Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:79540
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Received-SPF: softfail (pb1.pair.com: domain golemon.com does not designate 209.85.217.169 as permitted sender)
MIME-Version: 1.0
Sender: php@golemon.com
In-Reply-To: <98ABD377-A37A-4F25-8550-3E7684F22FDC@ajf.me>
References: <E21EA75D-69AA-408C-88B7-C8022E9FB2C7@ajf.me>
	<CAESVnVrfqA9EsK3jFs+jr4SiSgDHXr2+Ya3TkHJO8nH7mqh7hg@mail.gmail.com>
	<98ABD377-A37A-4F25-8550-3E7684F22FDC@ajf.me>
Date: Wed, 10 Dec 2014 11:16:05 -0800
Message-ID: <CAESVnVrqhr0oQ5sJMHY8vHvD8Hmz2C4P7gV7qQMrJK=FmgECGA@mail.gmail.com>
To: Andrea Faulds <ajf@ajf.me>
Cc: internals <internals@lists.php.net>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable
Subject: Re: [PHP-DEV] [VOTE][RFC] Unicode Codepoint Escape Syntax
From: pollita@php.net (Sara Golemon)

On Wed, Dec 10, 2014 at 10:59 AM, Andrea Faulds <ajf@ajf.me> wrote:
>> On 10 Dec 2014, at 18:55, Sara Golemon <pollita@php.net> wrote:
>> I was just updating my HHVM patch to match your PHP implementation and
>> an issue came up.  The following code, which is valid in PHP5:
>>
>> <?php
>> echo json_decode("\"ma\u00F1ana\"");
>>
>> Will throw a fatal compiler error as "\u00F1" is an invalid unicode
>> escape sequence.  Since this represents an unnecessary BC break, I'd
>> like to propose the error handling be modified to match \x, which is
>> to say: Pass the value through unmodified.
>
> I was wondering about that case. Previously, the patch just raised a warn=
ing
> but let it through unmodified. But then old code would be littered with w=
arnings,
> and I felt it was better just to throw an error. Part of the problem is t=
hat I=E2=80=99d rather
> mistakes in string literals be caught at compile time, to prevent someone
> accidentally echoing =E2=80=98\u00F1=E2=80=99 somewhere.
>
Well, consistency would be not raising a warning at all (like with \x
or unknown \ sequences), but I get your concern with helping coders
catch mistakes early.

I don't have the stats on how many projects are using \uXXXX sequences
(with the json_decode() or similar hack to parse them), but I know
FB's codebase is absolutely covered in them.

> A possible compromise might be to let =E2=80=98\u=E2=80=99 through but no=
t =E2=80=98\u{=E2=80=98.
>
Still don't like it from the inconsistency with existing escape
sequence handlers pov, but it'd cover the biggest set of BC issues, so
I'd be happy with it.

-Sara