Re: [RFC] Pipe Operator

On Thu, May 12, 2016 at 12:33 PM, Davey Shafik <davey@php.net> wrote: > On Thu, May 12, 2016 at 1:19 AM, Larry Garfield <larry@garfieldtech.com> > wrote: > > > On Mon, May 9, 2016, at 10:21 PM, Stanislav Malyshev wrote: > > > Hi! > > > > > > > |> seems like a common symbol to use, but it admittedly does look a > > > > > > So, usage in one semi-obscure language (F#) and one completely obscure > > > one (Elixir) - Clojure doesn't use |> - and one proposal for Javascript > > > now qualifies for "common". And that counting the fact that neither of > > > them actually uses the worst part of proposed syntax - magic variable > $$. > Is $0 being considered? It's not ideal but is used as "pattern match reference" in preg_replace, so we have a (sort of) precedent here. Same for $1 (first capturing subpattern). > > > > -- > > > Stas Malyshev > > > smalyshev@gmail.com > > > > If the issue is $$ feels too Perl like, what is the alternative? Is > > there another way to chain methods cleanly? > > > > In a sense, what we're really talking about here is continuations. > > Continuations (over-simplified) are a clean way of setting up "run this > > function, pass its result to this function, pass its result to this > > function, etc." That makes composition really easy. |> is essentially > > a continuation syntax. The $$ is to work around the fact that PHP > > function can have an arbitrary number of parameters, whereas > > continuations work best with single-parameter functions. > > > > Of course, with currying any multi-parameter function can be reduced to > > a series of single parameter functions. So what if we were to limit the > > concurrency syntax to single-parameter functions? And if you want to > > reduce a multi-parameter function to a single parameter function, yay > > closures. > > > > Would that limitation help or hinder? > > > > Either way, I firmly believe that more functional-friendly capabilities > > like continuations, promises, etc. are a direction that PHP needs to > > move, and syntax in that direction is valuable. > > > I wonder if it's possible to use $$ only when it's necessary to pass in the > argument positionally, that is, if it's not used on the right hand side, > it's passed in magically as the first argument, otherwise it's passed in > wherever the $$ (or whatever other placeholder you want) is used. > > _OR_ we just prepend it to the argument list used on the RHS. I noticed > that Closure also have ->> for passing in the argument last, vs -> for > first. I'm not sure I would like to have |>> for example. > > I think that $$ as a positional placeholder gives us the most flexibility. > $$ is super easy to type, apparently is possible with little conflict to > existing syntax, etc. > > I'm +1 on |> and $$. > > - Davey >