Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:118669
Content-Type: multipart/alternative;
 boundary="------------pTgJCsn3fHwd8Aco0leBgx4j"
Message-ID: <eeabffad-4765-dd2c-73af-ea81d39dec12@gmail.com>
Date: Tue, 20 Sep 2022 10:54:59 +0100
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:102.0) Gecko/20100101
 Thunderbird/102.2.2
Content-Language: en-GB
To: internals@lists.php.net
References: <CAJpLVh1tYVMSvfXjiuh5=V8Pki4QDFDB6vUQT=rE6k_boSdDNA@mail.gmail.com>
In-Reply-To: <CAJpLVh1tYVMSvfXjiuh5=V8Pki4QDFDB6vUQT=rE6k_boSdDNA@mail.gmail.com>
Subject: Re: [PHP-DEV] One-line heredoc for better syntax highlightning
From: rowan.collins@gmail.com (Rowan Tommins)

--------------pTgJCsn3fHwd8Aco0leBgx4j
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit

On 19/09/2022 20:10, Olle Härstedt wrote:
> More for users to learn? Don't you mean less, haha? This is an
> arbitrary limitation of heredoc, that we can remove (it might not be
> arbitrary from a yacc perspective, someone would have to explain that
> if so).


I don't think the rules for heredoc are arbitrary,  but they are quite 
specific: the end delimiter has to appear at the *start* of a line, so 
it can appear anywhere else in the text without prematurely ending the 
string:

$foo = <<<FOO
This is a normal heredoc
this is not the end FOO
this is:
FOO;

(Prior to PHP 7.3, only certain characters could appear *after* the end 
delimiter, but that's now more relaxed, and there's some clever rules 
around indenting.)


As far as I can see, a single-line heredoc would need to be a new 
syntax, with its own syntax rules. Something like

* If there is at least one non-whitespace character after <<<FOO but 
before the next newline, this is a one-line heredoc
* If so, the heredoc content is everything up to and including the last 
occurrence of "FOO", followed by an optional semicolon, and a required 
newline
* If "FOO" does not appear again before the next newline; or there is 
other text between "FOO" and the newline, that is a syntax error

For instance, the following would presumably be valid:

$foo = <<<FOO one-line heredoc ending at end of line FOO
;

$foo = <<<FOO one-line heredoc with semicolon FOO;

$foo = <<<FOO this is a heredoc containing FOO and other text FOO;

$foo = <<<FOO // I guess this is a string not a comment? FOO;


But the following would not:

$foo = <<<FOO one-line heredoc not ended

$foo = <<<FOO this heredoc contains FOO and then doesn't end

$foo = <<<FOO mixing one-line and multi-line syntax doesn't make sense
FOO;

$foo = <<<FOO this heredoc looks like it ends FOO . 'but actually it 
doesn't';


Perhaps you disagree with some of those rules and examples, but 
hopefully can see that the rules would need defining, and therefore 
learning.


>> I don't know about other editors, but PhpStorm at least also recognises an
>> in-line comment explicitly labelling any string, which if anything seems m
>> ore explicit:
>>
>> $query = /** @lang mysql */"SELECT * FROM foo";
> Pja. I wouldn't call it more or less explicit than heredoc notation, personally.


It's more explicit in the sense that it is syntax dedicated for that 
purpose, in a standard format. If I write <<<FOO hello FOO, there's no 
way to know if "foo" is the name of a language I'm hoping to label, or 
just a token which I know doesn't appear in my content; if I write /** 
@lang foo */"hello", there's no ambiguity, even if you've never heard of 
"foo" as a language.


Regards,

-- 
Rowan Tommins
[IMSoP]

--------------pTgJCsn3fHwd8Aco0leBgx4j--