http://bugzilla.spamassassin.org/show_bug.cgi?id=3841
------- Additional Comments From [EMAIL PROTECTED] 2004-09-28 17:57 -------
hmm. actually, both the before and after of that patch are less efficient than
they could be.
- before: is using closures (ie. "sub {
$varfromoutsidescope->method($othervars) }"), which are explicitly called out as
being slower than simple sub refs
- after: is using eval of a string, and closures, so double-slow
a better solution would be to do the following:
- replace the "sub { ... }" callbacks with function references of the methods
being called, e.g. "start_document => [ \&html_start ]".
- set a member on the HTML::Parser object to pass "out of band" any data that
needs to be passed, and then (in the callback) use that to get the
Mail::SpamAssassin::HTML object $self.
so e.g.:
my $hp = HTML::Parser->new(
api_version => 3,
handlers => [
start_document => [ \&html_start, "self" ],
start => [ \&html_tag, "self,tagname,attr,'+1'"],
end_document => [ \&html_end, "self" ],
end => [ \&html_tag, "self,tagname,attr,'-1'"],
text => [ \&html_text, "self,dtext"],
comment => [ \&html_comment, "self,text"],
declaration => [ \&html_declaration, "self,text"],
],
marked_sections => 1);
$hp->{cbobject} = $self;
and change those callback methods to get $self from the first argument passed
into them.
Would that work with ithreads? afaics it should, because they're "real"
functions defined in the source, not closures.
(PS: A better solution long term would be to use HTML::Parser in its subclassed
form, instead of with these callbacks; that way, $self in the HTML::Parser
callbacks is $self for Mail::SpamAssassin::HTML too ;) one for 3.1.0 maybe.)
(PPS: I guess this means we can't switch to using Singletons for certain objects
for speed, and assuming that there'd only be one Mail::SpamAssassin object per
interpreter, as we were just discussing last night ;)
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.
