Re: [XPATH] & [XQUERY] A.2.4 Lexical rules table

Hi Alberto.  Thanks for pointing this out.  The inclusion of NotNumber, 
and the NotOperatorKeyword non-token recognition was somewhat inadvertent, 
though, in the end, they may indeed be included. 

So, why are they there at all?  In the November 2003 draft, the expression 
"10divide3" would actually parse as "10 div ide3", because there was no 
rule that "div" would not be recognized as an operator in that state (i.e. 
no rules that said the operator had to have a space after it).  Clearly 
this would be unacceptable.  Also, the working group wanted to make a 
clean general rule that textual operators such as 'div' be space 
delimited, so we wanted to make "10div 2" be illegal.   So, after much 
discussion, we added the rules for delimiting and non-delimiting terminals 
found in http://www.w3.org/TR/xquery/#N157EE.  The use of NotNumber and 
NotOperatorKeyword is my attempt to implement these rules in a general way 
in the test parser, and these anti-tokens were reflected back into the 
spec because of our production process.  I think they probably do not 
belong in the spec itself, but, on the other hand, I hate to hide 
information that enables the test parser to work.  This is illustrative of 
why it is problematic to have these tables be normative, and why it would 
be better to make them non-normative.  If they can be made non-normative, 
then I would probably keep the anti-tokens and add explanations why they 
are there.

The inclusion of Unordered is a pure production bug.  It has been removed 
from the XPath spec.

-scott




Alberto Massari <amassari@progress.com> 
Sent by: public-qt-comments-request@w3.org
08/25/2004 12:49 PM

To
public-qt-comments@w3.org
cc

Subject
[XPATH] & [XQUERY] A.2.4 Lexical rules table







These lexical rules table have a few symbols that don't match any lexeme:

- in the DEFAULT state, "NotNumber"
- in the OPERATOR state, "Unordered" (only in XPath)
- in the OPERATOR state, "NotOperatorKeyword"

Do they have a special meaning (like "[NotOccurrenceIndicator]" in the 
OCCURRENCEINDICATOR state, but this is described in the same section) or 
should they be discarded?

Thanks
Alberto

Stylus Studio
http://www.stylusstudio.com

Received on Monday, 30 August 2004 16:41:54 UTC