OCaml User Survey 2020 Results
dropbox.comI'm a little disappointed that there's so little analysis on the free form questions.
I can't remember what I answered, so I was looking through some of these free form questions to try and identify my comments. I kept thinking something had to be me because it's word for word the same thoughts that have gone through my head 100s of times as a long-time OCaml user. Unfortunately, it feels like these are getting ignored because there's so little commentary on them (though I doubt that's actually the case).
I'm also a longtime Ocaml user, albeit a sporadic one.
I think the language is potentially becoming interesting as a statically typed data analysis platform. I don't see any good competitor in the area. I would love to be able to perform longer analyses in a statically typed platform, and transition to a production system without having to change languages. I think lots of people have similar needs.
The OWL libraries are surprisingly good despite having very little developer mass behind them. Of course the whole ecosystem is tiny in comparison with Julia, let alone Python. I wonder if large users like Jane Street have their own private Ocaml implementations of the equivalents to NumPy, SciPy, Matplotlib, Pandas, Keras.
Dunno, but one of their employee (Laurent Mazare) is actively making OCaml bindings for various data science libraries (such as https://github.com/LaurentMazare/ocaml-torch or https://github.com/LaurentMazare/ocaml-matplotlib)
reasonML / reasonML + bucklescript amount to about 25%, it's been out for 4 years while CAML was out for 30+.
Yet the wording makes it sound like it's disappointing "used by only 25%", "maybe the survey was not advertized enough in the Reason community".
Actually, the fact that someone can show up with a transpiler that makes the language curly braces friendly and get 25% market share just like that is just puzzling.
The majority of the Reason/ReScript community for sure did not fill out this survey. There is a large selection bias for OCaml. One of the obvious indicators: 191 of the respondents work in banking/finance sector.
So? Bucklescript was invented at Bloomberg IIRC
It was created originally at Bloomberg but I've never heard of that level of adoption internally in BBG. Nowadays from what I've seen the highest adoption is at various web devs in Europe and US.
Ya it's crazy how syntax can impact popularity like that. ReasonML is the same compiler, sample semantics, just a different coat of paint and it's already gaining more popularity than OCaml in the same timespan.
Have to consider the context though, I’d point to it being a front end NPM compatible language from the same FAANG that React is from (vs a language from French academics that’s used in less popular domains) before curly braces.
Elm uses a more traditional ML syntax and AFAIK is more popular than Reason, though that’s complicated by it not using React and both of them being dwarfed by Typescript.
A couple of the questions were multiple-choice, e.g.:
> If I was granted one new language feature today, I would ask for:
They commented about it:
> This was a tough question because only one choice was allowed, and the order in which the choices were listed may have influenced the replies. In retrospect, multiple choices presented in a randomized order would have been better.
This is actually one of my pet peeves about many surveys (and indeed, electoral systems)--they often use single-choice answer format (plurality voting i.e. radio button) when they should really be using multi-choice format (approval voting i.e. checkboxes). I said as much when the results were initially announced:
> There’s no reason to restrict the question to only one option, and allowing multiple choices (i.e., approval voting), allows people to express preferences in a more nuanced way so you get a more accurate overall view of what the community is interested in.
(from https://discuss.ocaml.org/t/ann-ocaml-user-survey-2020/6624/... )
I wish they would have collected more demographic information.
It would help both to determine how representative the survey was of the greater OCaml community, and also to ascertain the diversity of the community itself. Being rather niche, there is certainly going to be underrepresentation of certain groups.
It would be helpful to be able to compare the results to similar data covering the demographics of other language communities like Haskell and Rust, and the wider software/tech field.
I also feel like the distribution methods and sample size of the survey likely skewed the results significantly. The survey window was relatively short and the choice to distribute the survey through discuss.ocaml.org and some mailing lists definitely missed a lot of other segments of the community.
I find it funny that one of the questions asks where people interact with the community, and a large number of respondents answered discuss.ocaml.org. Well of course they do! That was one of the few noted places the survey was advertised!
In the end I agree with one of their conclusions:
> It would be ideal to delegate the formatting of the survey and the summarization of results to professionals.
But I would add that ‘distribution and administration’ to that list .
minor nit pick:
"Which types of software do you develop with OCaml?"
Others: .... Object Storage. object storage.
I think these 2 can be joined ;)
There's plenty of duplicates like this in the open-ended responses. They just printed them all as-is.
That's a (sanitised?) javascript error.