At 05:31 PM 3/1/2002 +0100, you wrote:
On Thu, 28 Feb 2002, Bruce Atherton <[EMAIL PROTECTED]> wrote:

> This raises an interesting question, though. Should Selector
> containers accept <patternset>s as elements and internally convert
> the <include>s and <exclude>s into equivalent selectors?

What about <patternset> is a selector container?

Well, of course it is, but I was using "selector container" to refer to <and>, <or>, and <not>. Sorry for being unclear. So you might have:


<patternset id="readonly">
    <permissionselect type="r"/>
</patternset>

<patternset id="javaclasses">
    <filenameselect name="**/*.java"/>
</patternset>

<patternset id="readonly-or-javaclasses">
    <or>
        <patternset refid="readonly"/>
        <patternset refid="javaclasses"/>
    </or>
</patternset>

Doing things this way, I'm not sure whether references make sense on the individual <and>, <or>, and <not> tags, but it requires that those three accept patternsets as children. It also means that the <or> has to interpret the <include>s and <exclude>s of the patternset, since the patternset has no built in methods to validate a file itself (not sure why not).

Or are you suggesting '<patternset type="and"|"or"|"not"><selector/></patternset>', with "and" being the default? Probably not, since it loses the clarity of Rob's proposal.

> Another question is whether I should still include the "inverted"
> attribute, given the existence of <not>?

Answer this on a case by case basis.  In general I'd say no, but some
selectors may become frequently used in both their normal and the
inverted form.

That seems sensible. Thanks.



--
To unsubscribe, e-mail:   <mailto:[EMAIL PROTECTED]>
For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>



Reply via email to