Hydrate pseudo columns on card output

[tahadx]

Fixes #202

Problem

Cards sitting in Fizzy's built-in lanes — Maybe, Not Now, Done — have no real column, so the API returns an empty column object ({"id":"","name":""}). The CLI passed that through verbatim, leaving --json consumers without a usable column id or name:

"column": { "created_at": "", "id": "", "name": "" }

Upstream behavior confirmed in basecamp/fizzy#2596 and basecamp/fizzy#3017; the API is unlikely to change soon, so this hydrates client-side.

Fix

When a card payload arrives with an empty column.id, infer its lane from fields the API already provides and emit the same pseudo-column shape that fizzy column list prints:

• closed == true → Done (done, kind closed)
• postponed == true → Not Now (not-now, kind not_now)
• otherwise published → Maybe (maybe, kind triage)
• drafts belong to no lane upstream and are left untouched

"column": { "id": "maybe", "name": "Maybe?", "kind": "triage", "pseudo": true }

The inference is safe because the states are mutually exclusive upstream: closing a card destroys its not_now record, and postponing clears its column. Hydrated ids round-trip back into the CLI (fizzy card column <n> --column maybe).

Applied to: card show, card list, search, and board stream/closed/postponed.

Testing

• New unit tests in internal/commands/pseudocolumns_test.go: per-lane inference, closed-over-postponed precedence, draft/real-column/unknown-payload no-ops, list handling, plus command-level tests for card show and card list.
• make fmt-check vet tidy-check and make race-test (includes TestSurfaceSnapshot) pass; no CLI surface change.

---

Summary by cubic

Hydrates pseudo columns on card output so JSON always includes a usable column id/name. Previously, cards in built-in lanes (Maybe, Not Now, Done) returned an empty column; now the CLI infers the lane and emits the same pseudo-column shape as column list.

• Inference: closed → Done (done, kind closed); postponed → Not Now (not-now, kind not_now); published → Maybe (maybe, kind triage). Drafts and cards with a real column are unchanged.
• Applied to: card show, card list, search, board stream, board closed, board postponed.
• Output shape example: {"id":"maybe","name":"Maybe?","kind":"triage","pseudo":true}; supports round-trip via fizzy card column <n> --column maybe.
• Safety: upstream states are mutually exclusive; closed takes precedence over postponed.
• Migration: none. JSON consumers can rely on column.id for built-in lanes.

^{Written for commit 4911fde. Summary will update on new commits.}

[Copilot]

Pull request overview

Hydrates missing card column metadata for Fizzy’s built-in lanes, resolving #202 for JSON consumers.

Changes:

• Infers Maybe, Not Now, and Done pseudo-columns.
• Applies hydration across card, search, and board outputs.
• Adds unit and command-level coverage.

Tip

If you aren't ready for review, convert to a draft PR.
Click "Convert to draft" or run gh pr ready --undo.
Click "Ready for review" or run gh pr ready to reengage.

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
internal/commands/pseudocolumns.go	Implements pseudo-column inference and hydration.
internal/commands/pseudocolumns_test.go	Tests inference, exclusions, lists, and card commands.
internal/commands/card.go	Hydrates card list and show output.
internal/commands/board.go	Hydrates closed, postponed, and stream output.
internal/commands/search.go	Hydrates search results.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[jeremy]

Fresh pass over 4911fde, checked against the fizzy model rather than the CLI alone.

Correctness. The inference matches upstream. cards/_card.json.jbuilder emits column only if card.column, so a card in any built-in lane arrives with no column at all; closed? is "a closure exists", postponed? is open? && published? && not_now.present?, and awaiting_triage? is active-with-no-column. That is exactly closed → done, postponed → not-now, published → maybe, drafted untouched. The closed-over-postponed precedence can't actually diverge (postponed? already requires open?), so it's a harmless belt.

Two shapes reach hydrateCardColumn: the SDK's non-pointer Column field survives normalizeAny as {"id":"","name":"","created_at":""} (omitempty doesn't drop a struct), while the --all paths go through jsonAnySlice over the raw pages, where the key is absent entirely. The nil-map read handles both, but no test covers the absent-key case — one more t.Run in TestHydrateCardColumn with no column key would keep the --all path honest.

Consistency. The mutation outputs still emit the raw column: card create (a fresh published card lands in Maybe), card update, card close, card reopen, card postpone, card resume, card untriage and card move all go through printMutation(normalizeAny(data)) without hydration. So fizzy card close 5 --json reports column.id: "" while fizzy card show 5 --json says done a second later. Same one-liner at those sites would close the gap; fine as a follow-up if you'd rather keep this PR to the read paths, but it's small enough to fold in.

One semantic note for the doc comment. close upstream doesn't clear column, so a card closed from a real column keeps that column in the API and is not hydrated — column.id == "done" is therefore not the way to detect closure; closed stays the signal. Worth a sentence on inferPseudoColumn so a consumer doesn't read column.id as the lane of record.

Verification. go test ./..., go test -race -count=1 ./internal/..., go vet, gofmt -l, go mod tidy -diff and make surface-check are all clean on this branch locally. The GitHub workflow runs for this head are sitting in action_required (first-time contributor), so the checks haven't run there yet; that's a maintainer button, not something I'm approving from here.

Fixes #202 as scoped; nothing here needs a change before merge, and I haven't pushed to the branch.