a more earnest attempt at renaming Void #25162

StefanKarpinski · 2017-12-18T15:34:52Z

StefanKarpinski · 2017-12-18T15:37:11Z

This is going to be massively conflict-prone, so I've separated it into the manual change to base/boot.jl which is fairly finicky and the automatic part that changes all the things.

StefanKarpinski · 2017-12-18T19:42:16Z

Ok, the real deal this time. Can be easily rebased since the large rename step is scripted.

Sacha0 · 2017-12-19T23:05:36Z

(Would you like a manual review here, or are you confident in your transformations? :))

StefanKarpinski · 2017-12-19T23:16:35Z

I'm pretty sure this is correct, we just need to make a decision.

JeffBezanson · 2017-12-19T23:49:50Z

What I recall from when we had Nothing was that people would mistakenly use it instead of nothing fairly often. The similarity to python's None might contribute to this. Python also uses NoneType as the type of None, which helps avoid this confusion.

StefanKarpinski · 2017-12-20T05:25:09Z

If we have any other name ideas, it's pretty easy to change this since it's automated now. I still don't understand why nothing :: Nothing is uniquely problematic when we use that convention all the time. We're doing missing :: Missing, for example, and this pattern is now fairly well established –
maybe it wasn't so common back then. Back then using Nothing instead of nothing often kind of worked since they're both singletons and we weren't using nothing except for its non-printing behavior in the REPL. Now we are actively using it to indicate a missing value – and passing Nothing where nothing is expected will fail, so that particular issue may well have already resolved itself. Using NoneType for the type of None in Python is palatable partly because you don't talk about types much in Python. We have to write Nothing pretty often, so I think some brevity matters.

JeffBezanson · 2017-12-20T06:49:22Z

There are very few cases where a value and its type have the same name except for case. There are several pairs, like filter and Filter, or adjoint and Adjoint, where the two names have a different relationship, typically a function and the type of value it returns.

StefanKarpinski · 2017-12-20T07:01:52Z

There's an awful lot of support for this rename on the original issue – and a bit of dissent, but it's unusually lopsided. Perhaps people who voted up or down over there can chime in here?

helgee · 2017-12-20T10:56:16Z

I concur with all arguments made in favor of the rename in the original issue and I want to add that Void is a very jargony term in my opinion. It is an uncommon word and many non-native speakers will encounter it the first time in the context of programming.

Calling it Nothing probably saves them a trip to the dictionary if they are learning programming with Julia and thus lowers the barrier of entry.

nalimilan · 2017-12-20T11:11:56Z

I think Nothing particularly improves clarity in docstrings where ::Union{T, Void} parameters and return values don't currently evoke the object nothing at all. The consistency with Missing is also really appealing.

rfourquet · 2017-12-20T11:34:49Z

I like terseness, so it's hard for me to support officially this rename, considering this evolution in length:

Nullable{Int}
Union{Int,Void}
Union{Int,Nothing}

But if I'm honest, I must admit to secretely hope Nothing will win, because it just feels right. Void by conflating two notions, is inconsistent in naming with both nothing and C's void. Nothing plus the Cvoid alias seems the least surprising on both fronts. And maybe we will have something like a Maybe{Int} = Union{Int,Nothing} type anyway for terseness :)

JeffBezanson · 2017-12-20T13:29:14Z

Ok, I'm not really set against this. The arguments for it are pretty reasonable.

We should certainly have a shorter way to write Union{T,Nothing}.

StefanKarpinski · 2017-12-20T14:40:18Z

Ready to go when CI passes.

StefanKarpinski · 2017-12-20T17:48:58Z

MacOS timeout; everything else has passed (again).

ararslan · 2017-12-20T19:11:37Z

Perhaps people who voted up or down over there can chime in here?

For whatever it's worth, I was one of the few dissenters, and I admit that Cvoid in ccall seems pretty nice. I actually dislike the trend of a singleton value thing having type Thing (and I rather like the name Void), which is mostly why I didn't like the proposal, but if we have a shorter way to write Union{T, Nothing} then that solves a significant portion of that concern for me.

Sacha0 · 2017-12-20T20:59:09Z

Perhaps people who voted up or down over there can chime in here?

Crossposting the relevant bit I posted in support:

The relationship between nothing and Void confused me when I came to Julia, and these changes completely eliminate that source of confusion. I find the Ptr{Void} -> Ptr{Cvoid} changes beautifully clarifying as well. ❤️!

git ls-files base doc examples src stdlib test | xargs perl -i -ple '/\b(ccall|cglobal|cfunction)\b/ && s/\bVoid\b/Cvoid/g' git ls-files base doc examples src stdlib test | xargs perl -i -ple 's/\bPtr\{Void\}/Ptr{Cvoid}/g' git ls-files base doc examples src stdlib test | xargs perl -i -ple 's/\Ref\{Void\}/Ref{Cvoid}/g' perl -i -ple 's/\bVoid\b/Cvoid/g' base/atomics.jl base/libgit2/status.jl base/random/dSFMT.jl doc/src/manual/calling-c-and-fortran-code.md test/llvmcall.jl test/threads.jl git ls-files base doc examples src stdlib test | xargs perl -i -ple 's/\bVoid\b/Nothing/g'

quinnj · 2017-12-21T02:17:12Z

I'd be in favor of making T? desugar to Union{T, Nothing}. I've come to think that this definition will be more used/common than a previously suggested idea of having it desugar to Union{T, Missing}. Not something we need to rush into though; we still need some time for all the Union improvements to ripple through the ecosystem.

StefanKarpinski · 2017-12-21T03:30:02Z

Let's see where there's more pain... Feel pain, apply syntax?

vchuravy · 2017-12-21T11:14:53Z

FemtoCleaner is going to have it's work cut out for it :P

StefanKarpinski · 2017-12-21T15:08:49Z

Eh, this is a dead easy one to automate (as proved by the PR itself).

ararslan · 2017-12-21T18:10:08Z

It's also now available in Compat.

ararslan added the deprecation This change introduces or involves a deprecation label Dec 18, 2017

StefanKarpinski force-pushed the sk/voider branch from d20461c to 418ed7c Compare December 18, 2017 19:41

StefanKarpinski force-pushed the sk/voider branch from 418ed7c to eccd466 Compare December 19, 2017 18:47

StefanKarpinski force-pushed the sk/voider branch from eccd466 to 7cb77f0 Compare December 20, 2017 14:39

StefanKarpinski added 3 commits December 20, 2017 16:54

const Cvoid = Void

84ce70b

NEWS and deprecation for Void => Nothing rename

0f4d5db

StefanKarpinski force-pushed the sk/voider branch from 7cb77f0 to 0f4d5db Compare December 20, 2017 21:57

StefanKarpinski merged commit 555264e into master Dec 20, 2017

StefanKarpinski deleted the sk/voider branch December 20, 2017 23:41

omus mentioned this pull request Jan 21, 2018

Added testsets and an upload function that can do retries invenia/FTPClient.jl#42

Merged

cormullion mentioned this pull request Mar 18, 2018

NEWS.md is getting a bit untidy #26508

Closed

tkf mentioned this pull request Sep 27, 2018

Make InterruptException capturable in a program #29383

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

a more earnest attempt at renaming Void #25162

a more earnest attempt at renaming Void #25162

StefanKarpinski commented Dec 18, 2017 •

edited

Loading

StefanKarpinski commented Dec 18, 2017

StefanKarpinski commented Dec 18, 2017

Sacha0 commented Dec 19, 2017

StefanKarpinski commented Dec 19, 2017

JeffBezanson commented Dec 19, 2017

StefanKarpinski commented Dec 20, 2017 •

edited

Loading

JeffBezanson commented Dec 20, 2017

StefanKarpinski commented Dec 20, 2017

helgee commented Dec 20, 2017

nalimilan commented Dec 20, 2017

rfourquet commented Dec 20, 2017

JeffBezanson commented Dec 20, 2017

StefanKarpinski commented Dec 20, 2017

StefanKarpinski commented Dec 20, 2017

ararslan commented Dec 20, 2017

Sacha0 commented Dec 20, 2017 •

edited

Loading

quinnj commented Dec 21, 2017

StefanKarpinski commented Dec 21, 2017 •

edited

Loading

vchuravy commented Dec 21, 2017

StefanKarpinski commented Dec 21, 2017

ararslan commented Dec 21, 2017

a more earnest attempt at renaming Void #25162

a more earnest attempt at renaming Void #25162

Conversation

StefanKarpinski commented Dec 18, 2017 • edited Loading

StefanKarpinski commented Dec 18, 2017

StefanKarpinski commented Dec 18, 2017

Sacha0 commented Dec 19, 2017

StefanKarpinski commented Dec 19, 2017

JeffBezanson commented Dec 19, 2017

StefanKarpinski commented Dec 20, 2017 • edited Loading

JeffBezanson commented Dec 20, 2017

StefanKarpinski commented Dec 20, 2017

helgee commented Dec 20, 2017

nalimilan commented Dec 20, 2017

rfourquet commented Dec 20, 2017

JeffBezanson commented Dec 20, 2017

StefanKarpinski commented Dec 20, 2017

StefanKarpinski commented Dec 20, 2017

ararslan commented Dec 20, 2017

Sacha0 commented Dec 20, 2017 • edited Loading

quinnj commented Dec 21, 2017

StefanKarpinski commented Dec 21, 2017 • edited Loading

vchuravy commented Dec 21, 2017

StefanKarpinski commented Dec 21, 2017

ararslan commented Dec 21, 2017

StefanKarpinski commented Dec 18, 2017 •

edited

Loading

StefanKarpinski commented Dec 20, 2017 •

edited

Loading

Sacha0 commented Dec 20, 2017 •

edited

Loading

StefanKarpinski commented Dec 21, 2017 •

edited

Loading