[Grid] Implement frozen first column in SWT (RCP), fixes #695#696
[Grid] Implement frozen first column in SWT (RCP), fixes #695#696PHMT-CW wants to merge 4 commits into
Conversation
Builds on the basic fixed column scaffolding from EclipseNebula#645 to make a frozen first column behave correctly across the rest of Grid: - getColumn(Point) resolves hits inside the fixed-column overlay to the frozen column on top instead of the scrolled column underneath. This automatically fixes getCell, overColumnHeader, overColumnFooter, and the mouse handlers that funnel through it. - getColumnHeaderXPosition and getOrigin are now fixed-aware, so GridEditor.computeEditorBounds and GridItem.getBoundsCorrected place cell editors on the frozen column instead of off-screen. - The fixed cell-painting pass now clips to the overlay rectangle so renderers cannot bleed into the scrolled area. - paintFooter got a fixed-aware re-pass so the frozen column footer stays put on horizontal scroll, mirroring the existing header pass. - The focus rectangle starts to the right of the frozen overlay so its left edge stays visible. - The header fixed pass detects when a frozen column shares a column group with scrollable members; it logs a one-time warning to stderr and degrades gracefully by drawing only the column header (no group band) in the overlay. - Updated GridColumn#setFixed Javadoc to document the supported case (single frozen first column) and the column-group restriction. - Replaced the stale "TODO: column freezing" note with a pointer to setFixed and the new snippet. Adds GridFrozenFirstColumnSnippet (plain frozen column with editor + group-spanning case) and GridFixedColumn_Test (setFixed, frozen-aware getOrigin, hit-testing under scroll, one-shot group-spanning warning). Fixes EclipseNebula#695 Signed-off-by: Cameron Weick <cpwei@phmtechnology.com>
Add a Tycho test-plugin pom for org.eclipse.nebula.widgets.grid.test and register the bundle as a child module of widgets/grid so CI runs the new GridFixedColumn_Test alongside the rest of the build. Surefire is intentionally scoped to **/GridFixedColumn_Test.java for now: the legacy *_Test classes in the bundle predate current SWT behavior, JUnit, and Mockito (1.8.4), and have been bit-rotting in the IDE-only .launch configuration for years. Restoring them is a worthwhile but orthogonal effort that should not block the frozen-first-column work this bundle was wired up to validate. While here, harden the freeze tests against a hidden ordering bug: ScrollBar.setSelection is silently clipped to the SWT default maximum of 100 until Grid.updateScrollbars() runs as part of a paint, so a setSelection(150) called before the first paint never actually scrolled. The new flushPaint / scrollHorizontallyTo helpers force a paint, set the requested selection, and return the actual clipped value; tests assert on the real delta and refuse to silently pass at scroll=0. Signed-off-by: Cameron Weick <cpwei@phmtechnology.com>
I'm a bit confused, I use this in SWT (Nebula has nothing to do with RCP) ... not sure whats needed for RAP. |
Hello, With regard to RAP, this comment specifically references the developer guide here. Currently, it remains somewhat unclear whether the SWT implementation was fully finalized. During our recent testing of the frozen column feature within our Eclipse RCP product, we encountered an issue where the frozen columns did not satisfy the hit test for expanding and collapsing tree items, alongside a few additional complications. |
On headless Linux GTK the SWT widget is never realized during the synthetic redraw/update/readAndDispatch cycle used by these tests, so Grid.updateScrollbars() never runs and the horizontal scrollbar stays at its defaults (visible=false, max=1, thumb=1). GTK treats setSelection on an invisible scrollbar as a silent no-op, which caused three frozen-column tests to fail on CI while passing everywhere else. Switch the scroll helper to ScrollBar.setValues with explicit setVisible(true), which atomically assigns selection/max/thumb and bypasses the realization dependency entirely. Fixes three CI failures in PR EclipseNebula#696. Signed-off-by: Cameron Weick <cpwei@phmtechnology.com>
PHMT-CW
force-pushed
the
grid/frozen-first-column-pr
branch
from
3bcced2 to
b549a0d
Compare
Would you mind to enhance the Snippet first to show the problems? That usually a bit easier to understand then. Also it is unclear why it is only affecting the first column? And yes this feature is quite new and would likely need some improvements or hardening, so thanks for working on this! |
Extend GridFrozenFirstColumnSnippet with a third section that demonstrates the freeze + tree interaction interactively: a frozen first column configured as the tree column, with expandable parent rows and a caption telling the reader what to do to verify the behaviour. This is the scenario that motivated the original bug report. Without freeze-aware hit-testing, clicking an expand/collapse toggle while the grid is horizontally scrolled routes the click to the column underneath the overlay and the toggle appears unresponsive. The snippet now lets a reviewer reproduce the original symptom and confirm the fix. The same scenario is also covered by an automated regression test, GridFixedColumn_Test.testTreeToggle_OnFrozenColumn_RespondsWhenScrolled. Refs EclipseNebula#695 Signed-off-by: Cameron Weick <cpwei@phmtechnology.com>
Yes of course. I've added a third section to The same scenario is also covered by an automated test, All three pass with the PR applied. Why first column only? On the broader question of whether Grid should eventually match the RAP build and support N fixed columns, my own preference would be to keep the SWT Grid relatively lean and point users with heavier needs at NatTable, which already handles multi-region freeze, fixed rows, viewport layers, and large datasets very well. That is only my opinion as a contributor though, so I am happy to defer to your judgement on the intended scope. If full parity with the RAP side is the goal then widening it to N columns afterwards should be a relatively small follow-up. |
|
@laeubi are you happy with these changes? |
3 participants
Closes #695.
Continues the scaffolding added in #645, which introduced the
setFixedAPI but only wired the behaviour for the RAP platform. On SWT (RCP) the TODO inGrid.javanoted that the RCP implementation was still pending; this PR delivers it for the common case of a single frozen first column.Summary
getColumn(Point),getOrigin, and header/footer hit tests now resolve to the frozen column under the fixed-overlay rectangle, so expand/collapse, selection, tooltips, andGridEditorplacement all behave correctly while the grid is scrolled horizontally.GridColumnGroup: a frozen column that shares a group with scrollable members skips the group header and logs a one-time warning to stderr. Documented in thesetFixedJavadoc.setFixedJavadoc documents the supported case and the group restriction.Tests
GridFixedColumn_TestcoveringsetFixed/isFixed, fixed-awaregetOrigin, hit-testing under scroll, origin invariants, and the one-shot group-spanning warning.org.eclipse.nebula.widgets.grid.testis now wired into the Tycho build. Surefire is scoped toGridFixedColumn_Testfor now — the legacy*_Testclasses in the bundle predate current SWT behaviour, JUnit and Mockito 1.8.4, and have been bit-rotting in the IDE-only.launchconfiguration for years. Restoring them is worthwhile but orthogonal; the commit message documents the rationale.Manual verification
GridFrozenFirstColumnSnippetdemonstrates the plain frozen column +GridEditor, and the group-spanning case.Screenshots / demo
GRID.Frozen.First.Column.-.2026-04-21.17-18-31.webm