Best way to add custom validation and warning messages

[evyasonov]

I want to use sqlglot as a validating tool. This is if a user writes a query

• which scans a huge table (tables, sizes and partitioned are predefined) and don't add partition field to it
• which uses has a depth of more than 3 subqueries
• which has "wrong" identifiers
  then the user should see a warning-comment after formatting the query.

What is the best way to implement this kind of custom logic?

[tobymao]

you can use the find, find_all, walk apis on expression. you can walk / search the AST to find these issues.

the first one is the trickiest because engines can do predicate push down which a simple AST search wouldn't find. for that case, you can first run the sqlglot optimizer to do predicate pushdown, before going back and looking for your custom rules.

[evyasonov]

okay, and what is the best way to to it technically?

I mean to create a function which will

• parse input sql to ast
• call optimize
• check each custom rule
• add comment to ast
• call generate to return output sql

Or to add this steps into a dialect / expand existing class?