Julia ranking trend, TIOBE, RedMonk


I collect some historical data of the Julia language ranking after it first enters Top-50 in Sep, 2016. Hope It could be helpful to understand the popularity of Julia.
(Data complement and correction are welcome )

================== Part 1: TIOBE Rank ==========

Month Rank Ratings Important Event Note
Sep, 2016 47 0.196% Release V0.5 First time entering Top50
Jun, 2017 Release V0.6
Nov, 2017 0.600%
Dec, 2017 47 0.439%
Jan, 2018 47 0.226%
Feb, 2018 50 0.189%
Mar, 2018 37 0.301% Highest rank till now
Apr, 2018 44 0.195%
May, 2018 46 0.342%
Jun, 2018 43 0.281%
Jul, 2018 >50
Aug, 2018 50 0.156

:grinning: As the marketing/percentage of Julia is still small, the ranking would not be so robust and fluctuation is normal.

The ranking and rating value graphs are as follows.

Figure 1 Julia Ranking (TIOBE)

Figure 2 Julia rating value (TIOBE)

================== Part 2: RedMonk Rank==========

The following table show the Julia Rank in RedMonk.

Quater Rank Important Event Note
2014-Q1 62
2014-Q2 57
2015-Q1 56
2015-Q2 52
2016-Q1 51
2016-Q2 52
2017-Q1 ? Data not found
2017-Q2 ? Data not found
2018-Q1 ? Data not found

Figure 3 Popularity trend according to GitHub and Stack Overflow
(Noted: the data is measured manually from the following figures from RedMonk. )

================== Part 3: Report from RedMonk ================

(a) 2013 - January

(b) 2013 - July

© 2014 - January

(d) 2014 - July

(e) 2015 - January

(f) 2015 - July

(g) 2016 - January

(h) 2016 - July

(i) 2017 - January

(j) 2017 - July

(k) 2018 - January

Figure 3 The graph of the ranking from 2014 (RedMonk)


So Julia is out of top50 in July? :confused:

Here are my additional/correctional data (it seems that you mistakenly copied data from February into April):

0.7.0-beta.189> df = DataFrame(Date=[Date(2018,4), Date(2018,5), Date(2018,6)], 
                               Rank=[44, 46, 43], 
                               Ratings=[0.195, 0.342, 0.281])
3×3 DataFrame
│ Row │ Date       │ Rank │ Ratings │
│ 1   │ 2018-04-01 │ 44   │ 0.195   │
│ 2   │ 2018-05-01 │ 46   │ 0.342   │
│ 3   │ 2018-06-01 │ 43   │ 0.281   │


Thanks! I have just corrected the data. In July, the index is out of top 50. Maybe, the decrease in July could be explained as mentioned by @ScottPJones :grinning:


One has to add that the TIOBE Index is calculated from the amount of search engine queries for the respective language. So if lots of people have to google in Stack Overflow about answers to programming related issues then that would boost the ranking. If, on the other hand, people get mostly along with a good included documentation, then the rating would suffer.
That being said, a placing among Crystal, Elixir, Elm, Euphoria, Forth, Groovy, Hack, Icon, IDL, Io, J, Julia, Korn shell, Ladder Logic,… does not IMHO reflect truly the growing popularity of Julia among scientific users.


Agree, Julia is quite mature for scientific computing now. The convenient parallelism and distributed computation attracts me to use Julia to replace some of the work previously developed using Python. Some libraries, such as JUMP and LightGraphs.jl, are very useful for me.


We should clearly make the documentation worse. :wink:


BTW name of that thread is suggesting that in November there was more high ranking than your “highest” :wink:


…or just have only online documentation :wink:
You show an important point.The flawed methodology TIOBE uses to determine the “most popular” programming languages. Contrary to that, e.g. Matlab has excellent documentation and is ranked 15.


Maybe even with excellent documentation, it’s hard to figure out how to do something well & performant in MATLAB?

Julia just seems to have a more consistent design (thanks, Jeff!!!), so that most of the time, I don’t even need documentation. It’s only when there are inconsistencies and warts still present, or I have to go back and check because the warts are in v0.6.x, and I have to use that for work, that I need the docs (and usually can find what I need via ?)


This is basically a reductio ad absurdum of the attitude of chasing language index ranks. The TIOBE index in particular has a huge amount of noise: beyond the top ten, the numbers are very close to each other and ranks vary hugely from month to month (applying a Kalman filter to it would be interesting). The best approach is to keep doing what is right for the language and its users—popularity will naturally flow from that and the indices will reflect that.


IEEE Spectrum has an interactive tool [1] that allows you to create your own ranking of languages by combining 12 different metrics (Google searches, Github repos,…) and then picking the type of languages (Web, Mobile, Enterprise, Embedded) that you want to compare against. Using the default weighting and comparing against all languages, Julia was 31 in 2017, 33 in 2016, and 39 in 2015. In 2017 Julia looked best when compared only against other ‘enterprise’ languages in which case if ranking only by Stack Overflow views Julia comes 10th and if ranking only by citations by papers in IEEE Xplore Julia comes 14th. My experience w publishing in IEEE publications is that there something like a 6-month delay between doing the work and publishing, so I guess the IEEE Xplore ranking is more indicative of work the authors were doing in 2016. The people publishing papers in IEEE conferences and journals are a target audience for Julia; I think the ranking is positive sign for Julia’s future.

Here is an image of the 2017 rankings of ‘enterprise’ languages ordered solely by IEEE Xplore citations.

[1] https://spectrum.ieee.org/computing/software/the-2017-top-programming-languages


Yes, I also found the IEEE ranking methodology much better, and with being able to select different weights, etc. actually useful, rather than just being silly entertainment (i.e. TIOBE).


I have added the ranking from RedMonk, which shows the increase trend of the popularity is quite significant.


It would be interesting to see the effect that one internet clickbot could do to the TIOBE since it’s just based on counting searches. Write it in Julia and make a blogpost :wink:


Redmonk rankings are interesting - and also the slope of the line.
That could possibly indicate (idle speculation here) that because Julia has a lot of official and semi-official channels for asking questions (previously Google groups, now Discourse and Gitter, plus even some Discord,
and IRC, nicely bridged), that fewer questions end up going out to StackOverflow.
It could also be an indication that the design of the language is good enough that even as more people are using it for their open source work on GitHub, questions are not increasing at the same rate.


Perhaps take some time out of looking at this data with huge statistical fluctuations and dubious collection methodology to take a look st some spurious correlations that become apparent when comparing time series comprised of a handful of data points. It always makes for fun reading.

(Extra credit for anyone who finds any fun spurious correlations with the Julia TIOBE ranking.)


I did say, it was totally idle speculation :wink: Something to entertain me while waiting for a bunch of unit tests to run!


To me it seems obvious that Julia is going to completely take over in scientific computing field.

It can only get better, a feedback loop is going to cause it to spread like wildfire. The initial release cycle with breaking changes required early adopters to be constantly ready to adapt to changes, but going forward with 1.0 the package ecosystem is going to grow uninterrupted at an even bigger scale, while the base language becomes optimized and embedded into other systems.


I think it’s going to do quite well as a tool for general computing, in a few more years.


4 years ago, when I have started to use Julia, I was sure “it’s going to do quite well as a tool for general computing, in a few more years”.

About one year ago I was expected Julia will be on the “Hall of Fame” of TIOBE Index.

Instead, Julia has disappeared from top 50 TIOBE Index (we can have an excuse as of “statistical fluctuation”) and we are waiting for “few more years”. I guess in 10 years we still will be waiting for “few more years”. To me, Julia is a great language but we need to convince other people in that. May be release 1.0 in “few more years” will help.