The Whiteboard

Matt D

12:04 AM

for me its more about problem solving and critical thinking

how do they solve problems?

but yeah. i've been asked a few good ones over the years.

the pi one's a good one :)

my question with the tailor series one, is how many iterations of the series do you use?

MetaFight

Some of the people I have worked with were able to implement algorithms. Some were able to design algorithms. Very few were able to design solutions, and only a handful factored in long term maintainability.

those are all points I'd test in an interview.

how?? beats me... but probably loads of pair/peer programming.

psr

12:31 AM

@MichaelT The MUMPS I hacked out on my terminal in 30 seconds: S PI4=1,Denom=3 F I=1:1 S PI4=PI4+((-1**I)/Denom) S Denom=Denom+2 w 4*PI4_":" Q:Denom>10000. Note the clever and efficient use of -1 exponentiated - much more elegant than fp.

One line, of course. To keep future maintainers eyes from getting tired.

(pausing for job offers...)

Jimmy Hoffa

1:55 AM

3

Q: Preventing a parser from turning into a (seemingly) god-sized object

So I have a program whose purpose is to take text files and parse them into a binary format that an embedded system understands. However, the text format I've inherited that I need to parse is sufficiently complex enough that after refactoring the main parse routine I'm left with a class with mo...

parsers eat another OO devotee...

psr

Yep. Procedural parser generators are the only way to go.

Jimmy Hoffa

4:56 AM

@psr There has got to be something to the idea that if someone can come up with code like that - regardless of the fact that it's terrible for such a great many reasons - must mean they've got some kind of overarching and large skill

not refering to you specificially, just all the "clever" in a bad way code in the world has to mean good things somehow..

gnat

6:24 AM

free spam flags...

-2

A: What if I have no good ideas to implement a feature?

Tell Us What’s Good We are always looking high and low for products, brands, individuals or organizations that are well aligned with our mission, and we definitely don’t want to leave any stone unturned. Social Good Network

why am I not surprised to see how this crappy "answer" makes a good match to an equally crappy question

huge thanks to guys who pulled it off from close vote queue for leaving this spam-invitation open

programmers.stackexchange.com/review/close/46713

gnat

7:07 AM

@Yannis @ChrisF @FrostEngineer @ThomasOwens I usually don't ask mods for stuff like that but today, newest questions tab seems to be really flooded with very low quality stuff, would you mind doing a bit of mod-cleanup?

gnat

8:03 AM

21 hours ago, by gnat

@MetaFight questions getting worse lately because it's day 3-4 after we've been "advertized" at Ars Technika - this is almost often the case that after some delay, site gets attacked by newcomers

user41796

12:46 PM

@gnat I voted to leave open because it wasn't a duplicate. And I didn't go back to VTC on the question because the question hadn't been touched in quiet a while. So I went with letting sleeping dogs lie. I now have a VTC as too broad. I would protect it, but the recent answers are from existing users.

gnat

1:05 PM

@GlenH7 yeah too broad makes sense, too. It was on my mind back then, too, I preferred dupe only hoping that this will help OP focus it to something more specific / answerable

Thomas Owens

1:26 PM

Hmm. I'm surprised I'm not finding research into what types of software development activities tend to generate the most defects. I thought this would be the easiest question to answer ever and come up with all kinds of empirical data to support my answer.

Matt D

because no one's done any research on it

bceause its hellishly subjective

how do you measure activities to defects?

Thomas Owens

It's actually easy. The outputs of each activity are fairly straightforward.

If your defect is in an output of an activity, it was injected by that activity.

I have plenty of internal data, but no industry data or cross-industry data to see how we stack up against other organizations.

Matt D

too many variables to draw a conclusion

Thomas Owens

Absolutely not. Since I just found a publication from Raytheon that does it.

Matt D

how do you factor in skill level, interruptions, complexity of solution, overall architexture complexity and dependencies?

Thomas Owens

1:36 PM

If your study is broad enough, you don't.

Matt D

familiarity with codebase, having a shitty day

family issues causing distracions

Thomas Owens

Nope. Don't care, doesn't matter.

Joe

Hello

Matt D

i dont agree

Thomas Owens

You can't measure any of those things.

Matt D

1:37 PM

that is exactly my point

Thomas Owens

Besides, if one guy having a bad day when doing requirements engineering means that you have a mistake in your requirements, then your process is screwed.

Matt D

we're not machines, were artists

Thomas Owens

No. We're engineers.

Matt D

process is repeatable

Thomas Owens

Yes, it is. It's measurable and quantifiable.

Matt D

1:38 PM

but thats only a framework

Thomas Owens

Software engineering is measurable and quantifiable. Or else I have no idea what I spent 5 years studying and the last 3 years doing, if not measuring software engineering.

Matt D

creativity is not quantfiable

Thomas Owens

I don't measure creativity. I measure cycle time, defects injected, on-time delivery, defect density.

Matt D

yes, and those things are pointless

Uhm, no?

1:40 PM

There are mathematical models to predict defect density of a software application. I can predict when testing is sufficient based on defects found.

Matt D

and what does that achieve?

Thomas Owens

It tells me when I can put software in a $16 million piece of hardware, put it in an airplane, and fly it.

Joe

why are you guys arguing this?

Matt D

its interesting

Thomas Owens

1:43 PM

Of course, all this measurement and modeling isn't necessary useful in all fields. I wouldn't expect it from Facebook, for example.

But I would expect it in the car that I drive, the plane I ride in, or the radiation treatment machine.

Matt D

i think, the best way to describe where i'm coming from is that, and this always irked me about SE (and i'm an SE) that metrics are really only useful when compared with themselves

and we can measure the results of things

but the act of programming tends to be a little more.. creative and flexible .. than the output may suggest

otherwise we'd have built a bunch of robots to program for us by now.

im just a very concrete type person

Thomas Owens

Solving problems does take creativity and has a human side. You can't measure everything.

Matt D

:)

Thomas Owens

For example, productivity.

I will never, ever, ever support anyone who tries to measure productivity. Or percentage completion of creative tasks, like design or implementation.

Joe

you can't measure productivity? That is a bold statement to make.

Matt D

1:46 PM

productivity is very very hard to measure

Thomas Owens

Not in creative tasks.

Joe

productivity is measured in output. It is easy to measure.

Thomas Owens

It is, but it's meaningless.

Matt D

@ThomasOwens in all honesty, and im really tired and rambling, and i think you're waaay more up to date with SE than i am.. i think we do software development totally fucking wrong

Thomas Owens

@MattD The thing is, there is no one right way.

Matt D

1:47 PM

@ThomasOwens well, i think there's at least a "better" way

@ThomasOwens i havent figured it out yet

Thomas Owens

Windows, Angry Birds, and avionics software are all different.

Matt D

@ThomasOwens im still trying to understand what the question is. to be able to answer it.

and yes. i totally agree.

Joe

I disagree. Productivity measures are important. It's the reason why Americans justify getting paid $30 an hour compared to a chinese worker getting $2 an hour

Matt D

@Joe whats the difference between 50, and 60 complete on a programming "task"

@Joe can you quantify that?

@ThomasOwens actually, you've inspired me. where'd you find that raytheon paper/

Joe

@MattD I don't understand that question

Mike

1:49 PM

@MattD The quantity of the difference between 50 and 60 is 10 :)

Thomas Owens

@MattD Crosstalk magazine. It's called "Advancing Defect Containment to Quantitative Defect Management"

"Bob has 10 classes to implement. He has implemented 6. How done is he?"

Matt D

^ * 1000000

Thomas Owens

If you say 60%, you're wrong. Because one of the classes is extremely complex and will likely require the time and effort of 3 other classes to test.

Matt D

and we can extend that and say. he does 1 class every 15 minutes, on averag

e

so, how productive is he being?

what the hell is "productivity" anyway

Joe

@ThomasOwens no statistician with experience would say 60%. they'd all weight the tasks according to difficulty and use a weighted completion

Matt D

1:51 PM

and you know what. they'd all be talking out of their asses

Joe

@MattD it's classified as output :)

Matt D

you can tell I had fun doing an SE degree cant you

I spent half the time arguing with the lectures over metrics that were largely meaningless unless you were comparing numbers generated with the same metric on other codebases

but this has a dobywhacker of 8, and this has a doobywhacker of 7,

but what the bloody hell does that mean anyway?

@Joe and what is that output? how do you classify it? is it classes? is it loc? is it complexity? is it defects? is it lack of defects? is it penguins?

Thomas Owens

Code measures are usually hints, imo. Things like cyclomatic complexity and the various readability measures are "hey, this probably needs a lot of unit tests" or "hey, can you look at how to refactor this to make it readable". Some people are like "you shall not have a cyclomatic complexity of >10 or else you must refactor".

Joe

Academics are trapped in theory, you can never expect them to know industry-level stuff (or at least the CS department I knew off - who were riding on their high horses, teaching Java because none of them probably knew python)

Matt D

@ThomasOwens i agree, and thats really the point i was trying to make earlier. although i probably came over as an asshole and i apologie for that :)

Joe

1:54 PM

@MattD Output would be classified as the final product. Measuring those in-betweens would be intermediary measures. those metrics are complex.

Thomas Owens

@MattD I get really intense myself when talking about empiricism in software engineering. And to top it off, I'm more of a centrist, realizing that empirical measures aren't the be-all-end-all, like some do.

Transcript for