1.8 KiB
August 15th, 2016 weekly
What went well?
@rewinfrey: Test generation branch got merged in. Completion of the syntax for JS! Progress on diff summaries <3
@joshvera: Deploying semantic-diff went pretty well. Added a bunch of issues for the remaining tasks. It’s a manageable chunk of work for this week.
@robrix: The p,q-gram precomputation branch got merged. Also resolved RWS bias, and got a chance to bounce that off @jbarnette ❤️
What went …less well?
@rewinfrey: We aren’t producing replacement patches, which is a bit of a question mark.
@joshvera: Some nils in Ruby causing some exceptions (tho they were easy to resolve). Almost botched a deploy, that could’ve gone better.
@robrix: Running into #683 at the last mile of the precomputing p,q-grams branch was frankly a bit of a shock.
What did we learn?
@rewinfrey: Learned a lot about ambiguity & contradiction in Haskell’s typechecking & type inference. Very cool seeing how unification is able to resolve types at the very last. Did some pairing coercing a monadic value into a monadic wrapper.
@joshvera: Composing effects in freer. Also that we rotate into platform support on Sunday, as I did, so I’ll be doing that this week.
@robrix: Discovered #683! Also learned that the RWS-Diff paper’s implementation suffers this problem, but they didn’t notice (or at least didn’t note it), probably due to some ameliorating factors (they match up equalities in an earlier pass which we lack, meaning that the most obvious effects of this wouldn’t be felt). This was also a good reminder that some of our correct behaviour is only incidentally captured in our tests (if at all), partly because it’s often difficult extracting a precise property to test given an intuitive grasp of what a correct diff should look like. Edit: Bonus round: .jbarnette me exists!