Add support for ClickHouse dialect

[tconbeer]

Add support for ClickHouse dialect

[tconbeer]

Awesome, thank you for the extra info. That should make this pretty straightforward. We don't have to touch parsing and can still lowercase lots of tokens- we'll just leave identifiers (incl. function names) alone.

CLI flag should be --dialect clickhouse

[tconbeer]

@Shlomixg I'm cutting the release for 0.10.0 right now -- if you upgrade, you can run that version with sqlfmt . --dialect clickhouse and we won't lowercase so many things. Would love to get any more feedback from you!

[tconbeer]

Thanks for the report, @Shlomixg !

I had no idea ClickHouse was case sensitive. (Why would they do that to us?!)

Should be easy enough to not lowercase the output, but...

Are SQL keywords (like select) also case sensitive, or is it just some/all function names?

I designed this with multiple dialects in mind, so it shouldn't be too bad to change the parsing logic to be case-sensitive. I'm just a little sad because we've gotten away with only one "polyglot" dialect so far.

[Shlomixg]

Thanks for your quick response!

I think only the CH unique functions are case-sensitive, meaning that select isn't case-sensitive.
You can find more information in the docs.
Note that aliases are also case-sensitive.

[tconbeer]

Implementation notes (this is bigger than I thought, to get the Dialect or Mode into the Node.capitalize() method):

• Add Dialect as a property of Analyzer
• Create a NodeManager (NodeBuilder?) class with Dialect as a property
• instantiate a NodeManager where needed in actions.py
• move Node.from_token() to NodeManager.create_node_from_token(); also move Node.capitalize() and Node.whitespace()
• update NodeManager.capitalize() to check for the Clickhouse dialect
• update Line.append_newline() so it doesn't use Node.from_token()

[Shlomixg]

Bravo :)
I'll test it soon and leave feedback.
Thanks!