- What cost the most: two avoidable rounds.
- My first clippy and test run reported almost nothing back, because I chained
export and S= with && before the &, which pushed the whole chain into the background. I had to rerun.
- Clippy then showed that four existing Rust
it files call Foreign::of, .library() and Foreign::default(). The brief’s caller list (project.rs, modules.rs) left out the tests, so I found them only after the build.
- What I looked up that the brief could have handed me:
- How tools files iterate a
BTreeMap (a plain for (k, v) in m; there is no .iter()), and that BTreeSet has no extend.
- That
fs::read returns Bytes and is async, so the moved function needs block_on on the Rust side.
- That
read_dir returns sorted names, not paths.
- That
Ledger::blank is already in Nikaia (ty.nika).
- Whether two generic host parameters (
[M: A, H: B]) lower; I wrote a probe for it.
- The signatures of
crate_root and read_beside.
- Checks run: two
lower-std runs to fix refusals, cargo check, one fast build, a fixpoint run, the brief’s it filters plus the guard set, the four adapted test modules, --lib project::, nikaia test --project tests/compiler, clippy, pre-push-check and the score.
- Wasted: the first clippy and test run (the shell mistake above) and
pre-push-check before badges.py, although builder-facts already warns about that badge.
- Missing: the commit script seems to have dropped my
Rs2nika: line from the report, which now ends at Disk freed:. The figure was 21 of 79 lines kept, under 60 %. The supervisor should add it back.
- One change that would have saved the most: have the brief’s Callers list grep
crates/nikaia/tests/it/ as well. Also add one line to builder-facts listing what tools code can do with collections: maps and sets are walked with for, there are no .iter() chains on them, and Bytes is copied into a Vec[u8] with a loop. That line would also have told the converter not to write those chains in its draft.
This site is open source.
Improve this page.
