🦊 ferox v0.3.1

Released: 2026-10-24

✨ What's New

▶ Run Selection

Select part of the editor and press Ctrl+Enter / F5 (or click Run / Explain) — only the selected SQL runs. While something is selected, the Run button reads ▶ Run Selection, so it's always clear what will execute. With no selection, the whole editor runs exactly as before.

✦ AI: Explain, Optimize & Fix Error

The AI button has a new ✦ ▾ menu next to it (works on the selection, if any):

• Explain query — a plain-language walkthrough of what the query returns, how tables are joined and filtered, and pitfalls worth knowing (NULL handling, join fan-out, missing LIMIT, unindexed filters)
• Optimize query — a rewrite that returns the same result faster, using your live schema and its indexes; if an index would help more than a rewrite, the suggested CREATE INDEX is in the explanation
• Fix error with AI — enabled after a query fails (also offered as a button in the Messages tab); sends the query plus the Postgres error and returns a minimal fix with the cause explained

AI-generated SQL is never executed automatically. It opens in a review window with Replace in editor / Copy SQL, and if the suggestion contains a write operation the original didn't have (DELETE, DROP, UPDATE, …) a red warning names it. Replacing the editor contents saves your original query to history first — press ↑ to get it back. Works with every configured provider (Claude, OpenAI, Groq, Ollama, custom).

🛠 Table Structure Editor

Right-click a table → Edit Table is now a full structure editor:

• Columns — rename, change type (with an automatic USING col::newtype cast), set/drop NOT NULL, set/drop DEFAULT, drop, add
• Indexes — drop existing ones (PK/UNIQUE-backed indexes correctly become DROP CONSTRAINT), add new ones with column picker, UNIQUE, and method (btree, hash, gin, gist, brin)
• Rename the table
• Live DDL preview — the exact SQL is always shown, syntax-highlighted, before anything runs; risky changes (dropped columns, type rewrites, SET NOT NULL, dropped indexes) are listed in red
• All-or-nothing — the whole change set runs as a single transaction, so a failing step leaves the table untouched
• Open in Editor puts the DDL into a new query tab instead of running it
• If the database rejects the change, the dialog stays open with the error — your edits aren't lost

💾 Schema Dump (no pg_dump needed)

Right-click a schema → Dump schema (structure + data)… or (structure only)… writes a plain .sql file you can run anywhere — including through ferox's own SSH tunnel, with no local pg_dump / version matching required.

• Covers enum & domain types, sequences (with current values), functions & procedures, tables (identity, generated & collated columns, partitions), CHECK / PK / UNIQUE / EXCLUDE / foreign keys, indexes, views, materialized views, and triggers — in restore-safe order (keys and triggers after the data)
• One consistent snapshot (REPEATABLE READ READ ONLY); rows stream through a server-side cursor straight to disk, so table size doesn't affect memory
• Values are written exactly as the server renders them — no type round-trip
• A failed or cancelled dump never leaves a half-written file behind
• Not included: owners, privileges, RLS policies, comments (like pg_dump --no-owner --no-privileges). Requires PostgreSQL 12+

📥 Export as INSERT Statements

New INSERT button next to CSV / JSON, plus Export Selected as INSERT… in the row right-click menu. Generates batched INSERT INTO … VALUES statements (100 rows each). The target table is taken from the browsed table, or inferred from a single-table query.

🦫 Import Connections from DBeaver

Connection → Import from DBeaver… finds your DBeaver workspace automatically (Windows, macOS, Linux, Flatpak, Snap) — or pick a data-sources.json by hand.

• Imports PostgreSQL connections with passwords, SSH tunnel, SSL mode, folders (→ groups) and DBeaver's prod/test connection type (→ color tag)
• Preview list with checkboxes; connections you already have are detected and unchecked
• Imported passwords go straight into the OS keychain

⬆ Update Check

ferox now tells you when a new release is out. One request to the GitHub releases API at startup, on a background thread (the UI never waits on the network). A banner offers Release notes / download, Skip this version, or dismiss. Nothing is downloaded or installed automatically.

• Settings → Check for updates on startup turns it off
• Settings → Check for updates now runs a check on demand

🐛 Fixes

⚡ Performance

• Schema dump streams rows through a cursor in 1,000-row batches — memory use doesn't scale with table size, and the 50k-row result cap doesn't apply
• The startup update check's HTTP/TLS stack is released again once the check finishes (an extra working-set trim), so idle RAM stays where 0.3.0 was
• Binary size: ~9.9 MB → ~10.1 MB (release, LTO). DBeaver password decryption reuses the aes / cbc crates already linked through the SSH stack — no new heavy dependencies

🧪 Testing

• 14 new unit tests — INSERT generation & quoting, single-table detection, DBeaver parsing & credential decryption, ALTER statement ordering, AI response parsing & write-operation detection, version comparison — run on every cargo test
• 3 new #[ignore]-gated integration tests (cargo test -- --ignored):
  • Dump round-trip — dumps a schema with every supported object kind, restores it under a new name on a live Postgres, and compares data, sequence positions, constraints, indexes and triggers
  • Structure editor — applies a generated ALTER batch on a live Postgres and verifies a failing batch leaves nothing half-applied
  • Update check — hits the real GitHub releases API

📦 Build Info