Time Spent Waiting for an Answer in StackOverflow (by language)(blog.guillermowinkler.com)
blog.guillermowinkler.com
Time Spent Waiting for an Answer in StackOverflow (by language)
http://blog.guillermowinkler.com/blog/2012/10/30/how-long-waiting-for-an-answer-in-stackoverflow/
15 comments
"StackOverflow was way more fun (and useful) at the beginning."
Wait, what? How could SO possibly be less useful today than it was a few years ago. Anytime I have any programming question I immediately google it and odds of a StackOverflow answer are very high.
I started iOS dev when it was fairly new and it sucked. So many bugs, gotchas, and misc shortcoming of the dev environment, Obj-C, or the SDK. Every single issue I ran into has now been cleanly answered on SO. Things I wasted dozens of hours on then would take me mere minutes as a new dev today.
I sympathize with it not being as fun to answer questions today, but I have no idea how you can say it was more useful then than it is now.
Wait, what? How could SO possibly be less useful today than it was a few years ago. Anytime I have any programming question I immediately google it and odds of a StackOverflow answer are very high.
I started iOS dev when it was fairly new and it sucked. So many bugs, gotchas, and misc shortcoming of the dev environment, Obj-C, or the SDK. Every single issue I ran into has now been cleanly answered on SO. Things I wasted dozens of hours on then would take me mere minutes as a new dev today.
I sympathize with it not being as fun to answer questions today, but I have no idea how you can say it was more useful then than it is now.
I meant useful for getting answers to new questions.
Meaning, if you're having a problem with something, and you googled it but didn't find answers ... I don't think you'll get useful results by posting the question to SO.
But of course it's still tremendously useful for the archive of questions and answers that it hosts.
Meaning, if you're having a problem with something, and you googled it but didn't find answers ... I don't think you'll get useful results by posting the question to SO.
But of course it's still tremendously useful for the archive of questions and answers that it hosts.
Posting with a 500 point bounty still does wonders. I once had a guy write a patch to the mercurial DVCS (small, but still) to answer a question.
Of course you have to wait for two days before you can offer a bounty.
Of course you have to wait for two days before you can offer a bounty.
"if you're having a problem with something, and you googled it but didn't find answers ... I don't think you'll get useful results by posting the question to SO."
I disagree. I just joined recently and that is exactly how I've used SO and in every instane I've gotten an answer to my question - and fast. Very fast.
I disagree. I just joined recently and that is exactly how I've used SO and in every instane I've gotten an answer to my question - and fast. Very fast.
He is clearly talking about it from the answering questions side, that is why he is talking about it as a game.
Clearly SO is a great resource.
But I am guessing you never found iphonedevsdk.com, it had all the answers before SO did. Although SO has better explained answers.
Clearly SO is a great resource.
But I am guessing you never found iphonedevsdk.com, it had all the answers before SO did. Although SO has better explained answers.
Yes. I just found ( http://stackoverflow.com/questions/9535809/best-cms-support-... ) that one of my questions recently voted up was closed. Beyond discussing if it was correctly closed or not, this situation reminds me of Wikipedia rules where you don't only contribute but lose your time arguing in favor of your changes.
For me, it is much more productive to write a blog article than waste time arguing in favor of your contributions.
For me, it is much more productive to write a blog article than waste time arguing in favor of your contributions.
> There was a time (first 6 months in the life of StackOverflow) where pretty much all questions would get answers within minutes.
This is now the case on younger sites of the StackExchange network. For instance, <http://french.stackexchange.com/unanswered>; is pretty much always empty.
This is now the case on younger sites of the StackExchange network. For instance, <http://french.stackexchange.com/unanswered>; is pretty much always empty.
SO used to have new questions every 20 minutes or so. The French Exchange seems to have (on average) few questions per day.
SO now has new questions every 20 seconds or so (not an exaggeration - an understatement if anything). A high percentage are still answered, but in absolute terms, the number of unanswered questions is obviously increasing.
"and best of all, I still get points from all my previous contributions. When I stopped, I had around 8k points; now I have 19k points"
Me too: it's the Bank of Stack Overflow. Like you, I was quite active a few years ago, back when iOS development was quite new (that's my speciality). But you're right, many questions now are very specific ("my code crashes at X", etc), which isn't inherently a bad thing but doesn't float my boat.
Me too: it's the Bank of Stack Overflow. Like you, I was quite active a few years ago, back when iOS development was quite new (that's my speciality). But you're right, many questions now are very specific ("my code crashes at X", etc), which isn't inherently a bad thing but doesn't float my boat.
The quality of the service, in the point of view of help to solve a problem we are encountering has significantly dropped. This is partly due to policy and partly due to the much lower quality and pertinence of the answers.
Contributors are there to earn points and just try whatever answer comes to their mind even if it's so vague or approximate that it is useless. Just trying, nothing to loose.
I guess the experts went away from it because the initial thrill has gone (sign of smartness) and they have better things to do then stare at a browser to catch new questions for no other rewards then getting points.
My impression is that stackoverflow was initially in a virtuous circle and now it's in a vicious circle.
I now have questions like how EXT4 files with holes impacts performance of an mmap ? Should I force a prior full fill of the file with dd before mmap ? I'll certainly get crap answers and I really don't know where to ask this to get a valuable answer. So I'll end up doing a benchmark, which has probably already been done many time or people knowing how EXT4 works could answer straight out of the book.
Contributors are there to earn points and just try whatever answer comes to their mind even if it's so vague or approximate that it is useless. Just trying, nothing to loose.
I guess the experts went away from it because the initial thrill has gone (sign of smartness) and they have better things to do then stare at a browser to catch new questions for no other rewards then getting points.
My impression is that stackoverflow was initially in a virtuous circle and now it's in a vicious circle.
I now have questions like how EXT4 files with holes impacts performance of an mmap ? Should I force a prior full fill of the file with dd before mmap ? I'll certainly get crap answers and I really don't know where to ask this to get a valuable answer. So I'll end up doing a benchmark, which has probably already been done many time or people knowing how EXT4 works could answer straight out of the book.
This.
Asking a question typically leads to a handful of "shotgunned" answers. Yes, they do come immediately -- but the real story is, they are templated ("I'll fix it later, just lemme be the first to answer"), incomplete, and/or simply wrong.
A dozen explanatory comments later, the SO gun slingers finally understand the question and its implications. No, I really was NOT asking that. Neither that. Yes, that condition I mentioned in the question IS important and you cannot ignore it in your answer. You have an error there. That won't work.
And once that happens... silence falls.
Asking a question typically leads to a handful of "shotgunned" answers. Yes, they do come immediately -- but the real story is, they are templated ("I'll fix it later, just lemme be the first to answer"), incomplete, and/or simply wrong.
A dozen explanatory comments later, the SO gun slingers finally understand the question and its implications. No, I really was NOT asking that. Neither that. Yes, that condition I mentioned in the question IS important and you cannot ignore it in your answer. You have an error there. That won't work.
And once that happens... silence falls.
Happened to me. And that person responding completely aside of the question voted to get my question closed and erased its answer. I was asking for axisting asynchronous signal slot library to avoid implementing mine. Another guy proposed to use QT and another one to use boost's signal which is not asynchronous. The later guy commented to the QT guy that boost was so much lighter and faster. Boost signal is known to be the less performant, but because its boost it must be the best. This guy also voted to close my question once I said I don't want to depend on boost. The question was finally closed and I implemented my self the signal slot lib with a perfect match to my requirements.
I'm curious to see an example of one of your questions like this.
The author filtered by questions with accepted answers only.
Strange that there is no Python there.
[Insert joke about Python being so intuitive and easy to use no one asks questions about it].
[Insert joke about Python being so intuitive and easy to use no one asks questions about it].
"[Insert joke about Python being so intuitive and easy to use no one asks questions about it]."
Hey, that must be true of Perl too! :)
Hey, that must be true of Perl too! :)
I'll bite, in response to the average time being just over half an hour.
> I think the time is higher than that of PHP or JavaScript because almost everything is just soooo easy with Python, so only questions with at least a bit of complexity get asked.
> I think the time is higher than that of PHP or JavaScript because almost everything is just soooo easy with Python, so only questions with at least a bit of complexity get asked.
It was kind of a relief for me - when frequenting HN, one might get the idea that everybody and their dog uses Python for everything, religiously. It's good that there's people out there who simply accidentally forget that it exists.
It's the tenth most used tag on Stack Overflow: http://stackoverflow.com/tags
From a visual quick scan of the tags here is a list of the top languages:
1 c# 371433
2 java 317047
3 php 295016
4 javascript 279976
5 c++ 156107
6 python 140451
7 objective-C 107083
8 c 74124
9 ruby 57577
10 vb.net 36981
11 perl 22101
12 dephi 17439
13 scala 11745
14 haskell 8696
15 actionscript 6102
16 groovy 4970
17 clojure 4046
.. f# 4046Additional to above, 11 months ago I also recorded some SO stats: http://news.ycombinator.com/item?id=3286348
Using these figures the % growth in questions asked in the last 11 months would be:
Using these figures the % growth in questions asked in the last 11 months would be:
java 76%
php 76%
javascript 85%
c++ 60%
python 70%
c# 716%
ruby 67%
perl 50%
scala 76%
clojure 77%
lua 77%
Really? I thought all of HN used Ruby; I of course use Python, C and C++.
Dear downvoter: What the hell? I can only assume that you clicked that button because I wrote something remotely-not-entirely-worshipful about Python, and in that case, have you really connected your identity so strongly to a single programming language that you feel comments unfriendly towards it should be invisible on a place like HN?
Here's the same query with Perl and Python added:
http://data.stackexchange.com/stackoverflow/query/edit/83730
Python comes in at 34, halfway between Java and Ruby, and Perl at 45, between Haskell and Scala.
http://data.stackexchange.com/stackoverflow/query/edit/83730
Python comes in at 34, halfway between Java and Ruby, and Perl at 45, between Haskell and Scala.
Many thanks for link. For posterity here is the list produced sorted on average:
avg std dev
php 24 41.2894351860362
javascript 26 43.0588803567294
c 27 45.4245721268946
c++ 28 45.2483939710062
c# 30 47.3345103589879
java 31 48.3674381995104
python 34 50.9682383156793
ruby 40 54.1612457522965
haskell 42 57.6626651794045
perl 45 57.7285412002331
scala 53 60.4111999775783
clojure 59 63.6228517085076
lua 60 64.9920293705815People rarely face a problem with the language itself. Most of the times its either the problem they are trying to solve, or its interface with the language.
> Hey, Haskell has pretty good answer times, at least considering its 33th position in the TIOBE Index.
I need to mention here that the Haskell community on Stack Overflow is AWESOME, they are helpful qualified people giving insightful answers (btw, Don Stewart is an active participant).
I need to mention here that the Haskell community on Stack Overflow is AWESOME, they are helpful qualified people giving insightful answers (btw, Don Stewart is an active participant).
I'll second this. Just browsing the questions in the Haskell tag on SO is a great way to learn the language. There's a really nice mix of practical questions and theoretical, mind-expanding ones there.
I second this. There are also qualified people asking the questions, which is also a good indicator (e.g. I recently saw questions from Edward Yang and Brent Yorgey).
By filtering out answers that arrive after more than 5 hours, you're really skewing the data, and in a very arbitrary way. Why not 4 hours, or 6?
Note that filtering also caused languages to change relative positions (e.g. PHP climbed from 5th place to 1st).
Well he could always compute the median. Less robust to outliers. 100 questions answered in 1 minute, with one question answered in 1 day will give you an avg of 15 minutes instead of the more representative 1 minute
It does, the 'clustering by hand' in categories according to hardness is completely unscientific.
I guess if done properly more categories would arise (e.g. bad formed questions) and possibly trends would change.
It's fun none the less ;)
I guess if done properly more categories would arise (e.g. bad formed questions) and possibly trends would change.
It's fun none the less ;)
"You see there, PHP running at front with 68 minutes average accepted answer time, either it’s too easy or there’re too many of them."
It's an average, the size of each data-set shouldn't really affect the results. If there's too many PHP questions then answer times would likely be slower. Seems just like an excuse to dig at PHP.
It's an average, the size of each data-set shouldn't really affect the results. If there's too many PHP questions then answer times would likely be slower. Seems just like an excuse to dig at PHP.
I think he means that duplicates might be causing a problem? PHP is very approachable, and you probably get a lot of new users asking similar questions, which would be handled quickly. I think this was his motivation in creating the plot of 'hard' questions which took more than a day.
Hi there! I was referring to the PHP programmers "them" :)
Pun intended.
Pun intended.
No Python, I'm starting to think OP is a troll.
Well, the OP has picked a completely arbitrary cutoff time, thus making any numbers obtained virtually meaningless. As suggested elsewhere, use the median and quartile ranges for non-normal distributions.
I doubt it's a troll, I'm sure it's well intentioned.
I doubt it's a troll, I'm sure it's well intentioned.
One alternative to leaving out the questions that take a long time is to calculate the average of log(minutes).
Harder to interpret, but it accounts for the questions with long delays without letting those questions dominate the calculation.
Harder to interpret, but it accounts for the questions with long delays without letting those questions dominate the calculation.
This technique is actually equivalent to the geometric mean: http://en.wikipedia.org/wiki/Geometric_mean#Relationship_wit...
I would have looked at removing the top and bottom 10% and/or fitting a sensible distribution.
Could (My|Postgre)SQL be considered a language too? How about RegEx?
I've created a query in the data browser, just select the tags you want:
http://data.stackexchange.com/stackoverflow/query/83626/how-...
http://data.stackexchange.com/stackoverflow/query/83626/how-...
I think another aspect is SO has the same problem as HN.
Same thing posted will get a different response depending on the time of day because of the signal:noise ratio.
Same thing posted will get a different response depending on the time of day because of the signal:noise ratio.
Rails?
Just change the tags in the set:
http://data.stackexchange.com/stackoverflow/query/83626/how-...
http://data.stackexchange.com/stackoverflow/query/83626/how-...
Hmmm. Seems like I'm an idiot: http://data.stackexchange.com/stackoverflow/query/83835/how-...
There's something I'm missing: Run query doesn't get me the result.
There's something I'm missing: Run query doesn't get me the result.
Actually they should mostly wait. The firehose of SO is largely filled with stupidity, lazyness, and people too lazy to learn or google what they need.
SO has become the tit to feed the stupid calf, and the milk has run dry because noone in their right mind should spend any time answering the questions of lazy, good for nothing idiots who think they could program someday, by merely posting every idiotic problem they encounter including syntax errors to SO.
If that was SO's mission. Mission accomplished. I and any other person who respects their own time and value will not spend 5 minutes on monkeys typing shakespeare.
SO has become the tit to feed the stupid calf, and the milk has run dry because noone in their right mind should spend any time answering the questions of lazy, good for nothing idiots who think they could program someday, by merely posting every idiotic problem they encounter including syntax errors to SO.
If that was SO's mission. Mission accomplished. I and any other person who respects their own time and value will not spend 5 minutes on monkeys typing shakespeare.
Then, somehow it got boring and complicated, and I stopped playing. I don't exactly remember what or how that happened, but it was a mixture of new rules and confusing features.
I haven't played there for more than 2 years, but I'm still in the top 2%, and best of all, I still get points from all my previous contributions. When I stopped, I had around 8k points; now I have 19k points.
StackOverflow was way more fun (and useful) at the beginning.
I think a major factor is that simply, all the basic and interesting questions already got asked. New questions these days tend to be very specific and complicated.