This week, the P5P summary will attempt to entertain you with several low-level hacks. (A weird kind of entertainment if any.) Read about printf(), optimisations, internals, perldoc, and other code stories below.
Nick Ing-Simmons changed (or restored, depending on the point of view) the
meaning of printf() in XS code. As of Perl 5.8.0, printf() was
#define'd to be an alias to PerlIO_printf(PerlIO_stdout(),...). Now,
it will be the bare C-level stdio printf(). Notably, when the standard
output is redirected from a Perl program, and when the perlio layer is
used (e.g. when the PERLIO environment variable has the value perlio),
printf() will continue to write to the pristine stdout stream. Thus, it's
now possible to have access to the perlio and to the stdio implementations
from XS.
This eventually caused some smoke tests to fail.
Mark-Jason Dominus wants a way to modify, at run-time, the prototype of a subroutine, and explained why (bug #20463). Rafael Garcia-Suarez said that this doesn't need to be in the core, and proposed a naive Inline::C-based implementation. However, as MJ wants this feature for Memoize, which is a core module, it needs to be in the core : perhaps in the Scalar::Util module ?
,= Joachim Breitner, wondering how to merge hashes efficiently, proposed to
introduce a new operator,,=, so that
%hash1 ,= %hash2;
would be semantically equivalent to
%hash1 = (%hash1, %hash2);
but faster (and it would work similary for arrays). Mark-Jason Dominus
pointed out that push() already merges arrays, and that some time ago
there was discussion of extending push to operate on hashes also.
Nicholas Clark objected to this new operator, as efficient merging of
hashes can be done in a XS module ; and he provided this XS module,
Hash::Append. (Though perhaps this code belongs to the core module
Hash::util.) Abigail, on the other hand, approved the creation of a
new operator, but dislikes the odd-looking,=.
Dave Mitchell fixed the debugger slowdown noticed last week.
Mark Mielke noticed that the array of back-references attached to a reference never shrinks, and he proposed a patch to improve the speed of the routine that weakens references.
There was a thread about the different
ways to modify the fileno value
of a FILE C structure on various platforms.
Dan Kogai released Encode v1.85, and, a few hours later, v1.86. Note that some tests are skipped on perl 5.8.0, since they rely on newest fixes.
Rafael Garcia-Suarez sent a fix for bug #18232 : the $1...$n
variables loose their utf8 flag when they're overriden by another
pattern match in an enclosed scope.
David Nicol reports (bug #20503) that perldoc produces confused output on
RedHat 8.0. Guess what, that's probably caused by UTF8 locales. Russ
Allbery blames groff. Adrian Enache suggests to pass the option
-Tlatin1 to [ng]roff (and to change
Adrian Enache also reports that reading from a filehandle in fixed-length record mode, and appending the results to a variable, doesn't work correctly since Perl 5.7.2. He also fixed it.
Alain Barbet posted some graphs about the smoke test results.
Sean M. Burke advices Windows users to upgrade their perldoc to the newest distribution (Pod::Perldoc) he put up on CPAN. The original perldoc distributed with 5.8.0 apparently has problems with some pagers (e.g. more on some flavours of Windows.)
This summary brought to you by Rafael Garcia-Suarez. Summaries are available on http://use.perl.org/ and/or via a mailing list, which subscription address is perl5-summary-subscribe@perl.org . Comments and corrections are welcome.