Route Search previews through the shared LHN alternate text pipeline by sosek108 · Pull Request #99062 · Expensify/App

sosek108 · 2026-08-20T11:54:16Z

Explanation of Change

Makes SearchRouter / Search row subtitles (alternateText) identical to LHN by construction.

Search options now compute their preview text via the shared getReportAlternateText - same as LHN
Deletes the interim helpers added in Fix alternate text prefix mismatch between LHN and search #98961 (getChatPreviewParts, isActionWithCustomAlternateText)

A new test file tests/unit/LhnSearchPreviewParityTest.ts will assert strict LHN === Search equality on identical Onyx state.

Fixed Issues

$ #98958
PROPOSAL:

Tests

npx jest tests/unit/LhnSearchPreviewParityTest.ts tests/unit/SidebarUtilsTest.ts tests/unit/OptionsListUtilsTest.tsx
Use large account with a lot of reports
Go through LHN and Search bar (CMD+K or magnifying glass icon) and compare subtitle of entries. These should be same for every entry

Offline tests

N/A

QA Steps

Use large account with a lot of reports
Go through LHN and Search bar (CMD+K or magnifying glass icon) and compare subtitle of entries. These should be same for every entry

PR Author Checklist

Screenshots/Videos

Android: Native

Android: mWeb Chrome

iOS: Native

Nagranie.z.ekranu.2026-09-1.o.12.49.39.mov

iOS: mWeb Safari

MacOS: Chrome / Safari

Nagranie.z.ekranu.2026-09-1.o.12.43.28.mov

…hn-alternate-text # Conflicts: # src/libs/OptionsListUtils/index.ts # src/libs/ReportAlternateTextUtils.ts

…, perf cleanups

…hn-alternate-text # Conflicts: # tests/ui/WorkspaceCompanyCardPageEmptyStateTest.tsx

codecov · 2026-09-01T09:23:33Z

Codecov Report

❌ Looks like you've decreased code coverage for some files. Please write tests to increase, or at least maintain, the existing level of code coverage. See our documentation here for how to interpret this table.

Files with missing lines	Coverage Δ
...c/components/CurrencyListContextProvider/index.tsx	`100.00% <100.00%> (ø)`
src/components/LocaleContextProvider.tsx	`85.48% <100.00%> (-0.24%)`	⬇️
.../components/Search/FilterComponents/InSelector.tsx	`90.00% <100.00%> (+0.52%)`	⬆️
src/components/Search/SearchAutocompleteList.tsx	`88.68% <100.00%> (+0.26%)`	⬆️
...rc/components/Search/SearchRouter/SearchRouter.tsx	`72.25% <100.00%> (+0.48%)`	⬆️
.../Search/SearchRouter/SearchRouterOptionsWarmer.tsx	`96.66% <100.00%> (ø)`
src/hooks/useAutocompleteSuggestions.ts	`73.91% <100.00%> (+0.64%)`	⬆️
src/hooks/useFilteredOptions.ts	`86.66% <100.00%> (+2.66%)`	⬆️
src/hooks/useGetExpensifyCardFromReportAction.ts	`100.00% <100.00%> (ø)`
src/hooks/useSearchSelector/base.ts	`88.28% <100.00%> (+0.47%)`	⬆️
... and 10 more
... and 7 files with indirect coverage changes

…hn-alternate-text

sosek108 · 2026-09-03T09:25:14Z

@MelvinBot review
@codex review

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 5f1c093e34

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

MelvinBot · 2026-09-03T09:42:05Z

Reviewed the diff. The refactor direction is right — collapsing getChatPreviewParts into the shared getReportAlternateText removes a whole class of drift. Three things to look at before merge, one of which undercuts the "identical by construction" claim.

1. Search and LHN still resolve `lastAction` with different algorithms

Everything downstream keys off lastAction, so parity holds only where the two resolvers agree — and they don't always.

LHN: OptionRowLHNData.tsx:105 calls getLastVisibleActionIncludingTransactionThread, which takes the newest visible action on the parent and the newest on the thread, then compares created.
Search: ReportAlternateTextUtils.ts:425-431 takes the first match in the pre-combined sortedActions[reportID] array.

Those differ because getCombinedReportActions drops IOU-create actions from the combined list: ReportActionsUtils.ts:895-898. isReportActionVisibleAsLastAction keeps them (:1436), so LHN sees them.

Failure case: a one-transaction expense report submitted with no follow-up comment. The parent's newest visible action is the IOU-create action and the thread only has CREATED. LHN resolves lastAction to the IOU action; Search's find skips it (filtered out of the combined array) and lands on an older action, or undefined. Different subtitle.

The cheap fix is to have resolveLastActionContext always use getLastVisibleActionIncludingTransactionThread — it's already the else branch — or move LHN onto the sortedActions path. One shared resolver is what makes the claim structural rather than incidental. Either way, please add a one-transaction-report case to LhnSearchPreviewParityTest.ts; the current suite doesn't cover one.

2. `invoiceReceiverPolicy` is declared but never supplied

OptionsListUtils/index.ts:237 adds it to GetAlternateTextConfig and :341 forwards it, but no caller passes it — it is always undefined. The PR encodes the resulting mismatch as an expected result (should diverge for empty invoice room…).

This looks closeable rather than permanent: prepareReportOptionsForDisplay already has policiesCollection in scope and already resolves policy from it, so report.invoiceReceiver.policyID → one more lookup gets you the rest of the way. Worth doing here, otherwise the test locks in a gap that the PR title says doesn't exist.

3. Dead plumbing — about 60 lines

cardList, workspaceCardList, localeCompare, formatPhoneNumber, and convertToDisplayString are threaded through createOption, createOptionFromReport, and processReport, but every call site of those three has showChatPreviewLine false, so they can't reach the branch that reads them:

processReport → createOption passes no config, and createFilteredOptionList never forwards localeCompare/formatPhoneNumber/convertToDisplayString anyway.
SearchRouter and InSelector call createOptionFromReport with config: {showPersonalDetails} only.
The InSelector getAlternateText call passes {} as config, so its nine new arguments are inert.

convertToDisplayString is on six type definitions and supplied by zero components. The only live consumer is prepareReportOptionsForDisplay. Trimming this keeps the surface honest about which inputs actually matter.

Minor

OptionsListUtils/index.ts:366-367 — the missing-report fallback returns formatReportLastMessageText(option.lastMessageText) with no Parser.htmlToText. The old code stripped HTML unless the last action was ADD_COMMENT, so raw markup can now render in that path (the case [Due for Payment 2026-04-09][$500] Search - HTML message is displayed different on LHN and on chat preview on search #82036 was about, inverted).
The PR body says the parity test "will assert" equality in future tense, and the all-platform checkboxes are ticked while the Android and mWeb sections are empty. Worth tidying before review.

Testing note

I tried to confirm the subtitle parity in a browser session, but the web session rendered a blank page (0 DOM nodes, dev server healthy and serving the bundle) and never reached Home, so I could not compare the two lists visually. The findings above are from reading the code. CI is green apart from Check independent approval.

view run

sosek108 · 2026-09-03T14:36:52Z

@MelvinBot review

MelvinBot · 2026-09-03T14:53:19Z

All three findings from my last pass are fixed at 6bc6cbaa, and CI is green apart from Check independent approval. Two things left, neither a blocker.

What got resolved

resolveLastActionContext now always calls getLastVisibleActionIncludingTransactionThread — the same resolver LHN uses — and the new should match for a one-transaction expense report whose newest action is the IOU create action test covers exactly the case that would have diverged.
invoiceReceiverPolicy is now resolved in prepareReportOptionsForDisplay with the same report-then-parent precedence as LHNOptionsList.tsx:77-84, and the invoice test now asserts equality instead of divergence.
The dead cardList / workspaceCardList / localeCompare / formatPhoneNumber / convertToDisplayString plumbing is gone, and Parser.htmlToText is back in the missing-report fallback.

1. The "clean up comment" commit stripped rationale that should stay

4d211eb removed roughly twenty comments. Most were restating the code and are fine to lose, but a few documented decisions that aren't recoverable from reading it:

ReportAlternateTextUtils.ts:1231-1232 — the || here sits directly under an eslint-disable for prefer-nullish-coalescing. The deleted comment explained that an empty precomputed string means "not computed yet," so ?? would silently kill the recompute fallback. Without it the next person reads a suppressed lint rule with no reason and switches it.
OptionsListUtils/index.ts:257-259 — the module-scoped collator looks like avoidable global state until you know new Intl.Collator loads locale data and can't be built per comparison inside a sort.
OptionsListUtils/index.ts:324 — privateIsArchived: !!isReportArchived passes one value for two distinct-looking parameters. The deleted comment explained both derive from the same NVP.
ReportAlternateTextUtils.ts:1651-1655 — the two-Onyx-key lookup in getExpensifyCardFromReportAction is not guessable; the Travel Billing key variant needs the note.

The author checklist asks for comments explaining "why" on code that isn't self-explanatory. These four qualify.

2. Preview resolution got slower per row — worth a manual sanity check

Moving to the LHN resolver was the right correctness call, but it changed the cost profile. The deleted getChatPreviewParts used sortedActions[reportID].find(...) on the pre-sorted derived array, which exits at the first visible action. resolveLastActionContext now runs getLastVisibleActionIncludingTransactionThread, which does two full Object.values scans over the raw action maps with an isNewerReportAction comparison per entry, and it runs unconditionally — before the || short-circuit that skips getLastMessageTextForReport.

That happens once per option inside prepareReportOptionsForDisplay, which re-runs on every keystroke in the search router. sortedActions is still threaded into getAlternateText for other uses, so the sorted array is right there if you want the cheap path on both surfaces.

baseline-perf-tests and branch-perf-tests both pass, so nothing is flagged. But those don't exercise the router's keystroke path, and your own test plan calls for a high-traffic account — worth typing into the router on one and confirming it still feels instant.

Web verification

🤖 Web test steps verification results: ✅ pass

Subtitle text matched between the LHN and the search router for the one chat present in both lists. Coverage is thin: the test account is fresh and has a single conversation, so this confirms the pipeline runs and produces identical output, not that the edge cases match.

Step	Status
LHN chat list subtitle captured One row, "Expensify Concierge", subtitle "Welcome to Expensify Hi! I'm Concierge. I can help you get set up, answer questions, and update your expenses with a quick chat. What can I do for you today?"	✅
Search router opened with empty query "Recent chats" showed the self-DM and Concierge, both with non-empty subtitles	✅
Subtitle identical between the two surfaces Concierge is the only chat in both lists; its subtitle string matched exactly	✅
No blank subtitle in the search router Both rows rendered subtitle text	✅

view run

mountiny

Nice cleanup — routing Search through getReportAlternateText is the right call, and passing the already computed reportMetadata and participantPersonalDetailListExcludeCurrentUser down from SidebarUtils.getOptionData removes real duplicated work per LHN row.

One request, plus a few non-blocking notes inline.

Please restore `tests/unit/ReportAlternateTextParityTest.ts` and its snapshot

I ran the deleted suite against this branch: 145 of 145 snapshots pass unchanged. Keeping it costs nothing and CI stays green:

git checkout dc10c9e0a0dac855412530ec08a68ac6798ecc05 -- tests/unit/ReportAlternateTextParityTest.ts tests/unit/__snapshots__/ReportAlternateTextParityTest.ts.snap

It pins the expected string for 115 action types. Only 25 of those appear in SidebarUtilsTest.ts, so 90 action types lose their only pinned expected output. LhnSearchPreviewParityTest asserts searchText === lhnText plus one toContain, so a change that moves both surfaces together is invisible to it.

That matters here specifically because this PR changes getLastActorDisplayNameFromLastVisibleActions to use getLastVisibleActionIncludingTransactionThread, and the LHN goes through that same function — the snapshot is the evidence that no LHN output shifted. The two suites cover different axes, so keep both.

Verified locally on `c16419e`

Check	Result
`tsc --noEmit`	clean
`LhnSearchPreviewParityTest` + `OptionsListUtilsTest` + `SidebarUtilsTest`	574 passing
`useSearchSelectorTest` + `useAutocompleteSuggestions` + `InSelectorTest`	56 passing
`eslint` on changed files	only pre-existing seatbelt patterns, nothing on new lines

Minor: the PR body still carries the // TODO: These must be filled out line under QA Steps.

mountiny · 2026-09-09T14:53:32Z

+                formatPhoneNumber: formatPhoneNumber ?? formatPhoneNumberPhoneUtils,
+                dateFnsLocale,
+                convertToDisplayString,
+                convertToDisplayStringWithoutCurrency: convertToDisplayStringWithoutCurrency ?? fallbackConvertToDisplayStringWithoutCurrency,


convertToDisplayStringWithoutCurrency is unreachable, so this fallback is always the implementation.

It is declared on GetAlternateTextConfig (line 251) but not on GetOptionsConfig (types.ts), SearchOptionsConfig, or prepareReportOptionsForDisplay's config. getAlternateText is only reached from createOption, prepareReportOptionsForDisplay, and InSelector, and none of them can supply it.

fallbackConvertToDisplayStringWithoutCurrency (line 283) also duplicates CurrencyListContextProvider/index.tsx lines 58-80 verbatim, comments included. Unlike sortedActions and currentUserAccountID, this param has neither a caller nor a TODO pointing at a refactor issue.

Either thread it down from useCurrencyListActions() alongside convertToDisplayString, or drop the param and the fallback.

The params thread is done and fallback duplicated code is gone.

It added some files to the diff as the change touched at least 7 callers

mountiny · 2026-09-09T14:53:33Z

+    const locale = IntlStore.getCurrentLocale();
+    if (!fallbackCollator || fallbackCollatorLocale !== locale) {
+        fallbackCollatorLocale = locale;
+        fallbackCollator = new Intl.Collator(locale, {usage: 'sort', sensitivity: 'variant', numeric: true, caseFirst: 'upper'});


This is the fourth copy of these collator options in the codebase:

LocaleContextProvider.tsx:97 (COLLATOR_OPTIONS)

ReportNameUtils.ts:269

SearchQueryUtils.ts:556

Worth exporting one shared helper so they cannot drift apart.

Extracted Collator to CollatorUtils

mountiny · 2026-09-09T14:53:33Z

              policyCollection: allPolicies,
              personalDetails,
              sortedActions,
+              cardList,


cardList and workspaceCardList cannot be read on this path.

Line 84 sets isUsedInChatFinder: false, which makes showChatPreviewLine false inside getSearchOptions, and both values are only consumed inside that branch (OptionsListUtils/index.ts:347). The getAlternateText(report, {}, ...) call in InSelector is the same story.

Safe to drop these two lines along with the useCardList() / useWorkspaceCardList() calls on lines 60-61.

Droped both lists

mountiny · 2026-09-09T14:53:33Z

+            const resolvedCurrentUserAccountID = currentUserAccountID ?? CONST.DEFAULT_NUMBER_ID;
+            const oneTransactionThreadReportID = transactionThreadIDs?.[report.reportID];
+            const {lastAction, lastActionReport, movedFromReport, movedToReport} = resolveLastActionContext(report, isReportArchived, visibleReportActionsData, oneTransactionThreadReportID);
+            const card = isCardIssuedAction(lastAction) ? getExpensifyCardFromReportAction({reportAction: lastAction, policy, cardList, workspaceCardList}) : undefined;


Worth a Reassure run before merge.

The LHN pays this resolution once per visible row (roughly 20), but prepareReportOptionsForDisplay is not memoised the way createFilteredOptionList is, so Search now runs the full getReportAlternateText per option up to maxElements.

tests/perf-test/OptionsListUtils.perf-test.ts and tests/perf-test/SearchRouter.perf-test.tsx both cover this path.

Good call. Two findings.

The two suites you named don't cover this path. createRandomReport (tests/utils/collections/reports.ts:13) sets no participants and no chatType, so processReport (OptionsListUtils/index.ts:988) drops all 5000 fixture reports before createOption. Both files use that factory, so getLastMessageTextForReport and getReportAlternateText never run in either. Local Reassure also can't resolve deltas this size: main vs main reported "significant" regressions up to +55%.

So I added covering scenarios (tests/perf-test/OptionsListChatPreview.perf-test.ts) — participant-bearing reports with report actions, deterministic fixtures, showChatPreviewLine: true, two result caps. Min of 4 interleaved runs per state:

Scenario main this branch branch + memo

getValidOptions previews, cap 20 26.74 ms 28.56 ms (+6.8%) 25.09 ms (−6.2%)

getValidOptions previews, cap 100 29.88 ms 34.82 ms (+16.5%) 24.94 ms (−16.5%)

createFilteredOptionList (build) 26.44 ms 21.78 ms (−17.6%) 21.49 ms (−18.7%)

Your read was right — the preview cost is real and scales with maxElements. The build path is ~18% faster because threading lastActions/transactionThreadIDs through createOption drops the per-report Onyx rescan.

Memo added: prepareReportOptionsForDisplay caches the preview per report, dropping the cache when any input feeding it changes identity (same shape as the createFilteredOptionList cache). One prerequisite — both it and getValidOptions defaulted visibleReportActionsData to a fresh {}, which the share-destination caller doesn't pass, so the guard invalidated every call (0 hits / 1320 misses); pointing both at the existing EMPTY_VISIBLE_REPORT_ACTIONS gives 1220 hits / 100 misses.

Caveat: those means are warm-cache dominated (1 cold + 10 warm), so the first call after a guarded input changes still costs the pre-memo amount.

…ithoutCurrencyForLocale(locale, amountInCents, currency, getCurrencyDecimalsImpl); the existing …EnLocale variant now delegates to it.

sosek108

@mountiny I've replied to your comments. I've also reintroduced Snapshot/parity tests

sosek108 · 2026-09-10T09:36:04Z

+                formatPhoneNumber: formatPhoneNumber ?? formatPhoneNumberPhoneUtils,
+                dateFnsLocale,
+                convertToDisplayString,
+                convertToDisplayStringWithoutCurrency: convertToDisplayStringWithoutCurrency ?? fallbackConvertToDisplayStringWithoutCurrency,


The params thread is done and fallback duplicated code is gone.

It added some files to the diff as the change touched at least 7 callers

sosek108 · 2026-09-10T09:58:36Z

+    const locale = IntlStore.getCurrentLocale();
+    if (!fallbackCollator || fallbackCollatorLocale !== locale) {
+        fallbackCollatorLocale = locale;
+        fallbackCollator = new Intl.Collator(locale, {usage: 'sort', sensitivity: 'variant', numeric: true, caseFirst: 'upper'});


Extracted Collator to CollatorUtils

sosek108 · 2026-09-10T10:06:07Z

              policyCollection: allPolicies,
              personalDetails,
              sortedActions,
+              cardList,


Droped both lists

sosek108 · 2026-09-10T11:06:12Z

+            const resolvedCurrentUserAccountID = currentUserAccountID ?? CONST.DEFAULT_NUMBER_ID;
+            const oneTransactionThreadReportID = transactionThreadIDs?.[report.reportID];
+            const {lastAction, lastActionReport, movedFromReport, movedToReport} = resolveLastActionContext(report, isReportArchived, visibleReportActionsData, oneTransactionThreadReportID);
+            const card = isCardIssuedAction(lastAction) ? getExpensifyCardFromReportAction({reportAction: lastAction, policy, cardList, workspaceCardList}) : undefined;


Good call. Two findings.

The two suites you named don't cover this path. createRandomReport (tests/utils/collections/reports.ts:13) sets no participants and no chatType, so processReport (OptionsListUtils/index.ts:988) drops all 5000 fixture reports before createOption. Both files use that factory, so getLastMessageTextForReport and getReportAlternateText never run in either. Local Reassure also can't resolve deltas this size: main vs main reported "significant" regressions up to +55%.

So I added covering scenarios (tests/perf-test/OptionsListChatPreview.perf-test.ts) — participant-bearing reports with report actions, deterministic fixtures, showChatPreviewLine: true, two result caps. Min of 4 interleaved runs per state:

Scenario main this branch branch + memo

getValidOptions previews, cap 20 26.74 ms 28.56 ms (+6.8%) 25.09 ms (−6.2%)

getValidOptions previews, cap 100 29.88 ms 34.82 ms (+16.5%) 24.94 ms (−16.5%)

createFilteredOptionList (build) 26.44 ms 21.78 ms (−17.6%) 21.49 ms (−18.7%)

Your read was right — the preview cost is real and scales with maxElements. The build path is ~18% faster because threading lastActions/transactionThreadIDs through createOption drops the per-report Onyx rescan.

Memo added: prepareReportOptionsForDisplay caches the preview per report, dropping the cache when any input feeding it changes identity (same shape as the createFilteredOptionList cache). One prerequisite — both it and getValidOptions defaulted visibleReportActionsData to a fresh {}, which the share-destination caller doesn't pass, so the guard invalidated every call (0 hits / 1320 misses); pointing both at the existing EMPTY_VISIBLE_REPORT_ACTIONS gives 1220 hits / 100 misses.

Caveat: those means are warm-cache dominated (1 cold + 10 warm), so the first call after a guarded input changes still costs the pre-memo amount.

… cache can hit

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

sosek108 · 2026-09-11T10:16:32Z

https://expensify.slack.com/archives/C08CZDJFJ77/p1789120225042869?thread_ts=1787056222.560989&cid=C08CZDJFJ77

Decided to split this one into five smaller:

PR 1 — Extract CollatorUtils
Files: 4 | +31 / −5

PR 2 — Generalize convertToDisplayStringWithoutCurrency for locale
Files: 2 | +23 / −26

PR 3 — Extract getExpensifyCardFromReportAction
Files: 2 | ~+32 / −19

PR 4 — useSortedActions -> useSortedReportActionsData
Files: ~10 | ~+32 / −23

PR 5 — #99062 replacement. This ships the core issued logic and tests
Files: ~21 (13 src + 8 test) | ~+2780 / −780  (src ~+796 / −411)

SEQUENCE
1,2,3,4 - should be straight forward and mergable in parallel
5 - can be merged after 1-4

Marking this one as Draft

…te-text

mountiny · 2026-09-16T07:09:01Z

I think we can close this now as we we are breaking it up to smaller prs

sosek108 added 5 commits August 20, 2026 11:52

Extract pure Expensify card lookup

fb19007

Unify lastAction sources via derived sorted actions

82f45d9

Route Search previews through getReportAlternateText

4769a84

Delete interim chat-preview helpers

57bf384

Add LHN-Search preview parity test

76db09f

melvin-bot Bot assigned sosek108 Aug 20, 2026

sosek108 mentioned this pull request Aug 26, 2026

Remove one-shot alternate text parity snapshot test #99545

Closed

40 tasks done

sosek108 added 3 commits September 1, 2026 09:00

Merge branch 'fix/lhn-search-alt-text-parity' into fix/search-adopt-l…

1b30e52

…hn-alternate-text # Conflicts: # src/libs/OptionsListUtils/index.ts # src/libs/ReportAlternateTextUtils.ts

Apply review fixes: shared last-action resolution, preview text reuse…

3f82630

…, perf cleanups

Merge branch 'fix/lhn-search-alt-text-parity' into fix/search-adopt-l…

3fedb19

…hn-alternate-text # Conflicts: # tests/ui/WorkspaceCompanyCardPageEmptyStateTest.tsx

sosek108 added 4 commits September 1, 2026 12:16

Clean up parity tests that are not required anymore

f45f490

clean cspell.json

43b2f91

Merge branch 'fix/lhn-search-alt-text-parity' into fix/search-adopt-l…

b2b7572

…hn-alternate-text

Merge branch 'main' into fix/search-adopt-lhn-alternate-text

5f1c093

chatgpt-codex-connector Bot reviewed Sep 3, 2026

View reviewed changes

Comment thread src/libs/ReportAlternateTextUtils.ts Resolved

clean up comment

4d211eb

clean up

7a66983

remove unused export

6bc6cba

sosek108 added 2 commits September 4, 2026 10:10

Merge branch 'main' into fix/search-adopt-lhn-alternate-text

4159248

added some comments

912cb4d

sosek108 marked this pull request as ready for review September 4, 2026 08:20

sosek108 requested review from a team as code owners September 4, 2026 08:20

melvin-bot Bot requested review from thelullabyy and removed request for a team September 4, 2026 08:20

mountiny reviewed Sep 9, 2026

View reviewed changes

sosek108 added 6 commits September 10, 2026 11:07

Merge branch 'main' into fix/search-adopt-lhn-alternate-text

5279f95

restore tests/unit/ReportAlternateTextParityTest.ts

61a0e52

extracted the formatting body into one shared convertToDisplayStringW…

d22ccce

…ithoutCurrencyForLocale(locale, amountInCents, currency, getCurrencyDecimalsImpl); the existing …EnLocale variant now delegates to it.

extract Collator to CollatorUtils

d77c50e

extract Collator to CollatorUtils

094012c

drop cardList and workspaceCardList from ShareTab

a209bd6

sosek108 commented Sep 10, 2026

View reviewed changes

sosek108 added 3 commits September 10, 2026 13:10

add perf scenarios covering the chat preview path

a9168f2

memoize chat preview text in prepareReportOptionsForDisplay

d78c915

keep the empty visibleReportActionsData default stable so the preview…

dc77b33

… cache can hit

sosek108 force-pushed the fix/search-adopt-lhn-alternate-text branch from 68cf36f to dc77b33 Compare September 10, 2026 11:11

sosek108 and others added 7 commits September 10, 2026 13:12

add report action constants to cspell dictionary

0722813

drop unused clearAlternateTextCache export

3271383

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

merge main into fix/search-adopt-lhn-alternate-text

0dd836f

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Merge branch 'main' into fix/search-adopt-lhn-alternate-text

f1ac928

Merge main into fix/search-adopt-lhn-alternate-text

f19d457

cspell fixes

0398020

Extract shared Intl.Collator into CollatorUtils

5b721e7

sosek108 mentioned this pull request Sep 11, 2026

Extract shared Intl.Collator into CollatorUtils #100944

Merged

39 tasks done

sosek108 marked this pull request as draft September 11, 2026 10:16

This was referenced Sep 11, 2026

Extract Expensify card lookup into pure function #100945

Merged

[Payment due @thelullabyy] Use useSortedReportActionsData instead of useSortedActions #100947

Merged

Generalize convertToDisplayStringWithoutCurrency for any locale #100948

Merged

sosek108 added 2 commits September 14, 2026 14:49

Merge branch 'main' into slice/collator-utils

beac33f

Merge branch 'slice/collator-utils' into fix/search-adopt-lhn-alterna…

7be2357

…te-text

mountiny closed this Sep 16, 2026

sosek108 mentioned this pull request Sep 16, 2026

Route Search row previews through the shared LHN alternate text #101298

Merged

45 tasks done

Scenario	main	this branch	branch + memo
`getValidOptions` previews, cap 20	26.74 ms	28.56 ms (+6.8%)	25.09 ms (−6.2%)
`getValidOptions` previews, cap 100	29.88 ms	34.82 ms (+16.5%)	24.94 ms (−16.5%)
`createFilteredOptionList` (build)	26.44 ms	21.78 ms (−17.6%)	21.49 ms (−18.7%)

Conversation

sosek108 commented Aug 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Explanation of Change

Fixed Issues

Tests

Offline tests

QA Steps

PR Author Checklist

Screenshots/Videos

Uh oh!

codecov Bot commented Sep 1, 2026 • edited by github-actions Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

sosek108 commented Sep 3, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

MelvinBot commented Sep 3, 2026

1. Search and LHN still resolve lastAction with different algorithms

2. invoiceReceiverPolicy is declared but never supplied

3. Dead plumbing — about 60 lines

Minor

Testing note

Uh oh!

sosek108 commented Sep 3, 2026

Uh oh!

MelvinBot commented Sep 3, 2026

1. The "clean up comment" commit stripped rationale that should stay

2. Preview resolution got slower per row — worth a manual sanity check

Web verification

Uh oh!

mountiny left a comment

Choose a reason for hiding this comment

Please restore tests/unit/ReportAlternateTextParityTest.ts and its snapshot

Verified locally on c16419e

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

sosek108 left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

sosek108 commented Sep 11, 2026

Uh oh!

mountiny commented Sep 16, 2026

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

sosek108 commented Aug 20, 2026 •

edited

Loading

codecov Bot commented Sep 1, 2026 •

edited by github-actions Bot

Loading

1. Search and LHN still resolve `lastAction` with different algorithms

2. `invoiceReceiverPolicy` is declared but never supplied

Please restore `tests/unit/ReportAlternateTextParityTest.ts` and its snapshot

Verified locally on `c16419e`