Hacker News Reader

I know some Haskell, but don't find code like this to be very readable:

    offsetParams :: M.HashMap ByteString NonnegRange
    offsetParams = M.fromList [(toS (replaceLast "limit" k), fromMaybe allRange (rangeGeq <$> (readMaybe =<< (toS <$> v)))) | (k,v) <- qParams, isJust v, endingIn ["offset"] k]

Not because it's something I don't understand at the concept level (there isn't anything complicated there), but because of the formatting and general lack of comments.

Maybe I'm talking about bike shed color, but it could be much better just by having some extra newlines that separate the pieces of list comprehension expression.

The "blub paradox" argument isn't a very good argument. It's effectively a gradeschool style ad hominem wrapped in fancier clothing.

Also, I don't think his criticism is of the "blub" mold anyhow. It seems to me more like it's criticism of the particular codebase's documentation, although it does seem to stretch into a hasty generalization about "all Haskell".

I don't agree that blub is an ad-hominem,or even a fallacy.

Blub basically says "you need to be proficient in X before you can see value in X, this if you are not proficient in X you won't see it's value". You can disagree with the statement, but there's no logical fallacy there.

> languages with other features/language constructs than you're used to just seem worthless, complex, without value and foreign to you.

Document the Haskell code and not stuff so much logic into one line. I think that's a fair thing to ask for in software engineering.

Anyone who complains about Haskell seems to have either a.) not programmed it b.) weren't programming "correctly" c.) a xenophobe/blub/NOS

I know what lambdas are, I know there are functions and folds. That in and of itself doesn't tell the story of the data that's actually being manipulated and the expected output. So that's why you have a 20-30 line comment block explaining what's going on in important API functions.

You could program python, JS or ruby similarly; wrapping multiple functions on the same line until you have to actually trial each function to figure out why your not getting an expected result. It feels clever, at the moment, but we don't do it because we're not helping colleagues understand it.

> You're not willing to learn the language because you don't understand it, and dismiss the language as having no value because you don't understand it. Is that really fair without at least having a basic understanding of the language?

The commenter was heralding the project as a "great example" of "real-world haskell" and recommended contributing to it. I was shocked to see 8k stars and a paltry 50 contributors; of which, only the top 10 have contributed over 100 lines of code. That's half the contributors, proportionally, to something like neovim or redis, and even worse compared to node.js projects.

And, I looked at the code of the project, which is open source, and didn't see code documentation.

I'm trying to be generous. My advice: stop packing the code so close and document it, probably more than normal, since the symbology in their is highly dense.

Here is a project with about the same number of stars/contributors/contributors with more then 100 lines/time since release

https://github.com/graphql/graphql-js/graphs/contributors?fr...

:) So by your logic, i could say:

"If this considered a real world javascript/node, it's no wonder there's not a lot of people using it." which is obviously wrong.

You are using the wrong metrics to judge the language popularity. Virtually every OSS project has only a handful of core contributors.

PS1: Note that i am not arguing PostgREST is a great example of real world haskell, it's not for me to say since i have bias :)

PS2: You are right that the source code is not commented and it could use a little of that but what you are missing (because you are not used to haskell) is that a lot of the documentation for haskell projects comes from the types and type signature of the functions. So just by looking at that, in a lot of cases, a haskell programmer can reliably tell what the function is doing without any documentation

It's indefensible to have code that tightly compacted, throwing functions around and not documented.

Don't care if it's Python (basically pseudocode), Javascript, Haskell or Erlang.

For all the time Haskell commenters turn criticism into a Socratic dialog, they could be considering ways to make their code digestable. It's been what, 20 years now?

Not just to non-haskell programmers, but all the haskell programmers I see swarming around threads like these on the internet; you'd think they'd actually be teaming up and collaborating with each other. I can't put my finger on it, I think there's a trend, a feeling, that it's not suitable for programmers focused on achieving business ends. It's more of a hobby thing

Programming in the enterprise means confronting the reality that there are design trade-offs and someone else has to read the code. Haskell programmers seem to go off the deep end trying to pull off a hack of making haskell work a certain way, then its "pencils down", sayonara. The real world doesn't accept that, programs have to be maintained by others.

I have a (non-contributor) experience with the PostgREST codebase. It's true PostgREST codebase is not great (lots of long functions, need some work to extract the main program into a library etc.) but the codebase definitely is easy to navigate and modify. For a transient project at work I've asked a dev to evaluate how much work it would be to port PostgREST to MSSQL. My dev had a working proof-of-concept in a few days. The bottleneck was not navigating and patching PostgREST codebase but dealing with MSSQL. We decided to go another route only for the lack of a native MSSQL driver in Haskell: there's an ODBC driver (which my dev used for his PoC) but I don't really trust ODBC+FreeTDS. Hence, we'll write a good-old C# binary with a Microsoft-blessed driver to expose the few ad-hoc queries our own team needs rather than a generic solution for the whole company and we'll move on. In short, I have some data invalidating your argument about the PostgREST code style being indefensible. Although I don't enjoy PostgREST codebase I love the product because it delivers what the website claims. Rather than diminishing the hard work of devs using a throwaway/slandering account, I just reckon that Haskell and "ship it culture" are reconcilable.

ODBC+FreeTDS is not generally a good idea. Microsoft's ODBC driver works fine though.

https://blog.afoolishmanifesto.com/posts/install-and-configu... explains how to get it going on debian.

I also did something similar to postgREST for MSSQL in java, not in haskell because of the same driver issues you mentioned. I managed to implement jwt auth pgREST style taking advantage of MSSQL user impersonation though I did jumped into some limitations, for example the lack of enviroment variables in transaction scope in MSSQL forces you to have to declare one db user for an app user.

I have little experience in Haskell despite that it was easy for me to understand the gist of what postgREST does and port it to java.

Some lines are indeed atrociously long. No Haskell programmer I know would advocate this. I think the project maintainer should insist on sane code formatting. All the critical lines I checked were from contributors. Other than that, I think the case you are trying to make here is largely imaginary. That is not to say that the project couldn't use some more source level documentation, but that you do not see how not actually knowing the language in syntax and semantic and never having written any code in it could possibly cloud your judgement on the comprehensibility, pretty much says all about your case, if you ask me. (I hope that sentence wasn't to long, if you ask nicely I'll add some documentation)

It's not like putting the code on one line. It's like complaining about using map instead of a for loop but you don't know the abstractions.

There's a line of too much abstraction, and lots of.haskell code crossed it, but the postgres code ive looked at doesn't seem to suffer from it.

Also I really hate the types are documentation people, but I mean it's hard to argue with it for asking about what transformation is happening.

Edit: also if be surprised if those numbers are all that bad if you consider the size of the codebase but yeah it's hard to get contributors to a Haskell code base.