Give me interfaces any day.

Ehh. Yer jes' lazy.

I look at this:

sp_rv
sp_generate_signature (sp_audio_info_t * song_audio_format,
sp_signature_t * signature,
guint (*get_pcm_data) (guint sample_offset,
guint num_samples,
guchar * output_buf,
void *user_data),
void *user_data)

and it makes my heart hurt. I'm so glad I only visit C world long enough to port it to Java world.

Then again, in the source to javazoom, a pure java MP3 decoder, there exists the line:

fs = (fs>32767.0f ? 32767.0f : (fs < -32767.0f ? -32767.0f : fs));

the ?: operator is the devil.

In reply to:

---

the ?: operator is the devil

---

Don't say things like that . I love the ?: operator. Just make sure it's commented properly and there's no problem. I even have ?: nested 4 and 5 deep.

I disagree about ?:. What is the devil is random magic numbers. He just couldn't make 32767 a constant, could he?

Of course, some indentation could be useful in that function declaration.

sp_rv
sp_generate_signature (sp_audio_info_t * song_audio_format,
sp_signature_t * signature,
guint (*get_pcm_data) (guint sample_offset,
guint num_samples,
guchar * output_buf,
void *user_data),
void *user_data)

C's insane type declarator syntax can usually be sanitised by introducing typedefs everywhere it gets hairy:

typedef guint (*get_pcm_data_fn) (guint sample_offset,
guint num_samples,
guchar * output_buf,
void *user_data);

sp_rv
sp_generate_signature (sp_audio_info_t * song_audio_format,
sp_signature_t * signature,
get_pcm_data_fn fn,
void *user_data)

but even so the typedef itself remains a bit icky.

Peter

I definitely agree about magic numbers ...

I just always prefer clarity over brevity ... I suppose if properly commented, ?: can be OK (certainly a SINGLE ?: does not return true from my devil function -- which, curiously enough, is: boolean isDevil = (hasQuestionColon) ? ((hasMoreThanOne) ? : true : false) : false, I guess I just find if/else's much easier to follow, and much easier to go back and modify later, whereas the syntax of nested ?:'s is so fragile that you have to really work at changing it.... But that's just me

Yeah -- It's definitely not fair to condemn a feature based on one of the more confusing examples of its use

In reply to:

---

boolean isDevil = (hasQuestionColon) ? ((hasMoreThanOne) ? true : false) : false

---

Using conditional operator with only true or false as results is just gratuitous obfuscation

More than one on the same line is bad, but I have no problem with multiple ?: operators if they are laid out as if-elseif-elseif-else (which is great because you don't need loads of parens), or as a binary chop or some other recognisable structure, with a line for each test and each result, and indented sensibly.

In reply to:

---

Using conditional operator with only true or false as results is just gratuitous obfuscation

---

I couldn't think of anything funny to put there that wasn't just true/false I should have made it a recursive infinite loop, so instead of true, i had isThisADevil()... Hindsight.

Sorry if I wasn't clear - I know you put true and false there as a joke, but I have seen (too much) real code that does that sort of thing.

Nah. You just have to wonder about some programmers when they do things like this:

(xyz ? func_1 : func_2)(arg1, xyz ? arg2 : arg3);

I've got a library for a expansion board that is full of things like that for some reason. They're actually a little more complicated but you get the general idea. Must have had an aversion to using if/else

- Trevor

You and me both buddy. There's got to be a better way.

Then again, I've been known to use a goto here and there.

Greg

(xyz ? func_1 : func_2)(arg1, xyz ? arg2 : arg3);

Hah! Took me a while to parse that. Pretty damn funny. I'm going to have to start using that all over the place, just to annoy folks.