[Haskell-cafe] Using HaXml filters for XML namespaces

[Graham Klyne]

I am (at last!) looking to implement an XML namespace processing filter for 
HaXml, having completed the various structural changes needed to carry 
namespace information in the content model.

The basic idea is to process a document such that:
(a) all xmlns:xx='xxuri' attributes are removed
(b) all attribute and element names of the form xx:foo are replaced by a 
URI xxuri and local name foo, corresponding to a currently in-scope 
namespace declaration.

This raises some questions, which may simply be that I'm overlooking some 
obvious way of using the CFilter combinator library:

(1) is there a combinator for updating attributes, in a fashion similar to 
the way 'chip' updates elements?  (I wrote my own "atip" for general entity 
substitution in attributes, but it seems that there should be a way to do 
this in the library.)

(2) it appears that I need the XML processing filter to carry a symbol 
table of in-scope namespaces, which will be updated through the traversal 
as new xmlnx:xx declarations come into scope.

I cannot see any way to do this with the existing combinators;  I think 
this would need a form of recursive combinator, along the lines of 
'foldXml' that uses a newly computed CFilter function for each successive 
recursion.

I imagine something like:

    mangleXml f mkf c = (f' `o` (chip (mangleXml f` mkf))) c
        where f' = mkf f c

(where I would use mkf to scan the attributes and create an augmented 
version of f that takes into account new xmlns declarations, and f would be 
a function that replaces names with qnames on an element and its attributes.)

Am I missing a trick with the existing combinators?

...

(An alternative approach I considered was to use a monadic form of CFilter 
function which could carry and update the namespace table as the traversal 
progresses, but that appears to require that (a) many of the existing 
combinators would have be re-written, and (b) it doesn't cleanly deal with 
issues of scoping xmlns declarations.)

...

#g


------------
Graham Klyne
For email:
http://www.ninebynine.org/#Contact