buildAiContext previously loaded table metadata sequentially in a
for-loop (3 API calls per table × up to 50 tables = 150 serial
round-trips). Now uses Promise.all to fetch all tables' columns,
indexes, and foreign keys concurrently, reducing total wait time
from O(n × latency) to O(latency).
Add a new "Export Database" feature accessible from the sidebar
right-click menu on database/schema nodes. Generates a .sql file
containing CREATE TABLE DDL, batched INSERT statements, and
view/procedure/function source code.
Backend (database_export.rs):
- Stream-writes SQL to file with configurable batch size
- Supports all SQL database types via existing DDL/INSERT generators
- Progress events with cancellation support
- Graceful error handling (continues on per-table failures)
Frontend (DatabaseExportDialog.vue):
- Connection/database/schema selector
- Checkboxes: structure, data, objects
- Real-time progress with current table and row count
- Cancel button
Wired up through Tauri commands, Web SSE routes, and API layer.
Schema Compare:
- Add checkbox per diff item for selective sync (with select-all)
- SQL preview now reactively updates based on checked items
- Add swap button to exchange source↔target
- Execute sync now runs statement-by-statement with progress indicator
- Show detailed error report for failed statements
Data Compare:
- Add swap button to exchange source↔target
- Execute sync now runs statement-by-statement with progress indicator
- Show error summary and per-statement failures
When multiple rows are selected, copyRowAsInsert now generates an
INSERT statement for each selected row instead of only the
right-clicked row.
Closes#166
The previous fix (8032c63) only persisted whereInput when the user
triggered reload/paginate/sort. Typing a condition and switching tabs
without executing still lost it because the local ref was destroyed
on component unmount.
Emit update:whereInput on every change so ContentArea writes the
value to activeTab.whereInput immediately.
Fixes#232
test_connection passed the original config to the JDBC plugin without
rewriting the connection URL to use the tunneled localhost address.
Native drivers (PG, MySQL, etc.) used the already-substituted host/port,
but the JDBC path bypassed this, so the plugin tried to connect directly
to the remote host — failing when only reachable through the tunnel.
Apply the same rewrite_jdbc_url_host logic used in get_or_create_pool.
Fixes#230
Exact and prefix matches now rank above substring and fuzzy matches
so searching for e.g. "core_flow" puts the exact table at the top
instead of burying it behind partial hits.
Add a shadcn-vue Tooltip to tree nodes that only activates when the
label text is truncated. Enable horizontal scrolling on the sidebar
tree container for deeply nested long names.
Expose Schema Diff and Data Compare dialogs from the top toolbar
so they can be opened without pre-selecting a connection in the
sidebar. Connections are chosen inside the dialog.
- Highlight the row containing the selected cell with a subtle
primary-color tint so the active row is easy to spot.
- Add first-page (« ) and last-page (») pagination buttons.
Last-page runs a COUNT(*) query to calculate the final page offset.
Add « and » buttons to the pagination bar. First-page jumps to page 1.
Last-page runs a COUNT(*) query to calculate total pages, then jumps
to the final page.
The save validator rejected NULL in NOT NULL columns even when the
column has a DEFAULT (e.g. serial/sequence). Skip those columns so
cloned rows with auto-increment PKs pass frontend validation.
When cloning or adding a row, null columns (e.g. auto-increment PKs)
were explicitly inserted as NULL, violating NOT NULL constraints.
Omit them so the database DEFAULT (sequence, now(), etc.) kicks in.
t8y2
Illuminated2020