Issues

• Bound the absolute pixel count a reader builds, before it builds the array

  enhancement
  oskar-j/walsh-hadamard-transform#81 · Filter by author oskar-j opened on Sep 27, 2026
  

• Coverage is 100% but three bounds are unobserved: mutations that survive the whole suite

  enhancement
  oskar-j/walsh-hadamard-transform#69 · Filter by author oskar-j opened on Sep 20, 2026

• Documentation drift: stale --coeff-removal tables, -v on compress, the coverage figure, and two more

  documentation
  oskar-j/walsh-hadamard-transform#66 · Filter by author oskar-j opened on Sep 20, 2026

• README says a pickled .npy is "refused from the header and never loaded"; it is unpickled and read

  documentation
  oskar-j/walsh-hadamard-transform#65 · Filter by author oskar-j opened on Sep 20, 2026

• Three public entry points accept inconsistent state and emit output nothing can read

  bug
  oskar-j/walsh-hadamard-transform#64 · Filter by author oskar-j opened on Sep 20, 2026

• Vectorizer: four rough edges around failed loads, empty channels and small pictures

  bug
  oskar-j/walsh-hadamard-transform#63 · Filter by author oskar-j opened on Sep 20, 2026

• Two hazards in the memo layer: a writable shared matrix and a lossy repr key

  bug
  oskar-j/walsh-hadamard-transform#62 · Filter by author oskar-j opened on Sep 20, 2026

• set_raw_data silently mangles input that set_array refuses by design

  bug
  oskar-j/walsh-hadamard-transform#61 · Filter by author oskar-j opened on Sep 20, 2026

• The per-element Python loops that remain on the pixel path

  enhancement
  oskar-j/walsh-hadamard-transform#60 · Filter by author oskar-j opened on Sep 20, 2026

• A thin PNG collapses the wavefront unfilter: a 662-byte file costs seconds of CPU

  bug
  oskar-j/walsh-hadamard-transform#59 · Filter by author oskar-j opened on Sep 20, 2026