Front page | perl.perl5.porters |
Postings from November 2002
Hello
I have been working further on the assertions patch.
Main change is that now assertion activation status is lexically
scooped. I have defined a new flag HINT_ASSERTING for PL_hints (AKA $^H).
> I'm slightly worried about using up a whole one-character command-line
> flag for assertions - such things are a scarce resource - but it may
> be a fair use for such a thing. I want to think further on Salvador's
> comments about the granularity of the -A switch.
In the new version, the -A switch is used with a list of tags:
perl -Afoo,bar ...
and what it really does is:
use assertions::activate split(/,/,{foo,bar});
in a similar fashion to '-M' or '-d:' switches.
I can also think about a more generic -A flag usage:
perl -Aassert=foo,bar -Adebug=cao ...
that would call
use Activate::assert split(/,/,{foo,bar});
use Activate::debug split(/,/,{cao};
...
(actually I almost prefer this second option to the one I have implemented)
Well, following with assertions::activate, calling
use assertion::activate @tags
convert @tags to a list of regular expressions and stores them in
@{^ASSERTING}.
then in the perl code, one can use the syntax
use assertions 'foo';
and assertions in its scope will be activated if 'foo' matches any of
the RE in @{^ASSERTING}.
You can also use several tags and assertions will be active if all of
them are selected. If no tag is given, the package name is used as the
tag. Finally there is
use assertions '&', @tags;
that checks for the new @tags in addition to the ones already in scope.
In the assertion package attached, there is an example.pl file showing
the different possibilities.
> And I need to look at the code, to see how complete and correct the
> implementation is; Yitzchak already mentioned one issue in this regard,
Following Yitzchak advice, I have updated affected modules (in the first
version of the patch, I had only changed things in the perl core).
> which suggests there'll be some more work to do to complete the patch.
The part where I am more dubious about the correctness of the
implementation is in Perl_ck_subr in 'op.c' when assertions are
optimized out if not activated. I am not completely sure if it is ok to
just free the assertion op subtree and put a new constant in its place.
Other issue is what happens to lexicals declared inside assertion args.
i.e.:
00 use strict;
01 use warnings;
02 no assertions;
03
04 sub out : assertion { print "@_\n" };
05
06 our $o="global";
07
08 sub test {
09 out (my $o="lexical");
10 print "o=$o\n";
11 }
12
13 test();
assertion call in line 9 is deleted but the my declaration is still
remembered.
Alternative behavior would be to open a new scope for the assertion
args, this could be easily done expanding assertion calls to
asserting and assertion(@args)
(being 'asserting' 0 or 1, depending on assertions activation).
Bye,
- Salva