Skip to content

feat(datasets): data-source UI overhaul for public datasets (DATA-4..7)#94

Merged
OmerShubi merged 2 commits into
mainfrom
claude/pedantic-sinoussi-c6586b
Jun 23, 2026
Merged

feat(datasets): data-source UI overhaul for public datasets (DATA-4..7)#94
OmerShubi merged 2 commits into
mainfrom
claude/pedantic-sinoussi-c6586b

Conversation

@OmerShubi

Copy link
Copy Markdown
Contributor

Summary

Reworks the Public datasets sidebar so it scales as the corpus catalogue grows (toward dozens of corpora), while keeping load a local/private dataset the primary path. Implements DATA-4 … DATA-7 from IMPROVEMENTS.md.

  • DATA-4 — Scalable picker. The flat dataset radio is now a searchable st.selectbox showing each corpus' short name + a language · size caption, one-line description, and Dataset home ↗ link. PUBLIC_DATASET_REGISTRY promoted to structured entries (short/language/size/description/link; loader+monitor preserved). Selection still rides public_dataset_choice (full label → deep-link/session round-trips).
  • DATA-5 — No per-source filtering. Removed PoTeC Texts/Readers and MultiplEYE Sessions/Stimuli. Each loader now reads the whole corpus; the global Narrow by trial filters scope it (per the request, default = all the data). Headless load_potec/load_multipleye keep their readers/texts/sessions/stimuli args.
  • DATA-6 — Expected files exposed. Each loader shows an "Expected files" expander listing the directory tree + file-name patterns it looks for.
  • DATA-7 — Download button + detection. Replaced the always-on Download if missing checkbox with a found-vs-Download status (datasets.potec_present / onestop_present / multipleye_inventory) + a one-click ⬇ Download button — an already-downloaded corpus never re-checks the network. Column-mapping unchanged.

Implementation notes

  • New datasets.potec_present / onestop_present; shared app helpers _dataset_dir_input / _dataset_access_status.
  • App-side cached wrappers dropped their readers/texts/sessions/stimuli/download args; the datasets-module functions are untouched.

Behavior change to flag

DATA-5 means PoTeC now loads all 900 trials (75 readers × 12 texts) up front instead of a default subset, then Narrow by scopes it.

Tests

  • potec_present / onestop_present detection unit tests.
  • test_each_public_dataset_loader_ui_renders smoke-tests all three real loaders; test_potec_source_renders updated for the selectbox.
  • Full suite: 585 passed, 3 skipped (one pre-existing unrelated failure: test_group_effect_size, a scipy/NaN env issue that fails identically on main).

Docs

CHANGELOG, scanpath_studio/CLAUDE.md, docs/onestop.md, docs/multipleye.md, IMPROVEMENTS.md (DATA-4…7 → Pending approval).

🤖 Generated with Claude Code

OmerShubi and others added 2 commits June 24, 2026 01:44
Rework the Public datasets sidebar so it scales as the corpus catalogue
grows while keeping local/private upload the primary path.

- DATA-4: replace the flat dataset radio with a searchable selectbox that
  shows each corpus' short name + language·size caption, description, and
  home link. Promote PUBLIC_DATASET_REGISTRY to structured entries
  (short/language/size/description/link; loader+monitor preserved).
- DATA-5: drop the per-source participant/text narrowing (PoTeC
  Texts/Readers, MultiplEYE Sessions/Stimuli). Each loader now reads the
  whole corpus; the global Narrow by filters scope it. Headless
  load_potec/load_multipleye keep their readers/texts/sessions/stimuli args.
- DATA-6: show an "Expected files" expander per loader listing the
  directory tree + file-name patterns it looks for.
- DATA-7: replace the always-on "Download if missing" checkbox with a
  found-vs-Download status (datasets.potec_present / onestop_present /
  multipleye_inventory) + a one-click Download button, so an already
  downloaded corpus never re-checks the network.

Adds datasets.potec_present / onestop_present; shared app helpers
_dataset_dir_input / _dataset_access_status. Tests: detection unit tests +
per-loader access-UI smoke test. Docs synced (CHANGELOG, CLAUDE.md,
docs/onestop.md, docs/multipleye.md, IMPROVEMENTS.md).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
… AOI writes

Review follow-ups for DATA-4..7:

- potec_present now requires every text's word + char AOI files (not just
  one). DATA-5 makes the loader read the whole corpus, so a partial tree
  would otherwise report "Found" and then crash mid-load (_potec_words
  raises on the first missing text) with no Download button to recover.
  Requiring all of them lets the status offer Download, which self-heals.
- download_potec writes each AOI file via a temp + atomic replace (like
  download_onestop), so an interrupted fetch can't leave a truncated file
  that is_file()/potec_present would treat as complete.
- Type the _dataset_access_status download callback param.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
@OmerShubi OmerShubi force-pushed the claude/pedantic-sinoussi-c6586b branch from 41de7d1 to f2169a5 Compare June 23, 2026 22:46
@OmerShubi OmerShubi merged commit 0750728 into main Jun 23, 2026
5 checks passed
@OmerShubi OmerShubi deleted the claude/pedantic-sinoussi-c6586b branch June 23, 2026 22:47
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant