Programmer productivity metrics

[NotAllThere]

I think it's easier for bad coders to get away with it. You can knock up something that looks half decent with a bit of drag and drop. You can write pages of inefficient code and bad database queries, but because the machines are powerful, performance isn't a problem.

I'm currently investigating performance issues for a client... the inefficient code rapidly becomes apparent. I've gathered evidence that the offshorers aren't even following their own inadequate standards.

[SunnyInHades]

#define fastcopy(x,y,z)

fastcopy(array1, array2, size);
fastcopy(array1, array2, size);
fastcopy(array1, array2, size);
fastcopy(array1, array2, size);
fastcopy(array1, array2, size);
fastcopy(array1, array2, size);
fastcopy(array1, array2, size);
fastcopy(array1, array2, size);
fastcopy(array1, array2, size);
fastcopy(array1, array2, size);

/* Copy'n'paste 'fastcopy(array1, array2, size);' which does nothing as many times as required.
Scatter liberally through files and within methods to taste -
no nosy line manager peeking at version control checkins will dare delete or question such lines */

[IR35FanClub]

Function points.

Did a course once. Useful. Some companies price their contracts using them.

They are language independent. But you have to work out how to convert function points to pounds using your own empirical evidence from the tool set you are using.

[thunderlizard]

As others have suggested, lines of code is not a useful metric. That's because, in any given function, usually only one line is actually coming up with the calculated value - the "return" statement. Getting a report of the number of "return" statements written would be much more useful.

I said "usually" there's only one return statement per function, but especially good programmers sometimes put multiple return statements in a single function. This is basically killing two birds with one stone, so those should count double.

hth
tl

[kingcook]

What is a good metric of productivity of programmers? I get the impression some of my dev team are slacking off and have asked the lead to send me a weekly report by developer of how many Lines Of Code each has written. However I am getting strong push back on this so I need an alternative.

What does the panel recommend. Also, what is so wrong with LOC as a metric. If I was running a potato factory I would want to know how many spuds each worker had peeled, same difference in my view.

Fecking prima donnas

Number of lines of code, definately. I've made a good living on it. Did I mention that I like to write lots of useless comments?

[chinglish]

There are enough code metric tools especially for Java - if you want metrics incorporate Checkstyle and Findbugs in your build script. Combined they will highlight duplicity cut n paste, inefficient code, security problems etc etc. By all means use LOC but only after they have passed checking by Checkstyle and FindBugs.

[eek]

There are enough code metric tools especially for Java - if you want metrics incorporate Checkstyle and Findbugs in your build script. Combined they will highlight duplicity cut n paste, inefficient code, security problems etc etc. By all means use LOC but only after they have passed checking by Checkstyle and FindBugs.

I think you've missed the entire point of this thread. LOC is about the worst measure you could use to test whether something is well written. Chances are the better code in a project is concise and clear. The long winded code will hide potential issues that aren't obvious until out in the wild.

[Lockhouse]

Very poor troll IMHO.

[Bunk]

As others have suggested, lines of code is not a useful metric. That's because, in any given function, usually only one line is actually coming up with the calculated value - the "return" statement. Getting a report of the number of "return" statements written would be much more useful.

I said "usually" there's only one return statement per function, but especially good programmers sometimes put multiple return statements in a single function. This is basically killing two birds with one stone, so those should count double.

hth
tl

What if the function's type is void?

[eek]

What if the function's type is void?

Make it a boolean (success / fail) and ignore the response.

I've seen entire programs with that probably for that exact reason.