LEMDO 2023-07-04 to 2023-04-07
to : Martin Holmes
Minutes: 340
Worked on sizes for some images in the QME repo, and confirmed that all plays except Leir have properly-sized graphics tags. Leir still has pointers to online images which will have to be localized.
On Wednesday, worked on the extremely thorny issue of the missing collation markers in some of the HTML output; the code which maps annotations and collations to sequences of text segments is very complicated, and it rather looks as though it was originally written for annotations only, with collations added on later. The basic problem seems to be that in some contexts, a span between two anchors which is actually within a line ends up scoped to one or more lines because of the way the two anchors are embedded in the encoding structure, and that means that the anchors are not processed in the normal way, so collation markers are not inserted. I’ve added a new template designed to catch exactly and only that situation, and I know it solves five out of six of my test cases; if there’s no additional fallout from that (duplicate ids are a slight possibility), then I can zero in on the one remaining test case and figure out how that differs.
On Thursday, I did zero in on that case, and devised a workaround for the rare case where we hit this problem; I’ve laid it out in a comment in the file, and I should be able to implement and test that solution fairly quickly.
On Friday, I implemented the plan from the day before, and it did actually work, but it also generated some duplicate app markers. Since the rare edge-cases that give rise to this (only six in the whole collection) are so difficult to predict or diagnose, I’ve now simply added a final pass which deletes duplicate ids. I’m aware of one case in which this results in a missing id, but so far only one; if that’s the only instance, we might solve that by a clever re-encoding of the source, but we won’t really know until we get a successful build.