[OP] Move Elementwise binary scalar ops to SimpleOp Registry

[tqchen]

No description provided.

[tqchen]

cc @pluskid @Javelinjs

This PR cause a backward compatible break in symbolic API.

• Overload of scalar - symbol
  • _DivScalar(lhs, scalar_on_left=true) is changed to _RDivScalar
• Same stands for _RPowerScalar, _RMinusScalar

I have updated the python API, but maybe changes in julia and scala is needed if you overloaded the operator. Since these symbols are not frequently used, I plan to merge this in when ok

[yzhliu]

I'll fix it this weekend.

[tqchen]

@Javelinjs please do a PR to my repo so the things can be fixed when I merge this in.

@thirdwing can you also do so for the R testcase?

[thirdwing]

OK. I will do that tonight.

[tqchen]

@thirdwing @Javelinjs any update on this?

[thirdwing]

Sorry for the late reply. Working on a conference paper. Will update this
soon.

On Sun, Apr 17, 2016 at 1:49 AM, Tianqi Chen notifications@github.com
wrote:

@thirdwing https://github.com/thirdwing @Javelinjs
https://github.com/javelinjs any update on this?

—
You are receiving this because you were mentioned.
Reply to this email directly or view it on GitHub
#1857 (comment)

Qiang Kou
qkou@umail.iu.edu
School of Informatics and Computing, Indiana University

[pluskid]

I think it would be good to set a goal to introduce versioning into libmxnet. Then this kind of breaking changes will not be a disaster for other language bindings.

[tqchen]

agreed

[tqchen]

@thirdwing @pluskid @Javelinjs this is now merged

[pluskid]

Thanks! I fixed the Julia side.

[yzhliu]

@tqchen Shall we also modify the divide and subtract function in python/mxnet/ndarray.py, as I mentioned in tqchen#1 ?

[tqchen]

@Javelinjs I changed the convention back, and changed the scala version as well.

[yzhliu]

Oh, I see.