LEMDO 2024-04-22 to 2024-04-26
to : Martin Holmes
Minutes: 970
On Monday, completed a first pass through the rewrite of the overlap code, and committed, but the build broke because (I think) at the end of the text I’m failing to clear any still-in-effect anchors.
On Tuesday, discovered that the actual problem was links whose endpoint was not after their beginning point, so I started by writing a diagnostic to find these. There were 33, and after fixing them, I was able to get a build to run locally. However, it ran out of memory doing staticSearch, so I switched the project over to the dev branch. I fixed a few other issues at the same time. However, the build time is still substantial, and I would like to look at other ways of cutting it down, including cutting file sizes by using shorter identifiers, and perhaps removing the HTML version of the A-Z index. We finally got a working build, and the results are really pleasing. There’s some fallout for Douai, but I can fix those fairly straightforwardly, I think.
On Wednesday, I followed up by fixing all the issues on Douai except two, on which I’m waiting for feedback from CS. Then I added the links from the annotations pages back into the texts, now that the markers are available in the texts to act as targets. Also worked through existing tickets to push forward with a couple.
On Thursday, builds were plagued with out-of-memory errors, so I started on strategies to mitigate this problem, via several GH tickets. I stopped building the individual person pages which are never linked from anywhere and are mostly empty, and we created a list of ids of documents which never need to be built, which are added to an exclusion list; we’ll see if that works, and if so, how much time it saves.
On Friday, I kept working on these issues, and by the end of the day I had a build on Jenkins that took just over two hours, and a partially parallelized build that would run in 42 minutes on my desktop. This is progress!