feat: Configurable object/column name formatting options for targets

[dluo-sig]

Feature scope

Targets (data type handling, batching, SQL object generation, tests, etc.)

Description

It would be helpful if there is a way to configure the column and object names that get mapped from the extractor.

1. direct mapping - names as-is
2. snake_case
3. PascalCase

Nice to have: some targets also have the ability to specify column names that are outside of the normal specs. For example, SQL Server can have column names with any character when qualified with square brackets, PostgreSQL/Snowflake can do the same with double quotes, etc. This would be an additional option to use the appropriate identifiers across the board when referencing objects/fields at the target. It would not be desirable to always force this option, as for example, Snowflake becomes case-sensitive when double quotes are provided.

[visch]

Yes, good part is there's already a function for this

sdk/singer_sdk/sinks/sql.py

Line 134 in 6334091

def conform_name( # noqa: PLR6301

I overrode this as I wanted "direct mapping" in your list to be the default not snake_case see https://github.com/MeltanoLabs/target-postgres/blob/1e59be2750961876d52b8e69cf05c5eb06cb13b4/target_postgres/sinks.py#L280-L282

There was a longer discussion about this here #1205

I think the idea of a config option to choose between them is a good one!

[dluo-sig]

@edgarrmondragon had proposed using the humps library for this.

[dluo-sig]

Related request #2545 (comment)

[lumenn]

I'm currently looking for same use case - i'd like to convert all column names to lower case (source is tap-mssql, and loader is target-postgres). In my case tables have around 100 columns, so making it manually per column doesn't seem inviting :)

[dluo-sig]

I think what could happen is that with this new configuration set, PluginMapper can take it into account in register_raw_stream_schema and insert key pairs into self.stream_maps_dict with the new mapping and a __NULL__ for the old column. We then need to handle __key_properties__ to take into account key columns that were renamed.