user54412
12:56 AM
Yes, NSA, I just spent the last 15 minutes looking up how to purchase radioisotopes online.

1:51 AM
@ChrisWhite It's worse than that. You've been known to answer questions about nuclear bombs.
Who knows what kind of mischief you might get up to?

2 hours later…
3:31 AM
Let's say you have a section of an answer that you've worked out, and is interesting in its own right, but you've discovered that it isn't directly relevant to the true answer. Do you scrap it, or put it in its own section at the end of the answer?

@alemi You could always ask a question it is the answer to yourself. If you put it in the other answer, it is unlikely to be found by people interested in only that part of the answer

If you wouldn't mind giving me some advice, the answer in question is my current answer to the racquetball question. I have the section discussing dispersion in a waveguide (the 2D analytic waveguide section), but I'm now convinced that has nothing really to do with it. I should probably just scrap it shouldn't I?

3:49 AM
Holy mother of simulation! That's a good answer! But you're probably right, it's not that interesting to read about the analytic waveguide solution in the middle of the answer. I'd probably move it to the end as a kind of extended footnote, though

Yeah, and also that answer is fairly long already. It's good to keep in mind that people are less likely to read a longer answer.

I'll have to rework the whole thing now. Turns out is just all about the arrival times of the pulses. Bah.

Posting a self-answered question is definitely an option to consider. We tend to do it less here than on some other sites, but as long as the question you ask is a good one, it's fine.

user54412
if anyone knows how to write a data.SE query for it, it'd be interested to know how answer scores vary with answer length
2

3:52 AM
@tpg2114 Ouch. I trust you have filed a bug report.

1 hour later…
5:03 AM
@ChrisWhite Still around?

user54412
@tpg2114 yep

user54412
I'm actually tinkering with the data explorer

I was about to plot the downloaded CSV as a scatter plot to see what it looked like, the graph from the explorer isn't all that great since it puts lines between it

user54412

user54412
less information, but easier to grok

5:07 AM
There ya go, I just gave the raw data and planned on using Python to analyze it :) Yours is better
Why did you cut short the length of the post?

That's a length in characters?

Yes
There is no clear way to get words:
7

I've recently been working on some database search functionality and wanted to get some information like the average words per document (e.g. text field in the database). The only thing I have found so far (without processing in language of choice outside the DB) is: SELECT AVG(LENGTH(content) -...

user54412
@tpg2114 You can extend it - the data's just really noisy out there

No worries, I was just curious

5 chars per word is a fair guess

5:10 AM
@alemi Equations and hyperlinks kind of screw with that though
$$E = m c^2$$ -- 15 characters, but how many words? Hard to say
Although it might be good to strip out whitespace from the length count, not that I expect the results to be any different

This is stackexchange as a whole, or restricted to physics?

Just the X-axis will change

user54412
@alemi just physics

user54412
but you can run the same query for any site

I wish I knew SQL better

user54412
5:15 AM
I'm working on one normalized by views, but apparently ViewCount only applies to question posts, I think

user54412
I don't really know what I'm doing

I want to see, on the same graph, 6 lines -- the three you have but limited to accepted answers only, and the three you have for not accepted answers
Which means looking for Id = ParentId and checking if the AcceptedAnswerID = <id of the answer> and joining that info together into one bigger table
I know it needs JOIN's but I have no idea how to do it
And yes -- ViewCount is only on question posts -- but you can pull the ViewCount from the Id == ParentId in the same way I want to get the accepted answer thing

user54412
joins.... why does the wikipedia article have 9 varieties?

wut
INNER, OUTER, and?

@ChrisWhite Yeah, exactly!

5:21 AM
oh right

@ManishEarth I'm sure you know exactly how to construct these things

@tpg2114 I know the basic ones. Never seen half of these

@ManishEarth Yeah, but are you dangerous enough to generate the 6 lines I want? :) Average +/- STD for accepted answers and Average +/- STD for not accepted ones

user54412
Got my query

user54412
5:28 AM

So what this shows is length doesn't matter and it's really just score due to views

Doesn't look like there's much of a relation between answer length and score, does it?

@tpg2114 hm. maybe :p

user54412
I forgot about the science long ago

user54412
must... query... more

5:32 AM
@ChrisWhite I'm sure I've mentioned that I wrote a SQL statement to compute vorticity from velocities and grids stored in a database

user54412
that's... awful

CREATE TABLE vorticty
AS SELECT a.i, a.j, a.k
, ((d.f3-e.f3)/(d.y-e.y)-(f.f2-g.f2)/(f.z-g.z)) AS vort_x
, ((f.f1-g.f1)/(f.z-g.z)-(b.f3-c.f3)/(b.x-c.x)) AS vort_y
, ((b.f2-c.f2)/(b.x-c.x)-(d.f1-e.f1)/(d.y-e.y)) AS vort_z
FROM scratch AS a JOIN scratch AS b ON a.i=b.i-1 AND a.j=b.j AND a.k=b.k
JOIN scratch AS c ON a.i=b.i+1 AND a.j=b.j AND a.k=b.k
JOIN scratch AS d ON a.i=b.i AND a.j=b.j-1 AND a.k=b.k
JOIN scratch AS e ON a.i=b.i AND a.j=b.j+1 AND a.k=b.k
JOIN scratch AS f ON a.i=b.i AND a.j=b.j AND a.k=b.k-1

@tpg2114 why
whyyy

We had access to a demo Netezza machine (they have since been bought out by IBM) and my advisor wanted us to get into the whole "big data" game and was convinced we could use these databases to do post processing work on a massively parallel scale
So I had to compute the vorticity for a 750 million grid point simulation from inside the database as a demo of potential
It only took about 20 seconds, it was actually kind of cool. The serial Fortran code took hours because I couldn't load very much of the file into memory at once

@tpg2114 ... he does know that big data isn't done with mysql, right?
big data uses a different type of object storage and blah. They use stuff like Hadoop and all

5:39 AM
@ManishEarth Have I mentioned that we didn't get funding to continue down that path (thank goodness)

However, for single tables (i.e. no JOIN statements), Dremel is a nice way to run huge queries
Google's Big Query lets you run queries very fast if it's on a single table

Plus Netezza had these crazy RAID setups and continual backup of every database activity, they had a lot of banks and stuff as customers. I kept trying to get them to disable it for me because I honestly don't really care if my row of variables is strided across 5 drives and backed up to 5 more. I want it to go fast
But they couldn't, their hardware wasn't set up any other way
I stayed up too late. Now I'm hungry with nothing to eat at home.

user54412

@ChrisWhite SQL genius in like an hour, nice work

user54412
today I also learned how to parse XML get python to parse XML

5:51 AM
There's an odd symmetry in the accepted and unaccepted averages
@ChrisWhite I've been doing that one for ages, you could have asked for help! Coincidentally if you need Fortran or C++ to do it, I can help there too

user54412
@tpg2114 you parse xml in fortran?!

There must not be many posts at 9000 characters -- average unaccepted is -19 and average accepted is 28...

@tpg2114 How many years ago are you talking about ?

@ChrisWhite We've had XML input files for our code for 6 years now
@hwlau 2008 I think

user54412
I suppose it would be nice for our code to have sophisticated input

5:54 AM

user54412
but there's no way we're changing it now, despite doing a full rewrite in a new language

It is horrendously buggy though and not maintained, so we've been maintaining our own version
@ChrisWhite Which new language?

user54412
C -> C++

Boost property trees are super easy and work a lot like Python ptree does (which is how we do our XML)
It would probably take a lot less time than you think to get it working

@ChrisWhite wut? you are still using it!

user54412
5:57 AM
maybe it could be done, but there are too many users who would complain if their input files had to be rewritten

@tpg2114 It is super-slow to compile it

user54412
@hwlau as in C++ == C? In our case the new code looks nothing like the old. Far cleaner, at least so far.

@hwlau I don't have any issues with it, my compilation time didn't change from libxml2 to the property trees
But all of my code is templates and stuff already so adding another template library wasn't noticible
@ChrisWhite That could be. We went through a painful "converter tool" phase and having both readers in the code for a few versions and marked as deprecated. Then we killed off the old way and told people "Too bad, you had warning for 2 years"
Which is way easier to do in C++ than in Fortran

@ChrisWhite sor, I read it wrogn

1 hour later…
7:09 AM
0

Recently David Z said in chat that It's good to keep in mind that people are less likely to read a longer answer. So I was wondering: How true is it that length can work against you in an answer? While we can't directly measure the number of users who do read individual answers, surely ther...

7:26 AM
^awesome

11 hours later…
6:23 PM
Food for thought: meta.stackexchange.com/a/215220 It's not exactly analogous to what we have here (some use-cases of 'too broad' come to mind), but the suggestion along the lines of "this question asks for more help than we're able to provide" is interesting.
3

6:55 PM
@Kyle I also like the "the asker should go out and read a book or ask someone, realize the problem with the question, then come back and fix it"
perhaps we should be more specific: "The asker should take an introductory course in special relativity"
2

@Jim Well, maybe not that far, but "perhaps the asker should try some questions on the basics of special relativity or introductory algebra before jumping to pathological variations of the twin paradox"... that should definitely be in our default close reasons ;)