LibWeb: Decouple layout from caret position and editable state - #10962
Merged
kalenikaliaksandr merged 4 commits intoAug 1, 2026
Merged
Conversation
The inner editor element of text-like <input> controls used white-space: nowrap, which collapses interior whitespace runs and trims trailing spaces from the rendered value. Input values are plain text and other engines render them verbatim by styling the inner editor with white-space: pre; with nowrap, <input value="a b "> lost the extra spaces, and a trailing space typed by the user only stayed visible because line layout special-cases trailing-whitespace trimming while the caret sits on the node. Switch the inner editor to white-space: pre so input values render verbatim. This also lets the caret advance past freshly typed trailing spaces without any help from layout, clearing the way for removing the cursor-position dependency from line layout.
Trailing-whitespace trimming skipped its work whenever the document cursor sat on the fragment's node, so that a trailing space typed into a text control stayed visible. Nothing invalidates layout when the caret moves, so this made geometry depend on whenever layout happened to run: the preserved space would linger after the caret left the node (or the element lost focus) until some unrelated relayout trimmed it, and through trailing_whitespace_inline_size() the caret position could even shift line breaking and float placement decisions. The behaviors this special case provided are all covered at the right layers now: text input inner editors preserve whitespace outright (white-space: pre), textarea editors inherit pre-wrap, and typing into an editing host canonicalizes a would-collapse space into a non-breaking space per the editing spec's whitespace canonicalization. A caret placed inside genuinely collapsed trailing whitespace renders clamped to the line end, since caret matching already accepts offsets into a fragment's trimmed trailing whitespace. Remove the document_cursor_is_on_node FFI callback and the trimming special case; layout geometry no longer depends on caret position, so the repaint-only invalidation done on caret moves is now sufficient by construction. Adds a regression test that the caret keeps advancing while trailing spaces are typed into a text input and an editing host.
Inline layout called back into the DOM mid-layout to ask whether an empty text node sits in an editable context (a text control's shadow tree or under an editing host), deciding whether to synthesize the zero-width fragment that keeps the line box alive with real font metrics so the caret has an anchor and the control keeps its baseline. That fact derives purely from DOM structure, so a query at layout time is the wrong shape for it: compute it when the layout node is created and stamp it as ProducesLineBoxFragmentWhenEmpty in NodeData, the same pattern already used for IsEditingHost. The existing editable-subtree walk refreshes the stamp when contenteditable or designMode change without a layout tree rebuild, and the layout engine now reads the flag with no knowledge of editing concepts. This removes the last DOM-querying FFI callback from inline layout, and the existing empty-editable layout expectations are unchanged by the conversion. The stamp intentionally drops the old is_mutable() gate: empty readonly and disabled text controls now produce the same zero-width fragment as mutable ones, so a focused readonly control shows its caret and empty controls keep the same baseline regardless of mutability, matching other engines.
Toggling contenteditable (or designMode) re-stamped the editing-host flag into layout NodeData but only requested a repaint, and nothing ever dirtied layout. The geometry that depends on those stamps -- the minimum block size of an empty editing host and the zero-width fragment of an empty editable text node -- would keep its stale shape until some unrelated relayout happened to run, so a div made editable at runtime showed no caret and kept zero height until then. Detect in the editable-subtree walk whether a node's stamped editing-host status or a text node's empty-text fragment behavior actually changed, and request a layout update for that node when it did. Adds a regression test that an empty text node in a freshly editable host immediately anchors a caret with line height, and that the host's minimum block size appears and collapses as contenteditable is toggled.
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (17)
💤 Files with no reviewable changes (4)
📝 WalkthroughWalkthroughThe change propagates editable-state updates into layout flags, synthesizes line-box fragments for qualifying empty text nodes, removes obsolete Rust callbacks, preserves input whitespace, and adds caret regression tests. ChangesEditable text layout
Estimated code review effort: 3 (Moderate) | ~20 minutes Sequence Diagram(s)sequenceDiagram
participant DOMNode
participant TextNode
participant NodeFacts
participant InlineLevelIterator
DOMNode->>TextNode: update empty line-box flag
TextNode->>NodeFacts: publish ProducesLineBoxFragmentWhenEmpty
InlineLevelIterator->>NodeFacts: read empty line-box fact
NodeFacts-->>InlineLevelIterator: return flag state
InlineLevelIterator->>InlineLevelIterator: synthesize zero-length text chunk
Possibly related PRs
Suggested reviewers: 🚥 Pre-merge checks | ✅ 3✅ Passed checks (3 passed)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Line layout queried the DOM mid-layout through two FFI callbacks: whether the
caret sits on a node (to skip trailing-whitespace trimming while typing) and
whether an empty text node is in an editable context (to synthesize the
zero-width fragment that anchors the caret). The first is deleted outright —
its job is now done at the right layers, by giving text input inner editors
white-space: pre like other engines and relying on the editing layer's
existing NBSP canonicalization — which also makes geometry independent of
caret movement instead of stale until an unrelated relayout. The second
becomes a NodeFlag stamped at layout-node creation and refreshed by the
editable-subtree walk, which now also triggers a relayout when a stamp
actually flips, fixing a pre-existing bug where an element made editable at
runtime showed no caret and kept zero height. Covered by new regression
tests for caret advance past typed trailing spaces and for contenteditable
toggling.