On Tuesday, November 26, 2013 3:45:19 AM UTC-6, Edward K. Ream wrote: Summary > > - Leo's sentinels are the biggest barrier to more widespread adoption of > Leo. Not documentation. Not anything else. > - @bool use_at_shadow_everywhere would allow us to test @shadow everywhere > without changing .leo files. > - @shadow could benefit by using Leo's existing caching mechanism. > - Some (relatively simple!) way must be found to allow @shadow to guess > correctly most of the time. >
Some recent thoughts: 1. Iirc (never a safe bet), @shadow uses more verbose old-style sentinels. Leo's new-style sentinels are provably the simplest thing that could possibly work. It's likely that adapting the @shadow algorithm to use new-style sentinels may have advantages. It's worth investigating. 2. The acid test of @shadow is how well it does, in fact, guess the location of inserted lines. An Aha: Leo's bzr log provides the ultimate test bed. We "mine" the bzr log for diffs. These diffs include sentinels, so they show the *optimal* operation of @shadow. Stripping sentinels from the "incoming" diffs provides the input to the @shadow algorithm. We can then compare the output of @shadow with the actual diffs. In other words, the contents of the bzr log provide the foundation for a huge unit test. 3. Given the importance of @shadow, any reasonable amount of work that improves its performance (making hard guesses) is justified. The present idea is to add features to the code that will clarify the operation of the algorithm and allow for experimentation in heuristic inferences. Edward -- You received this message because you are subscribed to the Google Groups "leo-editor" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at http://groups.google.com/group/leo-editor. For more options, visit https://groups.google.com/groups/opt_out.
