Releases

Changelog

Every release, newest first. Versions follow SemVer, and until 1.0 a minor version may change the settings format. The .vsix for each version is attached to its GitHub release.

Current 0.2.0 · 6 September 2026 Marketplace↗ Releases and .vsix downloads↗ CHANGELOG.md on main↗
0.2.0LatestCompare on GitHub↗

Added

  • Explain Plan and Explain Analyse, from the console toolbar's Explain dropdown, the editor's context menu, and the Tablecloth: Explain Plan and Tablecloth: Explain Analyse commands. The plan opens in a Plan tab next to the results: a tree of operations with what each works on, its cost and its row estimate, or the same nodes as a table with every figure. Explain Analyse runs the statement and adds the actual rows and the time per node, with a heat bar so the node that lied stands out. Every analyse is rolled back (a savepoint inside an open transaction, a transaction of its own otherwise), so a DELETE or UPDATE can be analysed without losing data on a transactional engine. PostgreSQL, MySQL and MariaDB (through its ANALYZE) give the full plan; SQLite has no analyse and shows its query plan instead, with a note in the Output.
    Explain Plan on a join: the operations as a tree, costs and row estimates, Explain Analyse a click away.
    Explain Plan on a join: the operations as a tree, costs and row estimates, Explain Analyse a click away.
    Explain Analyse: the real numbers, rows removed by the filter, and the time per node.
    Explain Analyse: the real numbers, rows removed by the filter, and the time per node.
  • A DELETE or UPDATE with no WHERE clause gets a warning squiggle in consoles and attached .sql files, and asks before it runs. The dialog names the table and how many rows it holds, counted on the console's own session so an open transaction's changes are included. Cancel is the default button, so a stray ⏎ after ⌘⏎ cancels; Run anyway runs it, and Don't ask again for this console silences it for that console. Run File on Data Source asks the same question with a Run all anyway answer for the rest of the run. The check follows IntelliJ's exemptions: a LIMIT, a JOIN whose condition constrains the table being written, and an UPDATE whose every assignment reads its own column are left alone.
    The whole statement is marked, comment excluded.
    The whole statement is marked, comment excluded.
    The dialog names the table and its row count; Cancel has focus.
    The dialog names the table and its row count; Cancel has focus.
  • Setting tablecloth.execution.warnWithoutWhere turns the question off everywhere. Changing it, in either direction, resets every console's "Don't ask again" choice.

Changed

  • tablecloth.inspections.enabled now also governs the DELETE or UPDATE without WHERE inspection.
0.1.1Compare on GitHub↗

Changed

  • The README and the Marketplace listing point at the docs site, sultanofcardio.github.io/tablecloth: the roadmap, the changelog and the known limits live there now, and the extension's homepage is the site.
    The docs site the README and the listing point at now.
    The docs site the README and the listing point at now.
0.1.0Phase 2Compare on GitHub↗

Phase 2 (daily driver) of the plan: edit data in the grid, and trust the SQL the console writes for you.

Added

  • Editable data grid: double-click a cell (or press F2) to edit it, and the change accumulates locally (edited cells blue, added rows green, deleted rows struck through). Submit shows the exact DML first and runs it as one atomic batch; Revert selected or Revert all throws it away.
    Two edited cells, a deleted row and a new row: four pending changes, nothing sent yet.
    Two edited cells, a deleted row and a new row: four pending changes, nothing sent yet.
    Submit shows the four statements before they run, as one batch.
    Submit shows the four statements before they run, as one batch.
  • Row operations in the grid toolbar and context menu: add a row, clone the selected row, delete rows, and Set NULL or Set DEFAULT on any cell.
    A new row with its auto and default placeholders, and the Submit button counting the changes.
    A new row with its auto and default placeholders, and the Submit button counting the changes.
  • Transaction mode per data editor, with isolation levels and commit or roll back, alongside the mode already available per console.
  • WHERE and ORDER BY fields above the grid, each with an IntelliJ-style completion lookup over the table's columns, the keywords valid in that clause, and functions. Nothing completes inside a string literal, and completing inside a quote you already typed keeps that quote.
    The WHERE lookup offering a column and a table after three letters.
    The WHERE lookup offering a column and a table after three letters.
  • Sorting and filtering from the header: click a column header to write the ORDER BY (Alt-click to sort by several columns), open a column funnel to pick from its distinct values, or use Filter by value on a cell.
    Sorted by created_at from its header, with the status funnel open.
    Sorted by created_at from its header, with the status funnel open.
  • Foreign-key navigation: a FK cell jumps to the referenced row, and the context menu opens the rows that reference the current one. View Query and Copy Query to Console show the SQL behind the grid.
  • Alternate views of a result: transpose, Table, Tree, and Text views, a value editor for long values (⇧⏎), the column list (⌘F12), and find in page (⌘F).
    The same page transposed, one column per row.
    The same page transposed, one column per row.
  • Console intelligence: completion for keywords and functions as well as objects, JOIN clauses and ON conditions inferred from foreign keys, live templates (sel, selw, ins, upd, del, tab, and friends), inspections for unresolved tables and columns with Change-to quick fixes, Format SQL (⌘⌥L, or Format Document), :name and ${name} parameters with a values dialog, and cancelling a running statement (⌘F2).
    Alias and keyword suggestions after a table name.
    Alias and keyword suggestions after a table name.
    An unresolved column flagged in the console, and a :min_total parameter waiting for a value.
    An unresolved column flagged in the console, and a :min_total parameter waiting for a value.
    The Change-to quick fix on the flagged column.
    The Change-to quick fix on the flagged column.
    The Parameters dialog on run.
    The Parameters dialog on run.
  • Import Data from File: delimiter detection, a column mapping step, creating the target table from the file, and batched inserts that stop or skip on error.
    Import Data in its own window: format settings, the column mapping, and the row count on the button.
    Import Data in its own window: format settings, the column mapping, and the row count on the button.
  • New extractors: HTML, JSON, Markdown, One-row, Pretty, Python-DataFrame, SQL-Insert-Multirow, and XML. Excel (.xlsx) is available through Export Data, and the extractor and Export menus follow IntelliJ's shape.
  • Go to Database Object (⌘⇧O) searches every introspected source, and Go to DDL opens the CREATE statement for tables, views, routines, sequences, and enum types.

Changed

  • Console result grids are editable only for single-table SELECTs whose key columns are in the result. Any other result opens read-only and says why.
  • The SQL Updates extractor keeps known primary-key values in the WHERE clause even when the key column is not among the exported columns, while SET still covers only the columns you selected.
  • An SQL Updates copy or export whose selected columns are all part of the key has nothing to set, so it produces a single comment naming the key columns and how to get UPDATE statements; the grid says the same in the status bar after a copy, or as a warning after an export to file, so the result is never silently empty.

Fixed

  • An isolation level is applied to the session that runs the statements it governs, and is reapplied to a fresh session after a reconnect.
  • Exports and client-side sorting keep 64-bit integers exact; large values no longer round through a floating-point value.
  • Generated SQL quotes reserved words and case-sensitive identifiers for the dialect in hand, so tables and columns named after keywords round-trip.

Known limits

  • The grid is no longer read-only. The limits this release still carries are listed in the known limits page.
0.0.3Compare on GitHub↗

Changed

  • New data sources default to Project scope (workspace settings) whenever a trusted workspace folder is open, so a source stays with the code it belongs to. Global remains a click away in the dialog, and is still the default with no folder open or in Restricted Mode.
    The Scope selector next to the name, defaulting to Project.
    The Scope selector next to the name, defaulting to Project.
  • The Project option in the Data Sources dialog is disabled in Restricted Mode instead of failing on save.
0.0.2Compare on GitHub↗

Added

  • Marketplace listing: preview flag, pricing, gallery banner, homepage and issue links.
    The listing on the Marketplace.
    The listing on the Marketplace.
  • Workspace trust support. In Restricted Mode, Project (workspace) data sources are hidden and cannot be created; they appear as soon as the workspace is trusted. Global data sources, consoles, and the grid are unaffected.
  • Virtual workspace support, limited to sources whose files (SQLite databases, SSH keys, CA certificates) live on the local disk.
  • GitHub Actions: CI on every push and pull request, and a tag-driven release that packages, publishes to the Marketplace, and creates the GitHub release with the vsix attached.

Changed

  • The datagrip keyword no longer appears in the manifest.
  • The vsix no longer ships the README screenshots or repository tooling files.
  • README rewritten for the Marketplace, with an Open in VS Code button and live Marketplace badges; the roadmap moved to its own page.

Security

  • The console webview now bundles DOMPurify from the npm package (3.4.14) instead of the older copy vendored inside Monaco, picking up the fixes for GHSA-c2j3-45gr-mqc4, GHSA-cmwh-pvxp-8882, GHSA-vxr8-fq34-vvx9, and GHSA-55q2-fjhq-7xh7.
0.0.1Phase 1Release on GitHub↗

First release: Phase 1 (MVP) of the plan.

Added

  • Connect: PostgreSQL, MySQL/MariaDB, and SQLite data sources with SSH tunnels, SSL modes, and pgpass. An IntelliJ-styled Data Sources dialog in its own floating window, with local validation, auto-derived names, and Test Connection. Global (user) and Project (workspace) scopes. Passwords stored only in the OS keychain. Per-source env colors and server-side read-only mode.
    The Data Sources dialog in its own floating window.
    The Data Sources dialog in its own floating window.
  • Explore: a Database tool window in IntelliJ's design language: vendor marks, env dots, introspection badges, schemas, tables with PK/FK keys, indexes, views, sequences, routines, and enum types. Toolbar row, anchored context menus, lazy introspection, and a data source Information tab in the Tablecloth panel.
    The Database tool window with a PostgreSQL source expanded to its columns.
    The Database tool window with a PostgreSQL source expanded to its columns.
  • Query: Monaco-based consoles under an IntelliJ toolbar. Run statement (⌘⏎) with the statement frame, run script, a schema switcher that really switches (search_path/USE), transaction mode (Auto/Manual) with isolation levels and commit/rollback, query history, and schema-aware completion that quotes identifiers when the dialect needs it. Consoles persist, rename, and reopen from the console dropdown.
    A console bound to acme.public, the statement frame on the caret's statement.
    A console bound to acme.public, the statement frame on the caret's statement.
  • Results: the Tablecloth panel mirrors consoles (Database → source → console) with per-console result tabs and Output logs. A DataGrip-style grid with the two-state floating pager, count on demand, sorting, and row selection. Extractors for SQL Inserts/Updates, Where Clause, and the CSV family. "Run File on Data Source…" for any .sql file.
    The Tablecloth panel: the console tree, a result tab per statement, the grid and the extractor picker.
    The Tablecloth panel: the console tree, a result tab per statement, the grid and the extractor picker.

Known limits

  • The grid is read-only.
  • MySQL DELIMITER blocks are not understood by the statement splitter.
  • SQLite empty results lose their column headers.
  • An isolation level is not reapplied after a silent reconnect.
  • Paste in the console uses the keyboard; Monaco's context-menu Paste is inert inside webviews.