The R2a function-library audit (Epic 19) needs one machine-generated, drift-proof inventory of every construct the elasticsql SQL dialect accepts - each row anchored to real source, each example proven parseable by the real parser, each ES translation derived from the AST - so the jOOQ-render estimate rests on a counted surface instead of a guess.
This lands three test-tree sources under sql/src/test/scala/app/softnetwork/elastic/sql/census/:
CensusModel.scala - the row model (CensusEntry, Kind, EsConstruct, Standard with shape-checked Rule-R evidence gates) and emitters (RFC-4180 CSV, provenance-stamped JSON object, GFM markdown).DialectCensus.scala - 259 hand-authored SYNTAX-FORM rows: 171 function rows covering all 95 SQLKeywords.functionTokens and all 149 accepted spellings (54 aliases, 22 of them computed at runtime by TimeField.words), 29 operators, 36 clauses, 6 literals, 17 units. Every ansi_adjacent citation was fetched from PostgreSQL 16 / MySQL 8.4 / DuckDB docs.DialectCensusSpec.scala - the generator/guard: wrap-tolerant anchors resolved to file:line at emit time (exactly-one-hit), every example parsed through Parser, ES constructs derived from the parsed AST with an es_construct_source column, two-way registry coverage diffs, the 95/95/149 pins, a phantom-help corpus walk (13 phantoms incl. the 3 wired json/ ones), and the emit of sql/target/epic-19/dialect-census.{csv,json,md}. Nothing is assume-guarded; runs on every sbt sql/test.
No production or published code changes (test artifacts are not published). Notable findings recorded in the census itself: || (Pipe) and bare E are DEAD tokens (declared, never consumed by any production); bare CROSS JOIN is grammar-accepted but rejected by StandardJoin.validate; LOG is the NATURAL log (PostgreSQL/DuckDB log is base-10); STRPOS reverses PostgreSQL/DuckDB argument order; DAYOFWEEK numbering differs across all three engines.
Green on Scala 2.12.20 + 2.13.16 (548/548 sql tests per leg); SQLKeywordsSpec untouched and green.
The R2a function-library audit (Epic 19) needs one machine-generated, drift-proof inventory of every construct the elasticsql SQL dialect accepts - each row anchored to real source, each example proven parseable by the real parser, each ES translation derived from the AST - so the jOOQ-render estimate rests on a counted surface instead of a guess.
This lands three test-tree sources under
sql/src/test/scala/app/softnetwork/elastic/sql/census/:CensusModel.scala- the row model (CensusEntry,Kind,EsConstruct,Standardwith shape-checked Rule-R evidence gates) and emitters (RFC-4180 CSV, provenance-stamped JSON object, GFM markdown).DialectCensus.scala- 259 hand-authored SYNTAX-FORM rows: 171 function rows covering all 95SQLKeywords.functionTokensand all 149 accepted spellings (54 aliases, 22 of them computed at runtime byTimeField.words), 29 operators, 36 clauses, 6 literals, 17 units. Everyansi_adjacentcitation was fetched from PostgreSQL 16 / MySQL 8.4 / DuckDB docs.DialectCensusSpec.scala- the generator/guard: wrap-tolerant anchors resolved tofile:lineat emit time (exactly-one-hit), every example parsed throughParser, ES constructs derived from the parsed AST with anes_construct_sourcecolumn, two-way registry coverage diffs, the 95/95/149 pins, a phantom-help corpus walk (13 phantoms incl. the 3 wiredjson/ones), and the emit ofsql/target/epic-19/dialect-census.{csv,json,md}. Nothing isassume-guarded; runs on everysbt sql/test.No production or published code changes (test artifacts are not published). Notable findings recorded in the census itself:
||(Pipe) and bareEare DEAD tokens (declared, never consumed by any production); bare CROSS JOIN is grammar-accepted but rejected byStandardJoin.validate; LOG is the NATURAL log (PostgreSQL/DuckDBlogis base-10); STRPOS reverses PostgreSQL/DuckDB argument order; DAYOFWEEK numbering differs across all three engines.Green on Scala 2.12.20 + 2.13.16 (548/548 sql tests per leg);
SQLKeywordsSpecuntouched and green.