Add filter negation feature by bernhardoj · Pull Request #93909 · Expensify/App

bernhardoj · 2026-06-18T05:17:58Z

Explanation of Change

Fixed Issues

$ #93208
PROPOSAL:

Tests

Same as QA Steps

Verify that no errors appear in the JS console

Offline tests

Same as QA Steps

QA Steps

Go to Spend > Expenses
Press the Filters button
Open Status. Workspace, To, From, Currency, Purchase currency, Merchant, and "Exported to" filter
Verify it has "Filter type" with "is" and "is not" options, and "is" is initially selected
Open Has filter
Verify it has "Filter type" with "has" and "doesn't have" options, and "has" is initially selected
Apply any value and filter type to the filters
If a negative filter type is applied (is not/doesn't have), verify the filter chip/bar label is prefixed with -
Close and reopen the filter
Verify the filter value and type are still selected with the correct value
Open the save search page
Verify it has the same label as in the filter chip/bar

For steps 3-7, if you want more test steps, follow this:
No filter is applied yet

Open any filter from the step above
Verify the positive filter type is selected
Change the filter type to negative
Verify the filter is not applied yet (because no value is selected)
Apply any value
Verify that a negation filter is applied
Change the filter type back to positive
Verify that a positive filter is applied
Unapply and reapply the value
Verify that a positive filter is applied

For filter chip/bar:

Press any applied filter chip/bar
Verify that the filter value and type are selected with the correct value
Play around with the test steps above

Verify that no errors appear in the JS console

PR Author Checklist

Screenshots/Videos

Android: Native

Android: mWeb Chrome

iOS: Native

iOS: mWeb Safari

MacOS: Chrome / Safari

web.mp4

melvin-bot · 2026-06-18T05:18:03Z

Hey, I noticed you changed src/languages/en.ts in a PR from a fork. For security reasons, translations are not generated automatically for PRs from forks.

If you want to automatically generate translations for other locales, an Expensify employee will have to:

Look at the code and make sure there are no malicious changes.
Run the Generate static translations GitHub workflow. If you have write access and the K2 extension, you can simply click: [this button]

Alternatively, if you are an external contributor, you can run the translation script locally with your own OpenAI API key. To learn more, try running:

npx ts-node ./scripts/generateTranslations.ts --help

Typically, you'd want to translate only what you changed by running npx ts-node ./scripts/generateTranslations.ts --compare-ref main

melvin-bot · 2026-06-18T05:18:05Z

@DylanDylann Please copy/paste the Reviewer Checklist from here into a new comment on this PR and complete it. If you have the K2 extension, you can simply click: [this button]

bernhardoj · 2026-06-18T05:18:57Z

Can I run the translation script myself?

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: d5746a1607

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".

bernhardoj · 2026-06-18T05:27:19Z

Btw, when we apply a negated HAS filter with a Tag value, it will be applied as "Tag: No tag".

w.mp4

It's from this code. It causes weird buggy behavior with the Has filter component, as shown in the video above.

App/src/libs/SearchQueryUtils.ts

Lines 1110 to 1112 in e5928ed

    
           if (hasNegatedTagFilter) { 
        
               filtersForm[FILTER_KEYS.TAG] = [CONST.SEARCH.TAG_EMPTY_VALUE]; 
        
           }

Is that expected?

github-actions · 2026-06-18T06:11:33Z

🚧 @dubielzyk-expensify has triggered a test Expensify/App build. You can view the workflow run here.

github-actions · 2026-06-18T06:17:03Z

🧪🧪 Use the links below to test this adhoc build on Android, iOS, and Web. Happy testing! 🧪🧪
Built from App PR #93909.

Android 🤖	iOS 🍎
https://ad-hoc-expensify-cash.s3.us-east-1.amazonaws.com/rock-artifacts/ad-hoc/rock-android-Adhoc-d5746a1607e-57a8e82-21a714aa9be6941a9a8919a860d2a71fc5cacdff/index.html	https://ad-hoc-expensify-cash.s3.us-east-1.amazonaws.com/rock-artifacts/ad-hoc/rock-ios-device-AdHoc-d5746a1-57a8e82-7c2e95c29b6978b3ea4fd4fbff344f6e1bf62777/index.html

Web 🕸️
https://93909.pr-testing.expensify.com

👀 View the workflow run that generated this build 👀

dubielzyk-expensify · 2026-06-18T06:29:33Z

This is looking great already @bernhardoj. Let's make this text label (13px) not body (15px)

dubielzyk-expensify · 2026-06-18T06:31:03Z

I'm not convinced we should use the is/is not in the chip itself:

I have to go and will have a noodle, but I think we might want an icon or do From: not Jon Dubby. But even that feels weird. I don't think it's crucial for it to be written fully out, but just be able to signal between is/isn't in some way, then when you open the filter you can see it more clearly there anyways

cc @Expensify/design @trjExpensify @JmillsExpensify

bernhardoj · 2026-06-18T07:27:03Z

Let's make this text label (13px) not body (15px)

Done

JmillsExpensify · 2026-06-18T14:46:50Z

I'm not convinced we should use the is/is not in the chip itself:
I have to go and will have a noodle, but I think we might want an icon or do `From: not Jon Dubby`. But even that feels weird. I don't think it's crucial for it to be written fully out, but just be able to signal between is/isn't in some way, then when you open the filter you can see it more clearly there anyways
cc @Expensify/design @trjExpensify @JmillsExpensify

I'm fine with what I'm seeing in the adhoc.

JmillsExpensify · 2026-06-18T14:47:50Z

@bernhardoj Looks like we're still missing support for Status and Workspace in the adhoc. Those are mentioned in the OP. Let me know if I'm missing something!

JS00001 · 2026-06-18T14:50:46Z

Status and workspace are still having their BE PRs worked on: https://github.com/Expensify/Auth/pull/22241

bernhardoj · 2026-07-01T06:05:21Z

Opened a seperate PR (because it's quite big) to allow workspace and status to be negatable.

cc: @aimane-chnaif

JmillsExpensify · 2026-07-01T10:14:22Z

Works for me.

aimane-chnaif · 2026-07-14T11:54:57Z

Is this ready for review? Lots of tests failing

bernhardoj · 2026-07-14T12:43:42Z

Not yet. I'll let you know when it's ready

bernhardoj · 2026-07-14T16:56:53Z

@aimane-chnaif ready now

JS00001 · 2026-07-14T17:41:46Z

lint is failing @bernhardoj

aimane-chnaif · 2026-07-14T18:17:41Z

+    if (isFilterNegatable(key)) {
+        const prefix = isFilterNegated(key) ? '-' : '';
+        return `${prefix}${translate(labelKey)}`;
+    }


chip label is now a raw - prefix, contradicting the PR's own QA steps

So an applied negated filter renders as -Merchant: Uber / -Has: Receipt.
This conflicts with three things:

The PR's own QA steps (steps 8–9) still specify Merchant is not: test and Doesn't have: Receipt.

The picker UI itself — NegatableFilter still labels the toggle buttons "is / is not" and "has / doesn't have", and those strings are still translated across all 9 languages. A user picks "is not" and gets back a chip reading "-Merchant", which doesn't echo the wording they chose.

The - is hardcoded and unlocalized — it bypasses translate entirely, so the search.filters.filterType.*.negative strings are now only used by the toggle, never by the chip.

chip label is now a raw - prefix, contradicting the PR's own QA steps

Looks like the AI still sees the old steps 😄

yup.
Btw non-blocking nit: the - is hardcoded in getLabelValue rather than coming from a translation, so the chip prefix isn't localized the way the picker toggle ("is not" / "doesn't have") is.

Can - syntax be localized? 🤔

No need localization but define in CONST?

Ah, that makes more sense. Updated.

aimane-chnaif · 2026-07-14T18:21:45Z

        default:
-            return null;
+            break;


default: case can return undefined.

The switch ends with default: break;, then unconditionally returns <NegatableFilter>{content}</NegatableFilter>.
For an unhandled key content is undefined, and NegatableFilter early-returns children for non-negatable keys — so the component returns undefined.
The switch is currently exhaustive so it's latent, but the old code returned null here.

Suggestion: have NegatableFilter return children ?? null.

Undefined also renders nothing, same as null

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: e483905be6

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".

chatgpt-codex-connector · 2026-07-14T18:24:39Z

+            policyID={getFilterNegatableValue(CONST.SEARCH.SYNTAX_FILTER_KEYS.POLICY_ID, values)}
+            ready={ready}
+            isNegated={isNegated}
+            onChange={(newValue, negated) => onChange(getFilterFormValues(baseFilterKey, newValue, negated))}


Preserve negated Has Tag selections

Negating the Has filter with the Tag option does not round-trip through the existing query normalization. This path writes hasNot: ['tag'], which serializes to -has:tag, but buildFilterFormValuesFromQuery translates -has:tag into tag: [No tag] instead of hasNot; after applying/reopening, the chip becomes Tag: No tag and the Has filter no longer shows doesn't have Tag selected. Special-case Tag here or prevent it from being written as a negated Has value so the new negation UI preserves its own state.

Useful? React with 👍 / 👎.

@JmillsExpensify @JS00001 currently, when we apply a negated Has with Tag value (-has:tag), it translates to {tag: ['none']}. Do we want to keep that behavior? (this is an existing behavior on prod)

no.tag.mp4

Or should we change it so it translates to {hasNot: tag}?

MelvinBot · 2026-07-14T18:26:08Z

🤖 Code review

Overall this is a clean, well-structured refactor — renaming filterKey → baseFilterKey, splitting the old FilterComponents into ListFilterContent + NegatableFilter, and threading an isNegated flag through the popups/pages. The negation logic itself checks out and is nicely covered by the new unit tests. One blocker (failing ESLint) and one minor robustness nit.

🔴 Blocker — ESLint is failing

The ESLint check is red. Two categories, both from this PR:

1. Parent-relative imports that must use the path alias (@dword-design/import-alias/prefer-alias):

SearchAdvancedFiltersContent.tsx:11 — '../ListFilterContent' → '@components/Search/FilterComponents/ListFilterContent'
TextInputFilterContent.tsx:20 — '../NegatableFilter' → '@components/Search/FilterComponents/NegatableFilter'
TextFilterPopup.tsx:17 — '../FilterComponents/NegatableFilter' → '@components/Search/FilterComponents/NegatableFilter'

2. New unsafe type assertions exceeding the eslint-seatbelt budget (@typescript-eslint/no-unsafe-type-assertion). The new as casts (e.g. removeNegation(filterKey) as SearchFilter['key'] in useSearchFiltersBar, as K in getFilterFormValues, and the ones flagged in SearchUIUtils.ts) tip each file over its grandfathered allowance ("N errors, only N-1 allowed"). Reduce the count back to the allowed number — prefer a type guard / narrowing over a raw assertion where possible, or confirm the seatbelt is intentionally being bumped.

🟡 Minor — `removeNegation` isn't anchored to the suffix

SearchQueryUtils.ts:2445 uses filterKey.replace(CONST.SEARCH.NOT_MODIFIER, ''), which strips the first Not anywhere in the string rather than the trailing modifier. It's safe for today's filter keys, but this PR makes it load-bearing (it now runs on every key in mapFiltersFormToLabelValueList). Consider anchoring it, e.g. filterKey.endsWith(NOT_MODIFIER) ? filterKey.slice(0, -NOT_MODIFIER.length) : filterKey, to match the already-correct isFilterNegated (which uses endsWith).

Things I verified look correct

getFilterFormValues correctly clears the opposite key (base vs. …Not) on every apply, so you never end up with both set from the UI.
getFilterNegatableValue keying off hasKey(values, negatedKey) preserves the is/is-not toggle state across reopen even when the value is empty — matches QA steps 10–11.
shouldShowFilter's new (isFilterNegatable(key) || !isFilterNegated(key)) clause is a sound guard: it prevents non-UI-negatable …Not keys (e.g. isNot, actionNot) from newly surfacing as chips now that labels resolve via removeNegation(key). No regression vs. old behavior (those keys weren't in FILTER_VIEW_MAP before either).
getDisplayValue …Not handling is complete for the negatable set (TO, FROM, HAS, CURRENCY, PURCHASE_CURRENCY, MERCHANT, EXPORTED_TO, POLICY_ID, STATUS) — the text/default fallthrough covers MERCHANT_NOT/EXPORTED_TO_NOT correctly, and IS isn't in NEGATABLE_FILTERS so the missing IS_NOT branch is intentional.
Changing HAS_NOT from string[] → HasFilterValues to match HAS is the right call for getFilterFormValues to stay type-consistent.
The FilterComponents/index.tsx → ListFilterContent.tsx rename is clean — no dangling imports of the old module or FilterComponentsProps/CommonPopup/CommonFilterContentWrapper remain.
typecheck, Bun tests, and all test jobs are green.

bernhardoj · 2026-07-15T03:49:21Z

All done, except #93909 (comment), waiting for confirmation.

add filter negation feature

d5746a1

bernhardoj requested review from a team as code owners June 18, 2026 05:17

melvin-bot Bot assigned bernhardoj Jun 18, 2026

melvin-bot Bot requested review from DylanDylann and JmillsExpensify and removed request for a team June 18, 2026 05:18

melvin-bot Bot removed the request for review from a team June 18, 2026 05:18

chatgpt-codex-connector Bot reviewed Jun 18, 2026

View reviewed changes

Comment thread src/components/Search/FilterComponents/AdvancedFilters/SearchAdvancedFiltersContent.tsx Outdated

remove unused export

a194ae7

bernhardoj added 2 commits June 18, 2026 15:22

simplify

c968a64

change font to 13px

9f46783

fix text based filter negated value is not shown

f335f57

JS00001 requested review from JS00001 and aimane-chnaif and removed request for DylanDylann June 18, 2026 12:12

lint

5d463f6

stephanieelliott removed their request for review July 10, 2026 20:21

bernhardoj added 2 commits July 14, 2026 19:22

Merge branch 'main' into feat/93208-filter-negation

98c7265

Merge branch 'main' into feat/93208-filter-negation

601db5f

bernhardoj added 2 commits July 14, 2026 20:30

fix error after conflicts

2b23c23

show the display value for negated status

11b7a2f

bernhardoj added 7 commits July 14, 2026 22:43

add back the merchant negation switch

c25bc0c

fix type

06dcde7

add back the merchant negation switch

d8f14c7

lint

f16fd16

rename

f797c1f

Merge branch 'main' into feat/93208-filter-negation

5df81ff

change submodule

e483905

bernhardoj requested a review from aimane-chnaif July 14, 2026 16:56

aimane-chnaif reviewed Jul 14, 2026

View reviewed changes

chatgpt-codex-connector Bot reviewed Jul 14, 2026

View reviewed changes

bernhardoj added 4 commits July 15, 2026 10:24

Merge branch 'main' into feat/93208-filter-negation

3e9b039

clear the negated status and has too

7bb704a

simplify

bcf345b

lint

ff4270f

bernhardoj added 2 commits July 15, 2026 11:56

lint

f394233

put the not prefix in a const

fac08fd

Uh oh!

Conversation

bernhardoj commented Jun 18, 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!

melvin-bot Bot commented Jun 18, 2026

Uh oh!

melvin-bot Bot commented Jun 18, 2026

Uh oh!

bernhardoj commented Jun 18, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

bernhardoj commented Jun 18, 2026

Uh oh!

github-actions Bot commented Jun 18, 2026

Uh oh!

github-actions Bot commented Jun 18, 2026

Uh oh!

dubielzyk-expensify commented Jun 18, 2026

Uh oh!

dubielzyk-expensify commented Jun 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

bernhardoj commented Jun 18, 2026

Uh oh!

JmillsExpensify commented Jun 18, 2026

Uh oh!

JmillsExpensify commented Jun 18, 2026

Uh oh!

JS00001 commented Jun 18, 2026

Uh oh!

bernhardoj commented Jul 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

JmillsExpensify commented Jul 1, 2026

Uh oh!

aimane-chnaif commented Jul 14, 2026

Uh oh!

bernhardoj commented Jul 14, 2026

Uh oh!

bernhardoj commented Jul 14, 2026

Uh oh!

JS00001 commented Jul 14, 2026

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!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

chatgpt-codex-connector Bot Jul 14, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

bernhardoj commented Jun 18, 2026 •

edited

Loading

dubielzyk-expensify commented Jun 18, 2026 •

edited

Loading

bernhardoj commented Jul 1, 2026 •

edited

Loading

🟡 Minor — `removeNegation` isn't anchored to the suffix