3 perlembed - how to embed perl in your C program
13 =item B<Use C from Perl?>
15 Read L<perlxstut>, L<perlxs>, L<h2xs>, L<perlguts>, and L<perlapi>.
17 =item B<Use a Unix program from Perl?>
19 Read about back-quotes and about C<system> and C<exec> in L<perlfunc>.
21 =item B<Use Perl from Perl?>
23 Read about L<perlfunc/do> and L<perlfunc/eval> and L<perlfunc/require>
26 =item B<Use C from C?>
30 =item B<Use Perl from C?>
42 Compiling your C program
46 Adding a Perl interpreter to your C program
50 Calling a Perl subroutine from your C program
54 Evaluating a Perl statement from your C program
58 Performing Perl pattern matches and substitutions from your C program
62 Fiddling with the Perl stack from your C program
66 Maintaining a persistent interpreter
70 Maintaining multiple interpreter instances
74 Using Perl modules, which themselves use C libraries, from your C program
78 Embedding Perl under Win32
82 =head2 Compiling your C program
84 If you have trouble compiling the scripts in this documentation,
85 you're not alone. The cardinal rule: COMPILE THE PROGRAMS IN EXACTLY
86 THE SAME WAY THAT YOUR PERL WAS COMPILED. (Sorry for yelling.)
88 Also, every C program that uses Perl must link in the I<perl library>.
89 What's that, you ask? Perl is itself written in C; the perl library
90 is the collection of compiled C programs that were used to create your
91 perl executable (I</usr/bin/perl> or equivalent). (Corollary: you
92 can't use Perl from your C program unless Perl has been compiled on
93 your machine, or installed properly--that's why you shouldn't blithely
94 copy Perl executables from machine to machine without also copying the
97 When you use Perl from C, your C program will--usually--allocate,
98 "run", and deallocate a I<PerlInterpreter> object, which is defined by
101 If your copy of Perl is recent enough to contain this documentation
102 (version 5.002 or later), then the perl library (and I<EXTERN.h> and
103 I<perl.h>, which you'll also need) will reside in a directory
104 that looks like this:
106 /usr/local/lib/perl5/your_architecture_here/CORE
110 /usr/local/lib/perl5/CORE
112 or maybe something like
116 Execute this statement for a hint about where to find CORE:
118 perl -MConfig -e 'print $Config{archlib}'
120 Here's how you'd compile the example in the next section,
121 L<Adding a Perl interpreter to your C program>, on my Linux box:
123 % gcc -O2 -Dbool=char -DHAS_BOOL -I/usr/local/include
124 -I/usr/local/lib/perl5/i586-linux/5.003/CORE
125 -L/usr/local/lib/perl5/i586-linux/5.003/CORE
126 -o interp interp.c -lperl -lm
128 (That's all one line.) On my DEC Alpha running old 5.003_05, the
129 incantation is a bit different:
131 % cc -O2 -Olimit 2900 -DSTANDARD_C -I/usr/local/include
132 -I/usr/local/lib/perl5/alpha-dec_osf/5.00305/CORE
133 -L/usr/local/lib/perl5/alpha-dec_osf/5.00305/CORE -L/usr/local/lib
134 -D__LANGUAGE_C__ -D_NO_PROTO -o interp interp.c -lperl -lm
136 How can you figure out what to add? Assuming your Perl is post-5.001,
137 execute a C<perl -V> command and pay special attention to the "cc" and
138 "ccflags" information.
140 You'll have to choose the appropriate compiler (I<cc>, I<gcc>, et al.) for
141 your machine: C<perl -MConfig -e 'print $Config{cc}'> will tell you what
144 You'll also have to choose the appropriate library directory
145 (I</usr/local/lib/...>) for your machine. If your compiler complains
146 that certain functions are undefined, or that it can't locate
147 I<-lperl>, then you need to change the path following the C<-L>. If it
148 complains that it can't find I<EXTERN.h> and I<perl.h>, you need to
149 change the path following the C<-I>.
151 You may have to add extra libraries as well. Which ones?
152 Perhaps those printed by
154 perl -MConfig -e 'print $Config{libs}'
156 Provided your perl binary was properly configured and installed the
157 B<ExtUtils::Embed> module will determine all of this information for
160 % cc -o interp interp.c `perl -MExtUtils::Embed -e ccopts -e ldopts`
162 If the B<ExtUtils::Embed> module isn't part of your Perl distribution,
163 you can retrieve it from
164 http://www.perl.com/perl/CPAN/modules/by-module/ExtUtils/
165 (If this documentation came from your Perl distribution, then you're
166 running 5.004 or better and you already have it.)
168 The B<ExtUtils::Embed> kit on CPAN also contains all source code for
169 the examples in this document, tests, additional examples and other
170 information you may find useful.
172 =head2 Adding a Perl interpreter to your C program
174 In a sense, perl (the C program) is a good example of embedding Perl
175 (the language), so I'll demonstrate embedding with I<miniperlmain.c>,
176 included in the source distribution. Here's a bastardized, nonportable
177 version of I<miniperlmain.c> containing the essentials of embedding:
179 #include <EXTERN.h> /* from the Perl distribution */
180 #include <perl.h> /* from the Perl distribution */
182 static PerlInterpreter *my_perl; /*** The Perl interpreter ***/
184 int main(int argc, char **argv, char **env)
186 PERL_SYS_INIT3(&argc,&argv,&env);
187 my_perl = perl_alloc();
188 perl_construct(my_perl);
189 PL_exit_flags |= PERL_EXIT_DESTRUCT_END;
190 perl_parse(my_perl, NULL, argc, argv, (char **)NULL);
192 perl_destruct(my_perl);
197 Notice that we don't use the C<env> pointer. Normally handed to
198 C<perl_parse> as its final argument, C<env> here is replaced by
199 C<NULL>, which means that the current environment will be used. The macros
200 PERL_SYS_INIT3() and PERL_SYS_TERM() provide system-specific tune up
201 of the C runtime environment necessary to run Perl interpreters; since
202 PERL_SYS_INIT3() may change C<env>, it may be more appropriate to provide
203 C<env> as an argument to perl_parse().
205 Now compile this program (I'll call it I<interp.c>) into an executable:
207 % cc -o interp interp.c `perl -MExtUtils::Embed -e ccopts -e ldopts`
209 After a successful compilation, you'll be able to use I<interp> just
213 print "Pretty Good Perl \n";
214 print "10890 - 9801 is ", 10890 - 9801;
221 % interp -e 'printf("%x", 3735928559)'
224 You can also read and execute Perl statements from a file while in the
225 midst of your C program, by placing the filename in I<argv[1]> before
228 =head2 Calling a Perl subroutine from your C program
230 To call individual Perl subroutines, you can use any of the B<call_*>
231 functions documented in L<perlcall>.
232 In this example we'll use C<call_argv>.
234 That's shown below, in a program I'll call I<showtime.c>.
239 static PerlInterpreter *my_perl;
241 int main(int argc, char **argv, char **env)
243 char *args[] = { NULL };
244 PERL_SYS_INIT3(&argc,&argv,&env);
245 my_perl = perl_alloc();
246 perl_construct(my_perl);
248 perl_parse(my_perl, NULL, argc, argv, NULL);
249 PL_exit_flags |= PERL_EXIT_DESTRUCT_END;
251 /*** skipping perl_run() ***/
253 call_argv("showtime", G_DISCARD | G_NOARGS, args);
255 perl_destruct(my_perl);
260 where I<showtime> is a Perl subroutine that takes no arguments (that's the
261 I<G_NOARGS>) and for which I'll ignore the return value (that's the
262 I<G_DISCARD>). Those flags, and others, are discussed in L<perlcall>.
264 I'll define the I<showtime> subroutine in a file called I<showtime.pl>:
266 print "I shan't be printed.";
272 Simple enough. Now compile and run:
274 % cc -o showtime showtime.c `perl -MExtUtils::Embed -e ccopts -e ldopts`
276 % showtime showtime.pl
279 yielding the number of seconds that elapsed between January 1, 1970
280 (the beginning of the Unix epoch), and the moment I began writing this
283 In this particular case we don't have to call I<perl_run>, as we set
284 the PL_exit_flag PERL_EXIT_DESTRUCT_END which executes END blocks in
287 If you want to pass arguments to the Perl subroutine, you can add
288 strings to the C<NULL>-terminated C<args> list passed to
289 I<call_argv>. For other data types, or to examine return values,
290 you'll need to manipulate the Perl stack. That's demonstrated in
291 L<Fiddling with the Perl stack from your C program>.
293 =head2 Evaluating a Perl statement from your C program
295 Perl provides two API functions to evaluate pieces of Perl code.
296 These are L<perlapi/eval_sv> and L<perlapi/eval_pv>.
298 Arguably, these are the only routines you'll ever need to execute
299 snippets of Perl code from within your C program. Your code can be as
300 long as you wish; it can contain multiple statements; it can employ
301 L<perlfunc/use>, L<perlfunc/require>, and L<perlfunc/do> to
302 include external Perl files.
304 I<eval_pv> lets us evaluate individual Perl strings, and then
305 extract variables for coercion into C types. The following program,
306 I<string.c>, executes three Perl strings, extracting an C<int> from
307 the first, a C<float> from the second, and a C<char *> from the third.
312 static PerlInterpreter *my_perl;
314 main (int argc, char **argv, char **env)
317 char *embedding[] = { "", "-e", "0" };
319 PERL_SYS_INIT3(&argc,&argv,&env);
320 my_perl = perl_alloc();
321 perl_construct( my_perl );
323 perl_parse(my_perl, NULL, 3, embedding, NULL);
324 PL_exit_flags |= PERL_EXIT_DESTRUCT_END;
327 /** Treat $a as an integer **/
328 eval_pv("$a = 3; $a **= 2", TRUE);
329 printf("a = %d\n", SvIV(get_sv("a", FALSE)));
331 /** Treat $a as a float **/
332 eval_pv("$a = 3.14; $a **= 2", TRUE);
333 printf("a = %f\n", SvNV(get_sv("a", FALSE)));
335 /** Treat $a as a string **/
336 eval_pv("$a = 'rekcaH lreP rehtonA tsuJ'; $a = reverse($a);", TRUE);
337 printf("a = %s\n", SvPV(get_sv("a", FALSE), n_a));
339 perl_destruct(my_perl);
344 All of those strange functions with I<sv> in their names help convert Perl scalars to C types. They're described in L<perlguts> and L<perlapi>.
346 If you compile and run I<string.c>, you'll see the results of using
347 I<SvIV()> to create an C<int>, I<SvNV()> to create a C<float>, and
348 I<SvPV()> to create a string:
352 a = Just Another Perl Hacker
354 In the example above, we've created a global variable to temporarily
355 store the computed value of our eval'd expression. It is also
356 possible and in most cases a better strategy to fetch the return value
357 from I<eval_pv()> instead. Example:
361 SV *val = eval_pv("reverse 'rekcaH lreP rehtonA tsuJ'", TRUE);
362 printf("%s\n", SvPV(val,n_a));
365 This way, we avoid namespace pollution by not creating global
366 variables and we've simplified our code as well.
368 =head2 Performing Perl pattern matches and substitutions from your C program
370 The I<eval_sv()> function lets us evaluate strings of Perl code, so we can
371 define some functions that use it to "specialize" in matches and
372 substitutions: I<match()>, I<substitute()>, and I<matches()>.
374 I32 match(SV *string, char *pattern);
376 Given a string and a pattern (e.g., C<m/clasp/> or C</\b\w*\b/>, which
377 in your C program might appear as "/\\b\\w*\\b/"), match()
378 returns 1 if the string matches the pattern and 0 otherwise.
380 int substitute(SV **string, char *pattern);
382 Given a pointer to an C<SV> and an C<=~> operation (e.g.,
383 C<s/bob/robert/g> or C<tr[A-Z][a-z]>), substitute() modifies the string
384 within the C<SV> as according to the operation, returning the number of substitutions
387 int matches(SV *string, char *pattern, AV **matches);
389 Given an C<SV>, a pattern, and a pointer to an empty C<AV>,
390 matches() evaluates C<$string =~ $pattern> in a list context, and
391 fills in I<matches> with the array elements, returning the number of matches found.
393 Here's a sample program, I<match.c>, that uses all three (long lines have
399 static PerlInterpreter *my_perl;
401 /** my_eval_sv(code, error_check)
402 ** kinda like eval_sv(),
403 ** but we pop the return value off the stack
405 SV* my_eval_sv(SV *sv, I32 croak_on_error)
412 eval_sv(sv, G_SCALAR);
418 if (croak_on_error && SvTRUE(ERRSV))
419 croak(SvPVx(ERRSV, n_a));
424 /** match(string, pattern)
426 ** Used for matches in a scalar context.
428 ** Returns 1 if the match was successful; 0 otherwise.
431 I32 match(SV *string, char *pattern)
433 SV *command = NEWSV(1099, 0), *retval;
436 sv_setpvf(command, "my $string = '%s'; $string =~ %s",
437 SvPV(string,n_a), pattern);
439 retval = my_eval_sv(command, TRUE);
440 SvREFCNT_dec(command);
445 /** substitute(string, pattern)
447 ** Used for =~ operations that modify their left-hand side (s/// and tr///)
449 ** Returns the number of successful matches, and
450 ** modifies the input string if there were any.
453 I32 substitute(SV **string, char *pattern)
455 SV *command = NEWSV(1099, 0), *retval;
458 sv_setpvf(command, "$string = '%s'; ($string =~ %s)",
459 SvPV(*string,n_a), pattern);
461 retval = my_eval_sv(command, TRUE);
462 SvREFCNT_dec(command);
464 *string = get_sv("string", FALSE);
468 /** matches(string, pattern, matches)
470 ** Used for matches in a list context.
472 ** Returns the number of matches,
473 ** and fills in **matches with the matching substrings
476 I32 matches(SV *string, char *pattern, AV **match_list)
478 SV *command = NEWSV(1099, 0);
482 sv_setpvf(command, "my $string = '%s'; @array = ($string =~ %s)",
483 SvPV(string,n_a), pattern);
485 my_eval_sv(command, TRUE);
486 SvREFCNT_dec(command);
488 *match_list = get_av("array", FALSE);
489 num_matches = av_len(*match_list) + 1; /** assume $[ is 0 **/
494 main (int argc, char **argv, char **env)
496 char *embedding[] = { "", "-e", "0" };
502 PERL_SYS_INIT3(&argc,&argv,&env);
503 my_perl = perl_alloc();
504 perl_construct(my_perl);
505 perl_parse(my_perl, NULL, 3, embedding, NULL);
506 PL_exit_flags |= PERL_EXIT_DESTRUCT_END;
508 text = NEWSV(1099,0);
509 sv_setpv(text, "When he is at a convenience store and the "
510 "bill comes to some amount like 76 cents, Maynard is "
511 "aware that there is something he *should* do, something "
512 "that will enable him to get back a quarter, but he has "
513 "no idea *what*. He fumbles through his red squeezey "
514 "changepurse and gives the boy three extra pennies with "
515 "his dollar, hoping that he might luck into the correct "
516 "amount. The boy gives him back two of his own pennies "
517 "and then the big shiny quarter that is his prize. "
520 if (match(text, "m/quarter/")) /** Does text contain 'quarter'? **/
521 printf("match: Text contains the word 'quarter'.\n\n");
523 printf("match: Text doesn't contain the word 'quarter'.\n\n");
525 if (match(text, "m/eighth/")) /** Does text contain 'eighth'? **/
526 printf("match: Text contains the word 'eighth'.\n\n");
528 printf("match: Text doesn't contain the word 'eighth'.\n\n");
530 /** Match all occurrences of /wi../ **/
531 num_matches = matches(text, "m/(wi..)/g", &match_list);
532 printf("matches: m/(wi..)/g found %d matches...\n", num_matches);
534 for (i = 0; i < num_matches; i++)
535 printf("match: %s\n", SvPV(*av_fetch(match_list, i, FALSE),n_a));
538 /** Remove all vowels from text **/
539 num_matches = substitute(&text, "s/[aeiou]//gi");
541 printf("substitute: s/[aeiou]//gi...%d substitutions made.\n",
543 printf("Now text is: %s\n\n", SvPV(text,n_a));
546 /** Attempt a substitution **/
547 if (!substitute(&text, "s/Perl/C/")) {
548 printf("substitute: s/Perl/C...No substitution made.\n\n");
552 PL_perl_destruct_level = 1;
553 perl_destruct(my_perl);
558 which produces the output (again, long lines have been wrapped here)
560 match: Text contains the word 'quarter'.
562 match: Text doesn't contain the word 'eighth'.
564 matches: m/(wi..)/g found 2 matches...
568 substitute: s/[aeiou]//gi...139 substitutions made.
569 Now text is: Whn h s t cnvnnc str nd th bll cms t sm mnt lk 76 cnts,
570 Mynrd s wr tht thr s smthng h *shld* d, smthng tht wll nbl hm t gt bck
571 qrtr, bt h hs n d *wht*. H fmbls thrgh hs rd sqzy chngprs nd gvs th by
572 thr xtr pnns wth hs dllr, hpng tht h mght lck nt th crrct mnt. Th by gvs
573 hm bck tw f hs wn pnns nd thn th bg shny qrtr tht s hs prz. -RCHH
575 substitute: s/Perl/C...No substitution made.
577 =head2 Fiddling with the Perl stack from your C program
579 When trying to explain stacks, most computer science textbooks mumble
580 something about spring-loaded columns of cafeteria plates: the last
581 thing you pushed on the stack is the first thing you pop off. That'll
582 do for our purposes: your C program will push some arguments onto "the Perl
583 stack", shut its eyes while some magic happens, and then pop the
584 results--the return value of your Perl subroutine--off the stack.
586 First you'll need to know how to convert between C types and Perl
587 types, with newSViv() and sv_setnv() and newAV() and all their
588 friends. They're described in L<perlguts> and L<perlapi>.
590 Then you'll need to know how to manipulate the Perl stack. That's
591 described in L<perlcall>.
593 Once you've understood those, embedding Perl in C is easy.
595 Because C has no builtin function for integer exponentiation, let's
596 make Perl's ** operator available to it (this is less useful than it
597 sounds, because Perl implements ** with C's I<pow()> function). First
598 I'll create a stub exponentiation function in I<power.pl>:
605 Now I'll create a C program, I<power.c>, with a function
606 I<PerlPower()> that contains all the perlguts necessary to push the
607 two arguments into I<expo()> and to pop the return value out. Take a
613 static PerlInterpreter *my_perl;
616 PerlPower(int a, int b)
618 dSP; /* initialize stack pointer */
619 ENTER; /* everything created after here */
620 SAVETMPS; /* ...is a temporary variable. */
621 PUSHMARK(SP); /* remember the stack pointer */
622 XPUSHs(sv_2mortal(newSViv(a))); /* push the base onto the stack */
623 XPUSHs(sv_2mortal(newSViv(b))); /* push the exponent onto stack */
624 PUTBACK; /* make local stack pointer global */
625 call_pv("expo", G_SCALAR); /* call the function */
626 SPAGAIN; /* refresh stack pointer */
627 /* pop the return value from stack */
628 printf ("%d to the %dth power is %d.\n", a, b, POPi);
630 FREETMPS; /* free that return value */
631 LEAVE; /* ...and the XPUSHed "mortal" args.*/
634 int main (int argc, char **argv, char **env)
636 char *my_argv[] = { "", "power.pl" };
638 PERL_SYS_INIT3(&argc,&argv,&env);
639 my_perl = perl_alloc();
640 perl_construct( my_perl );
642 perl_parse(my_perl, NULL, 2, my_argv, (char **)NULL);
643 PL_exit_flags |= PERL_EXIT_DESTRUCT_END;
646 PerlPower(3, 4); /*** Compute 3 ** 4 ***/
648 perl_destruct(my_perl);
657 % cc -o power power.c `perl -MExtUtils::Embed -e ccopts -e ldopts`
660 3 to the 4th power is 81.
662 =head2 Maintaining a persistent interpreter
664 When developing interactive and/or potentially long-running
665 applications, it's a good idea to maintain a persistent interpreter
666 rather than allocating and constructing a new interpreter multiple
667 times. The major reason is speed: since Perl will only be loaded into
670 However, you have to be more cautious with namespace and variable
671 scoping when using a persistent interpreter. In previous examples
672 we've been using global variables in the default package C<main>. We
673 knew exactly what code would be run, and assumed we could avoid
674 variable collisions and outrageous symbol table growth.
676 Let's say your application is a server that will occasionally run Perl
677 code from some arbitrary file. Your server has no way of knowing what
678 code it's going to run. Very dangerous.
680 If the file is pulled in by C<perl_parse()>, compiled into a newly
681 constructed interpreter, and subsequently cleaned out with
682 C<perl_destruct()> afterwards, you're shielded from most namespace
685 One way to avoid namespace collisions in this scenario is to translate
686 the filename into a guaranteed-unique package name, and then compile
687 the code into that package using L<perlfunc/eval>. In the example
688 below, each file will only be compiled once. Or, the application
689 might choose to clean out the symbol table associated with the file
690 after it's no longer needed. Using L<perlapi/call_argv>, We'll
691 call the subroutine C<Embed::Persistent::eval_file> which lives in the
692 file C<persistent.pl> and pass the filename and boolean cleanup/cache
695 Note that the process will continue to grow for each file that it
696 uses. In addition, there might be C<AUTOLOAD>ed subroutines and other
697 conditions that cause Perl's symbol table to grow. You might want to
698 add some logic that keeps track of the process size, or restarts
699 itself after a certain number of requests, to ensure that memory
700 consumption is minimized. You'll also want to scope your variables
701 with L<perlfunc/my> whenever possible.
704 package Embed::Persistent;
709 use Symbol qw(delete_package);
711 sub valid_package_name {
713 $string =~ s/([^A-Za-z0-9\/])/sprintf("_%2x",unpack("C",$1))/eg;
714 # second pass only for words starting with a digit
715 $string =~ s|/(\d)|sprintf("/_%2x",unpack("C",$1))|eg;
717 # Dress it up as a real package name
719 return "Embed" . $string;
723 my($filename, $delete) = @_;
724 my $package = valid_package_name($filename);
725 my $mtime = -M $filename;
726 if(defined $Cache{$package}{mtime}
728 $Cache{$package}{mtime} <= $mtime)
730 # we have compiled this subroutine already,
731 # it has not been updated on disk, nothing left to do
732 print STDERR "already compiled $package->handler\n";
736 open FH, $filename or die "open '$filename' $!";
741 #wrap the code into a subroutine inside our unique package
742 my $eval = qq{package $package; sub handler { $sub; }};
744 # hide our variables within this block
745 my($filename,$mtime,$package,$sub);
750 #cache it unless we're cleaning out each time
751 $Cache{$package}{mtime} = $mtime unless $delete;
754 eval {$package->handler;};
757 delete_package($package) if $delete;
759 #take a look if you want
760 #print Devel::Symdump->rnew($package)->as_string, $/;
771 /* 1 = clean out filename's symbol table after each request, 0 = don't */
776 #define BUFFER_SIZE 1024
778 static PerlInterpreter *my_perl = NULL;
781 main(int argc, char **argv, char **env)
783 char *embedding[] = { "", "persistent.pl" };
784 char *args[] = { "", DO_CLEAN, NULL };
785 char filename[BUFFER_SIZE];
789 PERL_SYS_INIT3(&argc,&argv,&env);
790 if((my_perl = perl_alloc()) == NULL) {
791 fprintf(stderr, "no memory!");
794 perl_construct(my_perl);
796 exitstatus = perl_parse(my_perl, NULL, 2, embedding, NULL);
797 PL_exit_flags |= PERL_EXIT_DESTRUCT_END;
799 exitstatus = perl_run(my_perl);
801 while(printf("Enter file name: ") &&
802 fgets(filename, BUFFER_SIZE, stdin)) {
804 filename[strlen(filename)-1] = '\0'; /* strip \n */
805 /* call the subroutine, passing it the filename as an argument */
807 call_argv("Embed::Persistent::eval_file",
808 G_DISCARD | G_EVAL, args);
812 fprintf(stderr, "eval error: %s\n", SvPV(ERRSV,n_a));
816 PL_perl_destruct_level = 0;
817 perl_destruct(my_perl);
825 % cc -o persistent persistent.c `perl -MExtUtils::Embed -e ccopts -e ldopts`
827 Here's an example script file:
830 my $string = "hello";
834 print "foo says: @_\n";
840 Enter file name: test.pl
842 Enter file name: test.pl
843 already compiled Embed::test_2epl->handler
847 =head2 Execution of END blocks
849 Traditionally END blocks have been executed at the end of the perl_run.
850 This causes problems for applications that never call perl_run. Since
851 perl 5.7.2 you can specify C<PL_exit_flags |= PERL_EXIT_DESTRUCT_END>
852 to get the new behaviour. This also enables the running of END blocks if
853 the perl_parse fails and C<perl_destruct> will return the exit value.
855 =head2 Maintaining multiple interpreter instances
857 Some rare applications will need to create more than one interpreter
858 during a session. Such an application might sporadically decide to
859 release any resources associated with the interpreter.
861 The program must take care to ensure that this takes place I<before>
862 the next interpreter is constructed. By default, when perl is not
863 built with any special options, the global variable
864 C<PL_perl_destruct_level> is set to C<0>, since extra cleaning isn't
865 usually needed when a program only ever creates a single interpreter
866 in its entire lifetime.
868 Setting C<PL_perl_destruct_level> to C<1> makes everything squeaky clean:
872 /* reset global variables here with PL_perl_destruct_level = 1 */
873 PL_perl_destruct_level = 1;
874 perl_construct(my_perl);
876 /* clean and reset _everything_ during perl_destruct */
877 PL_perl_destruct_level = 1;
878 perl_destruct(my_perl);
881 /* let's go do it again! */
884 When I<perl_destruct()> is called, the interpreter's syntax parse tree
885 and symbol tables are cleaned up, and global variables are reset. The
886 second assignment to C<PL_perl_destruct_level> is needed because
887 perl_construct resets it to C<0>.
889 Now suppose we have more than one interpreter instance running at the
890 same time. This is feasible, but only if you used the Configure option
891 C<-Dusemultiplicity> or the options C<-Dusethreads -Duseithreads> when
892 building perl. By default, enabling one of these Configure options
893 sets the per-interpreter global variable C<PL_perl_destruct_level> to
894 C<1>, so that thorough cleaning is automatic and interpreter variables
895 are initialized correctly. Even if you don't intend to run two or
896 more interpreters at the same time, but to run them sequentially, like
897 in the above example, it is recommended to build perl with the
898 C<-Dusemultiplicity> option otherwise some interpreter variables may
899 not be initialized correctly between consecutive runs and your
900 application may crash.
902 Using C<-Dusethreads -Duseithreads> rather than C<-Dusemultiplicity>
903 is more appropriate if you intend to run multiple interpreters
904 concurrently in different threads, because it enables support for
905 linking in the thread libraries of your system with the interpreter.
913 /* we're going to embed two interpreters */
914 /* we're going to embed two interpreters */
916 #define SAY_HELLO "-e", "print qq(Hi, I'm $^X\n)"
918 int main(int argc, char **argv, char **env)
920 PerlInterpreter *one_perl, *two_perl;
921 char *one_args[] = { "one_perl", SAY_HELLO };
922 char *two_args[] = { "two_perl", SAY_HELLO };
924 PERL_SYS_INIT3(&argc,&argv,&env);
925 one_perl = perl_alloc();
926 two_perl = perl_alloc();
928 PERL_SET_CONTEXT(one_perl);
929 perl_construct(one_perl);
930 PERL_SET_CONTEXT(two_perl);
931 perl_construct(two_perl);
933 PERL_SET_CONTEXT(one_perl);
934 perl_parse(one_perl, NULL, 3, one_args, (char **)NULL);
935 PERL_SET_CONTEXT(two_perl);
936 perl_parse(two_perl, NULL, 3, two_args, (char **)NULL);
938 PERL_SET_CONTEXT(one_perl);
940 PERL_SET_CONTEXT(two_perl);
943 PERL_SET_CONTEXT(one_perl);
944 perl_destruct(one_perl);
945 PERL_SET_CONTEXT(two_perl);
946 perl_destruct(two_perl);
948 PERL_SET_CONTEXT(one_perl);
950 PERL_SET_CONTEXT(two_perl);
955 Note the calls to PERL_SET_CONTEXT(). These are necessary to initialize
956 the global state that tracks which interpreter is the "current" one on
957 the particular process or thread that may be running it. It should
958 always be used if you have more than one interpreter and are making
959 perl API calls on both interpreters in an interleaved fashion.
961 PERL_SET_CONTEXT(interp) should also be called whenever C<interp> is
962 used by a thread that did not create it (using either perl_alloc(), or
963 the more esoteric perl_clone()).
967 % cc -o multiplicity multiplicity.c `perl -MExtUtils::Embed -e ccopts -e ldopts`
975 =head2 Using Perl modules, which themselves use C libraries, from your C program
977 If you've played with the examples above and tried to embed a script
978 that I<use()>s a Perl module (such as I<Socket>) which itself uses a C or C++ library,
979 this probably happened:
982 Can't load module Socket, dynamic loading not available in this perl.
983 (You may need to build a new perl executable which either supports
984 dynamic loading or has the Socket module statically linked into it.)
989 Your interpreter doesn't know how to communicate with these extensions
990 on its own. A little glue will help. Up until now you've been
991 calling I<perl_parse()>, handing it NULL for the second argument:
993 perl_parse(my_perl, NULL, argc, my_argv, NULL);
995 That's where the glue code can be inserted to create the initial contact between
996 Perl and linked C/C++ routines. Let's take a look some pieces of I<perlmain.c>
997 to see how Perl does this:
999 static void xs_init (pTHX);
1001 EXTERN_C void boot_DynaLoader (pTHX_ CV* cv);
1002 EXTERN_C void boot_Socket (pTHX_ CV* cv);
1008 char *file = __FILE__;
1009 /* DynaLoader is a special case */
1010 newXS("DynaLoader::boot_DynaLoader", boot_DynaLoader, file);
1011 newXS("Socket::bootstrap", boot_Socket, file);
1014 Simply put: for each extension linked with your Perl executable
1015 (determined during its initial configuration on your
1016 computer or when adding a new extension),
1017 a Perl subroutine is created to incorporate the extension's
1018 routines. Normally, that subroutine is named
1019 I<Module::bootstrap()> and is invoked when you say I<use Module>. In
1020 turn, this hooks into an XSUB, I<boot_Module>, which creates a Perl
1021 counterpart for each of the extension's XSUBs. Don't worry about this
1022 part; leave that to the I<xsubpp> and extension authors. If your
1023 extension is dynamically loaded, DynaLoader creates I<Module::bootstrap()>
1024 for you on the fly. In fact, if you have a working DynaLoader then there
1025 is rarely any need to link in any other extensions statically.
1028 Once you have this code, slap it into the second argument of I<perl_parse()>:
1031 perl_parse(my_perl, xs_init, argc, my_argv, NULL);
1036 % cc -o interp interp.c `perl -MExtUtils::Embed -e ccopts -e ldopts`
1040 use SomeDynamicallyLoadedModule;
1042 print "Now I can use extensions!\n"'
1044 B<ExtUtils::Embed> can also automate writing the I<xs_init> glue code.
1046 % perl -MExtUtils::Embed -e xsinit -- -o perlxsi.c
1047 % cc -c perlxsi.c `perl -MExtUtils::Embed -e ccopts`
1048 % cc -c interp.c `perl -MExtUtils::Embed -e ccopts`
1049 % cc -o interp perlxsi.o interp.o `perl -MExtUtils::Embed -e ldopts`
1051 Consult L<perlxs>, L<perlguts>, and L<perlapi> for more details.
1053 =head1 Embedding Perl under Win32
1055 In general, all of the source code shown here should work unmodified under
1058 However, there are some caveats about the command-line examples shown.
1059 For starters, backticks won't work under the Win32 native command shell.
1060 The ExtUtils::Embed kit on CPAN ships with a script called
1061 B<genmake>, which generates a simple makefile to build a program from
1062 a single C source file. It can be used like this:
1064 C:\ExtUtils-Embed\eg> perl genmake interp.c
1065 C:\ExtUtils-Embed\eg> nmake
1066 C:\ExtUtils-Embed\eg> interp -e "print qq{I'm embedded in Win32!\n}"
1068 You may wish to use a more robust environment such as the Microsoft
1069 Developer Studio. In this case, run this to generate perlxsi.c:
1071 perl -MExtUtils::Embed -e xsinit
1073 Create a new project and Insert -> Files into Project: perlxsi.c,
1074 perl.lib, and your own source files, e.g. interp.c. Typically you'll
1075 find perl.lib in B<C:\perl\lib\CORE>, if not, you should see the
1076 B<CORE> directory relative to C<perl -V:archlib>. The studio will
1077 also need this path so it knows where to find Perl include files.
1078 This path can be added via the Tools -> Options -> Directories menu.
1079 Finally, select Build -> Build interp.exe and you're ready to go.
1083 If you completely hide the short forms forms of the Perl public API,
1084 add -DPERL_NO_SHORT_NAMES to the compilation flags. This means that
1085 for example instead of writing
1087 warn("%d bottles of beer on the wall", bottlecount);
1089 you will have to write the explicit full form
1091 Perl_warn(aTHX_ "%d bottles of beer on the wall", bottlecount);
1093 (See L<perlguts/Background and PERL_IMPLICIT_CONTEXT for the explanation
1094 of the C<aTHX_>.> ) Hiding the short forms is very useful for avoiding
1095 all sorts of nasty (C preprocessor or otherwise) conflicts with other
1096 software packages (Perl defines about 2400 APIs with these short names,
1097 take or leave few hundred, so there certainly is room for conflict.)
1101 You can sometimes I<write faster code> in C, but
1102 you can always I<write code faster> in Perl. Because you can use
1103 each from the other, combine them as you wish.
1108 Jon Orwant <F<orwant@media.mit.edu>> and Doug MacEachern
1109 <F<dougm@covalent.net>>, with small contributions from Tim Bunce, Tom
1110 Christiansen, Guy Decoux, Hallvard Furuseth, Dov Grobgeld, and Ilya
1113 Doug MacEachern has an article on embedding in Volume 1, Issue 4 of
1114 The Perl Journal ( http://www.tpj.com/ ). Doug is also the developer of the
1115 most widely-used Perl embedding: the mod_perl system
1116 (perl.apache.org), which embeds Perl in the Apache web server.
1117 Oracle, Binary Evolution, ActiveState, and Ben Sugars's nsapi_perl
1118 have used this model for Oracle, Netscape and Internet Information
1119 Server Perl plugins.
1125 Copyright (C) 1995, 1996, 1997, 1998 Doug MacEachern and Jon Orwant. All
1128 Permission is granted to make and distribute verbatim copies of this
1129 documentation provided the copyright notice and this permission notice are
1130 preserved on all copies.
1132 Permission is granted to copy and distribute modified versions of this
1133 documentation under the conditions for verbatim copying, provided also
1134 that they are marked clearly as modified versions, that the authors'
1135 names and title are unchanged (though subtitles and additional
1136 authors' names may be added), and that the entire resulting derived
1137 work is distributed under the terms of a permission notice identical
1140 Permission is granted to copy and distribute translations of this
1141 documentation into another language, under the above conditions for