bd389d2f0d
* Part 1 - Parse `not` and `not-join` * Part 2 - Validate `not` and `not-join` pre-algebrization * Address review comments rnewman. * Remove `WhereNotClause` and populate `NotJoin` with `WhereClause`. * Fix validation for `not` and `not-join`, removing tests that were invalid. * Address rustification comments. * Rebase against `rust` branch. * Part 3 - Add required types for NotJoin. * Implement `PartialEq` for `ConjoiningClauses` so `ComputedTable` can be included inside `ColumnConstraint::NotExists` * Part 4 - Implement `apply_not_join` * Part 5 - Call `apply_not_join` from inside `apply_clause` * Part 6 - Translate `not-join` into `NOT EXISTS` SQL * Address review comments. * Rename `projected` to `unified` to better describe the fact that we are not projecting any variables. * Check for presence of each unified var in either `column_bindings` or `input_bindings` and bail if not there. * Copy over `input_bindings` for each var in `unified`. * Only copy over the first `column_binding` for each variable in `unified` rather than the whole list. * Update tests. * Address review comments. * Make output from Debug for NotExists more useful * Clear up misunderstanding. Any single failing clause in the not will cause the entire not to be considered empty * Address review comments. * Remove Limit requirement from cc_to_exists. * Use Entry.or_insert instead of matching on the entry to add to column_bindings. * Move addition of value_bindings to before apply_clauses on template. * Tidy up tests with some variable reuse. * Addressed nits, * Address review comments. * Move addition of column_bindings to above apply_clause. * Update tests. * Add test to ensure that unbound vars fail * Improve test for unbound variable to check for correct variable and error * address nits |
||
|---|---|---|
| .. | ||
| src | ||
| tests | ||
| Cargo.toml | ||
| README.md | ||
This crate turns an algebrized Datalog query into a domain-specific representation of a SQL query, and then uses mentat_sql to turn that into a SQL string to be executed.
This subsumes both planning and query construction, because in Mentat the SQL query is effectively a query plan.