Live data from Hacker News

Programming Trends on StackOverflow

dodgycoder.net

51–60 of 64 posts

Re: Programming Trends on StackOverflow

#51

I am a little surprised how comfortably PHP has remained as one of the top ranked tag. It's not consistent with the trend here on HN.

I guess that most PHP developers 're busying with their coding, not viewing HN threads like most of Ruby developers ;)

With a language as unstable as PHP, of course you'll be spending more time on StackOverflow than on HN ;)

Re: Programming Trends on StackOverflow

#52
post #26

As a happy Ruby/Rails dev I can't figure why Python has double the marketshare. I guess it's the Google backing and the history of Linux admin usage?

Python seems to be used in a lot of contexts, not limited to web development. While Ruby is a lovely language it appears to be used primarily with Rails in a web context.

That's right. Python is popular among sysops folks, and the scientific/mathematics people.

Re: Programming Trends on StackOverflow

#54

Interesting data. The founders (Joel Spolsky, Jeff Atwood) are strong proponents of Windows programming, and the initial popularity of .net and c# make a lot of sense in that context. As the site has become more and more popular, the audience now consists of many people who aren't even aware of who those founders are, let alone follow their work. As such, the language spread is much more representative. As someone wh…

I posit that Microsoft has more to do with this trend than anything else.

I was a Windows / Microsoft tools only developer for many many years and only recently switched to Javascript / Node.js and OSS tools for our business applications.

My usage of S.O. almost exactly parallels what you see in that graph over the time period and I'm personally aware of a great many other Windows developers moving away from it recently.

The reason for myself has absolutely nothing to do with Joel or Jeff and everything to do with Microsoft pushing out the smaller professional developers who supported them and by extension Windows for decades in favour of policies that benefit only the very big software shops and students / dabblers and then recently with Windows 8 basically giving us all the final push away in many, many different ways.

As for your last sentence; as a professional software developer of well over 20 years who 'goes near' Windows programming and everything else I can get my hands on I can't fathom how anyone can take pleasure in less opportunity and a smaller market and fewer interesting platforms to work with.

Re: Programming Trends on StackOverflow

#55

Interesting data. The founders (Joel Spolsky, Jeff Atwood) are strong proponents of Windows programming, and the initial popularity of .net and c# make a lot of sense in that context. As the site has become more and more popular, the audience now consists of many people who aren't even aware of who those founders are, let alone follow their work. As such, the language spread is much more representative. As someone wh…

Given StackOverflow's structure and operational methods, one would expect questions about products in the Microsoft stack to decline as searches return more relevant answers. To put it another way, as the site matures, more questions about x will tend to correlate to the inverse of x 's formal repositories of knowledge. A user who finds an answer to their ASP.NET question on StackOverflow, doesn't ask a new question.

This is a problem I run into. As a more junior developer, nearly all of my questions I haven't been able to answer with research have already been answered on stackoverflow. Employers want me to have tons of points, but I don't usually have to ask any questions.

Re: Programming Trends on StackOverflow

#56

As a happy Ruby/Rails dev I can't figure why Python has double the marketshare. I guess it's the Google backing and the history of Linux admin usage?

Python Libraries > Ruby Libraries, at least outside of the web development world. There is nothing even close to the quality of SciKit-Learn, Pandas, NumPy, SciPy, Mapnik, PyCUDA, Sage, PyQT, PyGTK+, PIL, Celery, etc.

Ruby is an amazing language (as far as mutability-by-default languages go), but it is far too web-centric...which seems to have turned into a self-fulfilling prophesy.

Re: Programming Trends on StackOverflow

#57

Earlier quoted context omitted.

Another option would be a stacked area graph like this (the one on the right) http://www.asymco.com/wp-content/uploads/2013/06/Screen-Shot...

Wow. Stacked area charts are so bad at conveying information efficiently or honestly - this one would be a good parody if it wasn't presumably an honest attempt. RE: dishonesty, see for example http://www.leancrew.com/all-this/2011/11/i-hate-stacked-area... RE: inefficiency, note how bad the trend lines become near the top as the noise accumulates from the curves below. e.g. the galaxy tab curve on the left. I am red…

"Here’s a fictitious example to show what I’m talking about."

Yes, the author's fictitious example sucks.

The chart I linked shows the data points as discrete, so it's actually the stacked column chart the author claims to prefer and eliminates the confusion between height and thickness (that said, I don't object violently to that specific issue since I know how to read the charts).

Why convert curves to numbers? If you just wanted numbers read tabular data.

Tufte uses the equivalent of stacked area charts in one of his most famous examples (Napoleon's attack on Russia). Is it the width or the height of the bar that connotes the strength of the army?

http://www.google.com/imgres?imgurl=http://cartographia.file...

Tufte's great. You can quote him on both sides of most arguments. Displaying information well requires judgment, not just a bunch of rules.

Re: Programming Trends on StackOverflow

#58
post #53
post #13

Same data as area chart, attempted to sort categories by age while keeping related technologies grouped: http://cl.ly/image/1U400y1l2t3r

This is probably something I should already know, but how do you read a chart like this?

Think of it as a number of streams[1] running in parallel. The wider the stream, the more important the associated tag. Since the actual values are scaled to add up to 100%, comparing between relative importance is all they're good for.

There's also aa absolute version of this, which is a simple stacked area chart. It's not so useful here, because the relative popularity of tags is much more interesting than the absolute numbers, which are largely a function of StackOverflow getting more users in total. See for yourself: http://cl.ly/image/101e1y1q3i19

[1] There's also a thing called a "stream graph", which is related but not the same: http://en.wikipedia.org/wiki/Streamgraph

Re: Programming Trends on StackOverflow

#59
post #53
post #13

Same data as area chart, attempted to sort categories by age while keeping related technologies grouped: http://cl.ly/image/1U400y1l2t3r

This is probably something I should already know, but how do you read a chart like this?

This chart plots the "share" of each question on the site, so the total is always 100%. If you read the chart from left to right you will see that the languages whose share increases become wider while those whose share decreases become narrower. It can be a bit difficult to read with this many options.

Re: Programming Trends on StackOverflow

#60

Earlier quoted context omitted.

Wow. Stacked area charts are so bad at conveying information efficiently or honestly - this one would be a good parody if it wasn't presumably an honest attempt. RE: dishonesty, see for example http://www.leancrew.com/all-this/2011/11/i-hate-stacked-area... RE: inefficiency, note how bad the trend lines become near the top as the noise accumulates from the curves below. e.g. the galaxy tab curve on the left. I am red…

"Here’s a fictitious example to show what I’m talking about." Yes, the author's fictitious example sucks. The chart I linked shows the data points as discrete, so it's actually the stacked column chart the author claims to prefer and eliminates the confusion between height and thickness (that said, I don't object violently to that specific issue since I know how to read the charts). Why convert curves to numbers? If…

The graph you linked is discrete only by technicality (all data samples are discrete, after all.) Compare how yours is 19 time slices squished together into a width (normalized for aspect ratio) almost half the size of the one I linked, which is 6 data points with adequate white-space.

Please note that in the (excellent) Napoleon chart, cross sectional thickness is used to convey the size of the army. Not only is this intuitive, but it is masterfully hinted by the way the line shrinks. Stacked area charts (as detailed in the article I linked) however are unintuitive - a slowly increasing trend stacked on a faster increasing trend will appear to decrease - this is very undesirable.

The chart you linked seems to want to say "The PC(+tablet) market is increasing, but Window's user base is stagnant". Either unstack the data or throw away/combine the irrelevant data. If you want to compare the Galaxy Tab to other Android tablets (as this chart seems to want to do as well) you will need a better chart or a second chart. The top of left chart is visual noise.

Post reply on HN