Jan-Åke Larsson wrote: > Angus Leeming wrote: > > 1. Here I had colour specials in the latex file. In future I will not > > need them, right? Instead I'll pass them direct to dvipng, as above. > > Yes. You can do either.
The DVI colors are preferred, of course. > > 2. I should prefer your -depth and -height flags, extracting the > > metrics info from the output above, rather than extracting it from > > the latex log file? Reason: the log file info is going to be wrong if > > I invoke dvipng with the '-T tight' option, right? (and I would like > > to do so so that I can strip off the front margin of a displaystyle > > equation). > > There are actually three ways to determine the boundingboxes. > > 1) "-T tight" which would make "-depth -height" necessary. You'd get > the boundingbox determined from the ink of the image. > > 2) From the output I see you have used the "tightpage" option to > preview, I have code lying around that will take the boundingboxes > from the specials instead, do you want that? > > 3) "dvipng <dvifile> -" will fire up the stdin interface where you > could input the dimensions yourself per page, as obtained from the > latex run. Once the stdin interface is up you give > "-T 1in,1in -O 40sp,30sp -pp1" and then "-T 3cm,3cm -O 3pt,5pt -pp2" > and so on. In case 2) and 3) you can of course give --depth and read the _pixel_ depth off dvipng's stdout. I was also thinking of doing 2b) and 3b) by allowing the bbox to grow from the (-T) specified bbox to include all ink produced by the DVI. This would eliminate shaved edges, while still keeping the specified bbox, whitespace and such. Opinions? /JÅ -- Linux: The choice of a GNU generation.
