Doing something every time is always faster than sometimes doing nothing...

PJH

...especially with your triangular wheel.

Spotted this out of the corner of my eye while fixing a problem elsewhere in the library. The language is C:

const char* ordSuffix(int num) {
	 // make num always positive. This is faster than having to check for
	 // negative cases.
	 num = num < 0 ? -num : num; // same as num = abs(num); but faster, so there.
	 // Numbers from 11 to 13 don't have st, nd, rd
	 if(10 < (num%100) && (num%100) < 14) return "th";
	 [snipped switch statement to return 'st', 'nd', 'rd' or ,'th']

boomzilla

Hmmm....negative ordinals. Taking Asimov to the next logical step, I suppose, but unorthodox.

DOA

Using a ternary when nothing happens in one of the cases is a pet peeve of mine. I swear they discover that operator and then they have to use it everywhere.

PJH

@DOA said:

Using a ternary when nothing happens in one of the cases is a pet peeve of mine. I swear they discover that operator and then they have to use it everywhere.

None of the suspects who could have written this is 'new' and ?: shouldn't be a novelty for any of them. In fact I'm inclined to think that the prime suspect simply copy-pasted this code from somewhere without thinking, and thought it might get used elsewhere so put in in the library. (As opposed to what should be happening: upon noticing something being used twice or more, refactor a version to the library, and change the originals when something else needs doing to them (rather than immediatly changing all of them just because of the refactoring) - the latter method tends to reduce library bloat and duplication. Usually.)

Apropos: There is, as I recall, an ex-(before I started)employee who was somewhat over-fond of using (nested, no less) ?: instead of if/then/else, and the resultant source was... well...

Not really fit for an environment where there is more than one programmer. (Though I get the impression there was only 2, possibly 3, programmers then - it was a few years ago.)

If I remember, and if it's still in the current codebase (though I'm certain it's obsoleted, as opposed to merely deprecated, and not actually being used) I'll dig some out and post it.

PJH

@boomzilla said:

Hmmm....negative ordinals. Taking Asimov to the next logical step, I suppose, but unorthodox.

In light of my last reply, and reviewing the thread, it appears my earlier thoughts in reply boomzilla went missing (or I simply didn't post it.. which is more likely.)

The library function concerned is used in a very small fraction of the total codebase that uses the whole library, and in the instances where it is used (including debug statements) are neither

time sensitive,
bottleneck/inner-loop suspects where the difference between ?: and abs() could ever be an issue or
where the parameter passed in would be (sensibly) negative

In short, it's used for user presentation for one small aspect of the system where you can't have a negative number of 'foo'.

Hence my copy/paste suspicion mentioned in my last post.

(And the aforementioned absence of my un-posted suggestion that "simply requiring the parameter to be unsigned instead of signed, and requiring callers that may require its use on negative values to figure out the signage before/when calling it would be better.")