- Part1: 326 (
326) - Part2: 381 (
381)
See: root README
- newtype pattern (e.g. to differentiate between origin values and diff-values)
- map_windows for sliding window (kernel convolution)
- working with those differencs ("discrete calculus") <-- though direct operation on raw values, recursively looking for alts, would have been simpler for part 2
- fuzzing (as there was a hard to find error case)
- Taking element property to lgoical operation too soon made code hard to read (compiler solves anyway?)
- unstable map_windows requires special allowance in root, not local file
- edge case was poorly handled in a couple ways, and not directly tested
- New types and sum/addition/ reference sum operation
- outputting type flow in iterators
- mapping with slices; Vec<Res<>> inversion
- quickcheck
- rand crate
.get()vs[_](actually off of separate traits, former being sealed; and latter deref's result)QUICKCHECK_TESTS=1_000_000(quick check env var)
- Finer trace-log targeting. (cont'd)
- write tests for edge cases as we go -- accumulates too much uncertainty otherwise
- decomposing code for testability