781 If it exists in L<perlapi>, it's part of the API. If it doesn't, and you
1782 think it should be (i.e., you need it for your extension), send mail via
1783 L<perlbug> explaining why you think it should be.
1785 Second problem: there must be a syntax so that the same subroutine
1786 declarations and calls can pass a structure as their first argument,
1787 or pass nothing. To solve this, the subroutines are named and
1788 declared in a particular way. Here's a typical start of a static
1789 function used within the Perl guts:
1792 S_incline(pTHX_ char *s)
1794 STATIC becomes "static" in C, and may be #define'd to nothing in some
1795 configurations in future.
1797 A public function (i.e. part of the internal API, but not necessarily
1798 sanctioned for use in extensions) begins like this:
1801 Perl_sv_setsv(pTHX_ SV* dsv, SV* ssv)
1803 C<pTHX_> is one of a number of macros (in perl.h) that hide the
1804 details of the interpreter's context. THX stands for "thread", "this",
1805 or "thingy", as the case may be. (And no, George Lucas is not involved. :-)
1806 The first character could be 'p' for a B<p>rototype, 'a' for B<a>rgument,
1807 or 'd' for B<d>eclaration, so we have C<pTHX>, C<aTHX> and C<dTHX>, and
1810 When Perl is built without options that set PERL_IMPLICIT_CONTEXT, there is no
1811 first argument containing the interpreter's context. The trailing underscore
1812 in the pTHX_ macro indicates that the macro expansion needs a comma
1813 after the context argument because other arguments follow it. If
1814 PERL_IMPLICIT_CONTEXT is not defined, pTHX_ will be ignored, and the
1815 subroutine is not prototyped to take the extra argument. The form of the
1816 macro without the trailing underscore is used when there are no additional
1819 When a core function calls another, it must pass the context. This
1820 is normally hidden via macros. Consider C<sv_setsv>. It expands into
1821 something like this:
1823 ifdef PERL_IMPLICIT_CONTEXT
1824 define sv_setsv(a,b) Perl_sv_setsv(aTHX_ a, b)
1825 /* can't do this for vararg functions, see below */
1827 define sv_setsv Perl_sv_setsv
1830 This works well, and means that XS authors can gleefully write:
1834 and still have it work under all the modes Perl could have been
1837 This doesn't work so cleanly for varargs functions, though, as macros
1838 imply that the number of arguments is known in advance. Instead we
1839 either need to spell them out fully, passing C<aTHX_> as the first
1840 argument (the Perl core tends to do this with functions like
1841 Perl_warner), or use a context-free version.
1843 The context-free version of Perl_warner is called
1844 Perl_warner_nocontext, and does not take the extra argument. Instead
1845 it does dTHX; to get the context from thread-local storage. We
1846 C<#define warner Perl_warner_nocontext> so that extensions get source
1847 compatibility at the expense of performance. (Passing an arg is
1848 cheaper than grabbing it from thread-local storage.)
1850 You can ignore [pad]THXx when browsing the Perl headers/sources.
1851 Those are strictly for use within the core. Extensions and embedders
1852 need only be aware of [pad]THX.
1854 =head2 So what happened to dTHR?
1856 C<dTHR> was introduced in perl 5.005 to support the older thread model.
1857 The older thread model now uses the C<THX> mechanism to pass context
1858 pointers around, so C<dTHR> is not useful any more. Perl 5.6.0 and
1859 later still have it for backward source compatibility, but it is defined
1862 =head2 How do I use all this in extensions?
1864 When Perl is built with PERL_IMPLICIT_CONTEXT, extensions that call
1865 any functions in the Perl API will need to pass the initial context
1866 argument somehow. The kicker is that you will need to write it in
1867 such a way that the extension still compiles when Perl hasn't been
1868 built with PERL_IMPLICIT_CONTEXT enabled.
1870 There are three ways to do this. First, the easy but inefficient way,
1871 which is also the default, in order to maintain source compatibility
1872 with extensions: whenever XSUB.h is #included, it redefines the aTHX
1873 and aTHX_ macros to call a function that will return the context.
1874 Thus, something like:
1878 in your extension will translate to this when PERL_IMPLICIT_CONTEXT is
1881 Perl_sv_setsv(Perl_get_context(), asv, bsv);
1883 or to this otherwise:
1885 Perl_sv_setsv(asv, bsv);
1887 You have to do nothing new in your extension to get this; since
1888 the Perl library provides Perl_get_context(), it will all just
1891 The second, more efficient way is to use the following template for
1894 #define PERL_NO_GET_CONTEXT /* we want efficiency */
1899 static my_private_function(int arg1, int arg2);
1902 my_private_function(int arg1, int arg2)
1904 dTHX; /* fetch context */
1905 ... call many Perl API functions ...
1910 MODULE = Foo PACKAGE = Foo
1918 my_private_function(arg, 10);
1920 Note that the only two changes from the normal way of writing an
1921 extension is the addition of a C<#define PERL_NO_GET_CONTEXT> before
1922 including the Perl headers, followed by a C<dTHX;> declaration at
1923 the start of every function that will call the Perl API. (You'll
1924 know which functions need this, because the C compiler will complain
1925 that there's an undeclared identifier in those functions.) No changes
1926 are needed for the XSUBs themselves, because the XS() macro is
1927 correctly defined to pass in the implicit context if needed.
1929 The third, even more efficient way is to ape how it is done within
1933 #define PERL_NO_GET_CONTEXT /* we want efficiency */
1938 /* pTHX_ only needed for functions that call Perl API */
1939 static my_private_function(pTHX_ int arg1, int arg2);
1942 my_private_function(pTHX_ int arg1, int arg2)
1944 /* dTHX; not needed here, because THX is an argument */
1945 ... call Perl API functions ...
1950 MODULE = Foo PACKAGE = Foo
1958 my_private_function(aTHX_ arg, 10);
1960 This implementation never has to fetch the context using a function
1961 call, since it is always passed as an extra argument. Depending on
1962 your needs for simplicity or efficiency, you may mix the previous
1963 two approaches freely.
1965 Never add a comma after C<pTHX> yourself--always use the form of the
1966 macro with the underscore for functions that take explicit arguments,
1967 or the form without the argument for functions with no explicit arguments.
1969 =head2 Should I do anything special if I call perl from multiple threads?
1971 If you create interpreters in one thread and then proceed to call them in
1972 another, you need to make sure perl's own Thread Local Storage (TLS) slot is
1973 initialized correctly in each of those threads.
1975 The C<perl_alloc> and C<perl_clone> API functions will automatically set
1976 the TLS slot to the interpreter they created, so that there is no need to do
1977 anything special if the interpreter is always accessed in the same thread that
1978 created it, and that thread did not create or call any other interpreters
1979 afterwards. If that is not the case, you have to set the TLS slot of the
1980 thread before calling any functions in the Perl API on that particular
1981 interpreter. This is done by calling the C<PERL_SET_CONTEXT> macro in that
1982 thread as the first thing you do:
1984 /* do this before doing anything else with some_perl */
1985 PERL_SET_CONTEXT(some_perl);
1987 ... other Perl API calls on some_perl go here ...
1989 =head2 Future Plans and PERL_IMPLICIT_SYS
1991 Just as PERL_IMPLICIT_CONTEXT provides a way to bundle up everything
1992 that the interpreter knows about itself and pass it around, so too are
1993 there plans to allow the interpreter to bundle up everything it knows
1994 about the environment it's running on. This is enabled with the
1995 PERL_IMPLICIT_SYS macro. Currently it only works with USE_ITHREADS
1996 and USE_5005THREADS on Windows (see inside iperlsys.h).
1998 This allows the ability to provide an extra pointer (called the "host"
1999 environment) for all the system calls. This makes it possible for
2000 all the system stuff to maintain their own state, broken down into
2001 seven C structures. These are thin wrappers around the usual system
2002 calls (see win32/perllib.c) for the default perl executable, but for a
2003 more ambitious host (like the one that would do fork() emulation) all
2004 the extra work needed to pretend that different interpreters are
2005 actually different "processes", would be done here.
2007 The Perl engine/interpreter and the host are orthogonal entities.
2008 There could be one or more interpreters in a process, and one or
2009 more "hosts", with free association between them.
2011 =head1 Internal Functions
2013 All of Perl's internal functions which will be exposed to the outside
2014 world are be prefixed by C<Perl_> so that they will not conflict with XS
2015 functions or functions used in a program in which Perl is embedded.
2016 Similarly, all global variables begin with C<PL_>. (By convention,
2017 static functions start with C<S_>)
2019 Inside the Perl core, you can get at the functions either with or
2020 without the C<Perl_> prefix, thanks to a bunch of defines that live in
2021 F<embed.h>. This header file is generated automatically from
2022 F<embed.pl>. F<embed.pl> also creates the prototyping header files for
2023 the internal functions, generates the documentation and a lot of other
2024 bits and pieces. It's important that when you add a new function to the
2025 core or change an existing one, you change the data in the table at the
2026 end of F<embed.pl> as well. Here's a sample entry from that table:
2028 Apd |SV** |av_fetch |AV* ar|I32 key|I32 lval
2030 The second column is the return type, the third column the name. Columns
2031 after that are the arguments. The first column is a set of flags:
2037 This function is a part of the public API.
2041 This function has a C<Perl_> prefix; ie, it is defined as C<Perl_av_fetch>
2045 This function has documentation using the C<apidoc> feature which we'll
2046 look at in a second.
2050 Other available flags are:
2056 This is a static function and is defined as C<S_whatever>, and usually
2057 called within the sources as C<whatever(...)>.
2061 This does not use C<aTHX_> and C<pTHX> to pass interpreter context. (See
2062 L<perlguts/Background and PERL_IMPLICIT_CONTEXT>.)
2066 This function never returns; C<croak>, C<exit> and friends.
2070 This function takes a variable number of arguments, C<printf> style.
2071 The argument list should end with C<...>, like this:
2073 Afprd |void |croak |const char* pat|...
2077 This function is part of the experimental development API, and may change
2078 or disappear without notice.
2082 This function should not have a compatibility macro to define, say,
2083 C<Perl_parse> to C<parse>. It must be called as C<Perl_parse>.
2087 This function is not a member of C<CPerlObj>. If you don't know
2088 what this means, don't use it.
2092 This function isn't exported out of the Perl core.
2096 If you edit F<embed.pl>, you will need to run C<make regen_headers> to
2097 force a rebuild of F<embed.h> and other auto-generated files.
2099 =head2 Formatted Printing of IVs, UVs, and NVs
2101 If you are printing IVs, UVs, or NVS instead of the stdio(3) style
2102 formatting codes like C<%d>, C<%ld>, C<%f>, you should use the
2103 following macros for portability
2108 UVxf UV in hexadecimal
2113 These will take care of 64-bit integers and long doubles.
2116 printf("IV is %"IVdf"\n", iv);
2118 The IVdf will expand to whatever is the correct format for the IVs.
2120 If you are printing addresses of pointers, use UVxf combined
2121 with PTR2UV(), do not use %lx or %p.
2123 =head2 Pointer-To-Integer and Integer-To-Pointer
2125 Because pointer size does not necessarily equal integer size,
2126 use the follow macros to do it right.
2131 INT2PTR(pointertotype, integer)
2136 SV *sv = INT2PTR(SV*, iv);
2143 =head2 Source Documentation
2145 There's an effort going on to document the internal functions and
2146 automatically produce reference manuals from them - L<perlapi> is one
2147 such manual which details all the functions which are available to XS
2148 writers. L<perlintern> is the autogenerated manual for the functions
2149 which are not part of the API and are supposedly for internal use only.
2151 Source documentation is created by putting POD comments into the C
2155 =for apidoc sv_setiv
2157 Copies an integer into the given SV. Does not handle 'set' magic. See
2163 Please try and supply some documentation if you add functions to the
2166 =head1 Unicode Support
2168 Perl 5.6.0 introduced Unicode support. It's important for porters and XS
2169 writers to understand this support and make sure that the code they
2170 write does not corrupt Unicode data.
2172 =head2 What B<is> Unicode, anyway?
2174 In the olden, less enlightened times, we all used to use ASCII. Most of
2175 us did, anyway. The big problem with ASCII is that it's American. Well,
2176 no, that's not actually the problem; the problem is that it's not
2177 particularly useful for people who don't use the Roman alphabet. What
2178 used to happen was that particular languages would stick their own
2179 alphabet in the upper range of the sequence, between 128 and 255. Of
2180 course, we then ended up with plenty of variants that weren't quite
2181 ASCII, and the whole point of it being a standard was lost.
2183 Worse still, if you've got a language like Chinese or
2184 Japanese that has hundreds or thousands of characters, then you really
2185 can't fit them into a mere 256, so they had to forget about ASCII
2186 altogether, and build their own systems using pairs of numbers to refer
2189 To fix this, some people formed Unicode, Inc. and
2190 produced a new character set containing all the characters you can
2191 possibly think of and more. There are several ways of representing these
2192 characters, and the one Perl uses is called UTF8. UTF8 uses
2193 a variable number of bytes to represent a character, instead of just
2194 one. You can learn more about Unicode at http://www.unicode.org/
2196 =head2 How can I recognise a UTF8 string?
2198 You can't. This is because UTF8 data is stored in bytes just like
2199 non-UTF8 data. The Unicode character 200, (C<0xC8> for you hex types)
2200 capital E with a grave accent, is represented by the two bytes
2201 C<v196.172>. Unfortunately, the non-Unicode string C<chr(196).chr(172)>
2202 has that byte sequence as well. So you can't tell just by looking - this
2203 is what makes Unicode input an interesting problem.
2205 The API function C<is_utf8_string> can help; it'll tell you if a string
2206 contains only valid UTF8 characters. However, it can't do the work for
2207 you. On a character-by-character basis, C<is_utf8_char> will tell you
2208 whether the current character in a string is valid UTF8.
2210 =head2 How does UTF8 represent Unicode characters?
2212 As mentioned above, UTF8 uses a variable number of bytes to store a
2213 character. Characters with values 1...128 are stored in one byte, just
2214 like good ol' ASCII. Character 129 is stored as C<v194.129>; this
2215 continues up to character 191, which is C<v194.191>. Now we've run out of
2216 bits (191 is binary C<10111111>) so we move on; 192 is C<v195.128>. And
2217 so it goes on, moving to three bytes at character 2048.
2219 Assuming you know you're dealing with a UTF8 string, you can find out
2220 how long the first character in it is with the C<UTF8SKIP> macro:
2222 char *utf = "\305\233\340\240\201";
2225 len = UTF8SKIP(utf); /* len is 2 here */
2227 len = UTF8SKIP(utf); /* len is 3 here */
2229 Another way to skip over characters in a UTF8 string is to use
2230 C<utf8_hop>, which takes a string and a number of characters to skip
2231 over. You're on your own about bounds checking, though, so don't use it
2234 All bytes in a multi-byte UTF8 character will have the high bit set, so
2235 you can test if you need to do something special with this character
2241 /* Must treat this as UTF8 */
2242 uv = utf8_to_uv(utf);
2244 /* OK to treat this character as a byte */
2247 You can also see in that example that we use C<utf8_to_uv> to get the
2248 value of the character; the inverse function C<uv_to_utf8> is available
2249 for putting a UV into UTF8:
2252 /* Must treat this as UTF8 */
2253 utf8 = uv_to_utf8(utf8, uv);
2255 /* OK to treat this character as a byte */
2258 You B<must> convert characters to UVs using the above functions if
2259 you're ever in a situation where you have to match UTF8 and non-UTF8
2260 characters. You may not skip over UTF8 characters in this case. If you
2261 do this, you'll lose the ability to match hi-bit non-UTF8 characters;
2262 for instance, if your UTF8 string contains C<v196.172>, and you skip
2263 that character, you can never match a C<chr(200)> in a non-UTF8 string.
2266 =head2 How does Perl store UTF8 strings?
2268 Currently, Perl deals with Unicode strings and non-Unicode strings
2269 slightly differently. If a string has been identified as being UTF-8
2270 encoded, Perl will set a flag in the SV, C<SVf_UTF8>. You can check and
2271 manipulate this flag with the following macros:
2277 This flag has an important effect on Perl's treatment of the string: if
2278 Unicode data is not properly distinguished, regular expressions,
2279 C<length>, C<substr> and other string handling operations will have
2280 undesirable results.
2282 The problem comes when you have, for instance, a string that isn't
2283 flagged is UTF8, and contains a byte sequence that could be UTF8 -
2284 especially when combining non-UTF8 and UTF8 strings.
2286 Never forget that the C<SVf_UTF8> flag is separate to the PV value; you
2287 need be sure you don't accidentally knock it off while you're
2288 manipulating SVs. More specifically, you cannot expect to do this:
2297 nsv = newSVpvn(p, len);
2299 The C<char*> string does not tell you the whole story, and you can't
2300 copy or reconstruct an SV just by copying the string value. Check if the
2301 old SV has the UTF8 flag set, and act accordingly:
2305 nsv = newSVpvn(p, len);
2309 In fact, your C<frobnicate> function should be made aware of whether or
2310 not it's dealing with UTF8 data, so that it can handle the string
2313 =head2 How do I convert a string to UTF8?
2315 If you're mixing UTF8 and non-UTF8 strings, you might find it necessary
2316 to upgrade one of the strings to UTF8. If you've got an SV, the easiest
2319 sv_utf8_upgrade(sv);
2321 However, you must not do this, for example:
2324 sv_utf8_upgrade(left);
2326 If you do this in a binary operator, you will actually change one of the
2327 strings that came into the operator, and, while it shouldn't be noticeable
2328 by the end user, it can cause problems.
2330 Instead, C<bytes_to_utf8> will give you a UTF8-encoded B<copy> of its
2331 string argument. This is useful for having the data available for
2332 comparisons and so on, without harming the original SV. There's also
2333 C<utf8_to_bytes> to go the other way, but naturally, this will fail if
2334 the string contains any characters above 255 that can't be represented
2337 =head2 Is there anything else I need to know?
2339 Not really. Just remember these things:
2345 There's no way to tell if a string is UTF8 or not. You can tell if an SV
2346 is UTF8 by looking at is C<SvUTF8> flag. Don't forget to set the flag if
2347 something should be UTF8. Treat the flag as part of the PV, even though
2348 it's not - if you pass on the PV to somewhere, pass on the flag too.
2352 If a string is UTF8, B<always> use C<utf8_to_uv> to get at the value,
2353 unless C<!(*s & 0x80)> in which case you can use C<*s>.
2357 When writing to a UTF8 string, B<always> use C<uv_to_utf8>, unless
2358 C<uv < 0x80> in which case you can use C<*s = uv>.
2362 Mixing UTF8 and non-UTF8 strings is tricky. Use C<bytes_to_utf8> to get
2363 a new string which is UTF8 encoded. There are tricks you can use to
2364 delay deciding whether you need to use a UTF8 string until you get to a
2365 high character - C<HALF_UPGRADE> is one of those.
2369 =head1 Custom Operators
2371 Custom operator support is a new experimental feature that allows you to
2372 define your own ops. This is primarily to allow the building of
2373 interpreters for other languages in the Perl core, but it also allows
2374 optimizations through the creation of "macro-ops" (ops which perform the
2375 functions of multiple ops which are usually executed together, such as
2376 C<gvsv, gvsv, add>.)
2378 This feature is implemented as a new op type, C<OP_CUSTOM>. The Perl
2379 core does not "know" anything special about this op type, and so it will
2380 not be involved in any optimizations. This also means that you can
2381 define your custom ops to be any op structure - unary, binary, list and
2384 It's important to know what custom operators won't do for you. They
2385 won't let you add new syntax to Perl, directly. They won't even let you
2386 add new keywords, directly. In fact, they won't change the way Perl
2387 compiles a program at all. You have to do those changes yourself, after
2388 Perl has compiled the program. You do this either by manipulating the op
2389 tree using a C<CHECK> block and the C<B::Generate> module, or by adding
2390 a custom peephole optimizer with the C<optimize> module.
2392 When you do this, you replace ordinary Perl ops with custom ops by
2393 creating ops with the type C<OP_CUSTOM> and the C<pp_addr> of your own
2394 PP function. This should be defined in XS code, and should look like
2395 the PP ops in C<pp_*.c>. You are responsible for ensuring that your op
2396 takes the appropriate number of values from the stack, and you are
2397 responsible for adding stack marks if necessary.
2399 You should also "register" your op with the Perl interpreter so that it
2400 can produce sensible error and warning messages. Since it is possible to
2401 have multiple custom ops within the one "logical" op type C<OP_CUSTOM>,
2402 Perl uses the value of C<< o->op_ppaddr >> as a key into the
2403 C<PL_custom_op_descs> and C<PL_custom_op_names> hashes. This means you
2404 need to enter a name and description for your op at the appropriate
2405 place in the C<PL_custom_op_names> and C<PL_custom_op_descs> hashes.
2407 Forthcoming versions of C<B::Generate> (version 1.0 and above) should
2408 directly support the creation of custom ops by name; C<Opcodes::Custom>
2409 will provide functions which make it trivial to "register" custom ops to
2410 the Perl interpreter.
2414 Until May 1997, this document was maintained by Jeff Okamoto
2415 E<lt>okamoto@corp.hp.comE<gt>. It is now maintained as part of Perl
2416 itself by the Perl 5 Porters E<lt>perl5-porters@perl.orgE<gt>.
2418 With lots of help and suggestions from Dean Roehrich, Malcolm Beattie,
2419 Andreas Koenig, Paul Hudson, Ilya Zakharevich, Paul Marquess, Neil
2420 Bowers, Matthew Green, Tim Bunce, Spider Boardman, Ulrich Pfeifer,
2421 Stephen McCamant, and Gurusamy Sarathy.
2423 API Listing originally by Dean Roehrich E<lt>roehrich@cray.comE<gt>.
2425 Modifications to autogenerate the API listing (L<perlapi>) by Benjamin
2430 perlapi(1), perlintern(1), perlxs(1), perlembed(1)