3 perltodo - Perl TO-DO List
7 This is a list of wishes for Perl. The tasks we think are smaller or
8 easier are listed first. Anyone is welcome to work on any of these,
9 but it's a good idea to first contact I<perl5-porters@perl.org> to
10 avoid duplication of effort, and to learn from any previous attempts.
11 By all means contact a pumpking privately first if you prefer.
13 Whilst patches to make the list shorter are most welcome, ideas to add to
14 the list are also encouraged. Check the perl5-porters archives for past
15 ideas, and any discussion about them. One set of archives may be found at:
17 http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/
19 What can we offer you in return? Fame, fortune, and everlasting glory? Maybe
20 not, but if your patch is incorporated, then we'll add your name to the
21 F<AUTHORS> file, which ships in the official distribution. How many other
22 programming languages offer you 1 line of immortality?
24 =head1 Tasks that only need Perl knowledge
26 =head2 Remove duplication of test setup.
28 Schwern notes, that there's duplication of code - lots and lots of tests have
29 some variation on the big block of C<$Is_Foo> checks. We can safely put this
30 into a file, change it to build an C<%Is> hash and require it. Maybe just put
31 it into F<test.pl>. Throw in the handy tainting subroutines.
33 =head2 POD -E<gt> HTML conversion in the core still sucks
35 Which is crazy given just how simple POD purports to be, and how simple HTML
36 can be. It's not actually I<as> simple as it sounds, particularly with the
37 flexibility POD allows for C<=item>, but it would be good to improve the
38 visual appeal of the HTML generated, and to avoid it having any validation
39 errors. See also L</make HTML install work>, as the layout of installation tree
40 is needed to improve the cross-linking.
42 The addition of C<Pod::Simple> and its related modules may make this task
45 =head2 merge checkpods and podchecker
47 F<pod/checkpods.PL> (and C<make check> in the F<pod/> subdirectory)
48 implements a very basic check for pod files, but the errors it discovers
49 aren't found by podchecker. Add this check to podchecker, get rid of
50 checkpods and have C<make check> use podchecker.
52 =head2 Parallel testing
54 (This probably impacts much more than the core: also the Test::Harness
55 and TAP::* modules on CPAN.)
57 All of the tests in F<t/> can now be run in parallel, if C<$ENV{TEST_JOBS}>
58 is set. However, tests within each directory in F<ext> and F<lib> are still
59 run in series, with directories run in parallel. This is an adequate
60 heuristic, but it might be possible to relax it further, and get more
61 throughput. Specifically, it would be good to audit all of F<lib/*.t>, and
62 make them use C<File::Temp>.
64 =head2 Make Schwern poorer
66 We should have tests for everything. When all the core's modules are tested,
67 Schwern has promised to donate to $500 to TPF. We may need volunteers to
68 hold him upside down and shake vigorously in order to actually extract the
71 =head2 Improve the coverage of the core tests
73 Use Devel::Cover to ascertain the core modules's test coverage, then add
74 tests that are currently missing.
78 A full test suite for the B module would be nice.
80 =head2 Deparse inlined constants
87 will currently deparse as
89 use constant ('PI', 4);
92 because the tokenizer inlines the value of the constant subroutine C<PI>.
93 This allows various compile time optimisations, such as constant folding
94 and dead code elimination. Where these haven't happened (such as the example
95 above) it ought be possible to make B::Deparse work out the name of the
96 original constant, because just enough information survives in the symbol
97 table to do this. Specifically, the same scalar is used for the constant in
98 the optree as is used for the constant subroutine, so by iterating over all
99 symbol tables and generating a mapping of SV address to constant name, it
100 would be possible to provide B::Deparse with this functionality.
102 =head2 A decent benchmark
104 C<perlbench> seems impervious to any recent changes made to the perl core. It
105 would be useful to have a reasonable general benchmarking suite that roughly
106 represented what current perl programs do, and measurably reported whether
107 tweaks to the core improve, degrade or don't really affect performance, to
108 guide people attempting to optimise the guts of perl. Gisle would welcome
109 new tests for perlbench.
111 =head2 fix tainting bugs
113 Fix the bugs revealed by running the test suite with the C<-t> switch (via
114 C<make test.taintwarn>).
116 =head2 Dual life everything
118 As part of the "dists" plan, anything that doesn't belong in the smallest perl
119 distribution needs to be dual lifed. Anything else can be too. Figure out what
120 changes would be needed to package that module and its tests up for CPAN, and
121 do so. Test it with older perl releases, and fix the problems you find.
123 To make a minimal perl distribution, it's useful to look at
124 F<t/lib/commonsense.t>.
126 =head2 Bundle dual life modules in ext/
128 For maintenance (and branch merging) reasons, it would be useful to move
129 some architecture-independent dual-life modules from lib/ to ext/, if this
130 has no negative impact on the build of perl itself.
132 =head2 POSIX memory footprint
134 Ilya observed that use POSIX; eats memory like there's no tomorrow, and at
135 various times worked to cut it down. There is probably still fat to cut out -
136 for example POSIX passes Exporter some very memory hungry data structures.
138 =head2 embed.pl/makedef.pl
140 There is a script F<embed.pl> that generates several header files to prefix
141 all of Perl's symbols in a consistent way, to provide some semblance of
142 namespace support in C<C>. Functions are declared in F<embed.fnc>, variables
143 in F<interpvar.h>. Quite a few of the functions and variables
144 are conditionally declared there, using C<#ifdef>. However, F<embed.pl>
145 doesn't understand the C macros, so the rules about which symbols are present
146 when is duplicated in F<makedef.pl>. Writing things twice is bad, m'kay.
147 It would be good to teach C<embed.pl> to understand the conditional
148 compilation, and hence remove the duplication, and the mistakes it has caused.
150 =head2 use strict; and AutoLoad
152 Currently if you write
155 use AutoLoader 'AUTOLOAD';
160 print join (' ', No, strict, here), "!\n";
163 then C<use strict;> isn't in force within the autoloaded subroutines. It would
164 be more consistent (and less surprising) to arrange for all lexical pragmas
165 in force at the __END__ block to be in force within each autoloaded subroutine.
167 There's a similar problem with SelfLoader.
169 =head2 profile installman
171 The F<installman> script is slow. All it is doing text processing, which we're
172 told is something Perl is good at. So it would be nice to know what it is doing
173 that is taking so much CPU, and where possible address it.
176 =head1 Tasks that need a little sysadmin-type knowledge
178 Or if you prefer, tasks that you would learn from, and broaden your skills
181 =head2 make HTML install work
183 There is an C<installhtml> target in the Makefile. It's marked as
184 "experimental". It would be good to get this tested, make it work reliably, and
185 remove the "experimental" tag. This would include
191 Checking that cross linking between various parts of the documentation works.
192 In particular that links work between the modules (files with POD in F<lib/>)
193 and the core documentation (files in F<pod/>)
197 Work out how to split C<perlfunc> into chunks, preferably one per function
198 group, preferably with general case code that could be used elsewhere.
199 Challenges here are correctly identifying the groups of functions that go
200 together, and making the right named external cross-links point to the right
201 page. Things to be aware of are C<-X>, groups such as C<getpwnam> to
202 C<endservent>, two or more C<=items> giving the different parameter lists, such
205 =item substr EXPR,OFFSET,LENGTH,REPLACEMENT
206 =item substr EXPR,OFFSET,LENGTH
207 =item substr EXPR,OFFSET
209 and different parameter lists having different meanings. (eg C<select>)
213 =head2 compressed man pages
215 Be able to install them. This would probably need a configure test to see how
216 the system does compressed man pages (same directory/different directory?
217 same filename/different filename), as well as tweaking the F<installman> script
218 to compress as necessary.
220 =head2 Add a code coverage target to the Makefile
222 Make it easy for anyone to run Devel::Cover on the core's tests. The steps
223 to do this manually are roughly
229 do a normal C<Configure>, but include Devel::Cover as a module to install
230 (see F<INSTALL> for how to do this)
238 cd t; HARNESS_PERL_SWITCHES=-MDevel::Cover ./perl -I../lib harness
242 Process the resulting Devel::Cover database
246 This just give you the coverage of the F<.pm>s. To also get the C level
253 Additionally tell C<Configure> to use the appropriate C compiler flags for
260 (instead of C<make perl>)
264 After running the tests run C<gcov> to generate all the F<.gcov> files.
265 (Including down in the subdirectories of F<ext/>
269 (From the top level perl directory) run C<gcov2perl> on all the C<.gcov> files
270 to get their stats into the cover_db directory.
274 Then process the Devel::Cover database
278 It would be good to add a single switch to C<Configure> to specify that you
279 wanted to perform perl level coverage, and another to specify C level
280 coverage, and have C<Configure> and the F<Makefile> do all the right things
283 =head2 Make Config.pm cope with differences between built and installed perl
285 Quite often vendors ship a perl binary compiled with their (pay-for)
286 compilers. People install a free compiler, such as gcc. To work out how to
287 build extensions, Perl interrogates C<%Config>, so in this situation
288 C<%Config> describes compilers that aren't there, and extension building
289 fails. This forces people into choosing between re-compiling perl themselves
290 using the compiler they have, or only using modules that the vendor ships.
292 It would be good to find a way teach C<Config.pm> about the installation setup,
293 possibly involving probing at install time or later, so that the C<%Config> in
294 a binary distribution better describes the installed machine, when the
295 installed machine differs from the build machine in some significant way.
297 =head2 linker specification files
299 Some platforms mandate that you provide a list of a shared library's external
300 symbols to the linker, so the core already has the infrastructure in place to
301 do this for generating shared perl libraries. My understanding is that the
302 GNU toolchain can accept an optional linker specification file, and restrict
303 visibility just to symbols declared in that file. It would be good to extend
304 F<makedef.pl> to support this format, and to provide a means within
305 C<Configure> to enable it. This would allow Unix users to test that the
306 export list is correct, and to build a perl that does not pollute the global
307 namespace with private symbols.
309 =head2 Cross-compile support
311 Currently C<Configure> understands C<-Dusecrosscompile> option. This option
312 arranges for building C<miniperl> for TARGET machine, so this C<miniperl> is
313 assumed then to be copied to TARGET machine and used as a replacement of full
316 This could be done little differently. Namely C<miniperl> should be built for
317 HOST and then full C<perl> with extensions should be compiled for TARGET.
318 This, however, might require extra trickery for %Config: we have one config
319 first for HOST and then another for TARGET. Tools like MakeMaker will be
320 mightily confused. Having around two different types of executables and
321 libraries (HOST and TARGET) makes life interesting for Makefiles and
322 shell (and Perl) scripts. There is $Config{run}, normally empty, which
323 can be used as an execution wrapper. Also note that in some
324 cross-compilation/execution environments the HOST and the TARGET do
325 not see the same filesystem(s), the $Config{run} may need to do some
326 file/directory copying back and forth.
330 Make F<pod/roffitall> be updated by F<pod/buildtoc>.
332 =head1 Tasks that need a little C knowledge
334 These tasks would need a little C knowledge, but don't need any specific
335 background or experience with XS, or how the Perl interpreter works
337 =head2 Weed out needless PERL_UNUSED_ARG
339 The C code uses the macro C<PERL_UNUSED_ARG> to stop compilers warning about
340 unused arguments. Often the arguments can't be removed, as there is an
341 external constraint that determines the prototype of the function, so this
342 approach is valid. However, there are some cases where C<PERL_UNUSED_ARG>
343 could be removed. Specifically
349 The prototypes of (nearly all) static functions can be changed
353 Unused arguments generated by short cut macros are wasteful - the short cut
354 macro used can be changed.
358 =head2 Modernize the order of directories in @INC
360 The way @INC is laid out by default, one cannot upgrade core (dual-life)
361 modules without overwriting files. This causes problems for binary
362 package builders. One possible proposal is laid out in this
364 L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2002-04/msg02380.html>.
368 Natively 64-bit systems need neither -Duse64bitint nor -Duse64bitall.
369 On these systems, it might be the default compilation mode, and there
370 is currently no guarantee that passing no use64bitall option to the
371 Configure process will build a 32bit perl. Implementing -Duse32bit*
372 options would be nice for perl 5.12.
374 =head2 Make it clear from -v if this is the exact official release
376 Currently perl from C<p4>/C<rsync> ships with a F<patchlevel.h> file that
377 usually defines one local patch, of the form "MAINT12345" or "RC1". The output
378 of perl -v doesn't report that a perl isn't an official release, and this
379 information can get lost in bugs reports. Because of this, the minor version
380 isn't bumped up until RC time, to minimise the possibility of versions of perl
381 escaping that believe themselves to be newer than they actually are.
383 It would be useful to find an elegant way to have the "this is an interim
384 maintenance release" or "this is a release candidate" in the terse -v output,
385 and have it so that it's easy for the pumpking to remove this just as the
386 release tarball is rolled up. This way the version pulled out of rsync would
387 always say "I'm a development release" and it would be safe to bump the
388 reported minor version as soon as a release ships, which would aid perl
391 This task is really about thinking of an elegant way to arrange the C source
392 such that it's trivial for the Pumpking to flag "this is an official release"
393 when making a tarball, yet leave the default source saying "I'm not the
396 =head2 Profile Perl - am I hot or not?
398 The Perl source code is stable enough that it makes sense to profile it,
399 identify and optimise the hotspots. It would be good to measure the
400 performance of the Perl interpreter using free tools such as cachegrind,
401 gprof, and dtrace, and work to reduce the bottlenecks they reveal.
403 As part of this, the idea of F<pp_hot.c> is that it contains the I<hot> ops,
404 the ops that are most commonly used. The idea is that by grouping them, their
405 object code will be adjacent in the executable, so they have a greater chance
406 of already being in the CPU cache (or swapped in) due to being near another op
409 Except that it's not clear if these really are the most commonly used ops. So
410 as part of exercising your skills with coverage and profiling tools you might
411 want to determine what ops I<really> are the most commonly used. And in turn
412 suggest evictions and promotions to achieve a better F<pp_hot.c>.
414 One piece of Perl code that might make a good testbed is F<installman>.
416 =head2 Allocate OPs from arenas
418 Currently all new OP structures are individually malloc()ed and free()d.
419 All C<malloc> implementations have space overheads, and are now as fast as
420 custom allocates so it would both use less memory and less CPU to allocate
421 the various OP structures from arenas. The SV arena code can probably be
424 Note that Configuring perl with C<-Accflags=-DPL_OP_SLAB_ALLOC> will use
425 Perl_Slab_alloc() to pack optrees into a contiguous block, which is
426 probably superior to the use of OP arenas, esp. from a cache locality
427 standpoint. See L<Profile Perl - am I hot or not?>.
429 =head2 Improve win32/wince.c
431 Currently, numerous functions look virtually, if not completely,
432 identical in both C<win32/wince.c> and C<win32/win32.c> files, which can't
435 =head2 Use secure CRT functions when building with VC8 on Win32
437 Visual C++ 2005 (VC++ 8.x) deprecated a number of CRT functions on the basis
438 that they were "unsafe" and introduced differently named secure versions of
439 them as replacements, e.g. instead of writing
441 FILE* f = fopen(__FILE__, "r");
446 errno_t err = fopen_s(&f, __FILE__, "r");
448 Currently, the warnings about these deprecations have been disabled by adding
449 -D_CRT_SECURE_NO_DEPRECATE to the CFLAGS. It would be nice to remove that
450 warning suppressant and actually make use of the new secure CRT functions.
452 There is also a similar issue with POSIX CRT function names like fileno having
453 been deprecated in favour of ISO C++ conformant names like _fileno. These
454 warnings are also currently suppressed by adding -D_CRT_NONSTDC_NO_DEPRECATE. It
455 might be nice to do as Microsoft suggest here too, although, unlike the secure
456 functions issue, there is presumably little or no benefit in this case.
458 =head2 Fix POSIX::access() and chdir() on Win32
460 These functions currently take no account of DACLs and therefore do not behave
461 correctly in situations where access is restricted by DACLs (as opposed to the
462 read-only attribute).
464 Furthermore, POSIX::access() behaves differently for directories having the
465 read-only attribute set depending on what CRT library is being used. For
466 example, the _access() function in the VC6 and VC7 CRTs (wrongly) claim that
467 such directories are not writable, whereas in fact all directories are writable
468 unless access is denied by DACLs. (In the case of directories, the read-only
469 attribute actually only means that the directory cannot be deleted.) This CRT
470 bug is fixed in the VC8 and VC9 CRTs (but, of course, the directory may still
471 not actually be writable if access is indeed denied by DACLs).
473 For the chdir() issue, see ActiveState bug #74552:
474 http://bugs.activestate.com/show_bug.cgi?id=74552
476 Therefore, DACLs should be checked both for consistency across CRTs and for
479 (Note that perl's -w operator should not be modified to check DACLs. It has
480 been written so that it reflects the state of the read-only attribute, even
481 for directories (whatever CRT is being used), for symmetry with chmod().)
483 =head2 strcat(), strcpy(), strncat(), strncpy(), sprintf(), vsprintf()
485 Maybe create a utility that checks after each libperl.a creation that
486 none of the above (nor sprintf(), vsprintf(), or *SHUDDER* gets())
487 ever creep back to libperl.a.
489 nm libperl.a | ./miniperl -alne '$o = $F[0] if /:$/; print "$o $F[1]" if $F[0] eq "U" && $F[1] =~ /^(?:strn?c(?:at|py)|v?sprintf|gets)$/'
491 Note, of course, that this will only tell whether B<your> platform
492 is using those naughty interfaces.
494 =head2 -D_FORTIFY_SOURCE=2, -fstack-protector
496 Recent glibcs support C<-D_FORTIFY_SOURCE=2> and recent gcc
497 (4.1 onwards?) supports C<-fstack-protector>, both of which give
498 protection against various kinds of buffer overflow problems.
499 These should probably be used for compiling Perl whenever available,
500 Configure and/or hints files should be adjusted to probe for the
501 availability of these features and enable them as appropriate.
503 =head2 Arenas for GPs? For MAGIC?
505 C<struct gp> and C<struct magic> are both currently allocated by C<malloc>.
506 It might be a speed or memory saving to change to using arenas. Or it might
507 not. It would need some suitable benchmarking first. In particular, C<GP>s
508 can probably be changed with minimal compatibility impact (probably nothing
509 outside of the core, or even outside of F<gv.c> allocates them), but they
510 probably aren't allocated/deallocated often enough for a speed saving. Whereas
511 C<MAGIC> is allocated/deallocated more often, but in turn, is also something
512 more externally visible, so changing the rules here may bite external code.
516 Several SV body structs are now the same size, notably PVMG and PVGV, PVAV and
517 PVHV, and PVCV and PVFM. It should be possible to allocate and return same
518 sized bodies from the same actual arena, rather than maintaining one arena for
519 each. This could save 4-6K per thread, of memory no longer tied up in the
520 not-yet-allocated part of an arena.
523 =head1 Tasks that need a knowledge of XS
525 These tasks would need C knowledge, and roughly the level of knowledge of
526 the perl API that comes from writing modules that use XS to interface to
529 =head2 safely supporting POSIX SA_SIGINFO
531 Some years ago Jarkko supplied patches to provide support for the POSIX
532 SA_SIGINFO feature in Perl, passing the extra data to the Perl signal handler.
534 Unfortunately, it only works with "unsafe" signals, because under safe
535 signals, by the time Perl gets to run the signal handler, the extra
536 information has been lost. Moreover, it's not easy to store it somewhere,
537 as you can't call mutexs, or do anything else fancy, from inside a signal
540 So it strikes me that we could provide safe SA_SIGINFO support
546 Provide global variables for two file descriptors
550 When the first request is made via C<sigaction> for C<SA_SIGINFO>, create a
551 pipe, store the reader in one, the writer in the other
555 In the "safe" signal handler (C<Perl_csighandler()>/C<S_raise_signal()>), if
556 the C<siginfo_t> pointer non-C<NULL>, and the writer file handle is open,
562 serialise signal number, C<struct siginfo_t> (or at least the parts we care
563 about) into a small auto char buff
567 C<write()> that (non-blocking) to the writer fd
573 if it writes 100%, flag the signal in a counter of "signals on the pipe" akin
574 to the current per-signal-number counts
578 if it writes 0%, assume the pipe is full. Flag the data as lost?
582 if it writes partially, croak a panic, as your OS is broken.
590 in the regular C<PERL_ASYNC_CHECK()> processing, if there are "signals on
591 the pipe", read the data out, deserialise, build the Perl structures on
592 the stack (code in C<Perl_sighandler()>, the "unsafe" handler), and call as
597 I think that this gets us decent C<SA_SIGINFO> support, without the current risk
598 of running Perl code inside the signal handler context. (With all the dangers
599 of things like C<malloc> corruption that that currently offers us)
601 For more information see the thread starting with this message:
602 http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2008-03/msg00305.html
604 =head2 autovivification
606 Make all autovivification consistent w.r.t LVALUE/RVALUE and strict/no strict;
608 This task is incremental - even a little bit of work on it will help.
610 =head2 Unicode in Filenames
612 chdir, chmod, chown, chroot, exec, glob, link, lstat, mkdir, open,
613 opendir, qx, readdir, readlink, rename, rmdir, stat, symlink, sysopen,
614 system, truncate, unlink, utime, -X. All these could potentially accept
615 Unicode filenames either as input or output (and in the case of system
616 and qx Unicode in general, as input or output to/from the shell).
617 Whether a filesystem - an operating system pair understands Unicode in
620 Known combinations that have some level of understanding include
621 Microsoft NTFS, Apple HFS+ (In Mac OS 9 and X) and Apple UFS (in Mac
622 OS X), NFS v4 is rumored to be Unicode, and of course Plan 9. How to
623 create Unicode filenames, what forms of Unicode are accepted and used
624 (UCS-2, UTF-16, UTF-8), what (if any) is the normalization form used,
625 and so on, varies. Finding the right level of interfacing to Perl
626 requires some thought. Remember that an OS does not implicate a
629 (The Windows -C command flag "wide API support" has been at least
630 temporarily retired in 5.8.1, and the -C has been repurposed, see
633 Most probably the right way to do this would be this:
634 L</"Virtualize operating system access">.
636 =head2 Unicode in %ENV
638 Currently the %ENV entries are always byte strings.
639 See L</"Virtualize operating system access">.
641 =head2 Unicode and glob()
643 Currently glob patterns and filenames returned from File::Glob::glob()
644 are always byte strings. See L</"Virtualize operating system access">.
646 =head2 Unicode and lc/uc operators
648 Some built-in operators (C<lc>, C<uc>, etc.) behave differently, based on
649 what the internal encoding of their argument is. That should not be the
650 case. Maybe add a pragma to switch behaviour.
652 =head2 use less 'memory'
654 Investigate trade offs to switch out perl's choices on memory usage.
655 Particularly perl should be able to give memory back.
657 This task is incremental - even a little bit of work on it will help.
659 =head2 Re-implement C<:unique> in a way that is actually thread-safe
661 The old implementation made bad assumptions on several levels. A good 90%
662 solution might be just to make C<:unique> work to share the string buffer
663 of SvPVs. That way large constant strings can be shared between ithreads,
664 such as the configuration information in F<Config>.
666 =head2 Make tainting consistent
668 Tainting would be easier to use if it didn't take documented shortcuts and
669 allow taint to "leak" everywhere within an expression.
671 =head2 readpipe(LIST)
673 system() accepts a LIST syntax (and a PROGRAM LIST syntax) to avoid
674 running a shell. readpipe() (the function behind qx//) could be similarly
677 =head2 Audit the code for destruction ordering assumptions
681 /* Need to check SvMAGICAL, as during global destruction it may be that
682 AvARYLEN(av) has been freed before av, and hence the SvANY() pointer
683 is now part of the linked list of SV heads, rather than pointing to
684 the original body. */
685 /* FIXME - audit the code for other bugs like this one. */
687 adding the C<SvMAGICAL> check to
689 if (AvARYLEN(av) && SvMAGICAL(AvARYLEN(av))) {
690 MAGIC *mg = mg_find (AvARYLEN(av), PERL_MAGIC_arylen);
692 Go through the core and look for similar assumptions that SVs have particular
693 types, as all bets are off during global destruction.
695 =head2 Extend PerlIO and PerlIO::Scalar
697 PerlIO::Scalar doesn't know how to truncate(). Implementing this
698 would require extending the PerlIO vtable.
700 Similarly the PerlIO vtable doesn't know about formats (write()), or
701 about stat(), or chmod()/chown(), utime(), or flock().
703 (For PerlIO::Scalar it's hard to see what e.g. mode bits or ownership
706 PerlIO doesn't do directories or symlinks, either: mkdir(), rmdir(),
707 opendir(), closedir(), seekdir(), rewinddir(), glob(); symlink(),
710 See also L</"Virtualize operating system access">.
712 =head2 -C on the #! line
714 It should be possible to make -C work correctly if found on the #! line,
715 given that all perl command line options are strict ASCII, and -C changes
716 only the interpretation of non-ASCII characters, and not for the script file
717 handle. To make it work needs some investigation of the ordering of function
718 calls during startup, and (by implication) a bit of tweaking of that order.
720 =head2 Organize error messages
722 Perl's diagnostics (error messages, see L<perldiag>) could use
723 reorganizing and formalizing so that each error message has its
724 stable-for-all-eternity unique id, categorized by severity, type, and
725 subsystem. (The error messages would be listed in a datafile outside
726 of the Perl source code, and the source code would only refer to the
727 messages by the id.) This clean-up and regularizing should apply
728 for all croak() messages.
730 This would enable all sorts of things: easier translation/localization
731 of the messages (though please do keep in mind the caveats of
732 L<Locale::Maketext> about too straightforward approaches to
733 translation), filtering by severity, and instead of grepping for a
734 particular error message one could look for a stable error id. (Of
735 course, changing the error messages by default would break all the
736 existing software depending on some particular error message...)
738 This kind of functionality is known as I<message catalogs>. Look for
739 inspiration for example in the catgets() system, possibly even use it
740 if available-- but B<only> if available, all platforms will B<not>
743 For the really pure at heart, consider extending this item to cover
744 also the warning messages (see L<perllexwarn>, C<warnings.pl>).
746 =head1 Tasks that need a knowledge of the interpreter
748 These tasks would need C knowledge, and knowledge of how the interpreter works,
749 or a willingness to learn.
751 =head2 lexicals used only once
755 $ perl -we '$pie = 42'
756 Name "main::pie" used only once: possible typo at -e line 1.
760 $ perl -we 'my $pie = 42'
762 Logically all lexicals used only once should warn, if the user asks for
763 warnings. An unworked RT ticket (#5087) has been open for almost seven
764 years for this discrepancy.
768 The handling of Unicode is unclean in many places. For example, the regexp
769 engine matches in Unicode semantics whenever the string or the pattern is
770 flagged as UTF-8, but that should not be dependent on an internal storage
771 detail of the string. Likewise, case folding behaviour is dependent on the
772 UTF8 internal flag being on or off.
774 =head2 Properly Unicode safe tokeniser and pads.
776 The tokeniser isn't actually very UTF-8 clean. C<use utf8;> is a hack -
777 variable names are stored in stashes as raw bytes, without the utf-8 flag
778 set. The pad API only takes a C<char *> pointer, so that's all bytes too. The
779 tokeniser ignores the UTF-8-ness of C<PL_rsfp>, or any SVs returned from
780 source filters. All this could be fixed.
782 =head2 state variable initialization in list context
784 Currently this is illegal:
786 state ($a, $b) = foo();
788 In Perl 6, C<state ($a) = foo();> and C<(state $a) = foo();> have different
789 semantics, which is tricky to implement in Perl 5 as currently they produce
790 the same opcode trees. The Perl 6 design is firm, so it would be good to
791 implement the necessary code in Perl 5. There are comments in
792 C<Perl_newASSIGNOP()> that show the code paths taken by various assignment
793 constructions involving state variables.
795 =head2 Implement $value ~~ 0 .. $range
797 It would be nice to extend the syntax of the C<~~> operator to also
798 understand numeric (and maybe alphanumeric) ranges.
800 =head2 A does() built-in
802 Like ref(), only useful. It would call the C<DOES> method on objects; it
803 would also tell whether something can be dereferenced as an
804 array/hash/etc., or used as a regexp, etc.
805 L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-03/msg00481.html>
807 =head2 Tied filehandles and write() don't mix
809 There is no method on tied filehandles to allow them to be called back by
812 =head2 Attach/detach debugger from running program
814 The old perltodo notes "With C<gdb>, you can attach the debugger to a running
815 program if you pass the process ID. It would be good to do this with the Perl
816 debugger on a running Perl program, although I'm not sure how it would be
817 done." ssh and screen do this with named pipes in /tmp. Maybe we can too.
819 =head2 Optimize away empty destructors
821 Defining an empty DESTROY method might be useful (notably in
822 AUTOLOAD-enabled classes), but it's still a bit expensive to call. That
823 could probably be optimized.
825 =head2 LVALUE functions for lists
827 The old perltodo notes that lvalue functions don't work for list or hash
828 slices. This would be good to fix.
830 =head2 LVALUE functions in the debugger
832 The old perltodo notes that lvalue functions don't work in the debugger. This
833 would be good to fix.
835 =head2 regexp optimiser optional
837 The regexp optimiser is not optional. It should configurable to be, to allow
838 its performance to be measured, and its bugs to be easily demonstrated.
840 =head2 delete &function
842 Allow to delete functions. One can already undef them, but they're still
845 =head2 C</w> regex modifier
847 That flag would enable to match whole words, and also to interpolate
848 arrays as alternations. With it, C</P/w> would be roughly equivalent to:
850 do { local $"='|'; /\b(?:P)\b/ }
852 See L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-01/msg00400.html>
855 =head2 optional optimizer
857 Make the peephole optimizer optional. Currently it performs two tasks as
858 it walks the optree - genuine peephole optimisations, and necessary fixups of
859 ops. It would be good to find an efficient way to switch out the
860 optimisations whilst keeping the fixups.
862 =head2 You WANT *how* many
864 Currently contexts are void, scalar and list. split has a special mechanism in
865 place to pass in the number of return values wanted. It would be useful to
866 have a general mechanism for this, backwards compatible and little speed hit.
867 This would allow proposals such as short circuiting sort to be implemented
870 =head2 lexical aliases
872 Allow lexical aliases (maybe via the syntax C<my \$alias = \$foo>.
874 =head2 entersub XS vs Perl
876 At the moment pp_entersub is huge, and has code to deal with entering both
877 perl and XS subroutines. Subroutine implementations rarely change between
878 perl and XS at run time, so investigate using 2 ops to enter subs (one for
879 XS, one for perl) and swap between if a sub is redefined.
883 Self-ties are currently illegal because they caused too many segfaults. Maybe
884 the causes of these could be tracked down and self-ties on all types
887 =head2 Optimize away @_
889 The old perltodo notes "Look at the "reification" code in C<av.c>".
891 =head2 Virtualize operating system access
893 Implement a set of "vtables" that virtualizes operating system access
894 (open(), mkdir(), unlink(), readdir(), getenv(), etc.) At the very
895 least these interfaces should take SVs as "name" arguments instead of
896 bare char pointers; probably the most flexible and extensible way
897 would be for the Perl-facing interfaces to accept HVs. The system
898 needs to be per-operating-system and per-file-system
899 hookable/filterable, preferably both from XS and Perl level
900 (L<perlport/"Files and Filesystems"> is good reading at this point,
901 in fact, all of L<perlport> is.)
903 This has actually already been implemented (but only for Win32),
904 take a look at F<iperlsys.h> and F<win32/perlhost.h>. While all Win32
905 variants go through a set of "vtables" for operating system access,
906 non-Win32 systems currently go straight for the POSIX/UNIX-style
907 system/library call. Similar system as for Win32 should be
908 implemented for all platforms. The existing Win32 implementation
909 probably does not need to survive alongside this proposed new
910 implementation, the approaches could be merged.
912 What would this give us? One often-asked-for feature this would
913 enable is using Unicode for filenames, and other "names" like %ENV,
914 usernames, hostnames, and so forth.
915 (See L<perlunicode/"When Unicode Does Not Happen">.)
917 But this kind of virtualization would also allow for things like
918 virtual filesystems, virtual networks, and "sandboxes" (though as long
919 as dynamic loading of random object code is allowed, not very safe
920 sandboxes since external code of course know not of Perl's vtables).
921 An example of a smaller "sandbox" is that this feature can be used to
922 implement per-thread working directories: Win32 already does this.
924 See also L</"Extend PerlIO and PerlIO::Scalar">.
926 =head2 Investigate PADTMP hash pessimisation
928 The peephole optimier converts constants used for hash key lookups to shared
929 hash key scalars. Under ithreads, something is undoing this work.
930 See http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-09/msg00793.html
932 =head2 Store the current pad in the OP slab allocator
935 I hope that I got that "current pad" part correct
937 Currently we leak ops in various cases of parse failure. I suggested that we
938 could solve this by always using the op slab allocator, and walking it to
939 free ops. Dave comments that as some ops are already freed during optree
940 creation one would have to mark which ops are freed, and not double free them
941 when walking the slab. He notes that one problem with this is that for some ops
942 you have to know which pad was current at the time of allocation, which does
943 change. I suggested storing a pointer to the current pad in the memory allocated
944 for the slab, and swapping to a new slab each time the pad changes. Dave thinks
945 that this would work.
947 =head2 repack the optree
949 Repacking the optree after execution order is determined could allow
950 removal of NULL ops, and optimal ordering of OPs with respect to cache-line
951 filling. The slab allocator could be reused for this purpose. I think that
952 the best way to do this is to make it an optional step just before the
953 completed optree is attached to anything else, and to use the slab allocator
954 unchanged, so that freeing ops is identical whether or not this step runs.
955 Note that the slab allocator allocates ops downwards in memory, so one would
956 have to actually "allocate" the ops in reverse-execution order to get them
957 contiguous in memory in execution order.
959 See http://www.nntp.perl.org/group/perl.perl5.porters/2007/12/msg131975.html
961 Note that running this copy, and then freeing all the old location ops would
962 cause their slabs to be freed, which would eliminate possible memory wastage if
963 the previous suggestion is implemented, and we swap slabs more frequently.
965 =head2 eliminate incorrect line numbers in warnings
973 } elsif ($undef == 0) {
976 used to produce this output:
978 Use of uninitialized value in numeric eq (==) at wrong.pl line 4.
979 Use of uninitialized value in numeric eq (==) at wrong.pl line 4.
981 where the line of the second warning was misreported - it should be line 5.
982 Rafael fixed this - the problem arose because there was no nextstate OP
983 between the execution of the C<if> and the C<elsif>, hence C<PL_curcop> still
984 reports that the currently executing line is line 4. The solution was to inject
985 a nextstate OPs for each C<elsif>, although it turned out that the nextstate
986 OP needed to be a nulled OP, rather than a live nextstate OP, else other line
987 numbers became misreported. (Jenga!)
989 The problem is more general than C<elsif> (although the C<elsif> case is the
990 most common and the most confusing). Ideally this code
1000 would produce this output
1002 Use of uninitialized value $undef in addition (+) at wrong.pl line 4.
1003 Use of uninitialized value $undef in addition (+) at wrong.pl line 7.
1005 (rather than lines 4 and 5), but this would seem to require every OP to carry
1006 (at least) line number information.
1008 What might work is to have an optional line number in memory just before the
1009 BASEOP structure, with a flag bit in the op to say whether it's present.
1010 Initially during compile every OP would carry its line number. Then add a late
1011 pass to the optimiser (potentially combined with L</repack the optree>) which
1012 looks at the two ops on every edge of the graph of the execution path. If
1013 the line number changes, flags the destination OP with this information.
1014 Once all paths are traced, replace every op with the flag with a
1015 nextstate-light op (that just updates C<PL_curcop>), which in turn then passes
1016 control on to the true op. All ops would then be replaced by variants that
1017 do not store the line number. (Which, logically, why it would work best in
1018 conjunction with L</repack the optree>, as that is already copying/reallocating
1021 (Although I should note that we're not certain that doing this for the general
1024 =head2 optimize tail-calls
1026 Tail-calls present an opportunity for broadly applicable optimization;
1027 anywhere that C<< return foo(...) >> is called, the outer return can
1028 be replaced by a goto, and foo will return directly to the outer
1029 caller, saving (conservatively) 25% of perl's call&return cost, which
1030 is relatively higher than in C. The scheme language is known to do
1031 this heavily. B::Concise provides good insight into where this
1032 optimization is possible, ie anywhere entersub,leavesub op-sequence
1035 perl -MO=Concise,-exec,a,b,-main -e 'sub a{ 1 }; sub b {a()}; b(2)'
1037 Bottom line on this is probably a new pp_tailcall function which
1038 combines the code in pp_entersub, pp_leavesub. This should probably
1039 be done 1st in XS, and using B::Generate to patch the new OP into the
1044 Tasks that will get your name mentioned in the description of the "Highlights
1047 =head2 make ithreads more robust
1049 Generally make ithreads more robust. See also L</iCOW>
1051 This task is incremental - even a little bit of work on it will help, and
1052 will be greatly appreciated.
1054 One bit would be to write the missing code in sv.c:Perl_dirp_dup.
1056 Fix Perl_sv_dup, et al so that threads can return objects.
1060 Sarathy and Arthur have a proposal for an improved Copy On Write which
1061 specifically will be able to COW new ithreads. If this can be implemented
1062 it would be a good thing.
1064 =head2 (?{...}) closures in regexps
1066 Fix (or rewrite) the implementation of the C</(?{...})/> closures.
1068 =head2 A re-entrant regexp engine
1070 This will allow the use of a regex from inside (?{ }), (??{ }) and
1071 (?(?{ })|) constructs.
1073 =head2 Add class set operations to regexp engine
1075 Apparently these are quite useful. Anyway, Jeffery Friedl wants them.
1077 demerphq has this on his todo list, but right at the bottom.