Fix issue #100, make work again with v0.5 #101

ScottPJones · 2016-04-03T04:15:02Z

This fixes deprecations due to removal of call overloading,
and also fixes a problem with handling subarrays,
due to JuliaLang/julia#15071 getting rid of the unexported index_generate function.

ScottPJones · 2016-04-03T04:23:23Z

@andreasnoack @ElOceanografo Could somebody from JuliaStats review this and possibly merge this?
@mbauman Does my change based on your suggestion look correct for subarrays? Thanks!!!

mbauman · 2016-04-03T19:19:52Z

src/subarrayn.jl

+const unsafe_getindex = Base.unsafe_getindex
+
+function Base.isnull{T,N,P<:NullableArray,IV,LD}(V::SubArray{T,N,P,IV,LD}, I::Int...)
+ @Base._inline_meta


Just use @inline function.... This private macro only exists for bootstrapping.

mbauman · 2016-04-03T19:26:57Z

There will be a performance regression for linear indexing with linear fast SubArrays here. That can be avoided with an additional method that uses the stride computation as in https://github.com/JuliaLang/julia/blob/57c53afba29a06493a2d993fb571a992d0ad733a/base/subarray.jl#L151.

mbauman · 2016-04-03T19:28:04Z

src/subarrayn.jl

+function Base.isnull{T,N,P<:NullableArray,IV,LD}(V::SubArray{T,N,P,IV,LD}, I::Int...)
+ @Base._inline_meta
+ @boundscheck checkbounds(V, I...)
+ @inbounds r = V.parent.isnull[Base.reindex(V, V.indexes, Base.to_indexes(I...))...]


Don't use to_indexes. They're already all Int.

Turns out, to_indexes for some reason is necessary, if I don't use it, it gets an error, saying that it can't find a method of reindex that matches. Any idea why?

Are you splatting I? Don't splat I.

nalimilan · 2016-04-03T20:05:48Z

Could you move the unrelated indentation changes to a separate commit?

nalimilan · 2016-04-03T20:06:25Z

src/subarrayn.jl

@@ -0,0 +1,25 @@
+const unsafe_getindex = Base.unsafe_getindex


This doesn't appear to be used anywhere.

That was just in the original code, so I left it in, didn't know if it was necessary. I'll remove it.

Turns out it is necessary for the v0.4 code.

ScottPJones · 2016-04-03T23:06:06Z

@mbauman @nalimilan I think this is all set to go now. Thanks to both of you for the review (learned some neat stuff about the new subarray changes to boot!)

mbauman · 2016-04-03T23:15:28Z

src/subarraynew.jl

+
+@inline function Base.isnull(V::NullableSubArray, I::Real...)
+ @boundscheck checkbounds(V, I...)
+ @inbounds return V.parent.isnull[Base.reindex(V, V.indexes, Base.to_indexes(I...))...]


Keep the signature the same between 0.4 and 0.5, and don't use to_indexes. Widening the signature is an unrelated change from what's required here, and it introduces a feature difference between code that's meant to be functionally the same.

It didn't work without to_indexes. I removed them, and then got an method error about reindex.

What did you try, and what error did you get? Don't splat I. Just V.parent.isnull[Base.reindex(V, V.indexes, I)...].

Never mind - the problem was an issue with splatting, I needed to remove both the to_index call and the splatting.

Why do you introduce these to_index calls in one commit, and remove them in the next one?

@nalimilan They were necessary to have the subscripts take Real, which would be more consistent with the code in Base/subarray.jl, however, I removed them based on @mbauman review comment above.
I do feel that it's incorrect to have these act differently than plain SubArrays, however, that could be fixed in a separate PR if people agree.

Presumably the commits will be squashed, removing the the addiction and
subtraction of that feature? I believe this is what @naliman was referring
to.

On Monday, April 4, 2016, Scott P. Jones notifications@github.com wrote:

In src/subarraynew.jl
#101 (comment)
:

@@ -0,0 +1,36 @@
+typealias NullableSubArray{T,N,P<:NullableArray,IV,LD} SubArray{T,N,P,IV,LD}
+
+@inline function Base.isnull(V::NullableSubArray, I::Real...)

@BoundsCheck checkbounds(V, I...)

@inbounds return V.parent.isnull[Base.reindex(V, V.indexes, Base.to_indexes(I...))...]

@nalimilan https://github.com/nalimilan They were necessary to have the
subscripts take Real, which would be more consistent with the code in
Base/subarray.jl, however, I removed them based on @mbauman
https://github.com/mbauman review comment above.
I do feel that it's incorrect to have these act differently than plain
SubArrays, however, that could be fixed in a separate PR if people agree.

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub
https://github.com/JuliaStats/NullableArrays.jl/pull/101/files/a2a4740c12a904f5bd11672d1190ef0eca59b02f..6a849974f7af5b75df58727dbe9f22dc5b978960#r58371197

@kmsquire I was asked to keep the fixing of the indentation as a separate commit, and I think the fixes related to adding Compat should probably be in a separate commit from the fixes related to JuliaLang/julia#15071, but I'll squash out that commit. Does that sound reasonable?

Perfect

On Monday, April 4, 2016, Scott P. Jones notifications@github.com wrote:

In src/subarraynew.jl
#101 (comment)
:

@@ -0,0 +1,36 @@
+typealias NullableSubArray{T,N,P<:NullableArray,IV,LD} SubArray{T,N,P,IV,LD}
+
+@inline function Base.isnull(V::NullableSubArray, I::Real...)

@BoundsCheck checkbounds(V, I...)

@inbounds return V.parent.isnull[Base.reindex(V, V.indexes, Base.to_indexes(I...))...]

@kmsquire https://github.com/kmsquire I was asked to keep the fixing of
the indentation as a separate commit, and I think the fixes related to
adding Compat should probably be in a separate commit from the fixes
related to JuliaLang/julia#15071
JuliaLang/julia#15071, but I'll squash out that
commit. Does that sound reasonable?

—
You are receiving this because you were mentioned.
Reply to this email directly or view it on GitHub
https://github.com/JuliaStats/NullableArrays.jl/pull/101/files/a2a4740c12a904f5bd11672d1190ef0eca59b02f..6a849974f7af5b75df58727dbe9f22dc5b978960#r58378574

Yes, that was what I meant.

mbauman · 2016-04-03T23:17:45Z

Besides #101 (comment) the SubArray commit looks good to me. I didn't review the other commits.

ScottPJones · 2016-04-04T00:22:08Z

OK, done now, fixed all of the subarray issues, the other issue was simply using @compat in 3 places to handle the deprecation of call overloading.

mbauman · 2016-04-04T00:32:10Z

LGTM. Thanks @ScottPJones.

ScottPJones · 2016-04-08T22:02:30Z

Bump! Anybody can merge this fix around? Thanks!

davidagold · 2016-04-09T15:01:40Z

@ScottPJones Thank you very much for this =) I have just one request -- could you please change the filenamesubarraynew.jl to something more specific, like subarray0_5.jl? Does that seem like a reasonable request?

Remove to_indexes/to_index, Int instead of Real indices Rename subarraynew.jl subarray0.5.jl

ScottPJones · 2016-04-09T16:58:47Z

@davidagold Very reasonable! (I should have thought of that myself! 😳) Thank you for the nice package!

mbauman reviewed Apr 3, 2016
View reviewed changes

nalimilan reviewed Apr 3, 2016
View reviewed changes

Fix issue JuliaStats#100 problems with v0.5, deprecated call() syntax

a2a4740

ScottPJones force-pushed the spj/fixv05 branch 2 times, most recently from fedd5cb to 144e57c Compare April 3, 2016 23:00

mbauman reviewed Apr 3, 2016
View reviewed changes

ScottPJones force-pushed the spj/fixv05 branch from 144e57c to d9505e3 Compare April 4, 2016 00:17

ScottPJones force-pushed the spj/fixv05 branch from d9505e3 to e92d84a Compare April 4, 2016 13:56

ScottPJones added 2 commits April 9, 2016 12:45

Fix problems with subarray code related to PR 10571 in JuliaLang/julia

69d81d9

Remove to_indexes/to_index, Int instead of Real indices Rename subarraynew.jl subarray0.5.jl

Make indentation match Julia style guide

09a616b

ScottPJones force-pushed the spj/fixv05 branch from e92d84a to 09a616b Compare April 9, 2016 16:56

davidagold merged commit 5de3aff into JuliaStats:master Apr 11, 2016

ScottPJones deleted the spj/fixv05 branch April 12, 2016 11:23

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Fix issue #100, make work again with v0.5 #101

Fix issue #100, make work again with v0.5 #101

ScottPJones commented Apr 3, 2016

ScottPJones commented Apr 3, 2016

mbauman Apr 3, 2016

mbauman commented Apr 3, 2016

mbauman Apr 3, 2016

ScottPJones Apr 3, 2016

mbauman Apr 3, 2016

nalimilan commented Apr 3, 2016

nalimilan Apr 3, 2016

ScottPJones Apr 3, 2016

ScottPJones Apr 3, 2016

ScottPJones commented Apr 3, 2016

mbauman Apr 3, 2016

ScottPJones Apr 3, 2016

mbauman Apr 3, 2016

ScottPJones Apr 4, 2016

nalimilan Apr 4, 2016

ScottPJones Apr 4, 2016

kmsquire Apr 4, 2016

ScottPJones Apr 4, 2016

kmsquire Apr 4, 2016

nalimilan Apr 4, 2016

mbauman commented Apr 3, 2016

ScottPJones commented Apr 4, 2016

mbauman commented Apr 4, 2016

ScottPJones commented Apr 8, 2016

davidagold commented Apr 9, 2016

ScottPJones commented Apr 9, 2016

		@@ -0,0 +1,25 @@
		const unsafe_getindex = Base.unsafe_getindex

Fix issue #100, make work again with v0.5 #101

Fix issue #100, make work again with v0.5 #101

Conversation

ScottPJones commented Apr 3, 2016

ScottPJones commented Apr 3, 2016

Choose a reason for hiding this comment

mbauman commented Apr 3, 2016

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

nalimilan commented Apr 3, 2016

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ScottPJones commented Apr 3, 2016

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mbauman commented Apr 3, 2016

ScottPJones commented Apr 4, 2016

mbauman commented Apr 4, 2016

ScottPJones commented Apr 8, 2016

davidagold commented Apr 9, 2016

ScottPJones commented Apr 9, 2016