[cxml-devel] Parsing yields TEXT elements with #\Newline #\Tab

[David Lichteblau]

Quoting Marco Antoniotti (marcoxa at cs.nyu.edu):
> I get all my actual RUNE-DOM::ELEMENTs interleaved with "bogus" TEXT  
> elements containing just #\Newline and #\Tab (or more #\Tab).
> This is obviously an artifact of parsing.  (See attached figure from a  
> 15 minutes CXML browser I whipped up)
>
> What I do not know it's (1) whether this makes sense or not, or (2)  
> whether it is dependent on my platform (LWM).

Certainly -- XML preserves whitespace in character data, except for CRLF
to LF normalization.

There are no universally correct rules for whitespace normalization in
XML, and in general, any change to whitespace could change  the meaning
of the document.


One rule that is relatively common is to consider whitespace
insignificant in "element content", e.g. in places where no
non-whitespace text nodes are allowed by the DTD.

This rule is implemented by CXML:MAKE-WHITESPACE-NORMALIZER
(see http://common-lisp.net/project/cxml/sax.html#misc), which may be
helpful in your case.

However, note that the limitation to element content means that you
actually need to write or find a DTD that matches your document.
Without a DTD, this approach doesn't work.


Other approaches are to use HTML rules for whitespace normalization
(which are a more tricky to get right though, and cxml does not provide
a ready-to-use function for this) or to discard all whitespace.  It
really depends on the schema and application.


(Note that we would like to have some support for this in cxml, because
whitespace rules also matter for indentation, and at some point we would
like to have more flexible/correct/useful indentation modes in our
serializer.  Whitespace stripping could be considered as a form of
indentation, in the sense that it is a "removal of all indentation".
But so far, I haven't found the time to implement anything in this
direction.)


d.