[xsl] Selecting nested unique values, Muench approach not a good fit?

[chrislott@xxxxxxxxxxx]

Dear XSL experts,

I hope you can please find a minute to look at my question.  I think it may
be a grouping question, but the Muench-style solutions I found so far have 
not done the trick (or I misunderstood how to apply them).

I would like to select a unique set of node values, basically a list of 
names, from a hierarchical structure in which the names might be duplicated.
Example XML is at the end.  To explain, a chunk has a collection of fields.
Every field has values.  A value may be decorated with one or more names
called "seeds".  The seeds are just simple strings.  I would like to select
the unique set of seed names from a chunk.  What seems to be complicating
this situation is the fact that I may have many <chunk>s in my XML file, and
each chunk may re-use the same seed names again.  For example, chunk 1 may
have fields "one" and "two", with seeds "0" and "1" attached to values in
fields one and two.  Then chunk 2 might well have seed "0" attached to some
values in chunk 2's seeds.  I hope I'm making sense; the example below will
probably help.

I read about and tried the Muench approach using xsl:key, the key function,
and generate-id to select the first node that matches a particular key value 
so as to avoid visiting duplicates.  In my case, an input file may have
many "seed" nodes, all with the same value.  Because they appear in different
chunks, I don't think there is any ambiguity -- seed "0" in one chunk 
seems quite different from seed "0" in another chunk -- but the Muench 
approach seems to want them to be globally unique (correct me where I'm wrong). 

In the example below, I'd like to select the distinct seed values for the
chunk named 'j'.  I.e., I want to write a select statement that yields 
"0" and "1".  A node-set is fine, a list of string would also be fine
(but maybe that's my procedural programming background showing through).
Once I have that list, I can wallk the fields, picking out values that
have the appropriate "seed" decorations without repeats.

It seems that if the seed names could be made unique across chunks, the 
Muench approach might work.  I.e., if the first chunk has seed names 
"chunk1-0", "chunk1-1", etc. and the next chunk has seed names 
"chunk2-0", "chunk2-1", etc.  If this is the only way to solve the problem,
I may be forced to go after it that way.

Any and all hints will be greatly appreciated.  Thanks in advance!

chris...

p.s. can you blame me for wanting a Perl hash right about now?

--

<chunks>

<chunk>
    <name>j</name>
    <usefields>
        <field>
        <name>field3</name>
        <valids>
            <validvalue>
                <value>three</value>
                <seeds>
                    <seed>0</seed>
                </seeds>
            </validvalue>
            <validvalue>
                <value>three</value>
                <seeds>
                    <seed>0</seed>
                </seeds>
            </validvalue>
            <validvalue>
                <value>drei</value>
                <seeds>
                    <seed>1</seed>
                </seeds>
            </validvalue>
        </valids>
        </field>
        <field>

        <name>field4</name>
        <valids>
            <validvalue>
                <value>four</value>
            </validvalue>
            <validvalue>
                <value>vier</value>
                <seeds>
                <seed>0</seed>
                <seed>1</seed>
                </seeds>
            </validvalue>
            <validvalue>
                <value>quatro</value>
            </validvalue>
        </valids>
        </field>
    </usefields>
</chunk>

( followed by another <chunk> with its own fields and seed names, and so on. )

</chunks>

 XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list