Repository navigation
Conversation
… columns After #7583 the interactive table shrinks Experiment and User to their floors, but the other columns keep fixed widths, so below about 134 columns the rightmost columns are still cut off. When the table is still too wide at the floors, two more tiers apply: 1. Compact: Started becomes `MM-DD HH:MM` and the MLflow column is sized to its widest stored label instead of a fixed 22 cells. Labels are not re-truncated, so they keep their end and retry note (#7582). 2. Hide: whole columns are dropped lowest priority first (Accelerators, User, MLflow, Progress) until the table fits. The hint line names the hidden columns; when the full key legend and that note do not both fit, the legend shortens to its keys. After each step the spare cells go back to Experiment and User. Enter still opens the row's MLflow link when the MLflow column is hidden. The one-shot table (piped, NO_COLOR, --limit) and terminals wide enough for today's table render exactly as before. Verified with unit tests at 80, 100, 120, 133, 160 and 200 columns and by running the CLI in a pty at those widths against e2-dogfood. CNXT-2619 Co-authored-by: Isaac <[email protected]>
Co-authored-by: Isaac <[email protected]>
Collaborator
✅ Integration test reportCommit: 74de95f
Top 3 slowest tests (at least 2 minutes):
|
…Lflow label width Review follow-ups for #7629. The hint line could still run past the terminal, and bubbletea cut off its end, which is where the loading / load-failed state sat. The hint now gives up detail in steps until it fits: the legend shrinks to its keys, then the hidden columns are counted instead of named, and as a last resort the line is truncated. The variant is chosen for the widest row counter and the loading state, so it does not flip as the cursor moves or a page loads. Tests now also cover the production case where every MLflow label is the full 22 cells, the Progress tier, the still-too-wide case, a page load that crosses a tier, the selected row in a reduced layout, and the hint at 70, 75, 80 and 100 columns in every paging state. Also: cap the compact MLflow width at mlflowColWidth, show a non-ISO start timestamp as is instead of slicing it, decide right alignment in one place, drop the duplicated link check in mlflowCell, and trim doc comments. Co-authored-by: Isaac <[email protected]>
…as no room for a legend Below the width where even the short legend fits, the hint now drops the legend before truncating, so the row counter and loading / load-failed state stay visible. Also test the MLflow width cap, compare whole hints in the stability check, and say that mlflowColWidth is the widest label list.go produces rather than the only one. Co-authored-by: Isaac <[email protected]>
maggiewang-db
approved these changes
Nov 14, 2026
maggiewang-db
left a comment
maggiewang-db
left a comment
Contributor
There was a problem hiding this comment.
Nit: Don't mention internal jira tickets in open source repo.
This branch has not been deployed
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
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.
Changes
Follow-up to #7583 for CNXT-2619. The interactive
databricks air listtable shrank Experiment and User to their floors, but every other column kept a fixed or content width, so below about 134 columns the rightmost columns were still cut off (Progress at 80 columns, Accelerators and most of User at 120).When the table is still too wide once Experiment and User reach their floors, two more tiers now apply, in order:
MM-DD HH:MM(11 cells instead of 19) and the MLflow column is sized to its widest stored label instead of a fixed 22 cells. The stored labels are not re-truncated, so they keep their end and their retry note (air: Shorten MLflow run names and IDs the same way inair list#7582).hidden: MLflow, User, Accelerators. The hint must itself fit the terminal, or bubbletea cuts off its end (where the loading / load-failed state sits), so it gives up detail in steps: the legend shortens to↑/↓ ←/→ ↵ i L q, then the hidden columns are counted (3 columns hidden), then the legend goes, and as a last resort the line is truncated. The row counter and paging state come first, so they survive every step. The variant is chosen for the widest row counter and the loading state, so it does not flip as the cursor moves or a page loads.After each step the spare cells go back to Experiment and User, so they widen again rather than staying at the floors. Enter still opens the row's MLflow link when the MLflow column is hidden. Lazily loaded pages are laid out against the same width; since a new page can only widen the natural columns, the layout moves one way as pages load (a column may compact or disappear mid-scroll, never reappear).
Unchanged:
NO_COLOR,--limit) is byte-for-byte the same.Implementation:
listColsgains alistColumnenum, anmlflowwidth, acompactStartedflag and ahiddenbitmask;layoutListCols(rows, width)runs the tiers; the header and row renderers iterate the enum and skip hidden columns.Before / after at 80 columns (e2-dogfood)
Before:
After:
At 120 columns only Accelerators is hidden; at 133 the compact tier alone fits with every column shown; at 134 and above the layout is unchanged.
Why
CNXT-2619 asked for a product decision between compacting fixed columns and hiding low-priority ones. Both are needed: with real data every MLflow label is already 22 cells, so compacting alone only covers terminals from 126 to 133 columns. Hiding is what gets 80 and 100 columns to a readable table.
Tests
TestLayoutListCols(0, 60, 80, 100, 120, 133, 160, 200 columns with short MLflow labels),TestLayoutListColsFullLabels(100, 120, 125, 126, 133 columns with 22-cell labels, the widest list.go produces and the usual case for MLflow-named runs),TestLayoutListColsMovesOneWayAsPagesLoad,TestLayoutListColsMLflowFloorIsHeader,TestLayoutListColsCapsMLflowAtColumnWidth,TestStartedCompact,TestHiddenNote,TestRenderHidesColumns,TestListModelNarrowTerminalHidesColumns,TestListModelHintNamesHiddenColumnsWithFullLegend,TestListModelHintFitsNarrowTerminal(40, 46, 70, 75, 80, 100 columns, 120 rows, idle / loading / load failed),TestListModelRelayoutsLoadedPageInNarrowTerminal,TestListModelEnterOpensMLflowLogsWhenColumnHidden. The existingTestStaticListTableIgnoresTerminalWidthandTestListColsFitWidthstill pass unchanged.air list --all-status --all-users: no table or hint line exceeds the terminal width at any size, and 134 and 160 match the main build.--limitoutput in a pty between the main build and this branch: identical.golangci-lint run ./cmd/air/...clean.This pull request and its description were written by Isaac.