Dan Hecht has posted comments on this change.

Change subject: IMPALA-3579: Strict handling of numeric overflow in text parsing
......................................................................


Patch Set 1:

(3 comments)

http://gerrit.cloudera.org:8080/#/c/3150/1//COMMIT_MSG
Commit Message:

Line 21:  However, integer and floating point parsing never returns
       :   an underflow today.
> Yeah for floating point we don't handle underflows. For integers I'm talkin
Is integer "truncation" the same as overflow?  I guess could you list an 
example with real numbers for what integer case we don't yet handle?


http://gerrit.cloudera.org:8080/#/c/3150/1/be/src/exec/text-converter.cc
File be/src/exec/text-converter.cc:

Line 249: (though the latter are only possible for
        :       // DECIMAL which isn't codegen'd)
> I thought we wanted to handle underflows in floats/doubles. Right now we do
What I was saying is that the wording was ambiguous. I first read to that mean 
that you were claiming underflow can only happen with DECIMAL, but what you 
mean is that the code only handles DECIMAL underflow.


http://gerrit.cloudera.org:8080/#/c/3150/1/testdata/workloads/functional-query/queries/QueryTest/strict-mode-abort.test
File 
testdata/workloads/functional-query/queries/QueryTest/strict-mode-abort.test:

Line 32
> Hm yeah, I think the best way to do that is introducing a new "bad data" fi
Oh yeah, okay, I didn't see that the other test did cover the negative case 
when i wrote this comment. I think that's fine.  Or you can add another column 
to the table, if that's easy.


-- 
To view, visit http://gerrit.cloudera.org:8080/3150
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-MessageType: comment
Gerrit-Change-Id: I7409c31ec0cb6fe0b2d9842b9f58fe1670914836
Gerrit-PatchSet: 1
Gerrit-Project: Impala
Gerrit-Branch: cdh5-2.6.0_5.8.0
Gerrit-Owner: Matthew Jacobs <[email protected]>
Gerrit-Reviewer: Dan Hecht <[email protected]>
Gerrit-Reviewer: Matthew Jacobs <[email protected]>
Gerrit-HasComments: Yes

Reply via email to