Hi.  Mostly off topic, but reading about this issue has finally prompted a
response.

On Wed, Aug 15, 2018 at 5:46 PM Paul Rogers <par0...@yahoo.com.invalid>
wrote:
<snip>

> If we provide schema hints ("field x, when it appears, will be a Double"),
> then Drill need not fight with the problem that the data will be Double in
> some files, Nullable Int in others. The schema hint removes the ambiguity.
>
</snip>

How hard would it be to have drill 'fix' the null Int problem by just
noting it is nullable with no other type information assumed until it
encounters that type, and then require that type from then on?  If an
entire file has only null's, then it doesn't define the field at all - only
files that have the type define the field, and then when they are combined
later, IF there are conflicting non-Null fields will the error be thrown?

Thanks,
Chris

Reply via email to