Added serde1 feature to Serialize/Deserialize WeightedIndex

[CGMossa]

It required adding serde as an optional dependency.
Also added a test using bincode like in the other crate.

Referering to this: #967

I hope this is up to standard, otherwise I'll take all of your feedback and ensure it is as correct
as possible.

[dhardy]

Thanks @CGMossa. Hmm, would it be possible to do the reformatting first in its own commit? Also please add a note to the changelog.

https://rust-random.github.io/book/contributing.html

[CGMossa]

I have the intellij auto use rustfmt. Which is not admittedly cool, but I thought that would be fine. I'll add a line in the changelog. I thought you'd suggest more things to add the derive attribute to. I'd be happy to add it everywhere needed. But I don't know which places.

…

On Fri, May 1, 2020, 19:29 Diggory Hardy ***@***.***> wrote: Thanks @CGMossa <https://github.com/CGMossa>. Hmm, would it be possible to do the reformatting *first* in its own commit? Also please add a note to the changelog. https://rust-random.github.io/book/contributing.html — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#974 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAIDVSCCVXXWLSJQCZ62MLLRPMBGPANCNFSM4MXET6CQ> .

[dhardy]

I thought you'd suggest more things to add the derive attribute to. I'd be
happy to add it everywhere needed. But I don't know which places.

It crossed my mind that this might be a good idea, but I didn't look into where yet. Probably would only be applicable to distributions.

[CGMossa]

These are the changes you've asked for. I will experiment with making all the distributions serde complaint.

…

On Fri, May 1, 2020, 20:31 Diggory Hardy ***@***.***> wrote: I thought you'd suggest more things to add the derive attribute to. I'd be happy to add it everywhere needed. But I don't know which places. It crossed my mind that this might be a good idea, but I didn't look into where yet. Probably would only be applicable to distributions. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#974 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAIDVSFNXMAM2WC6EOIEWR3RPMIQDANCNFSM4MXET6CQ> .

[vks]

It looks like the minimal supported serde version is incorrectly specified.

[CGMossa]

I wrote it exactly like it is written in rand_core and rand_pcg. Which version should I write instead?
Should I use this

serde = "1.0.106"

or

serde = "1.0.*"

?

[CGMossa]

I see that tha CI stuff is resulting in an error. I am unable to see the same error here. I've added a version number to one of the Cargo.toml and I'll wait and see. Even if I run the command that is making the error cargo test --features=serde1,log I don't get an error. But I have serde 1.0.105 installed and the latest is 1.0.106 and maybe the thing on the here has an earlier version?

[CGMossa]

This worked. I've no idea why, since the other crates in the rand-workspace also indicated serde with "1". Maybe they fixed a bug? Although, not this works.

[dhardy]

We test with the minimal version of Serde which crates say they support — but this must be unified across all dependencies, so will be the highest minimum version. The idea is to find the real minimum version, but I think most people only care that it's high enough to work. That said, 1.0.105 is less than two months old. If an older version suffices that would be preferable. You can test locally via cargo generate-lockfile -Z minimal-versions

[dhardy]

For completeness's sake (if you don't mind), we should also consider: Bernoulli, Uniform, UniformFloat, UniformDuration, StepRng, IndexVec (maybe).
The following structs have zero size (thus don't need serialising), but doing so might ease deriving of other types: OpenClosed01, Open01, Standard, Alphanumeric.

[CGMossa]

Yep, I'll take a look later today. Thanks for the follow up!

…

On Mon, 4 May 2020 at 09:40, Diggory Hardy ***@***.***> wrote: For completeness's sake (if you don't mind), we should also consider: Bernoulli, Uniform, UniformFloat, UniformDuration, StepRng, IndexVec (maybe). The following structs have zero size (thus don't need serialising), but doing so might ease deriving of other types: OpenClosed01, Open01, Standard, Alphanumeric. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#974 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAIDVSGSA6JM5OM46NQ4D33RPZWONANCNFSM4MXET6CQ> .

[CGMossa]

Minimal supported serde version is 1.0.103. Release note says:

Support deserializing untagged unit variants from formats that treat unit as None (#1668)

Test command:

rustup run nightly cargo generate-lockfile -Z minimal-versions & cargo test --features=serde1,log

[dhardy]

1.0.103

This version is almost six months old. Since we're developing against the master branch (will be a breaking release), this is fine IMO.

[CGMossa]

I've added the serialize and deserialize annotations. I also wanted to add tests using bincode, but some of these types have no fields I can test equality against, and PartialEq is not implemented for them so.. I don't know.

I've left some skeleton code, awaiting further feedback.

(this is like my biggest attempt at contributing yet! Thanks for the PR-mentoring!!)

[CGMossa]

Am I missing more things here?

[dhardy]

Sorry, I'll try to review in the next couple of days.

[dhardy]

Sorry, I haven't been paying this project much attention lately!

Your PR looks good. Those unfinished bits shouldn't be hard (I gave a couple hints).

[CGMossa]

I've added the tests based on your suggestions.

Do I resolve them or do I wait for you to click that they are resolved?

[dhardy]

Looks good to me. (There are some over-long lines, and you didn't need to match cases which can't occur in the tests, but it's still good enough.)

I'll wait a couple of days in case @vks wishes to review, then merge.

[vks]

Looks good!

[vks]

This should probably be squashed before merging. I think this is technically a breaking change.

[CGMossa]

I don't know how to squash?

[vks]

I think GitHub offers this option to the maintainers, but it seems like it is disabled for this repository.

Edit: I just enabled it, you don't have to do anything, we can let GitHub do it.

If you want to do that yourself, you can use git rebase -i to specify which commits should be squashed.

[dhardy]

master already contains breaking changes, so no problem there.

[CGMossa]

I've added the missing spaces. Thanks for enabling that feature, as I've got no training in using rebase or squashing. I'd appreciate a link to learn about this.

[dhardy]

Thanks for the PR!

git rebase is quite useful. Also git add -p. You may want to test them out on a dummy repo!