# Per-file escapes for sqlfluff, one entry per line, each with its reason.
# The org's sqlfluff config is delivered by the belt (lint:sql passes it with
# --config); this file is the repo-shaped half, on the same footing as
# deny.toml's skips and _typos.toml's vocabulary. Nothing goes in here to
# silence a finding that is simply inconvenient. Recorded upward as
# monumental-archive/.github#708.

# pgrx GENERATES this file, and .github/scripts/schema-snapshot.sh exists to
# diff freshly-generated SQL against it. sqlfluff wants LT01/RF06/LT05/CP05
# changes — quoted identifiers unquoted, whitespace moved, types upper-cased
# — and applying any of them means the snapshot is no longer what pgrx emits,
# so every later snapshot run diffs dirty forever. Same reasoning the canon
# already wrote into .editorconfig for the source-attest git fixture: a
# fixture reformatted away from what the tool emits stops being evidence.
crates/edtf-postgres/schema.snapshot.sql

# sqlfluff's postgres dialect cannot parse a namespaced GUC assignment.
# Reproduced on the single line, alone:
#   $ printf 'SET plpgsql.check_asserts = on;\n' > s.sql
#   $ sqlfluff lint --config <belt>/sqlfluff.cfg s.sql
#   L: 1 | P: 1 | PRS | Found unparsable section: 'SET ...
# One unparsable statement poisons the whole file, so the corpus's ASSERT
# lines go unlinted either way — the check is not merely wrong here, it is
# inert. The statement is deliberate: the corpus turns plpgsql assertions on
# so a run cannot silently pass because the default moved.
crates/edtf-postgres/tests/corpus.sql
