Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:91934
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Received-SPF: softfail (pb1.pair.com: domain golemon.com does not designate 209.85.217.170 as permitted sender)
MIME-Version: 1.0
Sender: php@golemon.com
In-Reply-To: <CAESVnVpMFOOeHgoFxz6RmET_3sbHwd0H4jkogNWNFNsagD5Dww@mail.gmail.com>
References: <CAESVnVqGWQbLRy9ctXh73=5X823F48HNAJuD_AZ=XBoFEfK1qg@mail.gmail.com>
	<CAESVnVpMFOOeHgoFxz6RmET_3sbHwd0H4jkogNWNFNsagD5Dww@mail.gmail.com>
Date: Thu, 24 Mar 2016 20:25:20 -0700
Message-ID: <CAESVnVqgMp6yeiZqWqk0b=72BJ3urcGcAXQmDmMm-GEWgSaBow@mail.gmail.com>
To: PHP internals <internals@lists.php.net>
Content-Type: text/plain; charset=UTF-8
Subject: Re: [RFC] [Vote] Short Ternary Assignment Operator
From: pollita@php.net (Sara Golemon)

On Thu, Mar 24, 2016 at 11:12 AM, Sara Golemon <pollita@php.net> wrote:
> After some discussion (and realizing the referenced implementation
> needed more work that a simple replacement of tokens), I've decided
> the close this vote, work with Midori on both implementations some
> more, and reopen at a later time with a complete implementation
> (possibly combined with ??=)
>
https://github.com/php/php-src/compare/master...sgolemon:short-ternary.coalesce

I've gone from scratch to implement this branch which doesn't
introduce any new opcodes, but it does add a new AST kind which is
compiled into something closely (but not quite) resembling a regular
short ternary.

It cheats slightly by assuming that since child[0] comes from a
`variable` in the parser that zend_compile_var() on it will always
yield IS_CV/IS_VAR, and I've got an assert in to guard on that, but a
second set of eyes would be nice there.

You'll notice there's a new runtime check in ZEND_JMP_SET to handle
the IS_INDIRECT case (which I see resulting from the dim/obj paths.
It's hidden behind an existing check for IS_VAR/IS_CV, so anything
producing a TMP won't hit it.  Hopefully it's not too harsh.

Based on feedback, I hope to add the ??= version of this
implementation as another commit and let Midori unify the two RFCs
into one. (unless someone objects to that)

-Sara