pecheny/dopesuite
Watch
2
0
Fork
You've already forked dopesuite
 0
Code Issues 11 Pull requests Projects Releases Packages Wiki Activity Actions

Architecture deepening: extract six deep modules #20

Merged
pecheny merged 6 commits from architecture-deepening into main 2026-07-23 00:14:50 +00:00
Conversation 0 Commits 6 Files changed 21 +1063 -406
pecheny commented 2026-07-23 00:14:38 +00:00
Owner
Copy link

Why

Six behaviour-preserving refactors. Each carves a testable deep module out of the frontend page monoliths or the match-edit path, following patterns the codebase already uses (sync.js, DopeTable, stage-cache.js, domain/resolver).

  • xy carddraft.js: the card-draft dirty kernel
  • dope stats-sync.js: the live stats loop shared by host and viewer
  • dope domain/matchedit: pure EK match-edit validation
  • dope entry-model.js: the od.js paste/coerce/invert kernel
  • xy handoutsession.js: the handout staging session
  • xy boardmembers.js: the members seam split out

Every extracted module has unit tests. Both apps' full suites pass.

🤖 Generated with Claude Code

# Why Six behaviour-preserving refactors. Each carves a testable deep module out of the frontend page monoliths or the match-edit path, following patterns the codebase already uses (sync.js, DopeTable, stage-cache.js, domain/resolver). - xy carddraft.js: the card-draft dirty kernel - dope stats-sync.js: the live stats loop shared by host and viewer - dope domain/matchedit: pure EK match-edit validation - dope entry-model.js: the od.js paste/coerce/invert kernel - xy handoutsession.js: the handout staging session - xy boardmembers.js: the members seam split out Every extracted module has unit tests. Both apps' full suites pass. 🤖 Generated with [Claude Code](https://claude.com/claude-code)
# Why
board.js held the card editor's draft/baseline as six module globals and
computed the Save-button dirty check inline, reachable only through the DOM —
so its trickiest, most bug-prone rule had no test. Move the draft state and the
dirty/normalize logic into carddraft.js behind a small interface; board.js keeps
the DOM and drives it. The kernel now has jstest coverage; behaviour is
unchanged (verified in-browser: the save/commit/edit/alias cycle is identical).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
# Why
host.js and viewer.js each held a byte-identical copy of the stats-page resync
loop (applyStatsMatchEvent + the throttle/debounce schedulers) over their own
timer globals, untested. Move it into stats-sync.js as a create(deps) factory —
the same injected-dependency shape DopeStageCache uses — so the
throttle/coalesce/seq-gap logic is unit-tested once and the duplicate is deleted.
Both pages now call statsSync.applyMatchEvent. Behaviour unchanged; verified by
the loop unit tests and the page-compile/selector-contract Go test (no seeded EK
fest available locally for a live-SSE browser pass).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
# Why
applyMatchEditTx mixed the pure match-edit rules (team bounds, action shape,
place/theme/answer/mark checks) with tx SQL in the server trunk, so the rules
were reachable only through a booted server + SQLite. Move the validation into a
pure domain/matchedit.Validate that returns a Plan; the server executes the plan
as SQL (which needs the tx, so it stays here — per ARCHITECTURE.md's domain/server
split, and mirroring domain/resolver). The rules now have direct unit tests;
behaviour is unchanged (the EK batch integration tests pass).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
# Why
od.js held its paste-parse, cell-coercion and column-invert logic as loose
functions tangled with the grid DOM, untestable (od.js runs getElementById at
load, so the harness can't import it). Move the three pure pieces —
parseClipboard, coerceValue, invertColumn — into entry-model.js and unit-test
them; od.js keeps the DOM/selection/persistence and calls in. Behaviour
unchanged (verified by unit tests + the page selector-contract Go test; no
seeded OD game available for a live browser pass). Wrapping od.js itself as a
loadable factory stays a follow-up.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
# Why
board.js held the handout image-staging session (session id, staged names,
in-flight guard, heartbeat timer) as loose globals and functions scattered
across the file, untested. Move the lifecycle — stage-once/dedup, single
in-flight, heartbeat, reap→clear, cleanup on close — into handoutsession.js as a
create(deps) factory with injected network + timer primitives; board.js keeps
the image gather/upload and DOM. The state machine now has unit tests;
behaviour unchanged (verified in-browser: the modal opens and closes cleanly).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
# Why
First seam-by-seam step of breaking up board.js: the members/sharing feature
(roster fetch/cache, the sharing overlay, add/remove) moves into boardmembers.js
as createBoardMembers(state, boardId) → { load, open }, owning its own DOM +
listeners and caching the roster onto the shared state. board.js drops ~90 lines
and the feature is now one file. Behaviour unchanged (verified in-browser: the
overlay opens, the roster renders, add-member errors surface). The remaining
seams (notifications, import, export, labels, timeline) are follow-ups.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Sign in to join this conversation.
Reviewers
No reviewers
Labels
Clear labels
No items
No labels
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
pecheny/dopesuite!20
No description provided.