Hi All,
I'm not sure whether oversee the problem or not, but from the profiling
results I got, caching the style sheet will improve the performance by
10%...
Loading the Style Sheet into the Tree structure --- 10%
Loading the Input Document into the Tree structure --- 10%
Processing & Transforming --- 80%
What I try to say is, if this is true, maybe we should start thinking of
other ways to improve the performance.
Thanks.
William.
-----Original Message-----
From: Kaiserovi [mailto:[EMAIL PROTECTED]]
Sent: November 6, 2000 1:23 PM
To: Sablotron Mailing List
Subject: Re: [Sab] Sablotron Performance - we want/need caching
Derek,
thanks a lot for the offer! Without such help, we'd still like to
implement caching ASAP, but then again, there's a lot of pressing stuff.
So if anyone agrees to do a part of the work, it's very welcome and
certainly will speed things up.
I guess that the approach suggested by David should suffice: there'd be
a function that would parse a document on demand and return a pointer.
The pointer could be handed on to SablotProcess instead of a URI (by
means of a struct that could hold one or the other).
This might or might not be complemented by the concept I described
earlier: on demand, a URI would be added to or removed from the list
containing things meant to be cached. This might be easier to use, but
probably would not be applicable to in-memory buffers. I'll appreciate
your opinions on whether this would be worth the work, as well as any
other comments.
The implementation work can start as soon as we reach an agreement about
the overall features.
Thanks again,
Tom
"Sayeau, Derek" wrote:
>
> Is there any plans to implement the discussed caching methods in near
future
> releases of sablotron? We would be willing to assist, if necessary.
>
> Thanks!
>
> -derek
>
> -----Original Message-----
> From: David Hedbor [mailto:[EMAIL PROTECTED]]
> Sent: Sunday, October 08, 2000 10:59 PM
> To: Sablotron Mailing List
> Subject: Re: [Sab] Sablotron Performance - we want/need caching
>
> Kaiserovi <[EMAIL PROTECTED]> writes:
>
> > Yes this sounds reasonable; there should be a function to parse a
> > document and return a pointer to its tree in any case. We'd have to have
> > several versions of each of the processing functions though:
> > SablotRunProcessor, SablotProcess etc. (you might want to pass in a
> > preparsed stylesheet, or the XML, or both) - or is there a more elegant
> > way?
>
> Pass a struct which can contain a filename/url, the data or
> the preprocessed data, with a union for the data and something to
> indicate the type of the data?
>
> --
> [ Below is a random fortune, which is unrelated to the above message. ]
> No matter where I go, the place is always called "here".
