One thing he said I absolutely agree with - when you're in a crunch situation, either in a startup or a new high-pressure project at BigCo, you don't pick a tool set with which you're not familiar. Even if it looks very nice. You're going to have enough frustration making deadlines without the additional hassle of discovering the warts (and they're always there) of a new language.
Where I work there are always projects that are a little bit under management radar - things that need to be done but don't need to be done in a big hurry. That's the right kind of project to scratch your chin and say "You know, I've never written anything in whitespace."
Someone once gave me the piece of advice that a project should only have one 'difficult' part. If it's a difficult ___domain problem or a difficult deadline, use an easy language (as in one your already very familiar with). If it's an easy ___domain problem and an easy deadline, then feel free to use a difficult language or framework (meaning one that you'll have to learn while you're doing the project.)
I've been learning & trying to use clojure(script) for ~2 years now.
A few things I miss in every other languages:
- The REPL: I usually don't need a debugger,
- Figwheel: that makes frontend dev a pleasure,
- reagent/reframe: are incredibly straightforward, there's no special syntax, no special constructs, just a few functions, a few concepts (dispatch / subscribed) and you're done,
- core.async: based on a simple channels / pipeline paradigm
But then you get into the practicalities of testing & writing code. Shady special cases[1][2], annoying testing tools[3], useless stack traces. And other issues everybody knows. Simple things that become a 30 min discussion with 3 different libraries and 2 blog articles to go through[4].
To me, there's this hardcore group of clojur'ist that seems to be hyper productive and keep introducing new concepts. And the "rest" (me at least), that are floundering.
You fight with useless stacktraces? Boom, transducers. You deal with buggy test runners? Bam, a new specification library with test generations.
I've started moving my code to ES6 and Go. The community is filled with helpful people, but it seems to me that if you're not writing Clojure professionally already, the language is becoming less and less relevant.
The examples you raised are all - I believe not by coincidence - related to how Clojure interact with the host platform (both Java and Javascript). It simply doesn't shield you from it. As a consequence tracebacks bubble up, and you're expected to know how to debug it, and hopefully use the debugger of the host platform. This is aggravated on JS since JS exceptions mean nothing most of the time.
clojure.spec (launching on 1.9) will allow better error reporting on some cases, but the philosophy of the language has always being about giving you a lot of power vs. shielding you from complexity.
Also, a lot of people being introduced to Clojure from a Ruby background (and others, ofc) are having to learn simultaneously about: functional programming, Lisp, immutable data structures, the internals of JS/JVM, advanced concepts like STM, CSP, and a whole host of new libraries. That is a steep learning hill, specially if one bought the argument that programmers are snowflakes who can't take frustration. Once you're on the other side though, the complex, messy real-world projects become more amenable, because Clojure was created thinking about those. That's what Rich meant by "Simple ain't easy".
IMHO, working at a company w/ a codebase of ~150 projects written in Clojure, the productivity has paid off w/ interest though.
> specially if one bought the argument that programmers are snowflakes who can't take frustration.
I think that characterising a (clearly knowledgeable, well-referenced) account of frustrations - especially ones that do not occur in other languages - with the dismissive term "snowflake" is exactly where the stereotype of the "smug Lisp weenie" comes from.
I have used and loved Clojure for years now (I wrote the patch that made Clojure vectors implement java.util.List, back in the pre-1.0 days). The fact that these problems exist doesn't itself give me existential worries about the future of the language. The fact that almost every Twitter thread in the OP, and the parent comment, contained this sort of dismissive reaction makes me very worried indeed.
If we replicate the community of Common Lisp, we will end up replicating its level of industry adoption.
Peter Norvig, Paul Graham, Dan Weinreb, Edi Weitz, Jans Aasman, Patrick Winston, Kent Pitman, the people from Clozure/Franz/LispWorks/..., and many many others.
And yet, even these luminaries were insufficient to propel significant industrial adoption in the face of a famously hostile/arrogant/condescending community. The same fate befalling Clojure would be terrible.
Exactly my cliché of the Clojure community: a discussion about simple things like testing and tooling becomes a place to brag about some random features of Clojure. Insulting my patience (or the original author) on the way is a new one.
You're just confirming my last point, professional Clojurist? Enjoy it. Others, be warned.
The problems you listed are about ClojureScript. I agree the ergonomics are still behind JVM Clojure, but then again the competition is significantly worse too in the JS space.
And if you end up concluding some other front-end language is better, Clojure is still a really good choice for the back end.
IMO The Clojure Community does care about developer happiness. I think "happiness" is very subject, Clojure makes me happy, i'm a dev, what more can i say on that? As Phillmv made his work about his feelings, i feel its appropriate todo the same
@phillmv blog and talk made me angry and sad at the same time.
I took a minute to try and understand why i felt this way and concluded it was because while his technical concerns were valid, his tone belittled my own experiences. He says he wasn't as productive in
a new paradigm, with a new language, on a tight deadline and because of this the Clojure Community doesn't care about Developer Happiness? He says he isn't talking about he community, but the langauge itself, but languaages dont have feelins so i'm confused on how
his complaint isn't being leveled at the people behind and using the language. From my perspective, Clojure enables people to do things that while possible in other languages, would require a lot more effort:
1. Projects like Onyx and Jepsen where a small team was able to accomplish something beautiful.
2. Property based testing and the awesomeness of ReFrame state charts?
What @phillmv tried todo was very ambitious, and its awesome that he gave it a shot,
but his overarching conclusion seems to put the blame on Clojure and that seems both unfair and unproductive.
Here are his specific clojure concerns and how i feel the community is addressing them in a way that makes me feel very happy:
1. StackTraces are hard to understand: Clojure Spec has been written about a ton in this regard
2. Debugger: Sayid Pro looks really promising, maybe a big company could pick it up and support it? But from my limited perspective things could be better for this story.
3. Startup Time: I don't know if anything is being done to address this, the current wisdom is that you work within the repl and structure your projects so they can be "reloaded" which has lot of nice properties.
Here are His Concerns i think are language agnostic:
1. It hard to Structure large apps : While Clojure applied has some suggestions i think every project is context specific.
2. New learning curve (lots of new ideas like reducers, map, transdecures, macros) : Their are a wealth of resources on all these topics, what in particular was he not getting answers about?
3. JVM requires deep editor integration: I feel this is true of every JVM language.
4. Dependency hell: True in every language? Clojure spec is actual designed to help with this eventually!
So his pain points are real, but thats not where he puts the emphasis. Rather he puts that on his feelings. And as John Oliver has warned us, that makes it easy to trivialize real problems, which i feel is what has been done here.
total agree, I hope Clojure Spec will eventually be applied to most Core Clojure functions, and that information will trickle up to user defined functions. But thats only a hope, and possible based on my limited knowledge of the tool.
However, aren't most languages with the best errors messages Static and Strongly Typed?
My point was more that I think Spec was created to help address the issue. Not necessarily that it was even a good solution.
Where I work there are always projects that are a little bit under management radar - things that need to be done but don't need to be done in a big hurry. That's the right kind of project to scratch your chin and say "You know, I've never written anything in whitespace."