[ENH] Update pip and conda packages

[kikocorreoso]

Related (#5 #3 #6).

Pypi and anaconda packages need to be updated in order to use lmoments3 with modern versions of scipy (see #6) and Python (see #3 and #5).

Only the people with permissions on pypi and anaconda could update the packages.

If the package is not maintained I could fork and maintain it but it would be better to keep the work in the original repo and channels (github, pypi, anaconda,...).

[neilnutt]

If you were to maintain, what is your preference? I can give you admin rights here but I don't think I can at pypi or Anaconda.

[kikocorreoso]

Hi, to update PyPI and anaconda we should have rights to upload packages on those channels:

• https://pypi.org/project/lmoments3/
• https://anaconda.org/OpenHydrology

or ping @faph to help on this.

If @faph is not available I think we could try these options:

• we could upload lmoments3 to the conda-forge channel but on Pypi we should change the name of the package to avoid conflicts (this would be a mess -> old lmoments3 on openhydrology anaconda channel, new lmoments3 on conda-forge anaconda channel, new lmoments3_name on PyPI, lmoments3 repo on github,...).
• We could fork lmoments3 and continue development (mainly maintenance) on other repository using a new name (lmom3 , py-lmoments3, ...).
• Other options...

Let's see if @faph can help on this or try to decide one of the potential alternatives.

@neilnutt Thanks for taking care on this.

[kikocorreoso]

It seems @faph is not available.

cc @neilnutt @monocongo

I have a lmoments3 dependency on scikit-extremes. It seems @monocongo also has this dependency.

I could just ship the necessary files from lmoments3 to make scikit-extremes work as it is a very tiny lib (with attribution and license).

But I think it would be better a fork.

Should we fork this repo, rename the library and continue the development there so we can update PyPI, conda,...? I think none of us have access or is a maintainer of lmoments3 on PyPI nor conda.

[monocongo]

I gave up on lmoments3 due to weird errors that I no longer get from my own implementations (also based on Hosking etc.). Your idea to fork and rename in order to move forward in the right way sounds like a winner.

[kikocorreoso]

@monocongo

Is your code in a standalone lib or is it integrated as part of other library?

Could your code be integrated in a potential fork of lmoments3 to make a better standalone lmoments estimation library?

[kikocorreoso]

Oh, maybe is this one? https://github.com/monocongo/climate_indices/blob/master/climate_indices/lmoments.py

[monocongo]

Yes, that's my code that uses L-moments. Bear in mind that I don't even pretend to be a statistician and have only a tentative grasp of what's being computed there, but it seems to match very closely with the lmoments3 results without resulting in dubious "invalid L-moment" errors for my climate indices calculations. If someone more knowledgable can go through that code and confirm its veracity I will be grateful.

[huard]

See https://github.com/scipy/scipy/issues/11782#issuecomment-636111912