Feature Request: Support for Duckdb

[PMassicotte]

duckdb is getting a lot of attention as a fast in-process analytical database. I found out that duckdb is not currently supported. I am currently using spark option, but it is not working every time.

Are there any plans to add it?

[nene]

Thanks for the feature request.

I'm open for a pull request, but I'm not really interested in implementing it by myself. Luckily for you, adding a support for a new dialect to sql-formatter is not that hard. You don't even need to fork sql-formatter to do so, you can just provide it a custom dialect definition.

Well, admittedly there's not that much of documentation there, but just copying an existing dialect implementation from sql-formatter source code and modifying it slightly should be pretty simple.

You can also look at the wiki, which documents syntax details of all the currently supported SQL dialects.

[PMassicotte]

Great, I will have a look at it!

[PMassicotte]

Just for reference:

select * from duckdb_keywords();
select * from duckdb_functions();

[hughcameron]

I've set up a PR for this.

It's based on the duckdb words:

COPY (
    SELECT DISTINCT upper(function_name) AS function_name
    FROM duckdb_functions()
    WHERE function_name SIMILAR TO '^[a-z].*'
    ORDER BY function_name
) TO 'duckdb_functions.txt' WITH (sep ',', header FALSE);

COPY (
    SELECT upper(keyword_name)
    FROM duckdb_keywords()
    ORDER BY keyword_name
) TO 'duckdb_keywords.txt' WITH (sep ',', header FALSE);

COPY (
    SELECT DISTINCT upper(logical_type)
    FROM duckdb_types()
    ORDER BY logical_type
) TO 'duckdb_types.txt' WITH (sep ',', header FALSE);

I've also leveraged the postgres tests, most are passing. I'm not clear on how to resolve the failing tests - it would be great to get some help.