Implement cmp() function

[hikari-no-yume]

Also see Implement <=> (spaceship/comparison) operator

[laruence]

why?

$a < $b? -1 : (($a > $b)? 1 : 0);

you can't change all user used codes into extension function..

[hikari-no-yume]

See my last three emails to internals (replying to Rasmus).

Essentially:

1. Users are bad at writing sort functions:

   function sort_func($a, $b) {
       return $a >= $b; // all too common
   }
   

2. This is already internal functionality, why not expose it?

3. It complements strcmp() well.

4. This means you could unify sorting and user sorting (make cmp() default comparison function).

5. It's a common use case. This would make writing sort functions easier, especially those that sort multiple things, and it's useful in other places.

6. Some other languages have an operator for this (notably Perl which has <=>)

[smalyshev]

Why not <=> then? Also, I'm a bit afraid taking "cmp" may be a BC problem. But maybe I'm too paranoid. I like the idea though.

[hikari-no-yume]

@smalyshev Yeah, we could just add the operator Perl has, actually. It reads pretty well. I'm sure some would object to the syntax, though, but function ($a, $b) { return $a <=> $b; } makes some sort of sense.

cmp() might cause conflicts, it's true. We actually have a sample function with that name on the usort manual page that does exactly what this built-in would do.

Maybe this'll end up just like the intdiv() RFC proposing a new operator, with a fallback option of a new function should the operator not get in.

[hikari-no-yume]

I suppose an advantage of cmp() is it can be easily polyfilled, but <=> does have a certain appeal...

[hikari-no-yume]

<=> would avoid the BC issue you're worrying about ^^

[hikari-no-yume]

Here's an argument against cmp(): in Perl, $foo cmp $bar is used for string comparison. I doubt familiarity to Perl programmers is an issue here, and people would almost certainly realise that it's not the same given we have strcmp() for that purpose, but it's worth noting.

[laruence]

@TazeTSchnitzel I've got a idea to fix that issue... running test, thanks

[hikari-no-yume]

@laruence Good, but it doesn't really remove the need for this.

[laruence]

@TazeTSchnitzel actually, I like @smalyshev 's <=> idea

[hikari-no-yume]

@laruence I do too, I'm also making a patch for that

[laruence]

@TazeTSchnitzel as I said, we are not baby-sitter, you can not change any user common used PHP codes into PHP internal function...

[hikari-no-yume]

This isn't baby-sitting, it's a basic language feature... Ruby, Perl, Groovy, Python 2 (but not 3?!), Haskell, OCaml, all of these have this.

[laruence]

@TazeTSchnitzel I was talking about cmp, not <=> :)

[hikari-no-yume]

@laruence They're exactly the same, though. If there's a case for an operator, I don't see why there isn't a case for a function.

[hikari-no-yume]

Ruby, Perl, Groovy have <=> (Perl also has cmp, the string comparison operator)
Python 2 has cmp()
Haskell and OCaml have compare()

[laruence]

I feel it is overhead, how different between function ($a, $b) { return $a <=> $b; } and a builtin cmp?

[hikari-no-yume]

Nothing, but that's assuming we have <=>. if that doesn't get in, surely we should have cmp().

[hikari-no-yume]

Made a <=> patch: #1007

[smalyshev]

Wow unit tests hate cmp()... so maybe operator is better. But RFC can propose both. Actually, even better to take over existing RFC if Davey Shafik doesn't mind (see link in other patch).