macbre / sql-metadata / 25284550376

100%

push

github

Version 3.0.0 Proposal draft with sqlglot as the parsing library (#617)

* wip - working version with sqlglot to refactor

* wip - extract bracketed names into new dialect

* wip - Rewrite REPLACE INTO → INSERT INTO in _ast.py._parse() before sqlglot parses it, so sqlglot produces a proper exp.Insert AST instead of exp.Command and parses it correctly without falling back to regex

* wip - Rewrite REPLACE INTO → INSERT INTO in _ast.py._parse() before sqlglot parses it, so sqlglot produces a proper exp.Insert AST instead of exp.Command and parses it correctly without falling back to regex

* add docstings, refactor to simplify most complex methods, add few tests from open issues to verify if it's handling the issues better than the old version, remove internal tokens and produce only list of strings if needed, remove compatibility layer to v1

* add tests from open issues that now passes and some small fixes to accommodate additional 3 tests

* accept capitalization and explicit as from sqlglot as opinionated defaults to simplify the bodies extraction

* simplify logic, refactor into classes with related functionalities

* additional simplification and cleanup

* remove unnecessary wrappers

* further simplification - add also architecture overview with charts and main notes

* next portion of cleanup, renaming files, update also agents.md file, switch to ruff for formating and linting

* refactor other functionalities from ast parser into separate classes

* change to ruff also in CI. add mypy and fix typing errors, add mypy to CI

* fix remaining mypy errors in untyped code

* further fixes and duplication cleanup

* fix unused code, bump coverage - add todo to revisit corner cases later for now mark nocover as unreachable from parser and this is the only entrypoint we want for majority of the tests

* add features to handle unnamed queries, extracting properly hive tables from queries with subscripts, some additional issues that were already fixed we... (continued)

1153 of 1153 new or added lines in 14 files covered. (100.0%)

1264 of 1264 relevant lines covered (100.0%)

1.0 hits per line