> On 14/05/2016, at 4:16 AM, Sara Golemon <poll...@php.net> wrote:
>
> On Fri, May 13, 2016 at 10:18 AM, François Laupretre <franc...@php.net> wrote:
>>> So we'd have to audit all 4k+ functions in the PHP runtime (and
>>> provide a mechanism for defining it on userspace functions)?
>>
>> That's right, except that, if we only consider functions accepting more than
>> 1 arg, we just need to check about 1,000 functions. It may seem a lot but,
>>
> Having been a large part of the audit for PHP6, I'd say "Yes, it
> definitely seems like a lot, and is." :p
>
>> for a potential solution to such a long-time issue as argument ordering
>> sadness, IMHO, it's worth the pain. I am currently doing it and I'll send
>> you the list when it is ready.
>>
> Awesome. Even if not used in this feature, it could potentially be
> useful elsewhere. I hope you put it in a public gist or similar for
> posterity.
>
>> For userspace functions, the logic can be quite simple : 0 arg -> not usable
>> in a pipeline, 1 arg or more -> substitute first arg. Then, we have 4
>> options :
>> - consider that, when using a userland function in a pipeline, the
>> lhs is always inserted in 1st position,
>>
> It would feel weird that internal functions can deviate from this
> rule, but user-space functions can't. :/
>
>> - provide a function to change the position on an existing function
>> (default = 0 when function is registered). A negative position would disable
>> using the function in a pipeline,
>>
> Feels like a hack, but I suppose if it were part of Reflection it
> might make sense...
>
> (new ReflctionFunction('my_func')->setPipeArgPosition(3);
>
>> - enrich the syntax of function declaration to designate the arg to
>> substitute.
>>
> This would be a good application for User Attributes:
>
> <<__PipeArg(1)>> // Zero based, obv
> function array_map(Callable $fn, array $arr) {...}
>
>
>> I would propose to start with the 1st choice. After people have the occasion
>> to use it for a while, we will see if an additional mechanism is requested.
>>
> That's not unreasonable if we're starting from a place of assuming
> that invisible substitution is the way to go. I'm not convinced, but
> I could be.
>
> Just to verify, you're suggesting an end-state something like this?
>
> $ret = array(1,2,3)
> |> array_map(function($x) { return $x * 2; }) // lhs implicitly
> passed as second arg
> |> array_sum(); // implicitly passed as only arg (first position)
> // $ret === int(12)
>
> -SaraI’m not fond of this approach. Take in_array as an example. I have, in the same file, piped an array in as the second argument and piped a string in as the first. To have the position of the piped variable be implicit, you’ll need multiple versions of in_array. Sure, you could try to use the type of the value being passed in, but that ends up much more magic and then you run into the same problem with strpos, or password_verify, or a variety of other functions that take multiple of the same type of argument where order matters. This approach will greatly reduce the power of this feature, making it much less useful for things beyond examples. -- Simon Welsh
smime.p7s
Description: S/MIME cryptographic signature
