What to read to understand finishing v0.5?

[Andreas Lobinger]

Hello colleagues,

i need a bigger picture of the status of v0.5, dates, timelines, missing features, missing testing, expected closing. Just go to github and select the v0.5 milestone gives me a diverse picture.

Wishing ahappy day,
        Andreas

[Scott Jones]

I'm curious about the when the following change will make it into v0.5, originally it was said that it would be done early after v0.4 split off of master:

julia> a = [[1],[2],[3]]

WARNING: [a,b,...] concatenation is deprecated; use [a;b;...] instead

 in depwarn(::ASCIIString, ::Symbol) at ./deprecated.jl:64

 in oldstyle_vcat_warning(::Int64) at ./abstractarray.jl:29

 in vect(::Array{Int64,1}, ::Vararg{Array{Int64,1}}) at ./abstractarray.jl:32

 in eval(::Module, ::Any) at ./boot.jl:267

while loading no file, in expression starting on line 0

3-element Array{Int64,1}:

 1

 2

 3

Instead of it acting like this:

julia> a = Vector{Int}[[1],[2],[3]]

3-element Array{Array{Int64,1},1}:

 [1]

 [2]

 [3]

[John Myles White]

I think it's fair to say that the reason your questions aren't already answered by GitHub is because there's no one who's made an executive decision about the answers to those questions.

 -- John

[David Anthoff]

At some point in the summer there was communication from the core team that v0.5 would be a short release cycle that was all about the arraypocalypse theme. My sense is that this original plan is off the table, and that (as John said) there is no new time plan to replace the original one, or at least none has been communicated from the core at this point. Take this info with a grain of salt, I’m largely reading tea-leaves here :)

[Stefan Karpinski]

The sticking point has been the decision – which seemed to have already been made as of last summer – about whether slices should be views or not. There has been fierce debate about that and no apparent consensus. Part of the problem is that we do not yet have all of the compiler support required to fully evaluate the potential performance of array views: we still need to heap allocate any object which refers to heap allocated objects, which means that we cannot stack allocate array views and the compiler cannot do much to optimize them.

At this point, I'm inclined to wrap up the rest of the array changes, including the one that Scott brings up and start the RC phase for 0.5. That would leave the array view change as a future potential change.

[Jeffrey Sarnoff]

With the ability to add methods to an abstract type, so we do not lose that important capability; makes sense to me.

[Jeff Bezanson]

The 0.5.0 milestone on github does in fact reflect our best
understanding of what's in the release. Perhaps there is a
presentation or UI problem here, but fundamentally I don't see why a
list of issues wouldn't work for this. One problem seems to be that
people too often add fairly minor items to the 0.5.0 milestone, so
randomly-chosen bugs get mixed in. Anything that could be addressed in
a point release should be in the 0.5.x milestone.

I think it could help if everybody actively critiques the milestone
issues, instead of looking at it and just concluding the list is junk.
For example, comment "this looks like a small bug, please move to
0.5.x", or "is anybody working on this?" etc.

[Jeffrey Sarnoff]

https://github.com/JuliaLang/julia/milestones/0.5.0

[Viral Shah]

This is about the right time to start the triage for the various 0.5 issues. With all the amazing compiler improvements, LLVM upgrade, Cxx readiness, thread-safety (will probably remain disabled for 0.5 potentially unless it receives significant testing) and a number of other cool things, I agree with Stefan here that we should not wait for array views for 0.5, but certainly include all the other array changes. Also, we are working hard towards having a debugger ready, which potentially could be the last major feature before calling feature freeze.

In my ideal world, we would have a 0.5 release right before JuliaCon, and focus on the next fun set of things in the hackathon. I do think this is possible.

Not relevant here, but I also feel that we are closing in on 1.0. Maybe another release or two after 0.5.

-viral

[Eric Forgy]

On Thursday, March 10, 2016 at 2:00:46 PM UTC+8, Viral Shah wrote:

Not relevant here, but I also feel that we are closing in on 1.0. Maybe another release or two after 0.5.

+1 (or should that be +1.0) :D

[DNF]

On Thursday, March 10, 2016 at 7:00:46 AM UTC+1, Viral Shah wrote:

This is about the right time to start the triage for the various 0.5 issues. With all the amazing compiler improvements, LLVM upgrade, Cxx readiness, thread-safety (will probably remain disabled for 0.5 potentially unless it receives significant testing) and a number of other cool things, I agree with Stefan here that we should not wait for array views for 0.5, but certainly include all the other array changes. Also, we are working hard towards having a debugger ready, which potentially could be the last major feature before calling feature freeze.

By debugger, do you mean Gallium.jl? I have occasionally taken a look at the repository, and got the impression that it was not very active.

A debugger for Julia would be the most amazing thing! It is the single most important missing feature that is holding me back from using Julia in my day-to-day work. Have you any idea if it would be long before the debugger could be integrated in the IDEs (Juno/Eclipse)?

[Andreas Lobinger]

Afaics Gallium.jl is a thin layer on some other debugging/compiling infrastructure. The main work is not in Gallium.jl iself, but in the lower layers.

[Viral Shah]

Work to prepare for the debugger has been going on in Base all along, DWARF.jl, and Keno also has stuff that he has not yet committed. He’ll send details when he has something for everyone to try out.

The first step is to get a working debugger. Once we have a debugger, the plan is to start integration with both Eclipse and Juno, but we haven’t yet figured out how easy or difficult it will be. I suspect it won’t be too difficult.

-viral

[DNF]

I had no idea. I guess it shows that unless one knows what to look for, it can be a bit hard to get an impression of the state of development in some cases.

Anyway, to me this is wonderful news, and the single most exciting thing I've heard for a while (about Julia, that is :)

[Tim Holy]

For those who are eager to see 0.5 materialize, there are things you can do to
help. Some are hard, but some are not. In particular, in

https://github.com/JuliaLang/julia/pull/15434

the thought is to audit every use of "classic" iteration (for i = 1:...) in
base, and see if there need to be changes. Since grep tells me there are 1745
of these, it's a big tedious job, but one that shouldn't require deep julia-
fu. Anyone who wants to pitch in can comment over at that issue about ways to
divvy up the work.

Best,
--Tim

[Sisyphuss]

I am monitoring the milestone constantly. I find the percentage does grow: 27% -> 28% -> 33%

[Daniel Carrera]

Hi Tim,

I'm leaving on a work trip on Monday, but let's see if I can manage to make a Julia contribution on a weekend. I need some help, as I've never submitted any code for Julia before. I left a comment in the PR. Let's start with the basics: How do I pull your modified branch / PR? How do I run the tests? How do I submit my modified code back to you?

Cheers,

Daniel.

[Andreas Lobinger]

Hello colleague,

On Friday, March 11, 2016 at 2:56:02 PM UTC+1, Daniel Carrera wrote:

Hi Tim,

I'm leaving on a work trip on Monday, but let's see if I can manage to make a Julia contribution on a weekend. I need some help, as I've never submitted any code for Julia before. I left a comment in the PR. Let's start with the basics: How do I pull your modified branch / PR? How do I run the tests? How do I submit my modified code back to you?

first of all: https://github.com/JuliaLang/julia/blob/master/CONTRIBUTING.md exists.
You really need some git experience and know how to organize your own work in local and github (your fork of julia) repositories. I found this helpful
http://stackoverflow.com/questions/27567846/how-can-i-checkout-a-github-pull-request

PRs you do (to julia) should be subject to automatic testing.

[Daniel Carrera]

On 11 March 2016 at 15:20, Andreas Lobinger <lobi...@gmail.com> wrote:

first of all: https://github.com/JuliaLang/julia/blob/master/CONTRIBUTING.md exists.

Thanks!

 

You really need some git experience and know how to organize your own work in local and github (your fork of julia) repositories. I found this helpful
http://stackoverflow.com/questions/27567846/how-can-i-checkout-a-github-pull-request

PRs you do (to julia) should be subject to automatic testing.

 

Yeah, I have some git experience. I use git on my own projects, but for me it's little more than running "git commit" from time to time. I used to know more. It's all slowly coming back to me. :-)

Cheers,

Daniel.