Repository navigation
Conversation
SQLQuack 0.22.1 analyses about a quarter faster, and SQLClean 0.11.1 lints on that same SQLQuack, so the lockfile keeps one copy. The new trees are read as their old meaning: - Adjacent string literals are one Literal::JoinedString, read as the joined string in the Rules query facts and the SQL tree's literals. BigQuery's new raw and adjacent bytes literals read as byte strings. - A Snowflake cast target can be a named type such as NUMBER(10, 2), read through DataType::unnamed as the type it denotes; a named type that does not deserialize is now an error instead of passing as written. - The BigQuery and Databricks type patterns ignore the new spelling field. BigQuery and Databricks now refuse nesting only past 512 levels, and the Databricks parser reads VACUUM, so those dialect tests nest 600 levels and use REPAIR PRIVILEGES as the construct it cannot read. SQLClean's formatter now declines a DuckDB body whose nested block comment holds an apostrophe, leaving it as written; that case keeps its Databricks check. Built-in rules gain guards for joins after a comma FROM item in DuckDB, PostgreSQL and Snowflake, and for adjacent string literals compared with an enum column. BREAKING CHANGE: SQLClean refuses PostgreSQL, BigQuery and Databricks SQL nested more than 512 levels, such as 600 UNION ALL operands, as it already did for Snowflake and DuckDB, so such a resource is now reported as rules-unevaluated unless its Rules are ignored.
SQLQuack 0.22 keeps each comma item's joins inside the item, as a joined table, in PostgreSQL, Databricks, DuckDB and SQL Server. Source drift read only the select's own joins, so a join USING a column the source gains, or an unqualified name in its ON clause, after a comma item no longer marked the model for analysis again. An unaliased joined FROM item now contributes its left side and its joins as the select's own joins do.
SQLClean's lint request takes max_function_call_depth. SQLBuild now sends its shared MAX_FUNCTION_CALL_DEPTH with every body it lints, as it already did when formatting, so the compiler and the Rules refuse the same nesting. The parse-guard check also fails when a source sends a SQLClean lint request or builds SQLClean's FormatRequest without the shared depth. With one limit, audits and SQL tests nested past it are compile errors rather than resources the Rules could not evaluate, so those end-to-end cases now check that the Rules lint them at the limit. A model nested one call past it is still reported unevaluated.
kvlonge
enabled auto-merge
November 20, 2026 03:33
…eclines SQLClean 0.11.1's formatter declines a body whose nested block comment holds an apostrophe, in DuckDB, Databricks and PostgreSQL alike, and leaves it as written with a format fault, so the case no longer formats in any dialect.
github-merge-queue
Bot
removed this pull request from the merge queue due to failed status checks
Nov 20, 2026
SQLQuack 0.22 scopes a join's ON clause to the tables before it, so J05, an ON clause naming a table joined later, is now a B004 compile error rather than a pending case.
github-merge-queue
Bot
removed this pull request from the merge queue due to failed status checks
Nov 20, 2026
github-merge-queue
Bot
removed this pull request from the merge queue due to failed status checks
Nov 20, 2026
Contributor
Author
|
Held for a batch PR: this branch is merged into the local integration branch and will land with the next batch. |
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
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.
Why
SQLQuack 0.22.1 analyses about a quarter faster, and SQLClean 0.11.1 lints on that same SQLQuack, so SQLBuild moves both pins together and keeps one SQLQuack in the build. SQLClean's lint now takes a function-call depth, so the Rules can use SQLBuild's shared 256 instead of SQLClean's own limit.
Changes
MAX_FUNCTION_CALL_DEPTH; the parse-guard check fails when a SQLClean lint or format request leaves it out.Verification
make check-ci.