Question: is it possible to perform static analysis on queries to detect errors? (= a linter)

[Lucas-C]

Hi,

I am looking for a tool to perform static query analysis,
and be able to detect errors like XXX must be an aggregate expression or appear in GROUP BY clause
without the need to deploy trino nor have a database with a vali schema.

It does not seem that trino CLI offers a query linter.

External tools like sqlfluff are not able to detect errors like XXX must be an aggregate expression or appear in GROUP BY clause:

$ sqlfluff lint query-with-error.zoql --dialect trino
All Finished 📜 🎉!

I think I found that this specific error is raised by AggregationAnalyzer.java.

Hence my question: is it possible to invoke some Java class/method to perform just query analysis on some sql query string input?

PS: The link to your Slack channel seems dead there: https://trino.io/slack

[nineinchnick]

The Trino parser is not a stable API, so using it for this purpose is not officially supported, but it was raised before:

• Add format command in the CLI #13044
• Add format_trino_query SQL scalar function #19433
• https://github.com/kokosing/trino-query-formatter

Obviously, only formatting a query is not the same as analyzing it, but it can also catch some errors.

[Lucas-C]

Thank you very much for your answer @nineinchnick

I tested https://github.com/kokosing/trino-query-formatter but it does not detect the issue in my query:

$ ./mvnw install -DskipTests=true
$ java -jar target/trino-query-formatter-*-executable.jar < query-with-error.zoql

It seems that it does not call the io.trino.sql.analyzer package.

PR #13044 seemed very interesting.
It looks like it tried to incorporate https://github.com/kokosing/trino-query-formatter into trinodb/trino
Out of curiosity, would a --format argument be a welcome addition to the trino CLI?

I'm a bit sad that both PRs were closed, but they are interesting starting points 👍

[nineinchnick]

The issues/PRs are closed, because the ideas were rejected.