RE: [PHP-DEV] [UPDATE] NGScan

From: Date: Mon, 09 Jul 2001 22:06:49 +0000
Subject: RE: [PHP-DEV] [UPDATE] NGScan
References: 1  Groups: php.dev 
Request: Send a blank email to php-dev+get-59453@lists.php.net to get a copy of this message 
At 00:52 10/7/2001, Sascha Schumann wrote:

On Tue, 10 Jul 2001, Zeev Suraski wrote:

At 00:26 10/7/2001, August Zajonc wrote:

You see this happen though all the time with other software.

Media players don't need to have 10 ways to decode wav files, but they do.


There's a very big difference here.  It does *exactly* the same
thing.  Unlike codecs, which are all different.  Here, it takes the same
text, and translates it to the same tokens


    It actually does not do that.


Then it is incompatible with the parser.  Unless you mean slightly different tokens in HTML blocks, which is obviously a negligible performance-related implementation detail.


    The language scanner performs some optimizations on a very
    low level.  The stream of tokens the new scanner produces is
    thus different from the Zend scanner's output.  And as I
    might have mentioned on Saturday, it is not a port of the
    Zend scanner to re2c.  It is a complete new implementation
    based on walking through zend_language_parser.h.


That certainly has disadvantages (stability/reliability), but may have some performance advantages with re2c.  I remember you mentioned using more states.

Zeev

Thread (66 messages)

« previous php.dev (#59453) next »