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 Smartmatch design issues
28 In 5.10.0 the smartmatch operator C<~~> isn't working quite "right". But
29 before we can fix the implementation, we need to define what "right" is.
30 The first problem is that Robin Houston implemented the Perl 6 smart match
31 spec as of February 2006, when smart match was axiomatically symmetrical:
32 L<http://groups.google.com/group/perl.perl6.language/msg/bf2b486f089ad021>
34 Since then the Perl 6 target moved, but the Perl 5 implementation did not.
36 So it would be useful for someone to compare the Perl 6 smartmatch table
37 as of February 2006 L<http://svn.perl.org/viewvc/perl6/doc/trunk/design/syn/S03.pod?view=markup&pathrev=7615>
38 and the current table L<http://svn.perl.org/viewvc/perl6/doc/trunk/design/syn/S03.pod?revision=14556&view=markup>
39 and tabulate the differences in Perl 6. The diff is
40 C<svn diff -r7615:14556 http://svn.perl.org/perl6/doc/trunk/design/syn/S03.pod>
41 -- search for C<=head1 Smart matching>
43 With that done and published, someone (else) can then map any changed Perl 6
44 semantics back to Perl 5, based on how the existing semantics map to Perl 5:
45 L<http://search.cpan.org/~rgarcia/perl-5.10.0/pod/perlsyn.pod#Smart_matching_in_detail>
48 There are also some questions that need answering:
54 How do you negate one? (documentation issue)
55 http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2008-01/msg00071.html
60 http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-12/msg00799.html
62 * Should smart matches be symmetrical? (Perl 6 says no)
64 * Other differences between Perl 5 and Perl 6 smart match?
68 Objects and smart match
69 http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-12/msg00865.html
73 =head2 Remove duplication of test setup.
75 Schwern notes, that there's duplication of code - lots and lots of tests have
76 some variation on the big block of C<$Is_Foo> checks. We can safely put this
77 into a file, change it to build an C<%Is> hash and require it. Maybe just put
78 it into F<test.pl>. Throw in the handy tainting subroutines.
80 =head2 POD -E<gt> HTML conversion in the core still sucks
82 Which is crazy given just how simple POD purports to be, and how simple HTML
83 can be. It's not actually I<as> simple as it sounds, particularly with the
84 flexibility POD allows for C<=item>, but it would be good to improve the
85 visual appeal of the HTML generated, and to avoid it having any validation
86 errors. See also L</make HTML install work>, as the layout of installation tree
87 is needed to improve the cross-linking.
89 The addition of C<Pod::Simple> and its related modules may make this task
92 =head2 merge checkpods and podchecker
94 F<pod/checkpods.PL> (and C<make check> in the F<pod/> subdirectory)
95 implements a very basic check for pod files, but the errors it discovers
96 aren't found by podchecker. Add this check to podchecker, get rid of
97 checkpods and have C<make check> use podchecker.
99 =head2 Parallel testing
101 (This probably impacts much more than the core: also the Test::Harness
102 and TAP::* modules on CPAN.)
104 All of the tests in F<t/> can now be run in parallel, if C<$ENV{TEST_JOBS}>
105 is set. However, tests within each directory in F<ext> and F<lib> are still
106 run in series, with directories run in parallel. This is an adequate
107 heuristic, but it might be possible to relax it further, and get more
108 throughput. Specifically, it would be good to audit all of F<lib/*.t>, and
109 make them use C<File::Temp>.
111 =head2 Make Schwern poorer
113 We should have tests for everything. When all the core's modules are tested,
114 Schwern has promised to donate to $500 to TPF. We may need volunteers to
115 hold him upside down and shake vigorously in order to actually extract the
118 =head2 Improve the coverage of the core tests
120 Use Devel::Cover to ascertain the core modules's test coverage, then add
121 tests that are currently missing.
125 A full test suite for the B module would be nice.
127 =head2 Deparse inlined constants
131 use constant PI => 4;
134 will currently deparse as
136 use constant ('PI', 4);
139 because the tokenizer inlines the value of the constant subroutine C<PI>.
140 This allows various compile time optimisations, such as constant folding
141 and dead code elimination. Where these haven't happened (such as the example
142 above) it ought be possible to make B::Deparse work out the name of the
143 original constant, because just enough information survives in the symbol
144 table to do this. Specifically, the same scalar is used for the constant in
145 the optree as is used for the constant subroutine, so by iterating over all
146 symbol tables and generating a mapping of SV address to constant name, it
147 would be possible to provide B::Deparse with this functionality.
149 =head2 A decent benchmark
151 C<perlbench> seems impervious to any recent changes made to the perl core. It
152 would be useful to have a reasonable general benchmarking suite that roughly
153 represented what current perl programs do, and measurably reported whether
154 tweaks to the core improve, degrade or don't really affect performance, to
155 guide people attempting to optimise the guts of perl. Gisle would welcome
156 new tests for perlbench.
158 =head2 fix tainting bugs
160 Fix the bugs revealed by running the test suite with the C<-t> switch (via
161 C<make test.taintwarn>).
163 =head2 Dual life everything
165 As part of the "dists" plan, anything that doesn't belong in the smallest perl
166 distribution needs to be dual lifed. Anything else can be too. Figure out what
167 changes would be needed to package that module and its tests up for CPAN, and
168 do so. Test it with older perl releases, and fix the problems you find.
170 To make a minimal perl distribution, it's useful to look at
171 F<t/lib/commonsense.t>.
173 =head2 Bundle dual life modules in ext/
175 For maintenance (and branch merging) reasons, it would be useful to move
176 some architecture-independent dual-life modules from lib/ to ext/, if this
177 has no negative impact on the build of perl itself.
179 =head2 POSIX memory footprint
181 Ilya observed that use POSIX; eats memory like there's no tomorrow, and at
182 various times worked to cut it down. There is probably still fat to cut out -
183 for example POSIX passes Exporter some very memory hungry data structures.
185 =head2 embed.pl/makedef.pl
187 There is a script F<embed.pl> that generates several header files to prefix
188 all of Perl's symbols in a consistent way, to provide some semblance of
189 namespace support in C<C>. Functions are declared in F<embed.fnc>, variables
190 in F<interpvar.h>. Quite a few of the functions and variables
191 are conditionally declared there, using C<#ifdef>. However, F<embed.pl>
192 doesn't understand the C macros, so the rules about which symbols are present
193 when is duplicated in F<makedef.pl>. Writing things twice is bad, m'kay.
194 It would be good to teach C<embed.pl> to understand the conditional
195 compilation, and hence remove the duplication, and the mistakes it has caused.
197 =head2 use strict; and AutoLoad
199 Currently if you write
202 use AutoLoader 'AUTOLOAD';
207 print join (' ', No, strict, here), "!\n";
210 then C<use strict;> isn't in force within the autoloaded subroutines. It would
211 be more consistent (and less surprising) to arrange for all lexical pragmas
212 in force at the __END__ block to be in force within each autoloaded subroutine.
214 There's a similar problem with SelfLoader.
216 =head2 profile installman
218 The F<installman> script is slow. All it is doing text processing, which we're
219 told is something Perl is good at. So it would be nice to know what it is doing
220 that is taking so much CPU, and where possible address it.
223 =head1 Tasks that need a little sysadmin-type knowledge
225 Or if you prefer, tasks that you would learn from, and broaden your skills
228 =head2 make HTML install work
230 There is an C<installhtml> target in the Makefile. It's marked as
231 "experimental". It would be good to get this tested, make it work reliably, and
232 remove the "experimental" tag. This would include
238 Checking that cross linking between various parts of the documentation works.
239 In particular that links work between the modules (files with POD in F<lib/>)
240 and the core documentation (files in F<pod/>)
244 Work out how to split C<perlfunc> into chunks, preferably one per function
245 group, preferably with general case code that could be used elsewhere.
246 Challenges here are correctly identifying the groups of functions that go
247 together, and making the right named external cross-links point to the right
248 page. Things to be aware of are C<-X>, groups such as C<getpwnam> to
249 C<endservent>, two or more C<=items> giving the different parameter lists, such
252 =item substr EXPR,OFFSET,LENGTH,REPLACEMENT
253 =item substr EXPR,OFFSET,LENGTH
254 =item substr EXPR,OFFSET
256 and different parameter lists having different meanings. (eg C<select>)
260 =head2 compressed man pages
262 Be able to install them. This would probably need a configure test to see how
263 the system does compressed man pages (same directory/different directory?
264 same filename/different filename), as well as tweaking the F<installman> script
265 to compress as necessary.
267 =head2 Add a code coverage target to the Makefile
269 Make it easy for anyone to run Devel::Cover on the core's tests. The steps
270 to do this manually are roughly
276 do a normal C<Configure>, but include Devel::Cover as a module to install
277 (see F<INSTALL> for how to do this)
285 cd t; HARNESS_PERL_SWITCHES=-MDevel::Cover ./perl -I../lib harness
289 Process the resulting Devel::Cover database
293 This just give you the coverage of the F<.pm>s. To also get the C level
300 Additionally tell C<Configure> to use the appropriate C compiler flags for
307 (instead of C<make perl>)
311 After running the tests run C<gcov> to generate all the F<.gcov> files.
312 (Including down in the subdirectories of F<ext/>
316 (From the top level perl directory) run C<gcov2perl> on all the C<.gcov> files
317 to get their stats into the cover_db directory.
321 Then process the Devel::Cover database
325 It would be good to add a single switch to C<Configure> to specify that you
326 wanted to perform perl level coverage, and another to specify C level
327 coverage, and have C<Configure> and the F<Makefile> do all the right things
330 =head2 Make Config.pm cope with differences between built and installed perl
332 Quite often vendors ship a perl binary compiled with their (pay-for)
333 compilers. People install a free compiler, such as gcc. To work out how to
334 build extensions, Perl interrogates C<%Config>, so in this situation
335 C<%Config> describes compilers that aren't there, and extension building
336 fails. This forces people into choosing between re-compiling perl themselves
337 using the compiler they have, or only using modules that the vendor ships.
339 It would be good to find a way teach C<Config.pm> about the installation setup,
340 possibly involving probing at install time or later, so that the C<%Config> in
341 a binary distribution better describes the installed machine, when the
342 installed machine differs from the build machine in some significant way.
344 =head2 linker specification files
346 Some platforms mandate that you provide a list of a shared library's external
347 symbols to the linker, so the core already has the infrastructure in place to
348 do this for generating shared perl libraries. My understanding is that the
349 GNU toolchain can accept an optional linker specification file, and restrict
350 visibility just to symbols declared in that file. It would be good to extend
351 F<makedef.pl> to support this format, and to provide a means within
352 C<Configure> to enable it. This would allow Unix users to test that the
353 export list is correct, and to build a perl that does not pollute the global
354 namespace with private symbols.
356 =head2 Cross-compile support
358 Currently C<Configure> understands C<-Dusecrosscompile> option. This option
359 arranges for building C<miniperl> for TARGET machine, so this C<miniperl> is
360 assumed then to be copied to TARGET machine and used as a replacement of full
363 This could be done little differently. Namely C<miniperl> should be built for
364 HOST and then full C<perl> with extensions should be compiled for TARGET.
365 This, however, might require extra trickery for %Config: we have one config
366 first for HOST and then another for TARGET. Tools like MakeMaker will be
367 mightily confused. Having around two different types of executables and
368 libraries (HOST and TARGET) makes life interesting for Makefiles and
369 shell (and Perl) scripts. There is $Config{run}, normally empty, which
370 can be used as an execution wrapper. Also note that in some
371 cross-compilation/execution environments the HOST and the TARGET do
372 not see the same filesystem(s), the $Config{run} may need to do some
373 file/directory copying back and forth.
377 Make F<pod/roffitall> be updated by F<pod/buildtoc>.
379 =head2 Split "linker" from "compiler"
381 Right now, Configure probes for two commands, and sets two variables:
387 This variable holds the name of a command to execute a C compiler which
388 can resolve multiple global references that happen to have the same
389 name. Usual values are F<cc> and F<gcc>.
390 Fervent ANSI compilers may be called F<c89>. AIX has F<xlc>.
394 This variable indicates the program to be used to link
395 libraries for dynamic loading. On some systems, it is F<ld>.
396 On ELF systems, it should be C<$cc>. Mostly, we'll try to respect
397 the hint file setting.
401 There is an implicit historical assumption, probably from Perl 1, that C<$cc>
402 is also the correct command for linking object files together to make an
403 executable. This may be true on Unix, but it's not true on other platforms,
404 and there are a maze of work arounds in other places (such as F<Makefile.SH>)
407 Ideally, we should create a new variable to hold the name of the executable
408 linker program, probe for it in F<Configure>, and centralise all the special
409 case logic there or in hints files.
411 A small bikeshed issue remains - what to call it, given that C<$ld> is already
412 taken (arguably for the wrong thing) and C<$link> could be confused with the
413 Unix command line executable of the same name, which does something completely
414 different. Andy Dougherty makes the counter argument "In parrot, I tried to
415 call the command used to link object files and libraries into an executable
416 F<link>, since that's what my vaguely-remembered DOS and VMS experience
417 suggested. I don't think any real confusion has ensued, so it's probably a
418 reasonable name for perl5 to use."
420 "Alas, I've always worried that introducing it would make things worse,
421 since now the module building utilities would have to look for
422 C<$Config{link}> and institute a fall-back plan if it weren't found."
425 =head1 Tasks that need a little C knowledge
427 These tasks would need a little C knowledge, but don't need any specific
428 background or experience with XS, or how the Perl interpreter works
430 =head2 Weed out needless PERL_UNUSED_ARG
432 The C code uses the macro C<PERL_UNUSED_ARG> to stop compilers warning about
433 unused arguments. Often the arguments can't be removed, as there is an
434 external constraint that determines the prototype of the function, so this
435 approach is valid. However, there are some cases where C<PERL_UNUSED_ARG>
436 could be removed. Specifically
442 The prototypes of (nearly all) static functions can be changed
446 Unused arguments generated by short cut macros are wasteful - the short cut
447 macro used can be changed.
451 =head2 Modernize the order of directories in @INC
453 The way @INC is laid out by default, one cannot upgrade core (dual-life)
454 modules without overwriting files. This causes problems for binary
455 package builders. One possible proposal is laid out in this
457 L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2002-04/msg02380.html>.
461 Natively 64-bit systems need neither -Duse64bitint nor -Duse64bitall.
462 On these systems, it might be the default compilation mode, and there
463 is currently no guarantee that passing no use64bitall option to the
464 Configure process will build a 32bit perl. Implementing -Duse32bit*
465 options would be nice for perl 5.12.
467 =head2 Make it clear from -v if this is the exact official release
469 Currently perl from C<p4>/C<rsync> ships with a F<patchlevel.h> file that
470 usually defines one local patch, of the form "MAINT12345" or "RC1". The output
471 of perl -v doesn't report that a perl isn't an official release, and this
472 information can get lost in bugs reports. Because of this, the minor version
473 isn't bumped up until RC time, to minimise the possibility of versions of perl
474 escaping that believe themselves to be newer than they actually are.
476 It would be useful to find an elegant way to have the "this is an interim
477 maintenance release" or "this is a release candidate" in the terse -v output,
478 and have it so that it's easy for the pumpking to remove this just as the
479 release tarball is rolled up. This way the version pulled out of rsync would
480 always say "I'm a development release" and it would be safe to bump the
481 reported minor version as soon as a release ships, which would aid perl
484 This task is really about thinking of an elegant way to arrange the C source
485 such that it's trivial for the Pumpking to flag "this is an official release"
486 when making a tarball, yet leave the default source saying "I'm not the
489 =head2 Profile Perl - am I hot or not?
491 The Perl source code is stable enough that it makes sense to profile it,
492 identify and optimise the hotspots. It would be good to measure the
493 performance of the Perl interpreter using free tools such as cachegrind,
494 gprof, and dtrace, and work to reduce the bottlenecks they reveal.
496 As part of this, the idea of F<pp_hot.c> is that it contains the I<hot> ops,
497 the ops that are most commonly used. The idea is that by grouping them, their
498 object code will be adjacent in the executable, so they have a greater chance
499 of already being in the CPU cache (or swapped in) due to being near another op
502 Except that it's not clear if these really are the most commonly used ops. So
503 as part of exercising your skills with coverage and profiling tools you might
504 want to determine what ops I<really> are the most commonly used. And in turn
505 suggest evictions and promotions to achieve a better F<pp_hot.c>.
507 One piece of Perl code that might make a good testbed is F<installman>.
509 =head2 Allocate OPs from arenas
511 Currently all new OP structures are individually malloc()ed and free()d.
512 All C<malloc> implementations have space overheads, and are now as fast as
513 custom allocates so it would both use less memory and less CPU to allocate
514 the various OP structures from arenas. The SV arena code can probably be
517 Note that Configuring perl with C<-Accflags=-DPL_OP_SLAB_ALLOC> will use
518 Perl_Slab_alloc() to pack optrees into a contiguous block, which is
519 probably superior to the use of OP arenas, esp. from a cache locality
520 standpoint. See L<Profile Perl - am I hot or not?>.
522 =head2 Improve win32/wince.c
524 Currently, numerous functions look virtually, if not completely,
525 identical in both C<win32/wince.c> and C<win32/win32.c> files, which can't
528 =head2 Use secure CRT functions when building with VC8 on Win32
530 Visual C++ 2005 (VC++ 8.x) deprecated a number of CRT functions on the basis
531 that they were "unsafe" and introduced differently named secure versions of
532 them as replacements, e.g. instead of writing
534 FILE* f = fopen(__FILE__, "r");
539 errno_t err = fopen_s(&f, __FILE__, "r");
541 Currently, the warnings about these deprecations have been disabled by adding
542 -D_CRT_SECURE_NO_DEPRECATE to the CFLAGS. It would be nice to remove that
543 warning suppressant and actually make use of the new secure CRT functions.
545 There is also a similar issue with POSIX CRT function names like fileno having
546 been deprecated in favour of ISO C++ conformant names like _fileno. These
547 warnings are also currently suppressed by adding -D_CRT_NONSTDC_NO_DEPRECATE. It
548 might be nice to do as Microsoft suggest here too, although, unlike the secure
549 functions issue, there is presumably little or no benefit in this case.
551 =head2 Fix POSIX::access() and chdir() on Win32
553 These functions currently take no account of DACLs and therefore do not behave
554 correctly in situations where access is restricted by DACLs (as opposed to the
555 read-only attribute).
557 Furthermore, POSIX::access() behaves differently for directories having the
558 read-only attribute set depending on what CRT library is being used. For
559 example, the _access() function in the VC6 and VC7 CRTs (wrongly) claim that
560 such directories are not writable, whereas in fact all directories are writable
561 unless access is denied by DACLs. (In the case of directories, the read-only
562 attribute actually only means that the directory cannot be deleted.) This CRT
563 bug is fixed in the VC8 and VC9 CRTs (but, of course, the directory may still
564 not actually be writable if access is indeed denied by DACLs).
566 For the chdir() issue, see ActiveState bug #74552:
567 http://bugs.activestate.com/show_bug.cgi?id=74552
569 Therefore, DACLs should be checked both for consistency across CRTs and for
572 (Note that perl's -w operator should not be modified to check DACLs. It has
573 been written so that it reflects the state of the read-only attribute, even
574 for directories (whatever CRT is being used), for symmetry with chmod().)
576 =head2 strcat(), strcpy(), strncat(), strncpy(), sprintf(), vsprintf()
578 Maybe create a utility that checks after each libperl.a creation that
579 none of the above (nor sprintf(), vsprintf(), or *SHUDDER* gets())
580 ever creep back to libperl.a.
582 nm libperl.a | ./miniperl -alne '$o = $F[0] if /:$/; print "$o $F[1]" if $F[0] eq "U" && $F[1] =~ /^(?:strn?c(?:at|py)|v?sprintf|gets)$/'
584 Note, of course, that this will only tell whether B<your> platform
585 is using those naughty interfaces.
587 =head2 -D_FORTIFY_SOURCE=2, -fstack-protector
589 Recent glibcs support C<-D_FORTIFY_SOURCE=2> and recent gcc
590 (4.1 onwards?) supports C<-fstack-protector>, both of which give
591 protection against various kinds of buffer overflow problems.
592 These should probably be used for compiling Perl whenever available,
593 Configure and/or hints files should be adjusted to probe for the
594 availability of these features and enable them as appropriate.
596 =head2 Arenas for GPs? For MAGIC?
598 C<struct gp> and C<struct magic> are both currently allocated by C<malloc>.
599 It might be a speed or memory saving to change to using arenas. Or it might
600 not. It would need some suitable benchmarking first. In particular, C<GP>s
601 can probably be changed with minimal compatibility impact (probably nothing
602 outside of the core, or even outside of F<gv.c> allocates them), but they
603 probably aren't allocated/deallocated often enough for a speed saving. Whereas
604 C<MAGIC> is allocated/deallocated more often, but in turn, is also something
605 more externally visible, so changing the rules here may bite external code.
609 Several SV body structs are now the same size, notably PVMG and PVGV, PVAV and
610 PVHV, and PVCV and PVFM. It should be possible to allocate and return same
611 sized bodies from the same actual arena, rather than maintaining one arena for
612 each. This could save 4-6K per thread, of memory no longer tied up in the
613 not-yet-allocated part of an arena.
616 =head1 Tasks that need a knowledge of XS
618 These tasks would need C knowledge, and roughly the level of knowledge of
619 the perl API that comes from writing modules that use XS to interface to
622 =head2 safely supporting POSIX SA_SIGINFO
624 Some years ago Jarkko supplied patches to provide support for the POSIX
625 SA_SIGINFO feature in Perl, passing the extra data to the Perl signal handler.
627 Unfortunately, it only works with "unsafe" signals, because under safe
628 signals, by the time Perl gets to run the signal handler, the extra
629 information has been lost. Moreover, it's not easy to store it somewhere,
630 as you can't call mutexs, or do anything else fancy, from inside a signal
633 So it strikes me that we could provide safe SA_SIGINFO support
639 Provide global variables for two file descriptors
643 When the first request is made via C<sigaction> for C<SA_SIGINFO>, create a
644 pipe, store the reader in one, the writer in the other
648 In the "safe" signal handler (C<Perl_csighandler()>/C<S_raise_signal()>), if
649 the C<siginfo_t> pointer non-C<NULL>, and the writer file handle is open,
655 serialise signal number, C<struct siginfo_t> (or at least the parts we care
656 about) into a small auto char buff
660 C<write()> that (non-blocking) to the writer fd
666 if it writes 100%, flag the signal in a counter of "signals on the pipe" akin
667 to the current per-signal-number counts
671 if it writes 0%, assume the pipe is full. Flag the data as lost?
675 if it writes partially, croak a panic, as your OS is broken.
683 in the regular C<PERL_ASYNC_CHECK()> processing, if there are "signals on
684 the pipe", read the data out, deserialise, build the Perl structures on
685 the stack (code in C<Perl_sighandler()>, the "unsafe" handler), and call as
690 I think that this gets us decent C<SA_SIGINFO> support, without the current risk
691 of running Perl code inside the signal handler context. (With all the dangers
692 of things like C<malloc> corruption that that currently offers us)
694 For more information see the thread starting with this message:
695 http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2008-03/msg00305.html
697 =head2 autovivification
699 Make all autovivification consistent w.r.t LVALUE/RVALUE and strict/no strict;
701 This task is incremental - even a little bit of work on it will help.
703 =head2 Unicode in Filenames
705 chdir, chmod, chown, chroot, exec, glob, link, lstat, mkdir, open,
706 opendir, qx, readdir, readlink, rename, rmdir, stat, symlink, sysopen,
707 system, truncate, unlink, utime, -X. All these could potentially accept
708 Unicode filenames either as input or output (and in the case of system
709 and qx Unicode in general, as input or output to/from the shell).
710 Whether a filesystem - an operating system pair understands Unicode in
713 Known combinations that have some level of understanding include
714 Microsoft NTFS, Apple HFS+ (In Mac OS 9 and X) and Apple UFS (in Mac
715 OS X), NFS v4 is rumored to be Unicode, and of course Plan 9. How to
716 create Unicode filenames, what forms of Unicode are accepted and used
717 (UCS-2, UTF-16, UTF-8), what (if any) is the normalization form used,
718 and so on, varies. Finding the right level of interfacing to Perl
719 requires some thought. Remember that an OS does not implicate a
722 (The Windows -C command flag "wide API support" has been at least
723 temporarily retired in 5.8.1, and the -C has been repurposed, see
726 Most probably the right way to do this would be this:
727 L</"Virtualize operating system access">.
729 =head2 Unicode in %ENV
731 Currently the %ENV entries are always byte strings.
732 See L</"Virtualize operating system access">.
734 =head2 Unicode and glob()
736 Currently glob patterns and filenames returned from File::Glob::glob()
737 are always byte strings. See L</"Virtualize operating system access">.
739 =head2 Unicode and lc/uc operators
741 Some built-in operators (C<lc>, C<uc>, etc.) behave differently, based on
742 what the internal encoding of their argument is. That should not be the
743 case. Maybe add a pragma to switch behaviour.
745 =head2 use less 'memory'
747 Investigate trade offs to switch out perl's choices on memory usage.
748 Particularly perl should be able to give memory back.
750 This task is incremental - even a little bit of work on it will help.
752 =head2 Re-implement C<:unique> in a way that is actually thread-safe
754 The old implementation made bad assumptions on several levels. A good 90%
755 solution might be just to make C<:unique> work to share the string buffer
756 of SvPVs. That way large constant strings can be shared between ithreads,
757 such as the configuration information in F<Config>.
759 =head2 Make tainting consistent
761 Tainting would be easier to use if it didn't take documented shortcuts and
762 allow taint to "leak" everywhere within an expression.
764 =head2 readpipe(LIST)
766 system() accepts a LIST syntax (and a PROGRAM LIST syntax) to avoid
767 running a shell. readpipe() (the function behind qx//) could be similarly
770 =head2 Audit the code for destruction ordering assumptions
774 /* Need to check SvMAGICAL, as during global destruction it may be that
775 AvARYLEN(av) has been freed before av, and hence the SvANY() pointer
776 is now part of the linked list of SV heads, rather than pointing to
777 the original body. */
778 /* FIXME - audit the code for other bugs like this one. */
780 adding the C<SvMAGICAL> check to
782 if (AvARYLEN(av) && SvMAGICAL(AvARYLEN(av))) {
783 MAGIC *mg = mg_find (AvARYLEN(av), PERL_MAGIC_arylen);
785 Go through the core and look for similar assumptions that SVs have particular
786 types, as all bets are off during global destruction.
788 =head2 Extend PerlIO and PerlIO::Scalar
790 PerlIO::Scalar doesn't know how to truncate(). Implementing this
791 would require extending the PerlIO vtable.
793 Similarly the PerlIO vtable doesn't know about formats (write()), or
794 about stat(), or chmod()/chown(), utime(), or flock().
796 (For PerlIO::Scalar it's hard to see what e.g. mode bits or ownership
799 PerlIO doesn't do directories or symlinks, either: mkdir(), rmdir(),
800 opendir(), closedir(), seekdir(), rewinddir(), glob(); symlink(),
803 See also L</"Virtualize operating system access">.
805 =head2 -C on the #! line
807 It should be possible to make -C work correctly if found on the #! line,
808 given that all perl command line options are strict ASCII, and -C changes
809 only the interpretation of non-ASCII characters, and not for the script file
810 handle. To make it work needs some investigation of the ordering of function
811 calls during startup, and (by implication) a bit of tweaking of that order.
813 =head2 Organize error messages
815 Perl's diagnostics (error messages, see L<perldiag>) could use
816 reorganizing and formalizing so that each error message has its
817 stable-for-all-eternity unique id, categorized by severity, type, and
818 subsystem. (The error messages would be listed in a datafile outside
819 of the Perl source code, and the source code would only refer to the
820 messages by the id.) This clean-up and regularizing should apply
821 for all croak() messages.
823 This would enable all sorts of things: easier translation/localization
824 of the messages (though please do keep in mind the caveats of
825 L<Locale::Maketext> about too straightforward approaches to
826 translation), filtering by severity, and instead of grepping for a
827 particular error message one could look for a stable error id. (Of
828 course, changing the error messages by default would break all the
829 existing software depending on some particular error message...)
831 This kind of functionality is known as I<message catalogs>. Look for
832 inspiration for example in the catgets() system, possibly even use it
833 if available-- but B<only> if available, all platforms will B<not>
836 For the really pure at heart, consider extending this item to cover
837 also the warning messages (see L<perllexwarn>, C<warnings.pl>).
839 =head1 Tasks that need a knowledge of the interpreter
841 These tasks would need C knowledge, and knowledge of how the interpreter works,
842 or a willingness to learn.
844 =head2 lexicals used only once
848 $ perl -we '$pie = 42'
849 Name "main::pie" used only once: possible typo at -e line 1.
853 $ perl -we 'my $pie = 42'
855 Logically all lexicals used only once should warn, if the user asks for
856 warnings. An unworked RT ticket (#5087) has been open for almost seven
857 years for this discrepancy.
861 The handling of Unicode is unclean in many places. For example, the regexp
862 engine matches in Unicode semantics whenever the string or the pattern is
863 flagged as UTF-8, but that should not be dependent on an internal storage
864 detail of the string. Likewise, case folding behaviour is dependent on the
865 UTF8 internal flag being on or off.
867 =head2 Properly Unicode safe tokeniser and pads.
869 The tokeniser isn't actually very UTF-8 clean. C<use utf8;> is a hack -
870 variable names are stored in stashes as raw bytes, without the utf-8 flag
871 set. The pad API only takes a C<char *> pointer, so that's all bytes too. The
872 tokeniser ignores the UTF-8-ness of C<PL_rsfp>, or any SVs returned from
873 source filters. All this could be fixed.
875 =head2 state variable initialization in list context
877 Currently this is illegal:
879 state ($a, $b) = foo();
881 In Perl 6, C<state ($a) = foo();> and C<(state $a) = foo();> have different
882 semantics, which is tricky to implement in Perl 5 as currently they produce
883 the same opcode trees. The Perl 6 design is firm, so it would be good to
884 implement the necessary code in Perl 5. There are comments in
885 C<Perl_newASSIGNOP()> that show the code paths taken by various assignment
886 constructions involving state variables.
888 =head2 Implement $value ~~ 0 .. $range
890 It would be nice to extend the syntax of the C<~~> operator to also
891 understand numeric (and maybe alphanumeric) ranges.
893 =head2 A does() built-in
895 Like ref(), only useful. It would call the C<DOES> method on objects; it
896 would also tell whether something can be dereferenced as an
897 array/hash/etc., or used as a regexp, etc.
898 L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-03/msg00481.html>
900 =head2 Tied filehandles and write() don't mix
902 There is no method on tied filehandles to allow them to be called back by
905 =head2 Attach/detach debugger from running program
907 The old perltodo notes "With C<gdb>, you can attach the debugger to a running
908 program if you pass the process ID. It would be good to do this with the Perl
909 debugger on a running Perl program, although I'm not sure how it would be
910 done." ssh and screen do this with named pipes in /tmp. Maybe we can too.
912 =head2 Optimize away empty destructors
914 Defining an empty DESTROY method might be useful (notably in
915 AUTOLOAD-enabled classes), but it's still a bit expensive to call. That
916 could probably be optimized.
918 =head2 LVALUE functions for lists
920 The old perltodo notes that lvalue functions don't work for list or hash
921 slices. This would be good to fix.
923 =head2 LVALUE functions in the debugger
925 The old perltodo notes that lvalue functions don't work in the debugger. This
926 would be good to fix.
928 =head2 regexp optimiser optional
930 The regexp optimiser is not optional. It should configurable to be, to allow
931 its performance to be measured, and its bugs to be easily demonstrated.
933 =head2 delete &function
935 Allow to delete functions. One can already undef them, but they're still
938 =head2 C</w> regex modifier
940 That flag would enable to match whole words, and also to interpolate
941 arrays as alternations. With it, C</P/w> would be roughly equivalent to:
943 do { local $"='|'; /\b(?:P)\b/ }
945 See L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-01/msg00400.html>
948 =head2 optional optimizer
950 Make the peephole optimizer optional. Currently it performs two tasks as
951 it walks the optree - genuine peephole optimisations, and necessary fixups of
952 ops. It would be good to find an efficient way to switch out the
953 optimisations whilst keeping the fixups.
955 =head2 You WANT *how* many
957 Currently contexts are void, scalar and list. split has a special mechanism in
958 place to pass in the number of return values wanted. It would be useful to
959 have a general mechanism for this, backwards compatible and little speed hit.
960 This would allow proposals such as short circuiting sort to be implemented
963 =head2 lexical aliases
965 Allow lexical aliases (maybe via the syntax C<my \$alias = \$foo>.
967 =head2 entersub XS vs Perl
969 At the moment pp_entersub is huge, and has code to deal with entering both
970 perl and XS subroutines. Subroutine implementations rarely change between
971 perl and XS at run time, so investigate using 2 ops to enter subs (one for
972 XS, one for perl) and swap between if a sub is redefined.
976 Self-ties are currently illegal because they caused too many segfaults. Maybe
977 the causes of these could be tracked down and self-ties on all types
980 =head2 Optimize away @_
982 The old perltodo notes "Look at the "reification" code in C<av.c>".
984 =head2 Virtualize operating system access
986 Implement a set of "vtables" that virtualizes operating system access
987 (open(), mkdir(), unlink(), readdir(), getenv(), etc.) At the very
988 least these interfaces should take SVs as "name" arguments instead of
989 bare char pointers; probably the most flexible and extensible way
990 would be for the Perl-facing interfaces to accept HVs. The system
991 needs to be per-operating-system and per-file-system
992 hookable/filterable, preferably both from XS and Perl level
993 (L<perlport/"Files and Filesystems"> is good reading at this point,
994 in fact, all of L<perlport> is.)
996 This has actually already been implemented (but only for Win32),
997 take a look at F<iperlsys.h> and F<win32/perlhost.h>. While all Win32
998 variants go through a set of "vtables" for operating system access,
999 non-Win32 systems currently go straight for the POSIX/UNIX-style
1000 system/library call. Similar system as for Win32 should be
1001 implemented for all platforms. The existing Win32 implementation
1002 probably does not need to survive alongside this proposed new
1003 implementation, the approaches could be merged.
1005 What would this give us? One often-asked-for feature this would
1006 enable is using Unicode for filenames, and other "names" like %ENV,
1007 usernames, hostnames, and so forth.
1008 (See L<perlunicode/"When Unicode Does Not Happen">.)
1010 But this kind of virtualization would also allow for things like
1011 virtual filesystems, virtual networks, and "sandboxes" (though as long
1012 as dynamic loading of random object code is allowed, not very safe
1013 sandboxes since external code of course know not of Perl's vtables).
1014 An example of a smaller "sandbox" is that this feature can be used to
1015 implement per-thread working directories: Win32 already does this.
1017 See also L</"Extend PerlIO and PerlIO::Scalar">.
1019 =head2 Investigate PADTMP hash pessimisation
1021 The peephole optimier converts constants used for hash key lookups to shared
1022 hash key scalars. Under ithreads, something is undoing this work.
1023 See http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-09/msg00793.html
1025 =head2 Store the current pad in the OP slab allocator
1028 I hope that I got that "current pad" part correct
1030 Currently we leak ops in various cases of parse failure. I suggested that we
1031 could solve this by always using the op slab allocator, and walking it to
1032 free ops. Dave comments that as some ops are already freed during optree
1033 creation one would have to mark which ops are freed, and not double free them
1034 when walking the slab. He notes that one problem with this is that for some ops
1035 you have to know which pad was current at the time of allocation, which does
1036 change. I suggested storing a pointer to the current pad in the memory allocated
1037 for the slab, and swapping to a new slab each time the pad changes. Dave thinks
1038 that this would work.
1040 =head2 repack the optree
1042 Repacking the optree after execution order is determined could allow
1043 removal of NULL ops, and optimal ordering of OPs with respect to cache-line
1044 filling. The slab allocator could be reused for this purpose. I think that
1045 the best way to do this is to make it an optional step just before the
1046 completed optree is attached to anything else, and to use the slab allocator
1047 unchanged, so that freeing ops is identical whether or not this step runs.
1048 Note that the slab allocator allocates ops downwards in memory, so one would
1049 have to actually "allocate" the ops in reverse-execution order to get them
1050 contiguous in memory in execution order.
1052 See http://www.nntp.perl.org/group/perl.perl5.porters/2007/12/msg131975.html
1054 Note that running this copy, and then freeing all the old location ops would
1055 cause their slabs to be freed, which would eliminate possible memory wastage if
1056 the previous suggestion is implemented, and we swap slabs more frequently.
1058 =head2 eliminate incorrect line numbers in warnings
1066 } elsif ($undef == 0) {
1069 used to produce this output:
1071 Use of uninitialized value in numeric eq (==) at wrong.pl line 4.
1072 Use of uninitialized value in numeric eq (==) at wrong.pl line 4.
1074 where the line of the second warning was misreported - it should be line 5.
1075 Rafael fixed this - the problem arose because there was no nextstate OP
1076 between the execution of the C<if> and the C<elsif>, hence C<PL_curcop> still
1077 reports that the currently executing line is line 4. The solution was to inject
1078 a nextstate OPs for each C<elsif>, although it turned out that the nextstate
1079 OP needed to be a nulled OP, rather than a live nextstate OP, else other line
1080 numbers became misreported. (Jenga!)
1082 The problem is more general than C<elsif> (although the C<elsif> case is the
1083 most common and the most confusing). Ideally this code
1093 would produce this output
1095 Use of uninitialized value $undef in addition (+) at wrong.pl line 4.
1096 Use of uninitialized value $undef in addition (+) at wrong.pl line 7.
1098 (rather than lines 4 and 5), but this would seem to require every OP to carry
1099 (at least) line number information.
1101 What might work is to have an optional line number in memory just before the
1102 BASEOP structure, with a flag bit in the op to say whether it's present.
1103 Initially during compile every OP would carry its line number. Then add a late
1104 pass to the optimiser (potentially combined with L</repack the optree>) which
1105 looks at the two ops on every edge of the graph of the execution path. If
1106 the line number changes, flags the destination OP with this information.
1107 Once all paths are traced, replace every op with the flag with a
1108 nextstate-light op (that just updates C<PL_curcop>), which in turn then passes
1109 control on to the true op. All ops would then be replaced by variants that
1110 do not store the line number. (Which, logically, why it would work best in
1111 conjunction with L</repack the optree>, as that is already copying/reallocating
1114 (Although I should note that we're not certain that doing this for the general
1117 =head2 optimize tail-calls
1119 Tail-calls present an opportunity for broadly applicable optimization;
1120 anywhere that C<< return foo(...) >> is called, the outer return can
1121 be replaced by a goto, and foo will return directly to the outer
1122 caller, saving (conservatively) 25% of perl's call&return cost, which
1123 is relatively higher than in C. The scheme language is known to do
1124 this heavily. B::Concise provides good insight into where this
1125 optimization is possible, ie anywhere entersub,leavesub op-sequence
1128 perl -MO=Concise,-exec,a,b,-main -e 'sub a{ 1 }; sub b {a()}; b(2)'
1130 Bottom line on this is probably a new pp_tailcall function which
1131 combines the code in pp_entersub, pp_leavesub. This should probably
1132 be done 1st in XS, and using B::Generate to patch the new OP into the
1137 Tasks that will get your name mentioned in the description of the "Highlights
1140 =head2 make ithreads more robust
1142 Generally make ithreads more robust. See also L</iCOW>
1144 This task is incremental - even a little bit of work on it will help, and
1145 will be greatly appreciated.
1147 One bit would be to write the missing code in sv.c:Perl_dirp_dup.
1149 Fix Perl_sv_dup, et al so that threads can return objects.
1153 Sarathy and Arthur have a proposal for an improved Copy On Write which
1154 specifically will be able to COW new ithreads. If this can be implemented
1155 it would be a good thing.
1157 =head2 (?{...}) closures in regexps
1159 Fix (or rewrite) the implementation of the C</(?{...})/> closures.
1161 =head2 A re-entrant regexp engine
1163 This will allow the use of a regex from inside (?{ }), (??{ }) and
1164 (?(?{ })|) constructs.
1166 =head2 Add class set operations to regexp engine
1168 Apparently these are quite useful. Anyway, Jeffery Friedl wants them.
1170 demerphq has this on his todo list, but right at the bottom.