Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:110761
To: PHP Internals <internals@lists.php.net>
Message-ID: <c11e2cfd-aa74-c64e-c8f7-7ddd5cde22ce@gmail.com>
Date: Sun, 28 Jun 2020 19:53:21 +0100
User-Agent: Mozilla/5.0 (Windows NT 10.0; WOW64; rv:68.0) Gecko/20100101
 Thunderbird/68.9.0
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Transfer-Encoding: 8bit
Content-Language: en-GB
Subject: Improving output of syntax errors
From: rowan.collins@gmail.com (Rowan Tommins)

Hi all,

During the discussion of "T_PAAMAYIM_NEKUDOTAYIM", there was broad 
agreement that internal token names should not be included in errors 
shown to users. I now have an implementation of this.

Note that we currently only customise the token descriptions placed into 
Bison's hard-coded template, so the format remains "syntax error, 
unexpected %s, expecting %s or %s or %s".


I have distinguished between two types of token:

  * Tokens which always represent the same text (e.g. keywords and
    operators) are represented by their standard form, e.g. 'unexpected
    token "static", expecting "function" ...'
  * Tokens which have variable content are given a user-friendly name,
    shown as well as the actual text when possible, e.g. 'unexpected
    identifier "foo", expecting quoted string ...'

As a special-case, quoted strings show the string's *content* in double 
quotes, e.g. 'unexpected quoted string "foo" ...' rather than 
'unexpected quoted string ""foo"" ...' or 'unexpected quoted string 
"'foo'" ...'.

A "..." is also included if the text is longer than 30 bytes (where 
previously it would have been silently truncated).


For example, given the following:

<<<<<<<<<<<<

The current 8.0 alpha will show this:

Parse error: syntax error, unexpected '<<' (T_SL), expecting identifier 
(T_STRING) or static (T_STATIC) or namespace (T_NAMESPACE) or \\ 
(T_NS_SEPARATOR) in filename.php on line N

The proposed patch will instead show this:

Parse error: syntax error, unexpected token "<<", expecting identifier 
or "static" or "namespace" or "\" in filename.php on line N


For more examples, see: 
https://rwec.co.uk/x/php-parse-errors/comparison.html

The patch can be reviewed at: https://github.com/php/php-src/pull/5722


I am happy to post a small RFC if people think this requires a vote.

Any other feedback is welcome.


(As an aside, the other commonly requested change was to include column 
numbers; this appears to be feasible, but definitely more complex, and 
with potential performance trade-offs. I hope to re-visit this later.)


Regards,

-- 
Rowan Tommins (né Collins)
[IMSoP]