EXPLAIN visualizer

The EXPLAIN visualizer turns the textual query plan into a tree of nodes you can read, compare, and color by metric. Replaces hand-reading EXPLAIN (ANALYZE, BUFFERS) and third-party services.

Run

Action	Shortcut
EXPLAIN (no execution, plan estimate)	`⌘`+`E`
EXPLAIN ANALYZE (executes for real)	`⌘`+`Shift`+`E`

On a production connection, EXPLAIN ANALYZE actually runs the query — meaning DELETE or UPDATE will execute. Use plain EXPLAIN (no ANALYZE) for estimates, or wrap in BEGIN; ... ROLLBACK;.

After running, ide99 opens a Plan tab with the node tree.

The plan tree

Nodes are stacked top to bottom: root (the last operation) on top, leaves (Seq Scan, Index Scan) at the bottom. Each node shows:

Type: Seq Scan, Hash Join, Sort, Aggregate, Limit, etc.
Table name / condition (if applicable)
Duration (Actual Total Time)
A bar indicator proportional to duration (you visually see the "hot" nodes)

Click a node — the side panel on the right opens with all metrics: cost (Startup Cost / Total Cost), rows (Plan Rows / Actual Rows), buffers (Shared Hit / Shared Read), disk read/write times.

Heatmap

Top-right of the diagram — a switch: none · duration · rows · cost. Changes node color and fill:

none — neutral (default).
duration — nodes shaded by Actual Total Time relative to the slowest. Red = your bottleneck.
rows — by Actual Rows × Loops. Shows where Postgres churns through millions of rows.
cost — by Total Cost. Useful for comparing "expensive" estimates.

Most common pattern: switch to duration, find the single red node — that's your problem.

Auto-insights

Above the diagram — a collapsible "Insights" block. If ide99 spots typical issues, they go here:

Seq Scan on a table with > 100K rows when a suitable index is missing
Sort spilling to Disk (work_mem overflowed)
Bitmap Heap Scan with a recheck mismatch (stale stats)
Hash Join with Batches > 1 (hash table doesn't fit in work_mem)
Nested Loop with a large outer relation (often a planner mistake, helped by ANALYZE)
Bad row estimate: Actual Rows / Plan Rows > 100× (hint at stale stats)

Each insight is clickable — jumps to the specific node.

Plan diff (before and after an index)

One of the most useful modes: compare two plans side by side.

Run EXPLAIN ANALYZE on the current query → plan A is saved.
Add an index or rewrite the query.
Run EXPLAIN ANALYZE again → a "Compare to previous" button appears.
A side-by-side opens: plan A on the left, plan B on the right. Nodes that got faster — green; slower — red.

Recent Plans

ide99 stores ~200 recent plans per connection locally. Open via the History icon in the EXPLAIN tab header or Cmd+P in the visualizer.

The list shows: query, duration, run time, node types. Search by query text or table name. Double-click loads the plan; you can diff against the current one.

If your connection has Exclude from recent plans on — that specific connection won't be logged. Useful for prod with PII.

Export

The Export button in the tab header — downloads a JSON of the plan. You can open it in pev2.dev, share on a forum, attach to a bug report.

Tip: what to read first in a plan

Make sure Heatmap = duration. Without it you read the plan blindly.
Find the red node. That's your bottleneck — usually Seq Scan or Sort.
Look at Plan Rows vs Actual Rows. If they differ by 100×+ — stats are stale, run ANALYZE <table>.
Check the Insights panel. Often it already tells you what to do.

Health screen — finds typical issues across the whole database, not just one query.
pg_stat_statements — top queries by time.
Live Ops — what's happening in the database right now.