feat(grid): filter a column by clipboard value (= clipboard, IN clipb… - #1112

Thanks for building this, filtering by a list pasted from somewhere else is a really practical addition. I like that the clipboard is read when the item is clicked rather than when the menu opens, that blank lines and duplicates are dropped, and that the 1000-value cap comes with a toast instead of silently cutting the list. Reusing formatSqlValueForColumnComparison for each IN item also keeps quoting and the PostgreSQL real cast consistent with the existing = filter. The tests and typecheck pass on the branch (cellValueFilter, DataGrid, i18n: 247 tests).

There's one problem I'd like fixed before merging, because it gives wrong results without any warning.

Non-numeric clipboard lines on a numeric column. A very common way to fill the clipboard is copying a column from a spreadsheet, header included. With the clipboard set to:

user_id
1
2

IN (clipboard) on an int column builds `user_id` IN ('user_id', 1, 2) (I checked it through parseClipboardInList + buildCellValueFilterClause with the mysql driver). Then:

• MySQL casts 'user_id' to 0, so the filter also returns every row where user_id = 0, with only a 1292 Truncated incorrect DOUBLE value warning that the user never sees.
• PostgreSQL rejects the whole query: ERROR: invalid input syntax for type integer: "id" (same clause shape run against a local PG table).

The same applies to = clipboard when the clipboard holds non-numeric text.

A fix that keeps the feature friendly: when the column is numeric (isNumericColumnType), drop the lines that aren't numbers and show a short toast with how many were skipped, and if nothing numeric is left, show the existing "no filter applied" message. A test with a header line in the clipboard would lock it in.

One coordination note, not something you need to change now: #1081 (still open, waiting for a rebase) moves the "Filter by this value" entries into a submenu and extends CellValueFilterOperator and CELL_VALUE_FILTER_LABEL_KEYS, so the two PRs conflict. Whichever lands second will need to rebase, and ideally these two clipboard entries would live in that submenu too.

Since this is your first PR here: if you feel like it, come say hi on our Discord, we'd be happy to have you around: https://discord.com/invite/K2hmhfHRSt

Thanks for the quick turnaround, and for going one step further than I asked. Dropping the header line in parseClipboardInList and reporting how many lines were skipped is exactly what I had in mind. Splitting "clipboard is empty" from "clipboard has no numeric values" makes the feedback much clearer, and the header and duplicate-header tests lock the behaviour in nicely. I checked the branch locally: 255 tests pass across cellValueFilter, DataGrid and i18n, tsc -b is clean, eslint is clean, and the new keys are in all 11 locales.

The fix itself is good. The only thing stopping the merge is that #1081 landed on main on the 15th, so the branch now conflicts in DataGrid.tsx, cellValueFilter.ts and every locale file. Could you rebase onto main? Since you'll be in there anyway:

• CellValueFilterOperator now also has >, >=, <, <= and contains, so IN needs to be added next to them, and also to CELL_VALUE_FILTER_LABEL_KEYS.
• The "Filter by this value" entries now live in a submenu. I'd like the two clipboard entries to go in there too, so the context menu stays short.

Two small things you can pick up during the rebase if you like, they're not blocking:

• filterClipboardNonNumericDropped uses "value(s)". We already use i18next _one / _other plurals elsewhere (databaseCount_one for example), so it would read better that way.
• isNumericColumnType is a substring match on keywords like int, so a PostgreSQL interval column counts as numeric. With this change, IN (clipboard) on that column drops every line and says there are no numeric values. It's an old helper and a rare case, so a follow-up is fine.

Once it's rebased I'll take another look and try it in the app before merging.

Thanks for sticking with this one. Putting both clipboard entries in the "Filter by this value" submenu keeps the context menu nice and short, and adding IN to CELL_VALUE_FILTER_BADGES and CELL_VALUE_FILTER_LABEL_KEYS keeps the operator maps exhaustive with the new >/</contains operators from #1081. I checked the branch locally: 323 tests pass across cellValueFilter, DataGrid and the i18n suite, and tsc -b and eslint on the changed files are clean.

I'm approving this. Before it can go in I need one more rebase, and I'm sorry about that: #1247 reformatted the whole codebase with Prettier right after your rebase, so DataGrid.tsx, cellValueFilter.ts and the test file conflict again. The conflicts are formatting only (4-space indent, single quotes, the @/types/plugins import). After resolving them please run pnpm format, because pnpm lint now also runs prettier --check.

While you're in there, one small i18n fix: the Russian filterClipboardNonNumericDropped only has _one and _other, but Russian needs _one, _few, _many and _other (see connectionCount_few in ru.json). As it is, 2 to 20 skipped values fall back to the English text, and 21 shows "Пропущено 1 ..." because _one hardcodes the 1 instead of using {{count}}.

A couple of things that are fine as follow-ups, nothing to do in this PR:

• Like interval, geometry types (POINT, LINESTRING) contain "int", so isNumericColumnType treats them as numeric and both clipboard entries would say there are no numeric values. isComparableColumnType already guards against this with isGeometricType, so the clipboard path could reuse the same check.
• On a MySQL FLOAT column the clipboard values arrive as strings, so they skip the Math.fround widening that formatSqlValueForColumnComparison applies to numbers, and IN (clipboard) with something like 1.1 probably won't match the stored value.

Once it's rebased I'll try it in the app on MySQL and PostgreSQL before merging.

Thanks for the quick turnaround on the last round. The Russian plurals now have all four forms with {{count}} in _one, which is exactly what I was after, and the rebase on top of the Prettier reformat was clean. I checked the branch again: 334 tests pass across cellValueFilter, DataGrid and the i18n suite, and pnpm build and eslint are clean.

I can't merge it yet though, and I'm sorry, because this time it's main that moved under you. The whole frontend now lives in packages/web-ui/ (src, tests and locales), so the same React app can run both in the desktop app and in the new tabularis web server. GitHub reports the branch as conflicting because of the move. Git's rename detection carries most of your changes across, but there's one real overlap: on main, DataGrid.tsx no longer imports readText from @tauri-apps/plugin-clipboard-manager. Clipboard access goes through the platform adapter now, and eslint.config.js blocks direct @tauri-apps/** imports in the web UI. I tried a throwaway merge and pnpm build fails with Cannot find name 'readText' on both clipboard entries (DataGrid.tsx:3392 and :3446 on your branch).

So when you rebase, please:

• move your changes to packages/web-ui/src/... and packages/web-ui/tests/utils/cellValueFilter.test.ts
• replace readText() in the two clipboard actions with platform.readClipboard(). platform from usePlatformCapabilities() is already in DataGrid, and pasteFromClipboard does the same thing. As a bonus the feature then works in the browser version too: if the user denies clipboard permission it throws, and your existing "Could not read the clipboard" toast handles it.
• keep only the new keys in the locale files. The last push also reformatted unrelated blocks in all 11 of them (environment.short, the whole themePackages section, "triggers":{), about 100 lines each. Locales are in .prettierignore, so I think an editor JSON formatter did it. Leaving those blocks alone keeps the diff readable and avoids conflicts with other PRs that touch them.

One small thing while you're in there: there's no test yet for buildCellValueFilterClause with IN. A few cases would be nice, for example a numeric column, a string with a quote like O'Brien, and one that passes a capabilities object instead of a driver id (.rules/plugins.md asks for that last one). It would also help to add a line to the PR description saying plugin drivers get the feature through activeCapabilities.

Something I found while testing, not for this PR: formatSqlValueForFilter only doubles single quotes, so on MySQL a clipboard line ending with a backslash, like C:\temp\, ends up as 'C:\temp\' and breaks the query. That helper is older than your change, so it's not on you, but clipboard text makes it easier to hit and I wanted you to know. The geometry and FLOAT follow-ups from last time are still fine to leave for later.

Once it's rebased I'll try it in the app on MySQL and PostgreSQL before merging.

Thanks for handling the frontend move and the clipboard adapter change. Both actions now call platform.readClipboard(), and the added IN cases cover quoted text and a capabilities object. I reran the filter, DataGrid and i18n suites on this head: 425 tests pass.

I haven't found a blocking issue. One nit for a follow-up: Prettier still reports DataGrid.tsx and cellValueFilter.test.ts. Please format those two files without reformatting the locale files.

Approved. I'm leaving this unmerged until my manual clipboard checks on MySQL and PostgreSQL, including denied browser permission and the transposed layout.

I have now done the manual Web app checks with the integrated frontend, the Tabularis 0.31.0 server container, PostgreSQL 16 and MySQL 8.4. Through the real context menu, clipboard id\n1\n2\n2 produced IN (1, 2) and returned rows 1 and 2 on both databases. Filtering a name with O'Brien generated O''Brien and returned the matching row. In the transposed layout, a clipboard list of 1 and 3 returned those two records. The shared browser denied clipboard access; that path displayed Could not read the clipboard and kept all three rows and the empty filter. For the successful cases I used local Chromium with clipboard permissions granted, without replacing the platform adapter or database client.

The formatting nit is fixed. I merged main, reran 429 filter/grid/i18n tests, and corrected the two inherited sidebar keyboard handlers flagged by the React linter; the 119 sidebar tests and source ESLint pass. The integrated production build passes. I have finished the manual checks and will merge once the refreshed CI passes. Desktop was not exercised here.