Clean out pre-0.4 cruft #323
Conversation
|
Looks like this will trigger warnings in StatsBase (re |
|
I modified the deprecation warning for Among the packages installed on my machine, the other remaining consumer of |
|
I'd like to avoid introducing deprecation warnings on Julia 0.4, since many packages aren't making new releases there. When we drop support for 0.4 would be a good time to do this, though this likely should be one of the last packages to do that. |
|
Very reasonable choice. I commented out the depwarns and renamed the files |
| end | ||
| export Libc, Libdl | ||
| else | ||
| import Base: Libc, Libdl |
There was a problem hiding this comment.
should be kept, or Compat.Libc moved to to-be-deprecated
There was a problem hiding this comment.
Not necessary.
julia> Compat.Libc
Base.Libc
| include("linspace.jl") | ||
| const linspace = LazyLinSpace.linspace | ||
| else | ||
| const linspace = Base.linspace |
There was a problem hiding this comment.
likewise should keep Compat.linspace available, possible future deprecation candidate
There was a problem hiding this comment.
julia> Compat.linspace == Base.linspace
true
| @test Compat.@Dict(1 => 1) == d | ||
| @test Compat.@Dict(1 => v) == d | ||
|
|
||
| @test typeof(@compat(Dict(1 => 1))) == Dict{Int,Int} |
There was a problem hiding this comment.
won't this be a no-op rather than a depwarn?
There was a problem hiding this comment.
They won't trigger a deprecation warning, but conversely the tests aren't required since Compat doesn't do anything to Dict syntax anymore. No point hanging on to no-op tests.
| end | ||
|
|
||
| if VERSION < v"0.4.0-dev+5322" | ||
| include("timer.jl") |
There was a problem hiding this comment.
delete this file then i guess?
|
|
But leave them commented out until we drop 0.4 support
|
Rebased. I'll merge later today unless I hear otherwise. |
|
Maybe merge after #326? |
|
That one is unquestionably of higher priority. This is just one of those PRs that shouldn't linger too long, since it's a magnet for merge conflicts. |
|
Actually I thought this removes 0.4 support which is not the case so nvm. I'll rebase this one if I want to merge #326 (probably early afternoon) before you merge this one. |
|
Is (most of) this code hurting anything by being here but inactive? |
|
It shouldn't hurt anything though it should be done right after 0.3 support is dropped... |
| calltypes[k] = v | ||
| end | ||
| elseif VERSION < v"0.5.0-dev+3831" | ||
| import Base.unsafe_convert |
There was a problem hiding this comment.
So what'll be the timeline for deprecating and removing this? (Assuming we don't want people to use Compat.unsafe_convert for ever, might also apply to a few other places).
I can see the argument of not giving new depwarn on 0.4 though waiting longer doesn't help since we always drop a version here only when virtually no one is using it anymore. Should we do that on the same release that drop 0.4 support assuming that's when 0.5 is still widely tested?
That was a while ago, and I argued against it at the time since leaving it in makes the possibility of restoring 0.3 support including (some of?) the latest features a little simpler, if a need ever arises for that. Code deletions could be reverted and people could use branches for such things, but if it's harmless to keep it... |
|
Proximal motivation was that I was entertaining the possibility of putting in the new |
|
So the old pre-0.4 linspace is hurting stuff by being here, I'm fine with deleting that if it's in your way. |
|
I should confess that I can't promise to get to this immediately. We could merge this when I'm sure I'll have time. |
|
There has been an important change since 0.3 support was dropped: with 0.6 on the horizon, I think that the chances of ever needing to backport anything to 0.3 are now very low. So deleting the code makes sense to me. |
|
OK, let's do it now just so there aren't rebase conflicts in the future. |
Puts Compat on a bit of a diet. I wasn't sure whether we should deprecate old functionality (do we need
CompatCompat.jl?), but I did so for safety. I am also not sure whether we want to run tests of deprecated functionality, but again included it for safety.