Yash Sharma 2013-11-20, 18:42
It would be great working with Jinfeng..
I have pushed code for (3) to look for NullExpressions on the same github repo:
Unfortunately was not able to test it because all the queries I hit on sqlline (with null parameters) were filtered by Optiq's Parsng exceptions. If anyone can share which query they are using I can try it on sqlline and test the flow.
For 1 & 2, I have a weighted-cost calculation in mind, based on the TYPE (required, optional repeated) . Open to suggestions..
From: [EMAIL PROTECTED] [[EMAIL PROTECTED]] on behalf of Jacques Nadeau [[EMAIL PROTECTED]]
Sent: Wednesday, November 20, 2013 8:26 PM
To: [EMAIL PROTECTED]
Subject: RE: hangout
There actually three different concepts to support with null.
1. non-nullable types should be first attempted to cast to their
non-nullable partners in precedence (such as non-nullable int to
non-nullable big int)
2. data mode casting should also be supported: non-nullable int to nullable
int to repeated int to get a function match.
3. Missing fields (currently represented my NullExpression) should be
replaced be null nullable constants of a particular minor type based on
function implementation introspection.
Drill-203 is focused on part three above. Agree that Jinfeng and Yash
should work together as this code is right on top of each other.
On Nov 20, 2013 6:39 AM, "Yash Sharma" <[EMAIL PROTECTED]> wrote:
> Copy That.
> Will add NULL as the least precedence type in our precedence map and will
> see if the getBestFunction() can pick up the closest implementation
> - Yash
> -----Original Message-----
> From: Timothy Chen [mailto:[EMAIL PROTECTED]]
> Sent: Wednesday, November 20, 2013 6:45 PM
> To: drill
> Subject: Re: hangout
> Btw Yash, we also have a bug around not handle nullable arguments well (
> and this should be also handled in the casting function search when null
> is passed.
> On Wed, Nov 20, 2013 at 2:17 PM, Yash Sharma <[EMAIL PROTECTED]
> > Sorry guys I dropped out of hangout yesterday due to connectivity issues.
> > I am working on Implicit Casting functionality. Have put a quick
> > prototype of the proposal here:
> > https://github.com/yashs360/incubator-drill-casting
> > Guys can have a look and let me know if its on right track.
> > Two prime things to mention:
> > - Am figuring out some way to convert the function call to new
> > function call.
> > - Next, I need to see how Optiq's classes can be used to determine the
> > type of casting (implicit/explicit)
> > -Regards, Yash
> > -----Original Message-----
> > From: Michael Hausenblas [mailto:[EMAIL PROTECTED]]
> > Sent: Tuesday, November 19, 2013 10:37 PM
> > To: Apache Drill Developer
> > Subject: Re: hangout
> > Yeah, regrets from me as well, on travel ...
> > Cheers,
> > Michael
> > --
> > Michael Hausenblas
> > Ireland, Europe
> > http://mhausenblas.info/
> > On 19 Nov 2013, at 17:52, Timothy Chen <[EMAIL PROTECTED]> wrote:
> > > Hey folks,
> > >
> > > Will be missing out hangout because of jetlag and time, but I just
> > > pushed the broadcast sender code to review.
> > >
> > > On the side still waiting for some response on the EMR side for more
> > > progress, but I'll try some more this week to see if I can figure
> > > out what's wrong with my script.
> > >
> > > Tim
> > ________________________________
> > NOTE: This message may contain information that is confidential,
> > proprietary, privileged or otherwise protected by law. The message is
> > intended solely for the named addressee. If received in error, please
> > destroy and notify the sender. Any use of this email is prohibited
NOTE: This message may contain information that is confidential, proprietary, privileged or otherwise protected by law. The message is intended solely for the named addressee. If received in error, please destroy and notify the sender. Any use of this email is prohibited when received in error. Impetus does not represent, warrant and/or guarantee, that the integrity of this communication has been maintained nor that the communication is free of errors, virus, interception or interference.