What the Linux app does, checked against main 8758a2b56 on 2026-11-14. 0.1.0-beta.3 is published on GitHub Releases, the beta PPA, the beta COPR project and the beta Flatpak repository (README). beta.2 was tagged but never published: release run 42382067978 failed before publishing, and #4093 fixed the failure for beta.3.
The work left and its order are in the repository's ROADMAP.md, and what 0.1.0 needs in the release criteria.
The app runs on GNOME 50 (ADR 0015) and connects to nine engines, each a crate linked in at compile time (ADR 0001): PostgreSQL, MySQL and MariaDB, SQL Server, ClickHouse, Cassandra and ScyllaDB, SQLite, DuckDB, Redis and MongoDB.
| Capability | Detail |
|---|---|
| Saved connections | Create and edit (the name included), duplicate with the secrets carried to the copy, delete with or without the connection's history |
| Groups, favourites and colours | Groups nest 3 levels deep and show as a tree in the library's order, empty ones too. New Group, New Subgroup, Rename and Delete from the menus; a delete asks first. A connection or a group moves by drag and drop or from its menu (Move to Group, Move Group To, Move Up, Move Down). Favourites and the 5 connections opened last lead the picker; favourites reorder the same way |
| Tags and icons | Several named tags per connection, each with a colour, from the row's menu or the connect dialog, shown after the name and read with the row's description; Manage Tags renames, recolours and deletes them. An icon from a fixed set shows in the row and beside the window title |
| Search | The connection list's search (Ctrl+F, or start typing) matches the name, host, database, user, driver, the names on the group's path and tag names |
| Import and export | A JSON file with no secrets in it, tags and icons included. On import, an entry with the same id and the same settings updates its name, colour, group, tags and icon; any other is added as a new connection. Tags are matched by id, then by name, so a second import adds none |
| Import from another app | Import from Another App in the main menu reads the connections of TablePlus, DBeaver, DataGrip, Beekeeper Studio and DBX found on this computer, or a Navicat .ncx export, and reviews each against the saved ones: Add, Add as Copy, Replace, or Keep Existing and add its queries. Passwords come along where the app keeps them in its own files, or in DataGrip's password file; one in the system keyring is entered at connect. DBeaver and DBX scripts and DataGrip consoles come along as saved queries; TablePlus and Beekeeper Studio keep theirs apart from any connection, so they are not imported |
| Connection URLs | A URL pasted at the top of a new connection fills the form: the engine, host, port, user, database and TLS mode, the authentication database, a DNS seed list and a replica set's hosts, and the password, kept in the Secret Service with Remember Password on. libpq, MySQL, MongoDB, Redis and Valkey, ClickHouse and Cassandra URLs, SQL Server's JDBC form and SQLite and DuckDB file URLs are read (core connection_url) |
| Database links | The desktop file registers the PostgreSQL, MySQL, MariaDB, SQL Server, ClickHouse, Cassandra, ScyllaDB, Redis, Valkey, MongoDB, SQLite and DuckDB URL schemes. A link for a saved connection asks before opening it; any other fills the connect dialog and waits for Connect, and the connection is saved once it connects (man page) |
| Unix sockets | PostgreSQL, MySQL and MariaDB on this computer through the server's socket, for peer, auth_socket and unix_socket logins: a folder for PostgreSQL, the socket file for MySQL, found in the distribution's usual place. The Flatpak sees the sockets under /run/postgresql and /run/mysqld (ADR 0035) |
| Test before saving | Test connects once and reports, then disconnects |
| Passwords | Kept in the Secret Service; a prompt asks when one is missing or refused, with Remember Password |
| Safe Mode | A level per connection, set in the connect dialog, or while connected from the header's Safe Mode button or the session menu, and saved back to the connection. Off asks before DROP, TRUNCATE, and DELETE or UPDATE without WHERE. Confirm Writes also asks before every other write from the editor, a grid save, a structure save, a CSV import, a SQL file import, or a drop, truncate or create from the sidebar, showing the SQL, or the file, and the connection, with Cancel the default. Read-Only refuses every editor statement that does not read, and opens the connection read-only: SQLite, DuckDB and ClickHouse refuse writes at the engine, PostgreSQL and MySQL run in read-only transactions, SQL Server relies on the login, with the app's own reads in a rolled-back transaction, and Cassandra and ScyllaDB rely on the login. A statement is a read only when its engine's grammar shows it reads, each statement of a script on its own. Moving to or from Read-Only reconnects the window |
| Query timeout per connection | Custom Query Timeout in the connect dialog sets the editor's timeout for that connection in place of the one in Preferences, from the next connect; 0 turns it off. The app's own calls keep their fixed limits: 15 seconds for metadata, 60 for a page or a count, 120 for a row save or a CSV import, none for a schema change or an export, and on the server 60 for a table-tab statement and 10 for a lock wait |
| Database files | SQLite and DuckDB: choose a file, create a new one, open one from Files or a terminal, where the file's header picks the engine, find a moved one with Locate, and choose read-only or read-write for one on a network or FUSE filesystem |
| Kerberos for SQL Server | From the ticket cache, in the Flatpak through the KCM socket |
| Health and reconnect | A banner while the connection is lost, reconnect with a backoff from 5 to 60 seconds, and metadata, page and count reads replayed once after it; writes are never replayed |
| Cancel on the server | Stopping a query stops the statement on the server (in process on SQLite); a ClickHouse mutation already running in a save is left to finish. CQL has no stop, so on Cassandra and ScyllaDB the app stops waiting and the server runs the statement to its own timeout |
| Capability | Detail |
|---|---|
| System OpenSSH client | One ControlMaster per connection (ADR 0013); empty fields come from ~/.ssh/config |
| Jump hosts | A ProxyJump chain, passed to ssh -J |
| Authentication | Agent, private key, password and keyboard-interactive; prompts go through the app's askpass helper |
| Host keys | An unknown or changed key is shown with its SHA256 fingerprint before connecting, even where ssh_config says accept-new or no. A changed key can be replaced in known_hosts, except under StrictHostKeyChecking yes, where the user fixes known_hosts outside the app and retries. A revoked key is refused |
| Capability | Detail |
|---|---|
| Modes | Disable, Prefer (PostgreSQL and MySQL), Require and Verify full; a new connection starts on the mode the engine's own client starts on (core ADR 0014) |
| Certificates | CA file, client certificate and key (every network engine but SQL Server), and a server name override (MySQL only through an SSH tunnel) |
| Remedies | A TLS failure says what went wrong. An unknown issuer jumps to the CA file, a name mismatch to the server name, and a server that refuses or requires TLS to the Encryption row |
| Capability | Detail |
|---|---|
| Data grid | A virtualized GtkColumnView with keyboard-focusable cells, pages of 100 to 10,000 rows, and a pager with an exact or estimated row count and the time the page took; a table of about 100,000 rows or fewer is counted on its own, a larger one offers Count Rows. Row numbers stay in view when the columns scroll sideways |
| Filters | Per-column operators, a raw WHERE fragment checked as one expression, match any or all; binary columns compare to 0x hex. Filter by Value and Exclude Value in a cell's menu add a rule. Kept per table, and per tab across a restart or a reopen. On Cassandra and ScyllaDB only the operators the server evaluates, matched all, with ALLOW FILTERING |
| Sorting | A header click sorts tables on the server and query results in the grid, one column at a time. Preferences can sort a table by its primary key or first column when it opens, ascending or descending. A Cassandra or ScyllaDB table keeps the server's order, and pages by its paging state |
| Cell editing | Editors by column kind: inline text, a multi-line editor, a calendar for dates and, in Row Details, a clock for times, a toggle for booleans, JSON checked before saving; values are edited from their exact text, never from what the cell displays. Tables with a primary key, in a table tab or as a query result that reads one table with its whole key (ADR 0036); key, generated and auto-increment columns, and binary, spatial, array and composite values, are read-only |
| Rows | Insert, duplicate and delete; Paste (Ctrl+V) adds tab-separated rows from the clipboard, filling the columns from the focused one or by a header's names, or edits the focused cell with one value, and asks first above 1,000 rows, up to 100,000; Set NULL, Empty, DEFAULT or a server expression on every selected row in one undo step; undo and redo; save or discard pending changes, and stop a save. On ClickHouse, Cassandra and ScyllaDB a save is not atomic, and Stop takes effect between statements |
| Server-computed values | A date, time or timestamp cell takes the engine's own NOW(), CURRENT_TIMESTAMP, CURRENT_DATE and the like, typed in any letter case or picked from the cell menu, and a saved row's cell takes DEFAULT where the engine has SET c = DEFAULT (PostgreSQL, MySQL and MariaDB, SQL Server, DuckDB). Only each engine's own list is offered, which on DuckDB is now() alone, since the bundled build has no ICU extension; the save writes the expression into the statement, so the Confirm Writes review and the history show it as written |
| Bulk delete | Deleting 10 rows or more asks first, unless Confirm Before Deleting Rows is off in Preferences |
| Copy | Copy; Copy As rows, with headers, CSV with or without headers, JSON, JSON Lines, Markdown, a text table, IN clause or one row of SQL values, and INSERT with or without generated keys in an editable table tab; Copy Column Name; Show Row as JSON. Preferences sets what Copy writes for several rows: TSV or CSV with or without headers, JSON or Markdown |
| Foreign keys | Open Referenced Row (Ctrl+J) on a key cell opens the table it points at in a new tab, filtered on every column of the key; Show Referencing Rows on a saved row opens a table whose key points at it. A tab already showing those rows comes to the front |
| Table views | Data, Structure and JSON share a table tab and its bottom bar, and switching keeps each view's unsaved changes. JSON shows the rows on screen; Columns in the bottom bar opens the column chooser |
| Row Details | Each field is edited in its own row, with a button that copies its whole value and a menu that sets it to Empty, NULL or DEFAULT or copies it. Primary keys, foreign keys and unsaved edits are marked and read out to screen readers. Search names or values, or show edited fields only. Preferences can open it whenever a row is selected |
| Grid display | The grid draws in the editor scheme's colours. Preferences set a cell font, row numbers, row height, alternate rows, the text for NULL, and dates as ISO, US or EU, with or without the time. Zero-width, control and special space characters are marked in cells and value fields unless turned off |
| Find in results | Ctrl+F in a query result or a Redis key's values finds the cells whose text holds the search, ignoring case, and Ctrl+G and Shift+Ctrl+G step through them; in a table tab Ctrl+F opens the filter |
| Selection summary | A readout in the bottom bar for the focused column of the selected rows: the count, how many are numbers, empty or not numbers, and the sum, average, minimum and maximum, each with Copy |
| Highlight rules | Highlight Rules in the bottom bar colour the rows or cells that meet a condition, kept per table across restarts, for database files and other open databases too |
| Column layout | Order, hidden columns and widths kept per table; a column added since goes last and a dropped one keeps its place. A header drag, or Move Column Left and Move Column Right in the cell menu (Alt+Left and Alt+Right), reorders them; Widen Column and Narrow Column (Shift+Alt+Right and Shift+Alt+Left) resize them; Reset Columns in the Columns dialog goes back to the table's own order |
| CSV import | Into an existing table, with a column mapping, in one write batch (one transaction where the engine has them), up to 64 MiB |
| Sidebar | Tables (partitioned ones included) by schema, with views, materialized views, foreign tables, sequences, procedures, functions and triggers each in a group above them; empty schemas listed. Search and Refresh sit in the sidebar's header, and the list is read again after the editor runs DDL. Open in New Tab and Show CREATE TABLE; Edit Structure, Import CSV, Truncate Table and Drop Table for tables, Delete All Documents for collections; New and Drop Schema, and New and Drop Database (Keyspace on Cassandra), on every engine that has them, never offered for a system object or the session's own database. Open Database on a connection opens another database of the server beside it. The bottom bar has an Add menu for a new table, schema or database, and Schemas, which picks the schemas listed and is kept per connection. Tables and Favorites switch what the sidebar lists. Dragging its edge, or the arrow keys on it, sets a width every window keeps |
| Routines and triggers | Functions, procedures, sequences and triggers, each trigger with the table it fires on, on every engine that has them. Opening one shows its definition read-only, with Copy and Open in Editor; every overload of a PostgreSQL function name is shown |
| ER diagram | Show ER Diagram in a table's menu or on a schema's heading opens the schema's tables as cards and their foreign keys as lines in crow's foot notation, read two tables at a time with Stop. Pan, zoom, Best Fit, drag tables, Reset Layout, and PNG and SVG export. Each table is a list item the keyboard and a screen reader reach: Tab and the arrow keys move between tables, Enter opens one, Alt+arrows move it, and its description names its columns and the tables it joins |
| Capability | Detail |
|---|---|
| Table definition | Columns, indexes, foreign keys and column comments |
| Index statements | Show CREATE INDEX and Copy CREATE INDEX on each index the table has, as the statement that makes it again: the server's own text on PostgreSQL, MySQL, MariaDB, SQLite and DuckDB, and written from the catalog, storage and options included, on SQL Server, ClickHouse, Cassandra and ScyllaDB. An index a constraint owns, such as a primary key or a UNIQUE constraint, a partition's index attached to its parent's and a ClickHouse sorting key offer none. Show CREATE TABLE uses the same statements |
| Changes | Create, rename and drop tables; add, drop and rename columns, and change their type, nullability, default, comment, primary key and auto-increment; add and drop indexes (column keys, unique, descending) and foreign keys. ClickHouse creates no tables, indexes or foreign keys and changes no primary key or auto-increment; SQL Server changes no auto-increment; SQLite keeps no column comments; Cassandra and ScyllaDB take no default or NOT NULL, and rename only key columns |
| Review | The SQL is previewed, and a save that loses data or can stop halfway is reviewed before it runs |
| SQLite rebuilds | Changes SQLite cannot ALTER rebuild the table in one transaction, keeping its rows, indexes, triggers, the views and triggers that name it and its AUTOINCREMENT counter, with foreign keys checked before commit. SQLite itself checks the views and triggers made again; a trigger that lists a dropped column or inserts by position is refused before Save |
| Capability | Detail |
|---|---|
| Key list | A Redis connection lists the keys of its numbered database with their types, read with SCAN; the search goes to the server, plain text matching anywhere in a key and text with *, ? or [ as a glob. A banner says when the list stops at its 5,000-key limit |
| Key tab | A key opens by type: a string's value, a hash's fields, a list by index, a set's members, a sorted set with scores and a stream's entries, 500 at a time with Load More, under its length and expiry |
| Commands | The editor takes Redis commands, one per line, and shows each reply as redis-cli lays it out; MULTI is tracked as a transaction. A waiting command stops with CLIENT UNBLOCK. Read-Only refuses every command that is not a read before it is sent (redis.md) |
| Capability | Detail |
|---|---|
| Connecting | A MongoDB connection reaches one host directly, a replica set from its hosts and name, a replica set or cluster through its DNS seed list (mongodb+srv://), or one host through SSH, with an optional authentication database |
| Collections | Databases list their collections and views. A collection opens in a table tab: its top-level fields as columns, _id first, nested documents and arrays as JSON. Paging, filters, the raw filter as a query document, sorting, editing a field, inserting and deleting run as mongosh calls with bound values, shown in the history and the Safe Mode review; Set to DEFAULT on a saved document removes the field (mongodb.md) |
| Statements | The editor takes mongosh statements (find, aggregate, the writes, runCommand, use, show) and shows documents one per row. A running statement stops with killSessions. Read-Only refuses every statement that is not a read, $out and $merge included, before it is sent |
| Capability | Detail |
|---|---|
| Editing | GtkSourceView with SQL highlighting, a scheme that follows the light or dark style, and toggle comment |
| Completion | Tables, views, columns, schemas, functions and keywords, with aliases resolved |
| Snippets | SELECT, INSERT, UPDATE, DELETE, a join (none on Cassandra and ScyllaDB), CREATE TABLE and CREATE INDEX per engine, and the user's own in GtkSourceView's format from a folder read again when it changes, a snippet scoped to one engine or to all; offered in completion, with Tab between stops (README) |
| Formatter | Format SQL, leaving a statement as typed when formatting would change its tokens |
| Scripts | Split per dialect (DELIMITER, GO and GO n, routine and trigger bodies) and sent a statement at a time, or a batch at a time on SQL Server; each result set, row count, notice or error is its own entry, and a failure stops the script except on SQL Server. Run Selection, and run the statement at the cursor |
| Find and replace | Ctrl+F and Ctrl+H, with case, whole word and regular expressions |
| Files | Open a .sql file into the editor (Ctrl+O) or drop one on it, and save to it (Ctrl+S, or Save As with Ctrl+Shift+S) |
| Results | Several result sets per run, a limit of 10,000 rows per result, 100,000 rows and 256 MB of values per run, with a banner and Export All, and Go to Error |
| Drafts | Each editor tab's text is kept, and a connection's tabs (table tabs with page and sort, structure and editor tabs) reopen when it opens |
| Multiple cursors | Ctrl+D for the next match, Ctrl+Shift+L for every match, Shift+Alt+Up and Down for the lines above and below, Ctrl+click; typing, deleting, pasting and moves at every cursor, one undo for all of them (README) |
| Vim mode | GtkSourceView's Vim emulation, off by default, with the mode and pending command under the text; :w, :e, :q, :wq and :x act on the tab, and the editor's Ctrl shortcuts work in every mode (README) |
| Capability | Detail |
|---|---|
| Query history | A SQLite database with full-text search, filters by connection, time and status, pins, export as SQL or CSV, and a retention period |
| What is recorded | Editor runs, and the SQL the app runs for grid saves, structure saves and successful imports; a run that can carry a password or connection secret, or whose text does not tokenize, is not recorded, and recording can be turned off |
| Show in Files | For the history database, from the history dialog and Preferences |
| Saved queries | Save Query keeps an editor tab's SQL under a name, for its connection. Favorites in the sidebar lists the saved queries of the connection in front, in folders and subfolders: a click opens one in a new editor tab, and its menu, F2 and Delete rename, delete, move it to a folder or share it. Deleting a folder asks first and keeps its queries. Saved Queries in the session menu lists every connection's, the active one first, with search, Open in New Tab, Replace Current Tab, Copy SQL, Rename and Delete. Saving over a saved query keeps its folder, keyword and place |
| Capability | Detail |
|---|---|
| Formats | CSV with its options, JSON, Markdown, HTML, XML, Excel, and SQL INSERT from a table tab |
| Scope | From a table tab, the current page with its pending edits, or every saved row under the filter and sort. From the editor, the rows shown, or every row by running the statement again on a separate read-only connection without the tab's USE or SET; a statement that writes is refused |
| Running | Off the GTK thread with progress and Stop; closing or quitting asks about a running export |
| Database | Export Database in the session menu writes a whole database or one schema to a .sql or .sql.gz file: tables, views, routines and triggers, and every table's rows as multi-row INSERTs, in the order pg_dump and mysqldump use, so psql, mysql, sqlcmd, sqlite3 and clickhouse-client load it as well as Import SQL File does. Read in process from the connection, not by calling those tools (core ADR 0011); not offered on Redis, MongoDB, Cassandra and ScyllaDB (README) |
| Capability | Detail |
|---|---|
| SQL files | Import SQL File in the session menu runs a .sql or .sql.gz file of any size on an editor session of its own, split as the engine's client splits it (DELIMITER, GO n, routine bodies, dollar quotes), with progress, Stop and a summary of what failed and where. pg_dump's COPY ... FROM stdin rows load through the copy protocol, text or CSV, and its \restrict lines are passed over |
| Errors and transactions | Stop at First Error, or run the rest and list each failure with its line. On PostgreSQL, SQL Server, SQLite and DuckDB the file can run as one transaction, so a failure or Stop undoes it all; MySQL, MariaDB and ClickHouse commit each statement as it runs. Safe Mode Read-Only offers neither import, and the import dialog is the one question Confirm Writes asks for the whole file |
| Capability | Detail |
|---|---|
| Sessions in a window | A window holds several connection sessions, their tabs side by side in one tab view, and a connection already open brings its window forward (ADR 0028) |
| One instance | A second launch presents the running app; New Window (Ctrl+Shift+N) opens a picker |
| Tabs | AdwTabView with the tab overview and split header bars, adapting down to 360×294 |
| Session restore | The windows of saved connections and database files open at quit open again at launch (Restore Session, on by default), also after New Window or a quit during the restore |
| Closing safely | Unsaved edits, a save in flight or of unknown outcome, an open editor transaction, or a running export ask before a window closes or the app quits; logout waits while any of those holds, and suspend while a save runs or an export writes |
| Notifications | Work that finishes in a tab you are not looking at sends a notification |
| Preferences | General: Restore Session, query history (on or off, retention, Clear, its location) and usage data. Data Grid: page size, alternate rows, row numbers, row height, Row Details on selection, invisible characters, date format, default sort, what Copy writes for several rows, NULL text, cell font and the bulk delete confirmation. Editor: font, scheme, Vim keybindings, the snippets folder and the query timeout. AI, MCP and Account |
| Workspace controls | The header has the sidebar toggle, Save Changes for table edits, structure and editable results, Refresh Tab, New Query, the tab overview, the assistant, and a lock that shows the Safe Mode level and changes it. Main Menu holds the app's commands; Session Menu holds the connection in front's: Open Quickly, Query History, Saved Queries, Export Database, Import SQL File, Disconnect and Safe Mode |
| Open Quickly | Ctrl+Shift+O finds tables and views across the connection's schemas, saved queries by name and successful past queries by their SQL; a query opens in a new editor tab |
| Shortcuts and About | A shortcuts dialog, and About from the metainfo with third-party licenses and debug info |
| Capability | Detail |
|---|---|
| Providers | Off until a provider is chosen in Preferences: Anthropic, OpenAI, or Ollama and other servers through the OpenAI-compatible protocol. Keys in the keyring; the model picked from the provider's list (ADR 0022) |
| Chat | A pane beside the tabs (Ctrl+Alt+A) with streamed replies, Stop, Retry and New Conversation; code blocks with Copy and Insert. A conversation carries the editor's SQL and the structure of the tables it names, and reads no rows |
| Editor | Explain, Review and Optimize the selection or the statement at the cursor, Fix with AI on a failed statement, and suggestions while typing (off by default) that Tab accepts |
| AI policy | Per connection: Ask Each Time (the default) asks before the first request in a window, Allow sends without asking, Never sends nothing from the connection |
| Capability | Detail |
|---|---|
| Server | Streamable HTTP on 127.0.0.1 only, off until turned on in Preferences, with a port preference and a bearer token kept in the keyring; it answers only programs running as the same user (ADR 0022) |
| External Clients | A level per connection: Blocked hides it, Read Only (the default) lets clients read, Read & Write lets them write after the user approves each statement in Safe Mode's dialog. Safe Mode Read-Only refuses every client write |
| Tools | 11 tools: list connections, open one in a window, list databases, schemas and tables, describe a table and show its DDL, browse and count rows with filters, run one statement, and export up to 20 tables as CSV, JSON or SQL INSERT in the answer, 100 rows a table unless asked for up to 1,000, within about 4 MB for the whole answer. Runs go into the query history |
| Client setup | Preferences > MCP lists the AI clients TablePro knows, found ones first, and whether each reaches TablePro. Install and Update write TablePro's entry into the client's own configuration and keep the rest, with a backup beside it; Claude Code is set up through claude mcp add, which the Flatpak copies instead of running. Copy gives a client's setup to paste and Remove takes it out; Copy Address, Copy Authorization Header and Regenerate Access Token work with the server on or off. A toast says when a client connects (core ADR 0023) |
| Capability | Detail |
|---|---|
| Opening | tablepro with files or links opens them in the running app, or starts it: a SQL file in an editor tab without running it, a SQLite or DuckDB file as a session, a database link as above |
| Commands | query runs one statement, or one SQL Server batch, from an argument, a file or standard input; connections, databases, schemas and tables list; describe prints columns, indexes and foreign keys, or the DDL; export prints 1 to 20 tables as CSV, JSON or SQL INSERT; status reports the app and its server; completion prints a bash, zsh or fish script, which the .deb and .rpm install. Rows print as CSV, TSV, JSON, JSON Lines or an aligned table, up to 1,000 per result (man page) |
| Access | A command that reads a database goes through the running app's MCP server, with its token from the keyring, and each connection's External Clients level; it opens a closed connection only from a terminal, and any prompt is answered in the app (core ADR 0018) |
| Capability | Detail |
|---|---|
| Sign-in | Preferences > Account signs in to a TablePro account through the browser, as on macOS; the sign-in is kept in the keyring (ADR 0038) |
| Plan, devices and teams | The plan and its status, each paid feature this platform has, included or not by the plan (ADR 0042), the account's devices with Remove, the device limit dialog, teams, and offline files for a computer with no internet |
| License ask | From the 30th day of use without a license, a toast asks whether you would buy one, then at most every 90 days |
| Usage data | On by default, as on macOS: the clients.md events go to PostHog until Preferences > General > Privacy turns it off (ADR 0039) |
| Capability | Detail |
|---|---|
| Share | Share Query and Share Result in the editor's menu, Share Result in a cell's menu and Share on a saved query make a public link through the TablePro account, with a title, note, expiry and Hide SQL, and on Pro a password, no expiry, downloads off and a team to share as. The sheet previews the rows sent (core ADR 0027) |
| Sensitive columns | Each column is kept, masked (Full, Last 4 Characters, Email, Stable Hash or NULL) or removed. A column that looks sensitive by name or by value starts masked, and the masking runs on this computer before anything is sent (core ADR 0028) |
| Opening links | A tablepro://share link opens the shared SQL in a new editor tab of the connection in front and never runs it. Manage Shared Links in the main menu opens the account's list in the browser |
| Capability | Detail |
|---|---|
| Flatpak | The Devel manifest builds on GNOME 51 for x86_64 and aarch64 with every Meson suite in the sandbox, and is linted, in the nightly Linux run; the release manifest builds the Flatpak a release signs into flatpak.tablepro.app and bundles for GitHub, offline from the release tarballs; CI tests it, signs it into test repositories, then installs, starts, updates and switches branches on Ubuntu 22.04, 24.04 and 26.04 (ADR 0041) |
| .deb and .rpm | Ubuntu 26.04 and 26.10 and Fedora 44 and 45, built offline in CI the way Launchpad and COPR build them, linted, then installed and started in fresh containers; a linux-v* tag publishes them on GitHub Releases, uploads to the PPA and starts the COPR build, a beta tag to the beta ones (ADR 0018, ADR 0041) |
| Desktop integration | Desktop file with Open With for SQLite and SQL files, the database URL schemes and tablepro://, D-Bus activation, a man page, metainfo with screenshots, and SVG icons |
| Translations | Vietnamese ships. The app's own strings go through gettext, and CI fails when POTFILES.in is stale or a translation loses a placeholder. CI builds the template with gettext 1.0 and publishes it from main and with each release; translators send pull requests against po/*.po (po/README.md) |
| Logs | stderr, which the session journals, with JSON lines on request. A panic also writes a crash report under $XDG_STATE_HOME, offered with Show Report on the next launch |
| Tests | Unit, gtk (with an accessibility audit of the picker, the connect pages, the workspace, the connect dialog and every preferences page), data and packaging suites; connection, DDL and TLS conformance, a CSV and SQL INSERT export round trip, a copy and paste round trip, and a database dump loaded back through Import SQL File and dumped again to the same file, on every SQL engine, keyspace and TLS conformance on Redis, and document and TLS conformance on MongoDB; Docker jobs for PostgreSQL with PgBouncer, MySQL and MariaDB, SQL Server, ClickHouse, Cassandra and ScyllaDB, Redis, MongoDB, SSH, the session and the grid save (testing.md); every critical audit finding has named guards (regression-tests.md) |