New package: Fminbnd v1.0.0

[JuliaRegistrator]

• Registering package: Fminbnd
• Repository: https://github.com/kagalenko-m-b/Fminbnd.jl
• Created by: @kagalenko-m-b
• Version: v1.0.0
• Commit: b52ae0c07af2413a1de3bc8901f20b534a2b1c65
• Reviewed by: @kagalenko-m-b
• Reference: kagalenko-m-b/Fminbnd.jl@b52ae0c#commitcomment-149298861

[github-actions]

Hello, I am an automated registration bot. I help manage the registration process by checking your registration against a set of AutoMerge guidelines. If all these guidelines are met, this pull request will be merged automatically, completing your registration. It is strongly recommended to follow the guidelines, since otherwise the pull request needs to be manually reviewed and merged by a human.

1. New package registration

Please make sure that you have read the package naming guidelines.

2. AutoMerge Guidelines are all met! ✅

Your new package registration met all of the guidelines for auto-merging and is scheduled to be merged when the mandatory waiting period (3 days) has elapsed.

3. To pause or stop registration

If you want to prevent this pull request from being auto-merged, simply leave a comment. If you want to post a comment without blocking auto-merging, you must include the text [noblock] in your comment.

Tip: You can edit blocking comments to add [noblock] in order to unblock auto-merging.

[goerz]

This seems to be the followup from #99531.

I'm not really clear on what the bnd part of the name should indicate.

[goerz]

I might tentatively suggest NetlibFmin as a name

[goerz]

I'd also strongly recommend leaving the name of the function exported by the package as fmin.

[kagalenko-m-b]

I might tentatively suggest NetlibFmin as a name

There’re widely used matlab/octave functions ‘fminbnd’ and python’s ‘fminbound’ that also derive from netlib’s fmin. “Bnd” means that it searches minimum on a bounded interval.

[goerz]

Hmm… okay, but you're very explicitly translating the Netlib code, which doesn't use that name 🤷

What I mean is: it very much seems like the name was chosen because it passes the automated name check, not because it's the most appropriate name for the package. But, I feel like a package name that indicates that this is the exact fmin function from Netlib would be a better route

[kagalenko-m-b]

it very much seems like the name was chosen because it passes the automated name check,

No. The name was chosen because matlab’s “fminbnd” is better known than netlib’s “fmin”.

[goerz]

Then you should rewrite the README to explain this. But I'm not sure it makes sense to say "Literal rewrite in Julia of Fortran routine fmin from Netlib." (which this is!) and then not have the function named fmin

[goerz]

I would also point out that your signature does not match the Matlab fminbnd function (wrong order of arguments), but it does match the signature of Netlib's fmin

[jaakkor2]

[noblock] Same Brent's method seems to be available in Optim.jl, see https://julianlsolvers.github.io/Optim.jl/stable/user/minimization/#Minimizing-a-univariate-function-on-a-bounded-interval

[kagalenko-m-b]

Same Brent's method seems to be available in Optim.jl, see https://julianlsolvers.github.io/Optim.jl/stable/user/minimization/#Minimizing-a-univariate-function-on-a-bounded-interval

Optim is a very large and complex package with many dependencies. For simple problems it may be desirable to have a standalone alternative.

[adienes]

rule 8 of package naming guidelines:

8. Avoid using a distinctive name that is already in use in a well known, unrelated project.

fminbnd is

• distinctive
• in use in MATLAB which is well known and
• "unrelated" because you are not using its implementation

also rule 4

4. Err on the side of clarity, even if clarity seems long-winded to you.

I think NetlibFmin seems much more clear to me what this package does given your testimony that this is a 1-1 rewrite of that function.

naming aside, and I'm sorry to be too critical, but I would be a bit hesitant to say that this meets the quality standards to register in General in the first place. This package consists of a single function that is completely illegible. Could you at least try to factor out some of the @gotos into traditional control flow? compare to for example scipy's implementation

[goerz]

I wouldn't come down quite as hard on this. I think if it implemented the same interface as Matlab and was documented as such (instead of a translation of the Netlib code), I'd be totally fine with Fminbnd.

Likewise, the way it is implemented now, NetlibFmin would be totally fine. I'm saying this coming from a Fortran background where Netlib is the go-to if you need some code that you don't want to write yourself (most of it being in the public domain). So I don't even object to a Julia package that translates an existing Netlib package, even if it borders on being trivially small in this case.

Could you at least try to factor out some of the @Gotos into traditional control flow?

With the direct reference to Netlib, it's kinda cute, and I don't think I'd mind. This is really as close of a translation of the original code as possible. It would be interesting to also write this with modern (haha) control flows and benchmark if there is any difference at all in performance.

[kagalenko-m-b]

“unrelated" because you are not using its implementation

Different implementation of the same algorithm is “unrelated”?

[kagalenko-m-b]

think if it implemented the same interface as Matlab and was documented as such (instead of a translation of the Netlib code), I'd be totally fine with Fminbnd

kagalenko-m-b/Fminbnd.jl@b52ae0c

[goerz]

Okay, that seems good to me… Just make sure that you retrigger the registration to really make sure that the correct commit gets tagged. Especially since you're using v1.0.0 for this