What apache/superset shipped
Written by FoxPlug from the repository's own releases; not affiliated with Apache Superset. Paste your own repository in the box below and FoxPlug writes the same weekly update for your product. This page is an automatic summary of the releases, pull requests and commits in github.com/apache/superset; Apache Superset did not write it and does not use or endorse FoxPlug. Every line links to the change it describes.
Get a weekly update like this for your product, free
Week of October 5, 2026
What shipped
- Group By native filters on dashboards can now be limited to a builder-curated allowlist of columns instead of any groupable column. Pull request #43736
- Typed geographic chart generation was added end to end, letting clients create native map charts rather than substitute unrelated ones. Pull request #44148
- Scheduled soft-delete purge and version-history prune now have a per-run cap so a large backlog is not removed in one pass. Pull request #45029
- Saving semantic-layer configuration now invalidates its metadata caches, and a Reload metadata action clears caches for upstream-only changes. Pull request #45100
- An opt-in compact native tools listing and trimmed tool guidance were added for tool discovery. Pull request #45084
- A shared datasource permission is preserved during soft-delete purge while another owner still uses it. Pull request #45034
- Malformed explicit time ranges are now rejected with the existing validation error instead of being turned into an upper-bound-only query. Pull request #45098
- Semantic-view chart data now preserves nullable integer precision instead of losing it through float conversion. Pull request #45126
- Cumulative normalized histograms now divide by the number of data points rather than the sum of running totals. Pull request #44652
- The versioning capture predicate is deferred until versioned work occurs, so unrelated writes no longer pay for it. Pull request #44928
Why it matters
This week tightened correctness in a few places where results were quietly wrong: semantic-view chart data no longer loses integer precision, cumulative normalized histograms divide correctly, and malformed time ranges are rejected instead of silently rewritten. It also made several failure modes less disruptive, from boolean filters on Databricks to shared datasource permissions during purge. Dashboards got a new allowlist for Group By filters and map charts can now be created natively.
Changelog entry
- Add a column allowlist to the Group By native filter on dashboards. Pull request #43736
- Add typed geographic chart generation end to end. Pull request #44148
- Cap soft-delete purge and version-history prune per run. Pull request #45029
- Invalidate semantic-layer metadata caches on configuration save and add a Reload metadata action. Pull request #45100
- Add opt-in compact native tools listing and trimmed tool guidance. Pull request #45084
- Preserve a shared datasource permission on soft-delete purge. Pull request #45034
- Reject malformed explicit time ranges instead of scanning everything. Pull request #45098
- Preserve nullable integer precision in semantic-view chart data. Pull request #45126
- Normalize before accumulating a cumulative histogram. Pull request #44652
- Resolve saved semantic-view chart sources by type in chart data and chart info. Pull request #45067
- Use a shared lock when re-reading a concurrently created permalink. Pull request #45102
- Filter special partition IDs from the BigQuery max partition query. Pull request #44156
This week: Group By native filters get a column allowlist, native map charts, capped soft-delete purge, metadata cache invalidation, and fixes for integer precision, cumulative histograms, Databricks boolean filters and malformed time ranges.
Changes this week include an allowlist for Group By native filters so builders can curate which columns viewers may group by, and typed geographic chart generation so native map charts can be created instead of substitutes. Scheduled soft-delete purge and version-history prune gained a per-run cap, saving semantic-layer configuration clears its metadata caches, and a shared datasource permission now survives purge while another owner uses it. Fixes landed for nullable integer precision in semantic-view chart data, cumulative normalized histograms, boolean filters on Databricks, malformed time ranges being rejected, and the versioning capture predicate being consulted only when versioned work happens.
Week of September 28, 2026
What shipped
- Dashboard Properties Modal now has a native GUI for managing label colors in the Styling tab alongside color scheme selection. Pull request #39434
- Storybook now publishes to Chromatic for visual regression testing in CI. Pull request #44103
- Added user-facing and admin documentation for the four semantic-layer MCP tools (
list_metrics,get_table,get_compatible_dimensions,get_compatible_metrics). Pull request #44130 - Fixed a silent TLS downgrade in MySQL connections where REQUIRED mode could fall back to cleartext when using MariaDB Connector/C. Pull request #44927
- Fixed cumulative product, min, and max operators treating data gaps as zero, which zeroed the remainder of series after any gap. Pull request #44828
- Heatmap charts now sort Y-axis by default when holding string categories instead of rendering in arbitrary query order. Pull request #44588
- Fixed SQLite DATE column time filters to write midnight as a bare date instead of a full timestamp literal. Pull request #44805
- Dataset editor now preserves manual edits across sort operations and properly syncs external SQL changes. Pull request #44858
- Native filter cascade dependency gates now stay in sync with live filter type changes. Pull request #44818
- CurrentUserRestApi now returns 401 for auth errors instead of 500 when JWT validation fails. Pull request #44213
Why it matters
This week shipped fixes for data integrity issues (cumulative operators, duplicate columns, heatmap sorting), improved database compatibility (SQLite dates, MySQL TLS, ClickHouse GROUP BY ALL coverage), and better error handling in authentication and API validation. Documentation for semantic-layer MCP tools and visual regression testing infrastructure also landed, supporting both developer experience and reliability.
Changelog entry
- Dashboard Properties Modal: native GUI for
label_colorsmanagement in Styling tab Pull request #39434 - CI: Storybook visual regression testing via Chromatic Pull request #44103
- Docs: semantic-layer MCP tools (
list_metrics,get_table,get_compatible_dimensions,get_compatible_metrics) Pull request #44130 - SQLite: DATE column time filters now write midnight as bare date Pull request #44805
- Post-processing: cumprod, cummin, cummax no longer treat gaps as zero Pull request #44828
- Heatmap: Y-axis now sorts string categories by default Pull request #44588
- MySQL: fixed silent TLS downgrade when using MariaDB Connector/C Pull request #44927
- API: CurrentUserRestApi returns 401 for auth errors instead of 500 Pull request #44213
- Dataset editor: preserves edits across sort and syncs external SQL changes Pull request #44858
- Native filters: cascade dependency gates sync with live filter type Pull request #44818
- SQL Lab: avoids duplicate generated column name collisions Pull request #44189
- CrateDB: testcontainers coverage for epoch-ms timestamp decoding Pull request #44904
This week: dashboard label_colors GUI, Chromatic visual regression testing, semantic-layer MCP docs, SQLite DATE fixes, cumulative operator corrections, heatmap sorting, MySQL TLS hardening, and auth error handling improvements.
Superset shipped significant improvements this week across data integrity, database compatibility, and developer experience. Dashboard styling now has a native GUI for label colors, Storybook integrated with Chromatic for visual regression testing, and semantic-layer MCP tools are now documented. Critical fixes address cumulative operators treating gaps as zero, SQLite DATE columns in time filters, silent TLS downgrades in MySQL, and authentication error responses. These changes strengthen reliability and compatibility across multiple database engines.
Week of September 21, 2026
What shipped
- Candlestick Chart visualization added to ECharts plugin for OHLC time series data with support for X-axis, Open/Close/High/Low metrics, and optional series dimensions. Pull request #43863
- Excel export now works directly without requiring S3 bucket configuration, previously returned 501 error on deployments without external storage. Pull request #44082
- Explore standalone modes now granular with separate UI hiding options, and dashboard fullscreen toggle made safe for iframe embedding. Pull request #44165
- Large CSV exports now stream without executing the full query first when row count exceeds streaming threshold. Pull request #44624
- Fixed Slice.form_data crash when params JSON decodes to non-dict types like strings or arrays. Pull request #44731
- MongoDB dataset queries now return rows by properly quoting table names in FROM clauses. Pull request #44583
- Dashboard edit-mode row resizing fixed to shrink properly and unblock resize handles above columns. Pull request #44359
- MCP dataset tools expanded with
create_dataset_metricanddelete_dataset_metricoperations. Pull request #44580 - Charts outside in-flight auto-refresh batches now display loading spinner when switching dashboard tabs. Pull request #44272
- Dataset column refresh now updates
changed_ontimestamp so chart cache invalidates properly. Pull request #43921
Why it matters
This week shipped visualization improvements with the new Candlestick Chart for time series analysis, fixed critical data export paths for deployments without external storage, and resolved several query execution bugs affecting MongoDB datasets, CSV streaming, and dashboard interactions. The changes improve both feature completeness and stability for core workflows.
Changelog entry
- Candlestick Chart visualization type added for OHLC time series Pull request #43863
- Excel export works directly without S3 bucket configuration Pull request #44082
- Explore standalone modes now granular; dashboard fullscreen safe in iframes Pull request #44165
- Large CSV exports stream without executing full query first Pull request #44624
- Fixed Slice.form_data crash on non-dict JSON params Pull request #44731
- MongoDB dataset queries return rows with proper table name quoting Pull request #44583
- Dashboard row resizing shrinks properly; resize handles unblocked Pull request #44359
- MCP tools add
create_dataset_metricanddelete_dataset_metricoperations Pull request #44580 - Loading spinner displays for charts outside auto-refresh batches Pull request #44272
- Dataset column refresh invalidates chart cache via
changed_onupdate Pull request #43921 - Table time comparison columns resolve auto-detected currency per row Pull request #44649
- DB2 time grain expressions improved with DATE_TRUNC function Pull request #43236
New Candlestick Chart visualization, Excel exports without S3, granular Explore standalone modes, and fixes for MongoDB queries, CSV streaming, and dashboard resizing.
This week brings visualization and stability improvements to Apache Superset: the new Candlestick Chart for OHLC time series, Excel export that works without S3 buckets, granular Explore standalone modes for iframe embedding, and critical fixes for MongoDB queries, large CSV streaming, dashboard resizing, and chart caching after dataset refresh.