[chrisclark]

Problem

The home page (query list) loaded the entire query log history on every view:

• get_queryset() did prefetch_related("querylog_set"), fetching every log row for every query just to display a run count and the last run.
• recently_viewed() looped over all of the user's logs in Python. Starting that loop evaluates the whole queryset.
• str(q.database_connection) ran one extra query per row.

So the page gets slower as history grows. On a local benchmark (125 queries, 300k logs) it took 9.7 s. demo.sqlexplorer.io, with 1.15M logs, takes ~5.5 s to first byte, while its other pages take ~0.3 s.

Changes

• Run count and last run (time and status) are computed in SQL, using Count plus Subquery annotations. A query that has never run still shows its creation time as a successful run, matching Query.last_run_log().
• Recently viewed is now one aggregate (latest run per query, top N) plus one fetch.
• select_related("created_by_user", "database_connection") replaces the prefetch and removes the per-row query.
• Explicit order_by("title"), because a model's Meta.ordering isn't applied once the query has an aggregate.

Also: N+1 on the query log page

Each row on /logs/ looked up its run_by_user and database_connection separately, which meant 25–35 SQL queries per page. select_related brings that to a constant 5. With this PR, the query list, logs, favorites, connections, query detail and playground pages each run a fixed number of queries as data grows.

Results (same 300k-log benchmark)

• 9.7 s → 0.9 s.
• A constant 7 SQL queries, regardless of how many queries or logs exist.
• The rendered list is identical to before: names, order, last-run dates and status, run counts, and recently viewed. I checked this by diffing the rendered HTML.

Tests

• test_last_run_status: a never-run query falls back to its creation time and success, and a failed last run shows as failed.
• test_recently_viewed_is_latest_run_of_distinct_queries: one entry per query, using the latest run, and only the user's own runs.
• test_query_count_does_not_grow_with_queries_or_logs: this fails on master (9 → 14 SQL queries as rows are added) and passes here.
• test_log_page_query_count_does_not_grow_with_rows: fails on master (17 vs 7) and passes here.
• Full suite: 291 tests pass under settings and settings_base. The list view and log view tests also pass on Postgres 16. ruff==0.16.8 is clean.

🤖 Generated with Claude Code