Add tooltips to icon-only buttons - #264
Conversation
|
Important
This repository does not receive automatic reviews because it has fewer than 10 stars. ⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Plus Run ID: 📝 WalkthroughWalkthroughThe change adds localized tooltips, enforces ChangesTooltip, focus, and phrase-strip enhancements
Estimated code review effort: 3 (Moderate) | ~30 minutes Merge Risk: 🟡 Moderate · up to At the current head, a same-group focus change can leave the continuous strip invisible, while some tooltips may still show an unresolved localization key and arc rendering may reserve excess space. These are bounded user-visible correctness issues, so the PR is not merge-ready until they are fixed or explicitly accepted. Sequence Diagram(s)sequenceDiagram
participant Interlinearizer
participant FocusProvider
participant ContinuousView
participant SegmentListView
Interlinearizer->>FocusProvider: provide shared focus state
FocusProvider->>ContinuousView: provide focus and navigation actions
FocusProvider->>SegmentListView: provide focus and selection actions
ContinuousView->>FocusProvider: update focused token
SegmentListView->>FocusProvider: select focused segment
Suggested reviewers: 🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
alex-rawlings-yyc
left a comment
There was a problem hiding this comment.
@alex-rawlings-yyc resolved 1 discussion.
Reviewable status: 0 of 25 files reviewed, all discussions resolved.
c66c342 to
62c8bac
Compare
alex-rawlings-yyc
left a comment
There was a problem hiding this comment.
@alex-rawlings-yyc made 5 comments.
Reviewable status: 12 of 25 files reviewed, 5 unresolved discussions (waiting on imnasnainaec).
src/components/ArcOverlay.tsx line 317 at r2 (raw file):
Previously, imnasnainaec (D. Ror.) wrote…
⛏️ I'm not sure but this might be overexplaining tw functionality.
Agreed — how tw-merge groups variants is Tailwind's to teach, not this file's. Cut to the constraint that actually has to hold, which is what a reader needs to know before touching the classes.
src/components/TokenChip.tsx line 391 at r2 (raw file):
Previously, imnasnainaec (D. Ror.) wrote…
⛏️ I'm not sure but this might be overexplaining tw functionality.
Same trim as the ArcOverlay one.
src/components/TokenLinkIcon.tsx line 289 at r2 (raw file):
Previously, imnasnainaec (D. Ror.) wrote…
⛏️ much of this comment appears to be pr-diff-justifying and lacking persistent value.
Agreed — the "the reason outranks the action" half was arguing for the choice rather than describing the code, which our comment rules call out directly ("the audience is the next reader of the code — never the reviewer of the current diff"). Kept the part that answers a real question for the next reader, namely which disabled states are deliberately silent.
src/components/TokenLinkIcon.tsx line 356 at r2 (raw file):
Previously, imnasnainaec (D. Ror.) wrote…
⛏️ putting a span between a tooltip and a disablable child is standard procedure and doesn't require a comment.
Agreed, dropped. It appeared twice in this file — the unlink trigger above had the same comment — so both are gone.
src/components/tooltip-delay.ts line 4 at r2 (raw file):
Previously, imnasnainaec (D. Ror.) wrote…
⛏️ "does not open on a different clock..." could be reworded away from the double-negative to a positive about what it does.
Reworded to the positive. Naming the two trees was also documenting consumers, which goes stale silently as trees are added or moved, so that went too.
alex-rawlings-yyc
left a comment
There was a problem hiding this comment.
@alex-rawlings-yyc made 1 comment.
Reviewable status: 8 of 25 files reviewed, 1 unresolved discussion (waiting on imnasnainaec).
src/utils/phrase-arc.ts line 10 at r3 (raw file):
Previously, imnasnainaec (D. Ror.) wrote…
⛏️ Unless this "HALF" is half of something that still exist, Devin suggests something functionally descriptive like
CONTROLS_ARC_OVERHANG_PXorCONTROLS_ARC_RISE_PX.
Done — renamed to CONTROLS_ARC_OVERHANG_PX, and swept the same stale "half" framing out of the computeStripTopPadding and computeStripRowGap docs.
It wasn't just vague — 12 hasn't been half of anything since CONTROLS_HEIGHT_PX = 28 landed as its own constant (28/2 is 14).
Left the value at 12; changing it to CONTROLS_HEIGHT_PX / 2 is a row-spacing change, not a naming fix.
7ffcd6a to
5f9b0b7
Compare
Every control that showed only an icon had its wording reachable by screen readers alone; each now names its action on hover. The mock `Tooltip` gained the provider requirement the real one enforces, which caught the view-options dropdown rendering outside every provider.
Its sibling branch already ran its label through `resolvedOrEmpty`, so a cross-segment slot could show a raw `%…%` key as hover text while PAPI's localization was still in flight.
12 stopped being half of anything once the pill was measured directly and CONTROLS_HEIGHT_PX = 28 became its own constant. Name the overhang for what it is and drop the stale framing from the two docs that echoed it.
An external focus move fades the strip out and defers the displayed-focus update until the fade completes. The reveal was then left to the scroll effect, which is keyed to focusPhraseIndex — so a move landing in the group already displayed never triggered it, and the strip stayed at opacity 0 indefinitely. Reveal from the focus effect instead, whenever the displayed focus has caught up with the live one. Gating that on having canceled a live fade does not work: a fade that reaches its timeout clears its own handle before committing the displayed ref, so there is nothing left to cancel by the time the effect re-runs. cancelPendingFade no longer reports whether it canceled anything, since no caller reads it.
5f9b0b7 to
604b1b2
Compare
The overhang constant's doc claimed the controls pill "sits centered on the arc top", which the pill's own positioning contradicts: it renders fully above its anchor. That framing also made 12 read as a stale half of the 28px pill height, when the two are independently tuned for two different reservations. Say what the value is for and that it is not a fraction of the pill. The mock Tooltip spread every prop it received onto the cloned trigger, including `open`, which addresses the real tooltip's visibility and means nothing on a DOM element. It was inert -- React drops false booleans, and the truncation-tooltip stub never opens -- but the pass-through is documented as serving outer `asChild` triggers, and `open` is not that. Drop it and note the exception.
The 14 per-component "shows no ... unresolved key" tests each check that one control resolves its label before handing it to TooltipContent. That property is checkable once, in the mock Tooltip, which every tooltip's text passes through on its way to the trigger. The check finds call sites no per-component test covers, and the existing 14 keep passing unchanged: they resolve the key to '' before it reaches TooltipContent, so the invariant never sees them. It caught three unguarded sites. SplitMarker and BaselineSplitGap render their label raw, so holding Alt during PAPI's localization window shows %interlinearizer_boundaryControl_split% as hover text; both now take the resolvedOrEmpty path their siblings use. CatalogRowView needs a different fix, since its gloss is visible cell text as well as tooltip text and blanking it would empty the column -- it falls back to an em dash until the lookup lands. renderOverlay now takes prop overrides rather than a built element.
"Token" is the parser's term for a unit of segmented text; readers
glossing scripture see words. Renames the visible English only --
localize keys and {token} format parameters keep their names, as neither
reaches a user.
alex-rawlings-yyc
left a comment
There was a problem hiding this comment.
Both reproduce, and there was a third. Running your check across the suite fails 3 suites / 29 tests with exactly two distinct messages — %interlinearizer_boundaryControl_split% and %interlinearizer_analysisCatalog_noGloss%.
PhraseStripParts.tsx:233 (SplitMarker) and CatalogRowView.tsx:89 are fixed. The third is SegmentView.tsx:200 — BaselineSplitGap renders <TooltipContent>{splitLabel}</TooltipContent> raw off the same %interlinearizer_boundaryControl_split% key, which is why it hid under that message in the dedupe. Both split sites now take the resolvedOrEmpty + tooltipContentOrUndefined path their siblings use; aria-label keeps the raw value as before.
You were right that CatalogRowView needs a different fix. glossLabel is visible cell text, so resolvedOrEmpty alone would blank the gloss column — it now falls back to an em dash, which reads as "no gloss" in any language and is replaced by the real placeholder once the lookup lands. One knock-on: an existing test asserted the bare key as the visible cell text, i.e. it pinned the bug. I changed it to expect the em dash and added a second test for the resolved placeholder.
BaselineSplitGap's new guard also got the two tests it earned rather than a coverage ignore, since the unresolved-key branch is the behavior being fixed.
@alex-rawlings-yyc made 4 comments.
Reviewable status: 7 of 32 files reviewed, 3 unresolved discussions (waiting on imnasnainaec).
src/__tests__/components/ArcOverlay.test.tsx line 45 at r4 (raw file):
Previously, imnasnainaec (D. Ror.) wrote…
Idea you can take or leave:
function renderOverlay(overrides: Partial<Parameters<typeof ArcOverlay>[0]> = {}) { return render(withTooltipProvider(<ArcOverlay {...requiredProps()} {...overrides} />)); }
Done — the invariant is in the Tooltip mock, essentially as sketched.
One correction worth recording, because it changes the argument in your favor: I expected the invariant to break these 14 tests and it doesn't. All 7 suites pass with it in place. The reason is the point — these components resolve the key to '' before it reaches TooltipContent, so the invariant never sees them. It checks the same property from the opposite side, non-conflictingly. So it's additive rather than a replacement, and I've kept all 14 along with the positive tests above them.
I confirmed it's load-bearing rather than decorative: reverting one of the fixes below makes it throw, restoring it makes the suite green again.
imnasnainaec
left a comment
There was a problem hiding this comment.
@imnasnainaec reviewed 5 files and all commit messages, made 2 comments, and resolved 3 discussions.
Reviewable status: 12 of 32 files reviewed, 2 unresolved discussions (waiting on alex-rawlings-yyc).
__mocks__/platform-bible-react.tsx line 1033 at r5 (raw file):
const text = tooltipContentText(tooltipText); if (UNRESOLVED_LOCALIZE_KEY.test(text)) { throw new Error(`Tooltip content carries an unresolved localize key: ${text}`);
Now that this throws, the per-component "shows no … unresolved key" tests can't fail through their own assertions. (A raw-label regression throws here during render, before any not.toHaveAttribute('title') runs, and a correct implementation never sets a title either way.) Let's delete them all.
src/__tests__/components/PhraseStripParts.test.tsx at r5 (raw file):
⛏️ Worth a test that SplitMarker's hover text reaches the trigger, mirroring names the split action on hover over the gap.
Every control that showed only an icon had its wording reachable by screen readers alone; each now names its action on hover. The mock
Tooltipgained the provider requirement the real one enforces, which caught the view-options dropdown rendering outside every provider.This change is
Summary by CodeRabbit
New Features
UI Improvements
Bug Fixes