Re: Alternative algorithm to restrict use of data/string in patterns.

[James Clark <jjc@jclark.com>]

You would need to change the check() function as well.  Instead of:


    case: p == <group> p1 p2 </group>
        check( p2, fpds );
        check( p1, calcFPDS(p2,fpds) )


case: p = <group> p1 p2 </group>
    check(p2, fpds);
    if (isNullable(p2))
       check(p1, max(fpds, calcFPDS'(p2));
    else
       check(p1, calcFPDS'(p2));


----- Original Message -----
From: "Kohsuke KAWAGUCHI" <kohsuke.kawaguchi@eng.sun.com>
To: "TREX ML" <trex@lists.oasis-open.org>
Sent: Thursday, May 03, 2001 11:59 AM
Subject: Re: Alternative algorithm to restrict use of data/string in
patterns.


>
> > > If calcFPDS can be made to the unary function, then what will be
> > > calcFPDS(<attribute/>) and calcFPDS(<empty/>)?
> >
> > I think you would return "element",  since "element"|X = X for any X.
>
> I still think calcFPDS cannot be an unary function.
>
> Consider checking the following pattern (with calcFPDS', which is the
> unary version of calcFPDS)
>
> <oneOrMore>
>   <group id="g1">
>     <string>abc</string>
>     <group id="g2">
>       <element />
>       <group id="g3">
>         <string>def</string>
>         <empty />
>       </group>
>     </group>
>   </group>
> </oneOrMore>
>
>
>
> check( <oneOrMore> ....... </oneOrMore>, "element" )
>
> => check( <group id=g1> ... </group>,
>        calcFPDS'( <oneOrMore>...</oneOrMore> )
>
> => check( <group id=g1> ... </group>, "text" )
>
> => check( <group id=g2> <element/> <group id=g3/> </group>, "text" )
>  & check( <string>abc</string>, calcFPDS( <group id=g2 /> ) )
>
>    consider the first check
>
> => check( <group id=g3> <string/> <empty/> </group>, "text" )
>  & check( <element/>, calcFPDS(<group id=g3 />) )
>
>    again consider the first one
>
> => check( <empty/>, "text")
>  & check( <string>def</string>, calcFPDS(<empty/>) )
>
>   consider the second check.
>   According to your suggestion, calcFPDS(<empty/>) is "element".
>
> => check( <string>def</string>, "element" )
> => OK
>
>   So it can't detect that the pattern is illegal. However, the pattern
> is actually illegal because it is equivalent to
>
> <oneOrMore>
>   <string>abc</string>
>   <element />
>   <string>def</string>
> </oneOrMore>
>
>
>
> --
> Kohsuke KAWAGUCHI                          +1 650 786 0721
> Sun Microsystems                   kohsuke.kawaguchi@eng.sun.com
>
>
>
>