The underlying agate library supports a force option where the types can be explicitly set for the columns. I'm hoping we can expose that capability in csvsql. Ideally, the types can be provided using the SQL that was created by csvsql when inference is turned on. This SQL can then be manually manipulated for the columns where the inference is wrong and fed back into csvsql for subsequent runs.
The underlying agate library supports a
force
option where the types can be explicitly set for the columns. I'm hoping we can expose that capability in csvsql. Ideally, the types can be provided using the SQL that was created by csvsql when inference is turned on. This SQL can then be manually manipulated for the columns where the inference is wrong and fed back into csvsql for subsequent runs.