Improving type detection

[David Tardon]

Hi,

On Thu, May 01, 2014 at 09:29:48AM -0400, Kohei Yoshida wrote:
> Hi Maxim,
> 
> On Thu, 2014-05-01 at 15:44 +0300, Maxim Monastirsky wrote:
> 
> > https://gerrit.libreoffice.org/9221
> > https://gerrit.libreoffice.org/9222
> > https://gerrit.libreoffice.org/9223
> > 
> > Any kind of feedback would be appreciated!
> 
> While we are on this subject, I've also received (quite some time ago)
> some feedback from the developer & maintainer of some of the external
> import filter libraries that we use.  Those libraries do handle type
> detection that are plugged into our type detection system, and they are
> all zip-based formats.  He was concerned about having to "detect" zip
> storage over and over again which he rightly said was not great for
> performance.
> 
> So I was thinking about perhaps introducing an additional boolean
> property with the MediaDescriptor - something like "ZipStorageDetected"
> which, when true, the subsequent detection code can choose to skip
> detecting for zip archive.

Yes, that would actually help, but not until we switch the import
libraries to librevenge. Currently the zip (and OLE2) detection must be
done inside the libraries. Alternatively, I could pass the
WPXSvInputStream instance in UserData, so subsequent detection functions
for external libs could re-use it. That might save even more time... It
would also be pretty well localized change: a few places in
writerperfect module.

D.