deprecate std.experimental.ndslice#4946
Conversation
wilzbach
left a comment
wilzbach
left a comment
There was a problem hiding this comment.
I have observed the pain resulting from syncing ndslice with its DUB packages for quite a while now, so I do understand your decision. In fact I was also asking for it in the past ;-)
Being a sole DUB packages makes the life of the maintainers a lot easier.
std/experimental/ndslice/internal.d
Outdated
| @@ -1,3 +1,4 @@ | |||
| deprecated("Please use appropriate Mir project at http://github.com/libmir") | |||
There was a problem hiding this comment.
Would you be so kind to
- add a comment similar to
/// @@@DEPRECATED_2017-04@@@(s.t. the deprecations can be automatically found) - add a timestamp for scheduled removal (idk what the exact deprecation time for std.experimental is - CC @jmdavis)
- link to the exact GitHub repo, s.t. users won't be lost - will there be
mir-ndslice?
There was a problem hiding this comment.
mir-std. BTW, ndslice will be completely reworked. Instead of existing numpy-like ndslice a new ndslice will be provided. 90% of code will be rewritten. New ndslice:
- Will be faster
- Will be integrated with ranges
- Will be integrated with algorithms
In addition,
- new Range API architecture
- new algorithms implementation and API
- 3 types of ndslices:
a. numpy-like: number of strides equals to number of dimensions
b. raw-contnious (CBLAS-like): number of strides equals to number of dimensions minus one
c. ideally continuous: number of strides equals to 0.
|
@wilzbach is current commit OK? |
Yes. thanks! I think a deprecation period of two releases is quite ok for As this is a big change in terms of policy, I want to give everyone a chance to notice (in particular @andralex). |
fix import
|
I'm okay with this. We shouldn't even need to worry with deprecations for experimental code, but it's nice. |
| @@ -1,3 +1,5 @@ | |||
| /// @@@DEPRECATED_2017-04@@@ | |||
| deprecated("Please use mir-algorithm DUB package: http://github.com/libmir/mir-algorithm") | |||
There was a problem hiding this comment.
@9il I just realized that the repo doesn't exist yet. Do you plan to create it or did you mean to link to https://github.com/libmir/mir-std?
There was a problem hiding this comment.
Okay - would you be so kind to create it, s.t. deprecation warnings don't go to a nirvana?
There was a problem hiding this comment.
There was a problem hiding this comment.
https is important in spite of the fact that github will redirect http to https - at least in theory a compromised network could intercept the http request and take it elsewhere...
|
Auto-merge toggled on |
3 participants
Including ndslice to Phobos was my mistake. It is better to distribute it as DUB package. Please merge this PR. Phobos synchronization terrificly slowdown ndslice packages development, also Phobos is a classic D library while Mir projects require strong BetterC guaranties. I has not enough resources to maintain both Mir and Phobos versions. Hope this will be merged and users would not be confused anymore by Mir and Phobos ndslices incompatibility.
ndslice author