[DISCUSSION] Should `CREATE EXTERNAL TABLE's create a new file if it doesn't exist?

[berkaysynnada]

CREATE EXTERNAL TABLE t STORED AS CSV LOCATION '~/files/foo.csv'; gives an error if the given file does not exist. When we are trying to create a new file and then insert some values into it, we are forced to create the file externally.

I wonder how it sounds if create external table's create the file indeed if it doesn't exist. If it seems like a useful behavior, then there will be a few more questions like

• at what instance exactly the file is generated (during table provider generation, or during scan, or during the next select or insert into query?)
• how are the stores other than local file system treated?

[alamb]

I think there is a usecase for DataFusion remaining "read only" and so some users will prefer to get errors if the table doesn't exist

Perhaps we can create a mode / config option, syntax to support "create file of not exists"

CREATE EXTERNAL TABLE t STORED 
AS CSV LOCATION '~/files/foo.csv'
CREATE FILE IF NOT EXISTS -- <-- Proposed Syntax

I don't like that particular syntax (hopefully someone could come up with something better) but I think it illustrates the idea