fix: keep the default table sorting alive past the first grid state save

[AStaroverov]

The problem

The default sorting a block passes to createPlDataTableV3({ sorting }) applies on the first render and then silently disappears: the sorted column stays in the table, the rows just stop being sorted.

resolveSorting (sdk/model/.../createPlDataTableV3.ts) reads the persisted user sorting three ways:

• null — the user never touched sorting → apply the block's default;
• [] — the user explicitly cleared the sorting → suppress the default.

The grid state never produced that null. AG Grid omits sort from its state while nothing is sorted, and convertAgSortingToPTableSorting turned the absent state into []. onStateUpdated fires as the grid initialises, so the first entry written to stateCache already stamped sorting: [] and the block's default was gone from that moment on.

Before @platforma-sdk/model@1.80.0 resolveSorting used isEmpty instead of isNil, so [] still fell through to the default and the bug stayed hidden. Going back to isEmpty is not the fix — it breaks clearing the sorting by hand.

The change

• convertAgSortingToPTableSorting returns null for an absent sort state, and [] only for a genuinely empty sortModel.
• normalizeSort in PlAgDataTableV2.vue promotes the absent state to an explicit { sortModel: [] } only once the grid has reported an explicit sort model before — so clearing the sorting by hand still suppresses the default. This is the same treatment the neighbouring normalizeColumnVisibility already gives hidden columns; both are now applied through normalizeGridState.
• stateForReloadCompare treats an empty sortModel as equivalent to an absent one, so the promoted { sortModel: [] } does not read as a state change and trigger a grid reload.

Compatibility

Existing projects recover on the next render without a state migration: their stored gridState.sort is absent, which now reads as "untouched".

Verification

• pnpm --filter @platforma-sdk/ui-vue run types:check — no errors in the changed files (the 18 errors elsewhere reproduce on a clean tree).
• formatter:check, linter:check — clean.
• New unit test table-state-v2.test.ts covering the three converter states, plus the package's existing node suites — green. PlDatasetSelector.jsdomtest.ts fails on a clean tree too in this environment.

Greptile Summary

The PR preserves a block’s default table sorting until the user explicitly changes or clears sorting, while retaining the distinction between untouched and intentionally empty state.

• Default sorting — the block-provided ordering used when no user sorting preference exists; it now survives initial grid-state persistence.
• Grid sort state — AG Grid’s optional sort object; an omitted object now remains distinguishable from an explicit empty sortModel.
• User sorting — persisted table ordering selected by the user; null means untouched, while [] means explicitly cleared.
• normalizeSort — the new transition normalizer that promotes an omitted sort to an explicit empty model only after an explicit sorting state previously existed.
• stateForReloadCompare — the reload comparison representation; it now treats absent and empty AG Grid sort models as operationally equivalent.
• Adds converter tests for absent, empty, ascending, and descending sort representations.

Confidence Score: 4/5

The PR appears safe to merge, with the non-blocking concern that its central sort-state transition logic is not directly covered by tests.

The nullable converter matches the receiving schema and model fallback semantics, and AG Grid initialization ordering contradicts the suspected sort-erasure path; only focused regression coverage for normalizeSort remains advisable.

Files Needing Attention: sdk/ui-vue/src/components/PlAgDataTable/PlAgDataTableV2.vue; sdk/ui-vue/src/components/PlAgDataTable/sources/table-state-v2.test.ts

Important Files Changed

Filename	Overview
sdk/ui-vue/src/components/PlAgDataTable/PlAgDataTableV2.vue	Adds transition-aware sort normalization and reload-comparison normalization; the behavior is coherent but lacks direct transition tests.
sdk/ui-vue/src/components/PlAgDataTable/sources/table-state-v2.ts	Converts absent AG Grid sort state to the nullable untouched representation expected by the table-state schema and model resolver.
sdk/ui-vue/src/components/PlAgDataTable/sources/table-state-v2.test.ts	Covers all converter representations but not the new normalizeSort state machine.
.changeset/pl-data-table-default-sorting.md	Accurately documents the absent-versus-empty sorting-state fix and compatibility behavior.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
  A[AG Grid state update] --> B{sort present?}
  B -- Yes --> C[Persist explicit sortModel]
  B -- No --> D{Previous sort present?}
  D -- No --> E[Keep sort absent]
  D -- Yes --> F[Persist empty sortModel]
  E --> G[Converter returns null]
  F --> H[Converter returns empty array]
  C --> I[Converter returns user sorting]
  G --> J[Use block default sorting]
  H --> K[Suppress block default]
  I --> L[Use user sorting]

Loading

Prompt To Fix All With AI

### Issue 1
sdk/ui-vue/src/components/PlAgDataTable/PlAgDataTableV2.vue:237-250
**Test sort-state transitions**

`normalizeSort` now carries the central three-state contract, but the added tests cover only the converter. Directly test that untouched sorting stays absent, explicit sorting is preserved, and clearing a previous sort produces an empty model; otherwise a regression can discard the block default or restore it after the user clears sorting without failing this test suite.

---

For each issue above, determine whether it is valid and should be fixed. If so, fix it directly.

_{Reviews (1): Last reviewed commit: "fix: keep the default table sorting aliv..." | Re-trigger Greptile}

Greptile also left 1 inline comment on this PR.

Context used:

• Knowledge Base — UI Vue SDK and Component Library

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 52.97%. Comparing base (c291fe2) to head (5070b32).
⚠️ Report is 20 commits behind head on main.
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1787   +/-   ##
=======================================
  Coverage   52.97%   52.97%           
=======================================
  Files         412      412           
  Lines       21322    21322           
  Branches     4761     4761           
=======================================
  Hits        11296    11296           
  Misses       8665     8665           
  Partials     1361     1361           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.