Improve SimplePie memory usage

[link92]

If we are to use SimplePie for importers (see #7400, for example, but also the RSS, Blogger, etc. importers), then the memory usage of the version included in 2.8 will be problematic. We should probably backport changes between r1032 and r1036 of SimplePie. Attached is a patch that does this.

[Denis-de-Bernardy]

see #7652

[Denis-de-Bernardy]

failed to find their revision log, for reference.

[Denis-de-Bernardy]

see also #7644

[link92]

​http://bugs.simplepie.org/repositories/revisions/sp1

[Denis-de-Bernardy]

Thanks. Out of curiosity, do you know why they're not using php's built-in parse_url() function?

​http://bugs.simplepie.org/repositories/diff/sp1/trunk/simplepie.inc?rev=1036

[Denis-de-Bernardy]

see also #9524

[link92]

Replying to Denis-de-Bernardy:

Thanks. Out of curiosity, do you know why they're not using php's built-in parse_url() function?

​http://bugs.simplepie.org/repositories/diff/sp1/trunk/simplepie.inc?rev=1036

We don't, because, uh… um, I wrote that code years ago? I think the answer is something along the lines of a few things in the manual:

• "Invalid characters are replaced by _." (This goes against what is needed for web compat.)
• "On seriously malformed URLs, parse_url() may return FALSE and emit a E_WARNING." (We want to break up everything before we try and make it valid, again for web compat.)
• "This function doesn't work with relative URLs." (Yeah, we parse those a lot, turning relative URIs into absolute ones.)

If the manual is wrong on all three things I'll remove our own parse_url. :)

[link92]

Also, the parse_url() function will go in SP 1.2 eventually, as everything will move over to the IRI class (actually, more likely the parse_url() will remain for backwards compatibility and just call the IRI class).

[link92]

Better patch, now including r1042

[ryan]

(In [11599]) Reduce SimplePie memory usage. Props link92. fixes #10147 for trunk

[ryan]

(In [11600]) Reduce SimplePie memory usage. Props link92. fixes #10147 for 2.8.1