On 31 Mar 2011, at 15:08, Dennis van Zoerlandt wrote: Hi Dennis
> In the meanwhile I have tested a few things. In the attachment you'll find a > FO file ( http://old.nabble.com/file/p31286241/fop1.0-5000-fo.zip > fop1.0-5000-fo.zip ) which has scrambled data because of confidentiality. > > I created the FO file with XMLspy and tried to create a PDF file with Apache > FOP 1.0 (fop.bat) on my Windows XP workstation. It produced (what it seems) > this error (see below). No PDF file was created. It seems like the classic "cram all content into one page-sequence" issue. With a file of that size, there is little or nothing you can do. The current architecture of FOP does not allow to render such documents without a sufficiently large heap. That said: I wrote the above while I was running your sample file (with FOP Trunk, using Saxon as XSLT/JAXP implementation), and it just completed on my end, with a heap of 1GB. It did take about 7 minutes, but still... I got a nice output file of 455 pages. I doubt that it is related to images, as there is only one fo:external-graphic. Do you have font auto-detection enabled, by any chance? That might consume an unnecessary amount of heap space, for example, if you only actually use a handful of custom fonts, but have a large number of those installed on your system. Another option is that some fixes for memory-leaks, applied to Trunk after the 1.0 release, are actually helping here. > Splitting the XML input file into several chunks is not a preferable option > for me, nevertheless it is a valid one. Note: it is, strictly speaking, not necessary to split up the input so that you have several FOs. What would suffice is to modify the stylesheet, so that the content is divided over multiple page-sequences. If you can keep the size of the page-sequences down to, say, 30 to 40 pages, that might already reduce the overall memory usage significantly. There are known cases of people rendering documents of +10.000 pages. No problem, iff not all of those pages are generated by the same fo:page-sequence. Regards Andreas --- --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
