Quote identifiers a database will not read back as itself - #25
Merged
Conversation
A mixed-case Postgres schema failed every query: an unquoted name folds to lower case and resolves to nothing, Oracle folds the other way, and MySQL on Linux compares table names case-sensitively. Table and column names are now quoted from the catalog before the statement is validated, and a name that is already correct is left alone. Reserved words come from each database rather than one shared guess, read through pg_get_keywords(), information_schema.KEYWORDS, V$RESERVED_WORDS and duckdb_keywords(). MySQL reserves 262 words where the guess had about a hundred. tools/generate-sql-keywords.mjs regenerates both this and the plugin's copy, and refuses to write if an engine is unreachable. Rewriting SQL text can change what a query means, so the rewriter leaves alone anything that is syntax rather than a name: string literals including escaped apostrophes and dollar-quoted bodies, comments, CAST types, EXTRACT fields, typed literals such as TIMESTAMP '2024-01-01', schema qualifiers, and a name spelled two ways across the catalog. Also: nested aggregates are repaired before they run, a set operation no longer trips the unknown-column floor, the prompt names the connected database so system-catalog filters are real, and an unescaped apostrophe is explained instead of looping until the attempts run out.
The plugin carries its own Kotlin engine, so the same rules have to exist twice and stay in step. The tests mirror the core suite case for case, which is what catches a divergence: the nested-aggregate branch here never incremented the repair counter, and the column floor read raw SQL where core reads text with literals blanked, so a value containing the word union disabled it.
…st account The screenshot was in no README or doc, so its stale disclaimer was never shown. The certification note answers the Edge policy that asks for test credentials: AskSQL has no accounts to issue, so it explains that and gives a reviewer a free path in with a local model and any spreadsheet.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
A mixed-case Postgres schema failed every query. An unquoted name folds to lower case and resolves to nothing; Oracle folds the other way; MySQL on Linux compares table names case-sensitively. Table and column names are now quoted from the catalog before the statement is validated, and names that are already correct are left untouched.
Reserved words now come from each database rather than one shared guess, read through
pg_get_keywords(),information_schema.KEYWORDS,V\$RESERVED_WORDSandduckdb_keywords(). MySQL reserves 262 words where the shared list had about a hundred.tools/generate-sql-keywords.mjsregenerates both the core and plugin copies and refuses to write if an engine is unreachable.Rewriting SQL text can change what a query means, and the AST guard only checks safety, so the rewriter leaves alone anything that is syntax rather than a name: string literals including escaped apostrophes and dollar-quoted bodies, comments, CAST types, EXTRACT fields, typed literals such as
TIMESTAMP '2024-01-01', schema qualifiers, and any name spelled two ways across the catalog.Also included: nested aggregates repaired before they run, set operations no longer tripping the unknown-column floor, the prompt naming the connected database so system-catalog filters are real, and an unescaped apostrophe explained rather than looping until the attempts run out.
Verification
Ships JetBrains 0.5.2 and VS Code 0.7.1, plus a core patch.