BatList.n_cartesian_product

[wistery-k]

Is there any reason for
BatList.n_cartesian_product [] = assert false
?
I think [[]] is good (or at least invalid_arg) instead of assert false.

[gasche]

Thanks for your report. It is clear that assert false is wrong: assertion failures are to be used in cases that cannot happen, and in particular it should never, ever be sent back to the user. This is clearly a bug.

After a bit of mathematical thinking, I think the One Right Answer in this case is [[]]. Would you care to provide a patch to change this behavior? BatList.n_cartesian_product currently has no tests at all. In your patch, could you write some unit tests verifying the behavior in the general case, for a list of singleton lists, for a list of a single list, and for the empty list?

[wistery-k]

I sended a pull request. (I'm a github beginner. I'm sorry for anything bad..)

[gasche]

Your contribution closes this bug. Thanks a lot, and I hope you'll find other bugs :-)

[murmour]

True story:

In 2012, young Max Mouratov wrote a bunch of code that used n_cartesian_product for "give me all combinations of items from the given lists, so I can do something useful to each of them".

In 2013, young Gabriel Scherer found the One Right Answer to "what should n_cartesian_product [] be", and wrote a deep blog post about it: http://gallium.inria.fr/blog/on-the-nary-cartesian-product/.

In 2023, the bunch of code that young Max wrote in 2013 suddenly broke on a particular input, and the now old Max discovered that n_cartesian_product [] returns [[]], and not [] as the young Max had expected (by intuition). But [] was not "something useful" in all the code that young Max wrote, and the old Max sympathized with the young one.

Then, the old Max read the deep blog post by the young Gabriel, and his IQ increased by one point. He understood the secrets of the universe, and the whole truth unfolded before him.

The blog post ended with a bit of infinite wisdom:

No theoretical considerations make as strong an argument as this one: the right answer is the one that makes the code elegant.

So, the old Max pondered about it for a while, and his mind got illuminated by the brightest idea ever. He copied the implementation of n_cartesian_product from Batteries and replaced [[]] with [] inside it, because it was an elegant and pragmatic solution that spared him the need to scatter if l = [] throughout his client code, across all of his programs.

There is no morale.

But the old Max from 2023 suggests that perhaps it'd be wise to document this corner case in the docstring of n_cartesian_product, because its behaviour is not quite intuitive, and not quite obvious to those programmers who aren't type theorists. He also wonders if the old and wise Gabriel from 2023 has any new thoughts on the matter now.

[UnixJunkie]

If you send an improvement for the ocamldoc of this function, I will gladly merge it.