Re: [xsl] problem with transforming mixed content

["Imsieke, Gerrit, le-tex gerrit.imsieke@xxxxxxxxx" <xsl-list-service@xxxxxxxxxxxxxxxxxxxxxx>]

A generic type (a) approach is descibed in my XML Prague 2019 paper 
"Splitting XML Documents at Milestone Elements Using the XSLT Upward 
Projection Method"

https://www.xmlprague.cz/day3-2019/#projection

The solution consists of three passes in distinct modes:

- preprocess: Turn the colon into a <sep/> element, remove the original 
namespace

- main: Use the upward projection splitter on title, splitting at sep, 
name the resulting chunks 'title' or 'subtitle' (the result is a 
split:chunks document with split:chunk children that have a name attribute)

- prostprocess: unwrap the resulting chunks and rename their top-level 
element (which is title, as in the preprocessed document) to the chunk names

Both files, the custom split-title.xsl and the split.xsl package, are 
here: https://gist.github.com/gimsieke/24acca99f88ff232ed453d9177700472

When invoking Saxon (at least 9.8 I think), you need to give 
-lib:split.xsl as a command line argument.

More info is in the paper (or in the slides).

Gerrit

On 15.08.2020 18:03, Wolfhart Totschnig wolfhart.totschnig@xxxxxxxxxxx 
wrote:
> Thank you, Martin, for the explanation of the unwanted namespace 
> declaration. I now know how to get rid of it.
>
> And thank you, Michael, for the detailed explanation of possible 
> approaches to the problem. Graydon's solution will work very well in my 
> case, I think, since I can test for most error-producing conditions 
> before applying the code and the probability of further errors seems 
> sufficiently low in my context and for my purposes. But I am still 
> curious: What would an approach of type (a) look like in my case? It 
> seems to me that implementing this approach would again face the 
> original problem: "turning the punctuation into markup" sounds like a 
> description of the original problem.
>
> Best,
> Wolfhart
>
>
> On 15.08.20 05:16, Michael Kay mike@xxxxxxxxxxxx wrote:
> > This problem comes up from time to time, and it's not easy.
> >
> > There seem to be three general approaches:
> >
> > (a) turn the punctuation into markup (e.g. turn ":" into <colon/>), 
> > then do the manipulation on a tree of nodes
> >
> > (b) turn the markup into punctuation, then do the manipulation on the 
> > resulting text.
> >
> > (c) do it all in one pass
> >
> > I see that Graydon's solution uses serialize() and parse-xml(), so 
> > that's a modern approach to doing (b); while Dimitre's solution does 
> > (c). In general I think the one-pass solution is often more 
> > complicated and runs the risk of not being extensible when the problem 
> > "evolves".
> >
> > One of the things that can cause the problem to "evolve" is error 
> > handling: dealing with situations where the input isn't quite as 
> > simple as in your example. For example, multiple colons, no colons, 
> > colons that are there for a different purpose, etc,. You haven't 
> > included any such cases in your requirements statement.
> >
> > If we ignore error handling, this example of the problem is simpler 
> > than some because the ":" is always going to be in an immediate child 
> > text node; we've seen other examples (like splitting a table) where we 
> > need to look for conditions much deeper in the structure. This is 
> > probably what makes a one-pass solution feasible in this case.
> >
> > Intuitively, my feeling is that (a) is the most rigorous approach, the 
> > one that is least likely to fail because of unanticipated input 
> > conditions. For example, Graydon's solution fails if the input 
> > contains tags with upper-case names, or if it contains comments with a 
> > colon in the text.
> >
> > Michael Kay
> > Saxonica
> >
> > > On 15 Aug 2020, at 03:16, Wolfhart Totschnig 
> > > wolfhart.totschnig@xxxxxxxxxxx 
> > > <mailto:wolfhart.totschnig@xxxxxxxxxxx> 
> > > <xsl-list-service@xxxxxxxxxxxxxxxxxxxxxx 
> > > <mailto:xsl-list-service@xxxxxxxxxxxxxxxxxxxxxx>> wrote:
> > >
> > > Dear list,
> > >
> > > I would like to ask for your help with the following mixed-content 
> > > problem. I am receiving, from an external source, data in the 
> > > following form:
> > >
> > > <title>THE TITLE OF THE BOOK WITH SOME <i>ITALICS</i> AND SOME MORE 
> > > WORDS: THE SUBTITLE OF THE BOOK WITH SOME <i>ITALICS</i></title>
> > >
> > > What I would like to do is
> > > 1) separate the title from the subtitle (i.e., divide the data at the 
> > > colon) and put each in a separate element node;
> > > 2) all the while maintaining the <i> markup;
> > > 3) and perform certain string manipulations on all of the text nodes; 
> > > for the purposes of this post, I will use the example of converting 
> > > upper-case to lower-case.
> > >
> > > So the desired output is the following:
> > >
> > > <title>the title of the book with some <i>italics</i> and some more 
> > > words</title>
> > > <subtitle>the subtitle of the book with some <i>italics</i></subtitle>
> > >
> > > How can this be done?
> > >
> > > I know that I can perform string manipulations while maintaining the 
> > > <i> markup with templates, i.e., <xsl:template match="text()"/> and 
> > > <xsl:template match="i"/>. But in this case I do not know how to 
> > > divide the data at the colon. And I know that I can divide the data 
> > > at the colon with <xsl:value-of select="substring-before(.,': ')"/>, 
> > > but then I loose the <i> markup. So I am at a loss.
> > >
> > > Thanks in advance for your help!
> > > Wolfhart