For example, according to the specification, what should happen if two referenced block following each other are in the same page- sequence as the index and after the index and at the beginning of a new page? When merging, one block would move to the previous page, therefore the page number would change, therefore no merging, so the block moves back one page, then the numbers could be merged, but the the block would move again. I think I'm getting a headache.
Apart from my earlier remarks, note that I personally do not yet consider merging to be a showstopper for a basic implementation. It would already be pretty cool if we could get to the point where the index just generates the sequence of page-numbers. Merging indeed magnifies some of the issues, which are quite unavoidable with FOP's current design. Without merging, if we model the index to what PageNumberCitationLM does, the reserved space will, in most cases, be larger than what is actually needed. I think, if one would take the time to artificially generate a first page-sequence with pages containing a lot of citations pointing towards the end of the document, you would already see side-effects to some extent. The actual page-numbers cannot be resolved before the line-breaks are computed, so... and here I'm not entirely certain. I have not yet run such test extensively myself. I think that we would end up with lines that are too short. Very visible to the naked eye in cases of justified alignment...
At any rate, the more citations in sequence, the more this issue is magnified. We could end up with a block of citations, with a layout of 3 lines, but the actual content could have been put on a bit more than one line.
The biggest obstacle remains: FOP does not revisit line-breaks. The correction is done in the area tree, where the area corresponding to the page-number is always a descendant of a line-area. I don't think the effect you describe above could occur with the current design. The result would rather be a set of line-areas with too little content. The blocks would remain on the same pages, but they could turn out to look rather ugly...
HTH! Andreas --------------------------------------------------------------------- To unsubscribe, e-mail: fop-users-unsubscr...@xmlgraphics.apache.org For additional commands, e-mail: fop-users-h...@xmlgraphics.apache.org
