Indeed, the TIOBE results fail the eyeball test so badly and in so many different ways that it serves better as a tool for optometrists.
6 Likes
It seems to be based on such (“Google, Bing, Yahoo!, Wikipedia, Amazon, YouTube and Baidu”) queries:
“<language> programming”, also other queries such as “programming with <language>”, “<language> development” and “<language> coding” should be tried out.
So it’s some measure of mindshare (not actual use, nor “about the best programming language”), likely explaining assembly so high (lots of stuff mentions “assembly programming”, even though most people don’t do any).
Job listings are of course not the only metrics we look at in Spectrum. A complete list of our sources is here, but in a nutshell we look at nine metrics that we think are good proxies for measuring what languages people are programming in. Sources include GitHub, Google, Stack Overflow, Twitter, and IEEE Xplore.
Julia there listed 28th, just after D (which we also use at work, mostly).
It just occurred to me, TIOBE seems to search for e.g. “Julia programming”, and it doesn’t find Julia discourse posts.
I tried to Google for a random post I found (and found it):
How to Plot Step Functions Correctly in Julia?
but not with:
“Julia programming” How to Plot Step Functions Correctly in Julia?
@logankilpatrick or whoever is responsible for Discourse (and Forem etc.), so can anyone add an invisible meta-tag (or needs it be visible, e.g. in a footer?) to all of Discourse? It seems appropriate.
Can people or need ask google (and bing etc.) to reindex your/this site?
2 Likes
I believe @viralbshah and others have looked at this before to make sure we are good but I’ll take a look again in a few hours.
6 Likes
I double checked this and it looks like all the Meta tags already say “Julia Programming” so we should be good.
There’s also “Julia programming language” in the page footer.
1 Like
But it is strange that a google search for "Julia programming" "Step Functions" doesn’t pull up that page. Maybe the footer doesn’t appear in the javascript-less version Google’s crawler sees? (edit: this is indeed the case)
9 Likes
While I share the exact same thought, the TIOBE Index is still very popular – at least the people I know of tend to use it as a measure for the success of a language. That being said, I still hope that we reach Top 20 soon since I am pretty sure that it will have a booster effect. No matter how bad the algorithm is and how often we point out its flaws.
3 Likes
Googling this doesn’t work:
“Julia language” “How to Plot Step Functions”
nor with “Julia tutorial” (while it’s a solved question so it kind of is) but "Julia programming” and "Julia programming language” works. I suppose that’s enough for TIOBE, but not sure for others, and I think at least “Julia language” additionally should be in meta tags, now that a new index is coming from them and others, in October, maybe not too late to ass meta tags?
I’m not sure it’s justified to add “tutorial” too as a meta tag (here, in general, possibly it can be done for solved questions conditionally?). If you do write one, or a blog, and it’s a tutorial please be specific about it.
At PYPL PopularitY of Programming Language index
Why do you use tutorial as Google Trends keyword ?
The following principles were used:
- just the language name, e.g. Python, would lead to inconsistent results, because Python has many other meanings;
- the same phrase should be used for all languages, for consistency.
- programming would not do : C programming is used much more than PHP programming, because PHP does not need the qualifier
- tutorial is a word used frequently by developers learning any new language : it makes a good leading indicator. What is a python tutorial, if not a tutorial on the programming language ?
- ios tutorial is used instead of the (less popular) objective-c tutorial to better reflect what iPhone developers search. This is confirmed by an analysis of language tag followers on StackOverflow, or of the visits of the Wikipedia language pages [s] : they give a 4-5% popularity to Objective-C, which is highly consistent with our estimate. It may flatter objective-c’s market share because it includes queries of iOS end users.
It does now, some someone changed something (post though from July 12th, and solved by an answer from same day), or just passing of time helped.
Right now Julia is 25th (0.28 %) at PyPL Oct index, but 12th in the UK (2.44 %, up 4 places, trend “+2.2 %”) vs. Matlab there 25th (0.24 %, trend: “-1.7 %”) so IF the index (and share) is meaningful as a “leading indicator”, then Julia 10x more popular there than Matlab (which dropped 15 places, since when?, and Lua up 9 places to 9th). And R is 5th just above “C/C++”… And Python has “-15.4 %” trend, though still on top.
To be fair, PyPL doesn’t have consistent ranking over time either, but (for Julia) it may actually be very understandable, (exponential?) smoothing would ideally be used for the seasonal fluctuation. And the rank is always going to be more unpredictable than plotting the “share” itself (or “ratings” at TIOBE).
It’s most obvious if you choose UK, and “show your favorite languages” by choosing Julia from the drop-down menu. It might indicate student use (or workplace), less in the summer. Note the graph is logarithmic, and the bottom cut off.
At least, on this leading indicator, we’re past PHP (and Swift, Haskell and Rust) which was, still is I think, the most common web programming language used (e.g. because of WordPress), though maybe not for new projects. I believe even on this index PHP was highest rated of all languages, and we’re also past where R was in 2010 (in the UK, and similar timelag for the US).
It’s unexplained why Matlab is falling off a cliff, in the UK, not elsewhere.
Hopefully Julia makes top 20 of TIOBE index, then we can get such a graph there.
1 Like
The most fishy thing about that PyPL chart is not Matlab falling off a cliff, but Python falling off a cliff. In just one year (autumn 2021 - autumn 2022) Python fell from 40% to about 25% (note the Y axis is logarithmic). UK only, but that still seems completely unreasonable to me. Unless there’s a simple explanation like a plotting bug or major change in methodology, I think the conclusion must be that PyPL is best ignored for individual countries.
I’m a R user for many years.julia is used in a similar way with R. But the market of R is large while julia is so small.Too little work is done is done to substitute R with julia.
search <space>R<space> and see if those experience applies to you, if not, keep using R, nobody said about substitute R; in fact, we almost never substitute languages: COBOL, Fortan still largely get run, instead, the world just find new area of computing and use new languages there.
I agree Julia could be a good substitute for R. Unfortunately the statistics libraries in Julia are fragmented with people writing packages to meet their own needs rather than building a composable suite. For example Vcov.jl supports FixedEffectModels.jl but not GLM.jl, and Effects.jl does the opposite, so if you need the features of both you have to roll your own. I’m not sure what a way forward would be.
my English maybe poor, however, to substitute is not to kill, you may misunderstand.As jar1 says, if more convenientlibraries and a well designed framework is built, it can attract more R users.
I have read some source code of FixedEffectModels.jl, So which pkg is the most recommended for learning?
While Python did top out at 41% in UK, I do not find it very credible as a share. If it’s wrong, then the “fall” is less.
If any language has 40% share (or of tutorial searches strictly), it means the rest, all combined, have only 60%, which is only 50% more. Can one language be that successful to have 40/60 = 66% the use of all languages combined? Since Python is commonly taught as a first language, I find it likely that people look for a Python tutorial more than when learning their 2nd, 3rd language. I.e., I think the top language is biased to be at least a little higher, and then the rest lower.
In France, Python topped out at 41.2% (now at 26.15%), in Germany at 36% (now 27.18%), in India 35.7% (now 25.77%), in the US 35.6% (30.46%) and worldwide 32.1% (now 28.3%), So the “fall” from the top is by 37% in France, down to 28% in India (not exactly a small country), and (lowest for any single country) in the US 14%, still the worldwide average “fall” is even lower but at only 12%. Likely the fall is real, or at least the overestimation was real. Overestimating the top language isn’t too bad, since it can’t go higher than rank 1…
How does TIOBE 15.74% “rating” square with PyPL 28.3% “share”? For TIOBE it always has 50 languages with ratings (plus “The Next 50 Programming Languages” unranked), so it seems to me the expected value rating would be 1/50 = 2%. That currently means only top9 higher than that…
PyPL has 28 languages, but sometimes 27, down to 25 for the UK, and 23 for France. Unexplained why not always the same number. There’s some cut-off I guess, and in part if dividing by 28 or (sometimes) 23 would matter. That alone is 22% difference. 1/28 = is 3.6% expected share (and 1/23 would mean 4.3% expected) vs 2% for TIOBE. The exact numbers for “share” or “rating” might not mean too much or be meaningful, mostly I guess meant to then rank the languages, and the rank will likely be relatively stable at the top, but less so way down, or lower you go.
I tried to look into it:
Note, Google trends normalizes to 100 (for the whole period), so Python is higher the week Mar 21 - 27, 2021 relative to the other week, while the share I calculated is lower:
100/(100+58+23+32) == 0.47 # Mar 21 - 27, 2021
88/(88+51+22+23) == 0.48 # Nov 21 - 27, 2021
but I only calculated for the top 4 languages. If you add more and some of them (or strictly the rest of them combined) had a good week) then Python’s shares will be lower.
It seems the data might be consistent, but it’s tedious to calculate. Also of course I doubt 40% share realistic (even for “Python tutorial” share, which likely is done without the quotes, that gives lower absolute numbers)?
I tried looking at Google trend to compare with PyPL, a long story, but I’m not sure there’s anything wrong with the plotting.
I tried to look at the code and issues, and found one intriguing: Rename "C" to "C/C++"? · Issue #2 · pypl/pypl.github.io · GitHub
“C++ has the same popularity as C on Google trends: to avoid duplication, it is not included in the PYPL index.” According to this, and to minimize confusion, i think it’s more proper to rename “C” to “C/C++”.
I think we should keep this thread about TIOBE and other indices, tracking them and discussing them, but not if Julia should be higher or R
In short, Julia could substitute (or also work with R or Python), the libraries/ecosystem could be improved, possibly larger libraries made that encapsulate other smaller. And Statistics module will no longer be part of Julia as of 1.9, and it will improve it (let’s not discuss it here, it seems like a done deal, and there’s a different discourse thread on it), for now done as:
(@v1.8) pkg> add https://github.com/JuliaStats/Statistics.jl
1 Like
Julia is 22nd at PyPL Nov 2022, up one place since a year ago (though trending down in recent months). 13th in the UK (up 5 places in a year, its highest ever, in case reliable) and 18th for the US:
There’s a question how reliable that index is, at least for individual countries, at least the smaller ones:
For Germany I got curious, why is Swift 12th and Scala following, both shooting up the ranks, Scala by 15 places. But they are only recovering to their previous places, and while in the dip a year ago, Julia was actually ahead of both…:
And why is Lua 4th in France, almost caught up with JavaScript, ahead of Go, and Kotlin next (and Scala briefly ahead of JavaScript a year ago)?
First the good news for PyPL Dec index, if reliable (this is meant to be a “leading indicator”)…, Julia is 9th in France at 3.95% (going up 12 places! or +3.4%). Is just after “C/C++” that still went up 6 places (if they were separate likely would be ahead of both, there!). Both Typescript and Matlab are up 11 places, and ahead of C#, PHP, and C/C++ there…
Maybe we shouldn’t celebrate, or celebrate while it lasts…
[While the share percentage goes up (and down, before in e.g. France), I think the ranking is less stable, so I write down both so I can track over time.]
In the US, largest country tracked (besides India), and a leading one we want to keep track of, Julia is up 6 places to 19th in the US (0.66% +0.2%) compared to R in 4th place (up two), head of C# and C/C++… while worldwide Julia is 27th (up one 0.24% +0.0%).
Julia is 17th in Germany (up two 1.69% -0.2%). In the UK 22th (down one 1.37% +0.6%).
Sadly (for Viral Shah) lowest rank for Julia is for India 27th (down one 0.17% -0.1%).
Now can anyone confirm Julia getting popular (in France), e.g. with job offerings there or subjective feelings, knows being taught in Universities or something?
Unlike at PyOPL where Java is (still) 2nd place:
For TIOBE Dec index where Julia is 2 (0.52%):
C++ surpassed Java for the first time in the history of the TIOBE index, which means that Java is at position 4 now. This is the first time that Java is not part of the top 3 since the beginning of the TIOBE index in 2001. Apart from all this, we see that Kotlin and Julia are getting closer and closer to the top 20.
Unlike for PyPL at TIOBE C and C++ are considered separately (as they should?!), if not, Java could be 3rd, but also for a long time now.
1 Like