In message <[EMAIL PROTECTED]>
          Adam Blokus <[EMAIL PROTECTED]> wrote:

> > Ok. That sounds useful. On the subject of performance, perhaps we need
> > some kind of progress indicator, if export's going to take some time.
> 
> The problem with performance that I noticed is that some pdf readers (like 
> kpdf, which I use mostly) need quite a lot of time to display a generated 
> file. Although - adobe reader works well.
> I have already some ideas to try later - playing less with the pdf stack, 
> changing modes (text, page description etc) less often etc. I am not sure 
> whether this will give any other improvement than a minor file size decrease, 
> so I didn't dedicate myself to it yet.

I would be surprised if the save/restore of PDF gstate would be the
performance killer for kpdf.  My initial bet is on spending time in
decompressing (especially image data) and detokinisation.

> Also - this changes seem to be related 
> with haru (kpdf complains sometimes about the pdf grammar in the file).

This is worthwhile to investigate.  If you're stuck why this is happening
feel free to pass on an example file.

Just be aware that Adobe Acrobat is very forgiving at syntactical and
semantic problems in PDF.  It even deals with xref tables missing (and
rebuilds it) and line ending screw ups.

> Generating a pdf file takes mostly not even enough time to notice it, 
> sometimes about 1 or 2 seconds - so I think there is no need to indicate the 
> progress of this process.
> I will take a deeper look at this issue either during the second or during 
> the 
> fourth week of my work. It is mostly not noticeable, so it doesn't have a 
> very high priority for me.

Again when we know of PDF readers complaining about the produced files, it
is important to get to the bottom of it as I would like to rule out as
much as possible the possibility of bad PDF file generation.  It could be
a problem of the PDF reader and then we can consider finding a workaround.

I wouldn't worry too much about performance of PDF readers dealing with
the files you're writing.  At least in first instance.

John.
-- 
John Tytgat
[EMAIL PROTECTED]

Reply via email to