I think we can divide the discussion to (a) whether or not we need
this feature, and (b) everything else.
Regarding (b), I agree completely with everything Stas said (no #xyz
syntax, no code generation in the scanner, etc.).
Regarding (a), we're likely to get the standard discussion of
simplicity/clarity vs. power. I think that in this particular case,
we can possibly come up with a solution that satisfies both camps completely.
The solution I propose is to create internal functions that will set
'virtual' filenames and line ranges, that will be saved in some sort
of a translation table, of real filename + line range to virtual
filename + lineno. The error callback can be made smarter to search
in the translation table, and in case of a match, report the virtual
filename:linenumber in addition to the real ones.
With that solution, there's no need to change any of the internal
structures and there are no compatibility concerns.
Thoughts?
Zeev
At 08:32 11/08/2006, Andi Gutmans wrote:
Some feedback from Stas on the subject. I agree with him completely but
wanted to get a second opinion.
Honestly I think while this is somewhat useful for auto-generated files,
it's not worth the risk and effort. It's a very marginal feature and very
few people will use it.
Supporting it in a non-intrusive way, at high quality and without making it
significantly harder for debuggers and profilers to work with this is just
not worth it...
Andi
> -----Original Message-----
> From: Stanislav Malyshev [mailto:[EMAIL PROTECTED]
> Sent: Tuesday, August 08, 2006 7:22 AM
> To: Andi Gutmans
> Subject: Re: FW: [PHP-DEV] Long awaited line directive
> >
> I'm afraid this one would confuse debuggers. Also, if there's
> any error in PHP code following line directive, it will be
> reported with new filename, even if there's actually no code
> causing this problem in this filename.
> Also, I don't like too much the #-syntax too. And I'm not
> sure it's healthy to generate code in scanner - AFAIK there
> was no code generated in the scanner previously. What would
> happen if someone runs tokenizer (which uses lex_scan) on
> such script? Or even highlighter?
>
> --
> Stanislav Malyshev, Zend Products Engineer [EMAIL PROTECTED]
> http://www.zend.com/ +972-3-6139665 ext.115
>
--
PHP Internals - PHP Runtime Development Mailing List
To unsubscribe, visit: http://www.php.net/unsub.php
--
PHP Internals - PHP Runtime Development Mailing List
To unsubscribe, visit: http://www.php.net/unsub.php
