post-rfc VS bytestring

That's what it's best for, but personally I use it for everything. If I ever get into low-level code I'll probably use Rust though.

You can confirm that parsers/tokenizers is ranked "best in class" here though:

https://github.com/Gabriella439/post-rfc/blob/main/sotu.md

Note that this is ported from here: https://github.com/Gabriella439/post-rfc/blob/main/sotu.md which comes with more exposition.

State of the Haskell Ecosystem

According to State of the Haskell ecosystem, Haskell is THE language of choice for implementing compilers, and THE language of choice for writing parsers. Thus, it is not surprising to see more Haskell projects from those particular categories than from other categories.

This is great for understanding what libraries to use in the Haskell ecosystem: https://github.com/Gabriella439/post-rfc/blob/main/sotu.md

In particular, I got comfortable reading hackage documentation to understand quickly how to use libraries (aeson, megaparsec, mtl, pipes, etc), got comfortable with the ecosystem (this helped: https://github.com/Gabriella439/post-rfc/blob/main/sotu.md), got comfortable with the main language idioms and features (https://smunix.github.io/dev.stephendiehl.com/hask/tutorial.pdf) and got comfortable with simple things that for some reason had confused me before (case, \case, let).

Gabriella Gonzalez has a great doc that is reasonably up-to-date, sounds similar to what you're looking for? https://github.com/Gabriella439/post-rfc/blob/main/sotu.md

With that being said, Python is without a doubt the best option, and I'd also be very interested to read the articles you found that say that Python is not a good choice because it's been the industry standard for a long time now. Data science and machine learning are one of the areas where the Haskell ecosystem is not as strong as other languages, but libraries and tools do exist. There's a great list of Haskell resources by domain here, and as you can see, there are Haskell bindings to tensorflow and pytorch, along with other libraries that support common data science programming.

The documentation is itself fairly funny, for those who don’t care to click ahead:

> This "function" has a superficial similarity to ‘unsafePerformIO’ but it is in fact a malevolent agent of chaos. It unpicks the seams of reality (and the IO monad) so that the normal rules no longer apply. It lulls you into thinking it is reasonable, but when you are not looking it stabs you in the back and aliases all of your mutable buffers. The carcass of many a seasoned Haskell programmer lie strewn at its feet.

> Witness the trail of destruction:

    https://github.com/haskell/bytestring/commit/71c4b438c675aa360c79d79acc9a491e7bbc26e7

If you bring in efficient strings from bytestring, densely packed arrays from vector, and an in-place sort from vector-algorithms, you can bring it down to 275ms (uses 19MB of mem).

bytestring

Note that this release is broken for Windows.

I've opened https://github.com/haskell/bytestring/issues/455 so the situation can be improved. You're very welcome to chime in on the discussion or to contribute some of the missing documentation yourself! :)

Highlights from the changelog:

Well, ByteString in particular should not have an IsString instance in a new report. That's pretty clear by https://github.com/haskell/bytestring/issues/140 : the concensus is that there is no good solution right now, but it should not have gotten an IsString instance in the first place. If a theoretical new Haskell Report 202x includes OverloadedStrings (as it should) to handle string literals analogously to numeric literals, I'd expect it to not give ByteString (which is really just a collection of octets) an IsString instance, with all it's issues and rattail due to the encoding question being implicitized.

Standard streaming libraries. They are being written by people that make the effort to understand performance and I have a hope that they make sure their streams run in linear space under any optimizations. It is curious and unsettling that we have standard lazy text and byte streams at the same time — and the default lazy lists, of course. I have been doing some work on byte streams and what I found out is that there is no way to check that your folds are actually space constant even if the value in question is a primitive, like say a byte — thunks may explode and then collapse over the run time of a single computation, defying any effort at inspection.