Developer surveys - hackerrank and stackoverflow

Nice to see Julia make it into the hackerrank 2019 survey. 10% of the respondents say they want to learn Julia in 2019!

Also, for those who missed it - the Stackoverflow 2019 survey is open.



Liked vs hated chart. Julia is the most hated programming language! Wonder why? Compilation latency?


I believe this could be related to the situation around the 1.0 release, where a lot of people tried out the language and found out lots of basic functionality wasn’t working (as packages had been broken by the update).

Well, last year’s SO survey it came up as one of the most liked and most hated. And that happens for pretty much anything new.

[Do you like the newest blockbuster film? Everyone likes it? No you think it it’s overproduced, but just by judging the trailer and you haven’t seen it? Oh…]

Also, if you couldn’t tell by taking the survey and the questions it asks, these surveys skew towards web development and database handling, both in terms of questions and people’s interests. Those don’t tend to be the strongest parts of Julia, so if the average person of this survey’s audience tries Julia, they would probably be like “wow, Julia’s web development frameworks are so underdeveloped compared to Python and Javascript! I can’t believe how much I have to do myself!”. But, well, yeah… Javascript’s ODE solvers suck.

1 Like

They explicitly ask on what language have you worked with that you would not like to work with again, don’t they?


Most dreaded means that a high percentage of developers who are currently using the technology express no interest in continuing to do so.

Basically that measure is just, if you used it for a bit this year, do you plan on using it next year? And you see Julia on there next to Haskell, MATLAB, and R as things that people tend to pick up and try for fun, but really only stick with if its use case really fits their day job. All I see in this is: because of the audience, people like to play with a mathematical language for a bit one year, but then plan to continue to use Rust, Kotlin, TypeScript, JavaScript, etc. (top of the loved lists) because their day job is a front-end/back-end developer. Python shows up with these languages because the vast majority of people using it are doing so for front-end/back-end work: the proportion of people using Python for ML (and actually the proportion of developers doing ML generally) is very very small.

Audience is crucial. To me, the audience that I would look at is at academic conferences. There, MATLAB is quite dominant (in applied math), with C++ and Fortran next (lot of legacy work, and this is where a lot of the real software is), and Python and Julia having similar numbers and both growing. But Julia growing quite rapidly in these circles, especially with the recognition it got from the Wilkinson Prize. For me, that’s the trend to watch.


This seems to claim that Python’s growth is explicitly due to ML/DataScience:

It’s growing because of it, but it’s still not close to the majority. The “Estimated Categories of Python Developers Over Time” chart shows that it had an astonishing growth, going from 1% of SO users (about 1/9 of Python SO users) to 5% of SO users (about 1/3 of Python SO users). Now the number of Python users who mainly view the pandas/numpy/matplotlib tags is equal to that who view the webdev ones, with other development closely behind which is how it ends up at 1/3. That’s an impressive growth of ML, but still not close to a majority.

What’s interesting about it to is that, when looking at the tags, the majority of the people in this group are not actually doing ML, but rather are plotting data from dataframes. The percentage of traffic to “machine learning”, “neural-network”, and “scikit-learn” tags pales in comparison to “pandas” and “matplotlib”. While the Tensorflow tag shows growth, it’s still not even as large at matplotlib. So, plotting got popular.

1 Like

It doesn’t matter so much if Julia is mainstream or not, because it already has a fantastic community and some of the best developers for math and science are developing for it. It is well known that Mathematics is the queen of the sciences (and Number Theory) but only a very small percentage of people become mathematicians, let alone number theorists. Does this unpopularity make Mathematics any less credible? No, math continues to be one of the most credible and long lasting subjects. Likewise, I am confident that Julia is growing healthy with its intended audience.


Nay, it is differential geometry :wink:

1 Like

Second this as someone working in a consultancy with a “Data Science” team - while there’s lots of chatter about ML/AI, in practice it mostly turns out they derive summary statistics from large-ish data sets and plot the results - so Python ends up being a fancier version of (or being combined with) Tableau/Qlikview/PowerBI


I think it’s fairly well established that Biology is the queen of the sciences?


I’m actually combining the study of both differential geometry and number theory with physics, so I agree with you on differential geometry (i’m making a package for differential geometry called Grassmann.jl), but you can only get so far without number theory. Number theorists have some of the coolest ideas, completely overlooked by other scientists.

1 Like

It is clear that you should start non-representative web survey about this, the result of which we should then over- and misinterpret :wink: