Skip to content

DataTray 0.7.0

Choose a tag to compare

@github-actions github-actions released this 29 Jul 23:25
bea7201

Added

  • DuckDB is a supported engine. A store-only provider plugin — install it from Tools › Plugin Store; it
    is not bundled with the app. It opens a .duckdb file, or an in-memory database (under Advanced) when you
    only want a scratchpad, and puts its schemas, tables, views, sequences, columns and indexes in the tree,
    with row counts and column counts on the nodes. Queries, scripts, EXPLAIN, View Definition, and DDL
    Create for schemas and tables work as they do for the other SQL engines. The download is ~82 MB: DuckDB is
    an embedded engine, so the plugin carries the engine itself for every platform DataTray runs on.
  • A query can read a file directly, which is the reason to reach for DuckDB in the first place:
    SELECT * FROM 'events.parquet', read_csv_auto('data.csv') and read_json_auto(…) all just work in a
    query tab, against an in-memory connection with no database file at all. Such a result set is read-only —
    there is no table behind it to write to.
  • Editing works for a plain browse. DuckDB's driver does not tell us which table a result set came from,
    so the provider works it out from the query itself: double-click a table, filter it, sort it or page
    through it and the grid stays editable, provided the table has a primary key and the query still selects
    it. Anything more — a join, an aggregate, a view, a file — comes back read-only rather than risking a save
    against the wrong rows.
  • Cancelling a long query genuinely stops it. DuckDB is an analytics engine, so a query that runs for a
    minute is normal rather than a mistake, and its driver ignores the usual cancellation signal; the provider
    bridges that so the Cancel button does what it says.
  • ClickHouse is a supported engine. A store-only provider plugin — install it from Tools › Plugin
    Store
    ; it is not bundled with the app. It connects over ClickHouse's HTTP interface
    (port 8123, or 8443 with the protocol set to https for ClickHouse Cloud) and puts its databases, tables,
    views, columns and data-skipping indexes in the schema tree, with on-disk sizes and row counts on the
    nodes. Queries, scripts, EXPLAIN, the database switcher, View Definition, DDL Create for databases and
    tables, the Activity Monitor and user/role management all work as they do for the other SQL engines. Spin
    a local server up from the Local Containers plugin like any other engine — the provider ships its own
    container recipe.
  • Three things follow from ClickHouse being columnar rather than row-oriented, and are worth knowing rather
    than discovering: result grids are read-only (the HTTP protocol carries no primary-key metadata to
    trace a result back to one table, and ClickHouse edits rows through asynchronous ALTER TABLE … UPDATE
    mutations, which a generated UPDATE … WHERE key = … would not express); a multi-statement script is sent
    one statement per request, because the server rejects a batched body outright; and a batch is not
    transactional
    , since ClickHouse has no transaction for ordinary MergeTree work — a failure leaves the
    statements before it applied.
  • Generated CREATE TABLE DDL declares ENGINE = MergeTree with an ORDER BY over the columns marked as
    primary key (tuple() when none are), and expresses a nullable column as Nullable(T) — ClickHouse
    columns are NOT NULL unless the type says otherwise, the opposite of every other engine here.

Changed

  • Containers DataTray starts are now labelled kontena.source=datatray, not sqlexplorer. The
    Kontena desktop app shows them under the new name; containers created before this release keep the
    old label and are still recognised. Anything else filtering on the old value needs updating.

These builds are unsigned / not notarized.
macOS: after moving to /Applications, clear quarantine: xattr -cr "/Applications/DataTray.app".
Windows: pick the -setup.exe (per-user, no admin) or the .zip; SmartScreen may warn on first launch.