query_authoring_pglisted

# query_authoring_pg — diagnostic SQL against the app's Postgres The pg kit's query half: `dev_kit` = the sidecar, `test_authoring_pg` = the test infra, this = ad-hoc SQL against the fork's app DB. Use when diagnosing data issues, verifying a migration's effect, or checking an invariant by query. ## Know which DB you're pointed at | DB | Where | What its data proves | |---|---|---| | Sandbox sidecar (`$DATABASE_URL`) | inside your container | your dev/test copy — only what you or the tests put there. Empty/missing rows here prove NOTHING about the FnB's data. | | The FnB's stack DB (dev/prod) | on the host, outside your container | the data actually being diagnosed — reachable only by handoff (below) | Name the DB in every finding. An FnB-reported data issue lives in THEIR DB — never confirm or refute it against your sidecar. Reproduce the shape locally if useful; the verdict query runs on their side. ## psql mechanics SQL is not a shell command — `SELECT …` pasted at a fish/bash prompt dies at the shell. Run it through psql: ```bash psql "$DATABASE_URL" -X -P pager=off -c "SELECT count(*) FROM users;" # one-shot psql "$DATABASE_URL" -X -v ON_ERROR_STOP=1 -f diag.sql # scripted ``` - `-X` skips any psqlrc; `-P pager=off` keeps output capture-friendly. - `\x auto` (in-session) for wide rows. - Schema truth = `\dt` + `\d <table>` — check a column's actual type before writing predicates against it; never guess from habit. ## Parameters — `:name` is