Skip to content

docs(spec): Table's header row sticks in its own viewport; every column keeps the width a person sets - #10131

Draft
cixzhang wants to merge 9 commits into
mainfrom
spec/table-sizing-delta
Draft

cixzhang wants to merge 9 commits into
mainfrom
spec/table-sizing-delta

Conversation

@cixzhang

@cixzhang cixzhang commented Nov 5, 2026 •

Copy link
Copy Markdown
Contributor

Summary

People scrolling a long table lose the column headings. The last column of a resizable table snaps back when they resize it. This draft keeps the header row in view inside Table's own viewport and lets every column keep the width a person sets.

Spec explainer (human view): kept internally with the owner review.

Owner decisions (draft until decided; parked, not yet queued)

  1. Sticky header row (component:Table FR26, DEC-8). Recommended A: with a bounded block size, Table's own Scroll region keeps the header row at its block start while rows scroll. This builds on DEC-5 and FR19. Alternatives: B, drop Table's scroll wrapper so a caller's scroller hosts the header; C, no sticky header.
  2. Column widths (useTableColumnResize FR11, DEC-8; component:Table FR27 for the filler track). Recommended A: every column, including the last, keeps its width; leftover space goes to a filler track; callbacks report every key. Alternatives: B, keep the last column as the remainder; C, an opt-in prop.

Current-state impact

  • component:Table: FR26 (two-axis scroll when the block size is bounded), FR27 (the filler track), a verification row (FR26, FR27), and DEC-8.
  • module:Table/useTableColumnsVirtualization: FR14 is amended for FR26.
  • useTableColumnResize: FR11, a verification row, and DEC-8. The implementation sets the last column to the remainder today.
  • No code changes.

Open #11287 (by @bezabyte) proposes a page-scroll sticky header behind a hasStickyHeader prop for Tables without a bounded height. This record keeps the header inside Table's own viewport (DEC-8); the unbounded-page case is recorded as a separate needs-spec gap. This builds on @bezabyte's sticky-header work in #11287: the record names its opaque header surface and a divider that moves with the pinned header as the implementation's starting point. Thanks to @bezabyte for the report and the reference implementation.

Writing rubric results

Run over component:Table, module:Table/useTableColumnResize and module:Table/useTableColumnsVirtualization (this change's lines):

  • R1 narration: none; column resize DEC-8's context states the system in the present tense.
  • R2 research in the record: none.
  • R3 evidence naming a change: none; no evidence, basis or verification cell names a pull request, issue or commit.
  • R4 rejected alternatives: one Rejected: line per alternative, inside Table DEC-8 and column resize DEC-8.
  • R5 pull request or issue references remaining: none (searched # + digits, /pull/, /issues/).
  • R6 ruled open questions still listing options: none; this change adds no open question.
  • R7 project status: none in this change's lines. FR11's "phases" are render phases; FR15's existing "does not glide yet" is unchanged text from main.

@vercel

vercel Bot commented Nov 5, 2026 •

Copy link
Copy Markdown

The latest updates on your projects. Learn more about Vercel for GitHub.

Project Deployment Actions Updated
astryx Ready Ready Preview Nov 21, 2026 3:34pm UTC

Request Review

@meta-cla meta-cla Bot added the CLA Signed This label is managed by the Meta Open Source bot. label Nov 5, 2026
@astracat-bot

astracat-bot Bot commented Nov 5, 2026 •

Copy link
Copy Markdown
Contributor

AI review status for this pull request.

Review status Updated
🔄 In progress (for maintainers only) Nov 21, 2026, 3:36 PM UTC

@github-actions github-actions Bot added the needs:spec-owner-review Current knowledge records await owner approval label Nov 5, 2026

@astracat-bot astracat-bot Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This pull request needs a maintainer check before review continues.

[Automated review]

# Conflicts:
#	docs/specs/components/Table/plugins/columnResize/useTableColumnResize.spec.md

@astracat-bot astracat-bot Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This pull request is on hold for an owner decision on the two draft requirements it adds: component:Table FR23 with DEC-7 (the header row stays in view in Table's own Scroll region whenever its height is bounded) and module:Table/useTableColumnResize FR11 with DEC-8 (every column, the last included, keeps the width a person sets, and leftover space goes to a filler track). An engineering owner records that decision by commenting /approve-spec <full head SHA> on this pull request, and the author may do so as an engineering owner. Whatever the decision, these need fixing:

  • FR23 leaves the Table claims it contradicts unamended: FR1, the accessibility contract and the Scroll region design row still make inline the only scroll axis without windowing, Compatibility and Builder impact still promise unchanged scrolling, and column-resize FR2 still reserves the region for horizontal overflow.
  • FR11 leaves the column-resize record's Intent, Compatibility ("every width outcome", no API change) and Ownership non-goals unamended, its defect note still expects a trailing column to absorb the delta, and neither record states the compatibility class and migration that INV9 requires for the new onColumnResizeEnd key.
  • FR23 and its verification row name a maxHeight Table prop that does not exist, and DEC-7's premise that Table owns a both-axis viewport holds only under windowing.
  • FR11's filler track is new Table anatomy with no owning record and no rule hiding it from assistive technology.
  • DEC-8 repeats the Open questions word for word and lacks its Reference line, and DEC-7 ends with a stray "None.".
  • The FR23 verification row puts its failure text under "Representative states" and leaves the last two columns empty.
  • The merge from main adds an unrelated brace-only edit to apps/sandbox/perf/scenes.perf.spec.ts although the description says there are no code changes, so restore main's line here or land the edit on its own.

[Automated review]

sticky header, which breaks Table's own viewport ownership. Rejected: no
sticky header, which leaves every long table to lose its headings.

None.

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Stray "None." after the rejected alternatives; delete it.

| FR8 | Resizing stays available without hover. A boundary remains in tab order with its accessible name and values, keyboard resize is unchanged, and the reveal adds discovery only — it is never a precondition for use. This holds on every device whose input can operate the control, including a coarse-primary device with a keyboard. | Current source, WAI-ARIA splitter pattern, DEC-3 | Keyboard path verified on hover-capable pointers; coarse-primary parity pending. |
| FR9 | A boundary with a visible focus indicator is `emphasized` alone. Focus MUST NOT raise the row-level hint, so focusing one boundary never draws its neighbors. Only a hover-capable pointer inside the header row does that. | Owner direction, 2026-09-21 (DEC-5) | Owner-directed; focused coverage pending. |
| FR10 | While the Table's scroll region is scrolling, header hover resolves no new disclosure: a boundary the pointer reaches only because content moved under it stays `hidden`, and boundaries already `hinted` when the scroll began settle back to `hidden`. Disclosure resumes when scrolling stops. A drag in progress and a visible focus indicator are unaffected. | Owner direction, 2026-09-21 (DEC-6) | Owner-directed; implementation and coverage pending. |
| FR11 | Every resizable column, including the last, keeps the width a person sets, and the width callbacks report every column's key. Space the columns do not fill goes to an empty filler track after the last column, not into the last column; when the columns are wider than the table, the Scroll region scrolls inline (DEC-8). | Proposed, pending | Proposed, pending |

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This contracts width resolution and the commit callback, which the Intent ("remain uncontracted"), Compatibility ("every width outcome") and the Ownership non-goals still exclude. The filler track is also new Table anatomy: say how it renders in a native table and that it is hidden from assistive technology, and record it in component:Table.

Comment on lines +396 to +406
- **OQ1 — Forced colors.** How do `hinted` and `emphasized` resolve under forced-colors
and high-contrast modes? (`checkable`)
- **OQ4 — Draft design record.** `design:user-states` is draft. When it is promoted,
its conditional-hover requirement should be cited here instead of restated.
- **OQ5 — Sticky offset reconciliation.** After a drag writes rendered pixel widths,
sticky pin offsets computed from declared widths are stale. Which record owns the
reconciliation, and does it belong to the width contract or to sticky columns?
(`checkable`)

OQ2 (theme selection) is closed by DEC-7 and OQ3 (coarse pointers) by DEC-3 and
DEC-4. Their numbers are retained rather than reused.

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This repeats the Open questions section below word for word; delete it from DEC-8 and add the **Reference:** module:Table/useTableColumnResize/DEC-8 line the other decisions carry.

Comment thread apps/sandbox/perf/scenes.perf.spec.ts Resolved
@github-actions

github-actions Bot commented Nov 9, 2026 •

Copy link
Copy Markdown
Contributor

PR Analysis Report

Current run: The current change has no visual or preview-producing scope. Visual checks are not applicable; preview links are not shown.


Generated by PR Enrichment workflow | View current CI run

@astracat-bot astracat-bot Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This pull request is on hold for an owner decision on the two draft requirements it adds:

  • component:Table FR23 with DEC-7: the header row stays in view inside Table's own Scroll region whenever its block size is bounded.
  • module:Table/useTableColumnResize FR11 with DEC-8: every column, including the last, keeps the width a person sets, and leftover space goes to a filler track.

An engineering owner records that decision by commenting /approve-spec <full head SHA> on this pull request. The author may do so as an engineering owner.

Either way, these need fixing before approval:

  • FR23 makes the Scroll region scroll on both axes without windowing, but these claims still apply the block axis only under windowing: Table FR1, its accessibility contract, the Scroll region design row, Compatibility and Builder impact, and the column-resize record's FR2 and Ownership. Amend them or narrow FR23.
  • FR11 contracts width resolution and the commit callback, which the column-resize Intent, Compatibility, Ownership non-goals and defect note still exclude or contradict. It also needs a compatibility class and a migration under architecture:public-component-api INV9.
  • FR23 and its verification row bound the Table with maxHeight, which TableProps does not have. Bound it by its parent's height instead, or record a new prop.
  • Regression: DEC-7 says Table bounds its block size for windowing (FR19), but FR19 on main now says the block size is the height its parent gives it. Update the branch and restate DEC-7's premise.
  • The filler track is new Table anatomy with no owner and no rendering rule, and nothing hides it from assistive technology. Record it in component:Table.
  • Miss: the description has no ## Writing rubric results section, which architecture:knowledge-contracts requires of a pull request that changes a record.
  • DEC-8 repeats the Open questions word for word and lacks its **Reference:** line, and DEC-7 ends with a stray "None.".
  • The FR23 verification row puts the failure text under "Representative states" and leaves the failure and audit columns empty.

The unrelated perf-test edit from the earlier revision is gone.

[Automated review]

Comment on lines +553 to +556
A person scrolling a long table loses the column headings, and wrapping Table
in another scroller traps the sticky header inside Table's own scroll wrapper.
Table already owns its viewport on both axes (DEC-5) and bounds its block size
for windowing (FR19). Giving that viewport a bounded height keeps the header

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Regression: FR19 on main now reads "Its block size is the height its parent gives it", so Table no longer bounds its own block size for windowing. Restate this as "a windowed Table already scrolls on both axes inside the height its parent gives it (FR19)". The first sentence's trap holds only once the Table overflows inline. A fitting Scroll region clips on both axes, so an outer sticky header still pins (spec:AST-025 FR21).

# Conflicts:
#	docs/specs/components/Table/Table.spec.md

@astracat-bot astracat-bot Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This pull request is on hold for an engineering owner's decision on the two draft requirements it adds: component:Table FR24 with DEC-7 (the header row stays in view inside Table's own Scroll region whenever its block size is bounded) and module:Table/useTableColumnResize FR11 with DEC-8 (every column, including the last, keeps the width a person sets, and leftover space goes to a filler track). An engineering owner records that decision by commenting /approve-spec <full head SHA> on this pull request, and the author may do so as an engineering owner.

Either way, these need fixing before approval:

  • FR24 makes the Scroll region scroll on both axes without windowing, but Table FR1, its accessibility contract, the Scroll region design row, Compatibility, Builder impact, and the column-resize record's FR2 and Ownership still apply the block axis only under row windowing, so amend them or narrow FR24.
  • Regression: column windowing FR14, now current on main, says only a rows capability gives Table a parent-sized block viewport, which FR24's container-height bound contradicts, so amend FR14 and Table FR19 with FR24 or scope FR24 to row-windowed Tables.
  • FR11 contracts width resolution and the commit callback, which the column-resize Intent, Compatibility, Ownership non-goals and defect note still exclude or contradict, and it needs a compatibility class and a migration under architecture:public-component-api INV9.
  • FR24 and its verification row bound the Table with maxHeight, which TableProps does not have, so bound it by its parent's height or record a new prop.
  • Regression: DEC-7 says Table bounds its block size for windowing (FR19), but FR19 on main now gives row windowing the height its parent gives it and keeps a columns-only Table's block sizing, so restate DEC-7's premise.
  • The filler track is new Table anatomy with no owner, no rendering rule, nothing hiding it from assistive technology and no place in column windowing's row structure or aria-colcount, so record it in component:Table.
  • Miss: the description has no ## Writing rubric results section, which architecture:knowledge-contracts requires of a pull request that changes a record.
  • DEC-8 repeats the Open questions word for word and lacks its **Reference:** line, and DEC-7 ends with a stray "None.".
  • The FR24 verification row puts the failure text under "Representative states" and leaves the failure and audit columns empty.

Automated review (maintainer link)

main now holds Table FR24 and DEC-6 (children-mode wrapping columns). The sticky-header rule moves to FR25 and its decision to DEC-8 (DEC-7 is the open plugin-admission spec), with the verification row and reference renamed.

@astracat-bot astracat-bot Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This pull request is on hold for an engineering owner's decision on the two draft requirements it adds: component:Table FR25 with DEC-8 (the header row stays in view inside Table's own Scroll region whenever its block size is bounded) and module:Table/useTableColumnResize FR11 with DEC-8 (every column, including the last, keeps the width a person sets, and leftover space goes to a filler track). An engineering owner records that decision by commenting /approve-spec <full head SHA> on this pull request, and the author may do so as an engineering owner.

Either way, these need fixing before approval:

  • FR25 makes the Scroll region scroll on both axes for any bounded Table, but Table FR1, its accessibility contract, the Scroll region design row, Compatibility, Builder impact, and the column-resize record's FR2 and Ownership still apply the block axis only under row windowing, so amend them or narrow FR25.
  • Regression: column windowing FR14 says only a rows capability gives Table a parent-sized block viewport, which FR25's container-height bound contradicts, so amend FR14 and Table FR19 with FR25 or scope FR25 to row-windowed Tables.
  • FR11 contracts width resolution and the commit callback, which the column-resize Intent, Compatibility, Ownership non-goals and defect note still exclude or contradict, and it needs a compatibility class and a migration under architecture:public-component-api INV9.
  • FR25 and its verification row bound the Table with maxHeight, which TableProps does not have, so bound it by its parent's height or record a new prop.
  • Regression: Table DEC-8 says Table bounds its block size for windowing (FR19), but FR19 now gives row windowing the height its parent gives it and keeps a columns-only Table's block sizing, so restate DEC-8's premise.
  • The filler track is new Table anatomy with no owner, no rendering rule, nothing hiding it from assistive technology and no place in column windowing's row structure or aria-colcount, so record it in component:Table.
  • The description still names Table FR24 and DEC-7 and says no open pull request claims these lines, but the head adds FR25 and DEC-8 and #11287 proposes a competing sticky-header requirement, so update the description and name #11287 as the alternative.
  • Miss: the description has no ## Writing rubric results section, which architecture:knowledge-contracts requires of a pull request that changes a record.
  • The column-resize DEC-8 repeats the Open questions word for word and lacks its **Reference:** line, and Table DEC-8 ends with a stray "None.".
  • The FR25 verification row puts the failure text under "Representative states", leaves the failure and audit columns empty, and sits above FR1-FR3.
Claims and files checked

Review mode: full; reason: spec-record

  • Claim: people scrolling a long table lose the headings, and a resized last column snaps back. Verified: Table has no sticky header path, and the resize plugin makes the last column the remainder and leaves its key out of the callback.
  • Claim: the draft keeps the header row in view inside Table's own viewport and lets every column keep its width. Verified as what the new Table and column-resize rows propose.
  • Claim: the sticky-header decision is Table FR24 with DEC-7. False: the head adds FR25 and DEC-8; FR24 is the children-mode wrap rule.
  • Claim: the decision builds on DEC-5 and FR19. Partly false: DEC-8 cites both, but FR19 no longer says Table bounds its own block size.
  • Claim: the column-width decision is column-resize FR11 with DEC-8, with every key reported and leftover space in a filler track. Verified.
  • Claim: Table gains an FR, a verification row and DEC-7, numbered after another draft's DEC-6. False: the head adds FR25 and DEC-8, and that draft now adds DEC-7.
  • Claim: the column-resize record gains FR11, a verification row and DEC-8. Verified.
  • Claim: the implementation sets the last column to the remainder today. Verified in the column-resize plugin's width computation.
  • Claim: no code changes. Verified: two Markdown records change.
  • Claim: no open pull request claims these lines. False: #11287 adds a sticky-header requirement row to the same table.
  • Claim: the human-view explainer is kept with the owner review. Not checkable; the description carries no link.
  • File: docs/specs/components/Table/Table.spec.md — findings (FR25 contradictions, maxHeight, DEC-8 premise and stray line, verification row).
  • File: docs/specs/components/Table/plugins/columnResize/useTableColumnResize.spec.md — findings (FR11 contradictions, filler track, copied Open questions).

Automated review (maintainer link)

main now holds Table FR25 (missing cell values). The sticky-header rule moves to FR26, with its verification row renamed.

@astracat-bot astracat-bot Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This pull request is on hold for an engineering owner's decision on the two draft requirements it adds: component:Table FR26 with DEC-8 (the header row stays in view inside Table's own Scroll region whenever its block size is bounded) and module:Table/useTableColumnResize FR11 with DEC-8 (every column, including the last, keeps the width a person sets, and leftover space goes to a filler track). An engineering owner records that decision by commenting /approve-spec <full head SHA> on this pull request, and the author may do so as an engineering owner.

Either way, these need fixing before approval:

  • FR26 makes the Scroll region scroll on both axes for any bounded Table, but Table FR1, its accessibility contract, the Scroll region design row, Compatibility, Builder impact, and the column-resize record's FR2 and Ownership still apply the block axis only under row windowing, so amend them or narrow FR26.
  • Regression: column windowing FR14 says only a rows capability gives Table a parent-sized block viewport, which FR26's container-height bound contradicts, so amend FR14 and Table FR19 with FR26 or scope FR26 to row-windowed Tables.
  • FR11 contracts width resolution and the commit callback, which the column-resize Intent, Compatibility, Ownership non-goals and defect note still exclude or contradict, and it needs a compatibility class and a migration under architecture:public-component-api INV9.
  • FR26 and its verification row bound the Table with maxHeight, which TableProps does not have, so bound it by its parent's height or record a new prop.
  • Regression: Table DEC-8 says Table bounds its block size for windowing (FR19), but FR19 now gives row windowing the height its parent gives it and keeps a columns-only Table's block sizing, so restate DEC-8's premise.
  • The filler track is new Table anatomy with no owner, no rendering rule, nothing hiding it from assistive technology and no place in column windowing's row structure or aria-colcount, so record it in component:Table.
  • The description still names Table FR24 and DEC-7 and says no open pull request claims these lines, but the head adds FR26 and DEC-8 and #11287 proposes a competing sticky-header requirement, so update the description and name #11287 as the alternative.
  • Miss: the description has no ## Writing rubric results section, which architecture:knowledge-contracts requires of a pull request that changes a record.
  • The column-resize DEC-8 repeats the Open questions word for word and lacks its **Reference:** line, and Table DEC-8 ends with a stray "None.".
  • The FR26 verification row puts the failure text under "Representative states", leaves the failure and audit columns empty, and sits above FR1-FR3.
Claims and files checked

Review mode: full; reason: spec-record

  • Claim: people scrolling a long table lose the headings, and a resized last column snaps back. Verified: Table has no sticky header path, and the resize plugin makes the last column the remainder and leaves its key out of the callback.
  • Claim: the draft keeps the header row in view inside Table's own viewport and lets every column keep its width. Verified as what the new Table and column-resize rows propose.
  • Claim: the sticky-header decision is Table FR24 with DEC-7. False: the head adds FR26 and DEC-8; FR24 is the children-mode wrap rule.
  • Claim: the decision builds on DEC-5 and FR19. Partly false: DEC-8 cites both, but FR19 no longer says Table bounds its own block size.
  • Claim: the column-width decision is column-resize FR11 with DEC-8, with every key reported and leftover space in a filler track. Verified.
  • Claim: Table gains an FR, a verification row and DEC-7, numbered after another draft's DEC-6, because the main branch took FR23. False: the head adds FR26 and DEC-8, the main branch has also taken FR24 and FR25, and that draft now adds DEC-7.
  • Claim: the column-resize record gains FR11, a verification row and DEC-8. Verified.
  • Claim: the implementation sets the last column to the remainder today. Verified in the column-resize plugin's width computation.
  • Claim: no code changes. Verified: two Markdown records change.
  • Claim: no open pull request claims these lines. False: #11287 adds a sticky-header requirement row to the same table.
  • Claim: the human-view explainer is kept with the owner review. Not checkable; the description carries no link.
  • File: docs/specs/components/Table/Table.spec.md — findings (FR26 contradictions, maxHeight, DEC-8 premise and stray line, verification row).
  • File: docs/specs/components/Table/plugins/columnResize/useTableColumnResize.spec.md — findings (FR11 contradictions, filler track, copied Open questions).

Automated review (maintainer link)

| FR23 | While columns window, Table owns `aria-colcount`, the full column count, and each mounted header and body cell's full-table `aria-colindex`; spacer cells carry neither, and without column windowing neither attribute is added. | `module:Table/useTableColumnsVirtualization` FR5; DEC-5 | Specified; release verification required |
| FR24 | In children mode with wrapping cells (`textOverflow="wrap"`, the default without windowing), each column takes the width its content needs: a cell breaks a word only when that word alone cannot fit its column, and a header label stays on one line at its full width. When the columns need more inline space than the Table has, the Scroll region (FR1) scrolls them; columns that fit follow their content instead of sharing the width equally. Under `textOverflow="truncate"`, header and body cells keep to the Table's box and end in an ellipsis. Data-driven column widths follow FR13 and do not change with cell content. | DEC-6 | Verified by `TableHeaderCell.test.tsx` and the `ChildrenModeNarrow` story play assertions in Chromium |
| FR25 | Table treats `null`, `undefined` and an Invalid Date (a `Date` whose time is `NaN`) as missing cell values. The default cell renderer, used for a column without `renderCell`, renders a missing value as an empty cell and never throws for one; a valid `Date` renders as its ISO text. `useTableSortableState`'s default comparator sorts missing values, and `NaN`, after every other value in ascending order and before them in descending order, and never throws for an Invalid Date. A column's `renderCell` and a `comparators` entry replace these defaults for their key. | Owner decision, 2026-11-19 | Implemented; unit-tested |
| FR26 | With a bounded block size (`maxHeight`, or a filling height in a container that sets one), Table's own Scroll region scrolls on both axes and the header row stays in view at its block start while rows scroll under it. Sticky columns (`useTableStickyColumns`) keep their place in both directions. Table adds no second scroll region, and a caller does not wrap Table in another scroller to get this (DEC-8). | Proposed, pending | Proposed, pending |

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

TableProps has no maxHeight (packages/core/src/Table/Table.tsx:75-136), so this names an input that does not exist. The row also contradicts current claims left in place: FR1 ("both inline/block with admitted row windowing"), the accessibility contract, the Scroll region design row, Compatibility and Builder impact, and column windowing FR14 ("only a rows capability gives the Table a parent-sized block viewport"). Amend them with this row, or scope it to row-windowed Tables.


| Contract | Verification | Representative states | Mutation or failure expectation | Audit section |
| ---------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | --------------------------- |
| FR26 | Real-browser check: a 200-row Table with `maxHeight`, with and without sticky columns, scrolled on both axes | Header row leaves view, a second scroll region appears, or sticky columns lose their place | — | — |

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The failure text sits under "Representative states" and the last two columns are "—". Move it to "Mutation or failure expectation", put the states (200 rows, with and without sticky columns, both axes) in their column, name an audit section, and place the row after the existing ones instead of above FR1-FR3. maxHeight is not a Table prop here either.

@cixzhang

Copy link
Copy Markdown
Contributor Author

/approve-spec ef95413

@github-actions github-actions Bot removed the needs:spec-owner-review Current knowledge records await owner approval label Nov 21, 2026
@cixzhang

Copy link
Copy Markdown
Contributor Author

/approve-spec c71f7f0

@astracat-bot astracat-bot Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Changes requested. An engineering owner has approved the two new requirements at this head, so the spec direction is settled; the records still need these fixes before they can merge:

  • component:Table FR26 makes the Scroll region scroll on both axes whenever the block size is bounded, but FR1, the accessibility contract, the Scroll region design row and the Compatibility and Builder impact rows still limit block scrolling to row windowing, and the column-resize record's FR2 and Ownership still reserve the region for horizontal overflow.
  • Regression: FR26's "filling height in a container that sets one" contradicts column windowing FR14, which gives a parent-sized block viewport only to a rows capability, so amend FR14 and Table FR19 with FR26 or scope FR26 to row-windowed Tables.
  • Column-resize FR11 contracts width resolution and the commit callback while the record's Intent, Compatibility ("every width outcome", no public API change), Ownership non-goals and defect note still exclude or contradict it, and neither record states the compatibility class and migration that architecture:public-component-api INV9 requires for these released-behavior changes.
  • FR26 and its verification row name a maxHeight input that TableProps does not have, so bound the Table by its parent's height only or record the new prop.
  • Regression: Table DEC-8 says Table "bounds its block size for windowing (FR19)", but FR19 gives a row-windowed Table the height its parent gives it.
  • FR11's empty filler track is new Table anatomy with no owner, no rule hiding it from assistive technology, and no place in column windowing's column structure, aria-colcount or the detail panel's colSpan.
  • The description names FR24 and DEC-7 and says no open pull request claims these lines, but the head adds FR26 and DEC-8, and open #11287 proposes a competing sticky-header requirement.
  • Miss: the description has no ## Writing rubric results section, which architecture:knowledge-contracts requires for a pull request that changes a record.
  • Column-resize DEC-8 repeats the Open questions section word for word and lacks its **Reference:** line, and Table DEC-8 ends with a stray "None.".
  • The FR26 verification row puts its failure text under "Representative states", leaves the last two columns empty, and sits above the FR1-FR3 row.

Amending the records changes their text, so an engineering owner approves the new head again with /approve-spec <full head SHA>.

Claims and files checked

Review mode: incremental; reason: none

  • Claim: people scrolling a long table lose the column headings: verified, Table renders no sticky header row today.
  • Claim: the last column of a resizable table snaps back: verified, the resize plugin sets it to the remaining width (useTableColumnResize.tsx:208-218).
  • Claim: the change keeps the header row in Table's own viewport and lets every column keep its width: verified, FR26 (Table.spec.md:255) and FR11 (useTableColumnResize.spec.md:130).
  • Claim: a human-view explainer is kept with the owner review: not checkable in the repository; no finding.
  • Claim: the sticky-header decision is Table FR24 with DEC-7: false, the head adds FR26 and DEC-8.
  • Claim: the recommended option builds on DEC-5 and FR19, with two rejected alternatives: partly true, DEC-8 records both alternatives, but its FR19 premise is stale.
  • Claim: the column-width decision is column-resize FR11 with DEC-8, with a filler track and every key reported: verified (useTableColumnResize.spec.md:130, :383).
  • Claim: Table gains FR24, a verification row and DEC-7, numbered after another draft's DEC-6: false, it gains FR26 and DEC-8, and #10107 adds DEC-7.
  • Claim: column resize gains FR11, a verification row and DEC-8: verified (:130, :268, :383).
  • Claim: the implementation sets the last column to the remainder today: verified (useTableColumnResize.tsx:208-218).
  • Claim: no code changes: verified, only the two records change.
  • Claim: no open pull request claims these lines: false, #11287 adds a competing sticky-header row.
  • No changeset entries.
  • File docs/specs/components/Table/Table.spec.md: FR1 and accessibility contradiction, FR14 contradiction, maxHeight, DEC-8 premise, filler anatomy, stray line, shifted verification row.
  • File docs/specs/components/Table/plugins/columnResize/useTableColumnResize.spec.md: FR2 and Ownership contradiction, FR11 scope and compatibility, filler anatomy, copied Open questions and missing Reference line.

Automated review (maintainer link)

@github-actions github-actions Bot added the needs:spec-owner-review Current knowledge records await owner approval label Nov 21, 2026

This branch was successfully deployed

1 active deployment
Preview — 579282eb Deployed Nov 21, 2026 by vercel[bot]
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

CLA Signed This label is managed by the Meta Open Source bot. needs:spec-owner-review Current knowledge records await owner approval

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant