xmlreader is fast XML pull parser ?

[Michael Meeks]

On Tue, 2016-05-10 at 13:49 +0200, Eike Rathke wrote:
> On Monday, 2016-05-09 11:38:25 +0800, LGH710681042 at gmail.com wrote:
> > I know xmlreader is currently used by configmgr, when we start up LO, configmgr 
> > will parse *.xcd files in /share/registry directory. It will takes several seconds.
> > so i guess that xmlreader is not real fast, i want to know how can we optimize the speed of parsing files? 
> 
> Run with callgrind and investigate with kcachegrind where the actual
> bottle necks are. 

	Having done this in the past (and make sure you don't use a dbgutil
build - since this makes xmlreader very much slower than other pieces
due to the nice assertions in it) - there are a few obvious things:

	* configmgr XML uses multiple namespaces
		+ this makes the XML bigger and much slower to parse
		+ supposedly this helps with DTDs (or something) but ...

	* configmgr XML parsing is single threaded
		+ it would be far more ideal to parse in multiple
		  threads, even phones have spare un-used threads these
		  days.
		+ however this needs some care with code-structure and
		  layering.

	Beyond that, I've spent a little while reading the code there, side by
side with the profile and looking for obvious sillies without great
success =)

	HTH,

		Michael.

-- 
 michael.meeks at collabora.com  <><, Pseudo Engineer, itinerant idiot