<xsl:sort>

This seems to be a good point to introduce <xsl:sort>. This element can be used inside <xsl:apply-templates> or <xsl:for-each> to alter the order in which nodes are processed. In most cases, when either of these statements selects these nodes, they will be processed in document order – that is, the order in which they were encountered when processing the document from top to bottom. This order can be altered using <xsl:sort>.

At the start of this chapter, we introduced a book catalog with just two books in it and a simplified structure. Now is a good time to use the full catalog, catalog.xml, which is available in the code download. Here are the first two books from this version:

We will be using this quite a bit later in the chapter, but for now we will use it to see the effect of <xsl:sort>. Firstly, let's get a list of the book titles using ShowTitles.xsl:

When applied to catalog.xml, this gives the result:

If you look through the catalog, you will see that this is the order in which the books are listed. Now let's add an <xsl:sort> statement to the stylesheet (and re-save it as CatalogSort.xsl):

The books are now listed in the order of their id attributes. These are their ISBN numbers without the hyphens. We could have just used the ISBN numbers, but the sort would then get confused if the style of hyphenation was not consistent. This is the result of using the stylesheet above on catalog.xml:

Inspection of the catalog will show that the titles are now listed in ISBN order.

There are various optional attributes of <xsl:sort> that control the sorting. We have already used select. Had we omitted this, the string value of the current node would have been used, rather than the id attribute. In this case, since the node has element content, this is not really a safe option (the string value would be the concatenation of the string values of the descendant nodes), but had we instead created a template for the <Title> element, we could have sorted on this by using the default value for the select attribute. The stylesheet might then have looked like this (CatalogSort2.xsl):

And the result on transforming catalog.xml would have looked like this:

Because it is now applied to the <Title> elements, the line:

causes the titles to be output in alphabetical order.

Using the data-type attribute of <xsl:sort>, we can specify whether the data on which we are sorting is text (the default) or number. We sorted our ISBN values as text, which works in this case of fixed length strings, but in other cases, there would be a difference. The sequence of numbers from 1 to 20 sorted as text would be:

1,10,11,12,13,14,15,16,17,18,19,2,20,3,4,5,6,7,8,9

The order attribute has two possible values: ascending (the default) and descending, which
are self-explanatory.

The lang attribute controls language-specific sorting orders. For example, in German, it is conventional to include the letter äafter the lettera, while in Swedish it would come afterz. The possible values are the same as those of the xml:lang attribute and are specified inIETF RFC 1766 (http://www.cis.ohio-state.edu/htbin/rfc/rfc1766.html), "Tags for the Identification of Languages".

When sorting textual information, we can choose whether upper-case or lower-case values come first using the case-order attribute. This has values of upper-first and lower-first. If the former is used, when values start with the same letter, values starting with upper-case letters will occur in the output before any starting with lower-case letters. The default for this is language-dependent, and not specified in the recommendation. With no language specified, both XT and SAXON ignore case when sorting.

Sorting can be achieved on multiple criteria by using <xsl:sort> elements sequentially. In this case, sorting will occur first using the criteria defined by the first element, then sorted within these results using the later elements in order. If, for some strange reason, we wanted to sort all the speeches of Hamlet first in order of speaker, then by number of lines in the speech, then in alphabetical order of the text, we could use the following stylesheet, HamletSort.xsl:

Using this just on our extract of the first scene (HamletExtract.xml), all Bernardo's speeches come before Francisco's, then for each speaker, all the one line speeches come before the two line speeches and within these criteria, the speeches are listed in alphabetical order. Notice that the apostrophe comes before any letters, as it does in English-language dictionaries: