Idiom for templating similar to other templating engines like Velocity? - templates

Question

I have an XSLT(2.0) file; which takes an input XML data file and creates DDL/SQL Statements.
It works just fine. But it is a bit difficult to maintain, as it contains a lot of formatting information in 'concat' statements like this:
<xsl:value-of select="concat('CREATE USER ',$username,' IDENTIFIED BY ',$password,';',$nl)"/>
What I would prefer to do would be to encode my SQL Statements in a manner like this instead:
<some-enclosing-elements>[...]CREATE USER <username/>, identified by <password/>; [literally a newline here][...]</some-enclosing-elements>
I would perhaps keep this format above in the XML data file itself in a 'lookup' table at the top of the either the XSLT or the data document iself (I can't work out which yet).
Is there a standard idiom that would allow this kind of templating ?
Any ideas ?
By the way; the data document contains many different users to create of course

Answer 1

The AVT approach is just a little bit too devious for my taste. I tend to rely on the implicit concatenation done (in 2.0) by xsl:value-of:
<xsl:value-of select="'CREATE USER', $username, 'identified by', $password"/>
Another approach which I have used in applications where this kind of text templating is significant is to essentially write my own templating engine within XSLT; have a "message file" containing message templates in the form
<message nr="1">CREATE USER <p:user/> IDENTIFIED BY <p:password/></message>
and then write template rules to expand the messages by substituting the parameters.

Answer 2

#xiaoyi is right, showing the main alternative to using concat(). However that's even more notation-heavy than the concat(), since you have to keep repeating <xsl:value-of select="..." />.
A nice alternative would be to use attribute value templates (AVTs):
[...]CREATE USER {username}, identified by {password};
[...]
But ATVs are only available for (certain) attributes, not for text nodes (directly). How do you use them for this purpose?
One way in XSLT 2.0 would be to use an AVT to create a new literal result element with an attribute; specify the value of that attribute using an AVT; and then select the value of the new attribute:
<xsl:variable name="query">
<dummy val="[...]CREATE USER {username}, identified by {password};
[...]" />
</xsl:variable>
<xsl:value-of select="$query//#val" />
Yes that's some significant overhead per formatted string, but there's very little overhead per field within the string. You could do several strings together like this:
<xsl:variable name="queries">
<q val="[...]CREATE USER {username}, identified by {password};
[...]" />
<q val="[...]CREATE TABLE {tablename}, blah blah;
[...]" />
</xsl:variable>
<xsl:value-of select="$queries/q[1]/#val" />
<xsl:value-of select="$queries/q[2]/#val" />
You could use position indices as above, or use an id attribute to identify each string.
I have not seen this method advocated elsewhere, so I'd be curious to hear what others think about it.
Never mind, except...
Given the simpler approach shown by Michael Kay's answer, I don't think there's any point in doing it this way. I guess that explains why others haven't advocated this method. :-)
The only situation I can think of where this approach might still be of use is if you can't use XSLT 2.0, but you do have access to the nodeset() extension function (e.g. in IE or .NET environment). In that case you would need to wrap nodeset( ) around $queries wherever you used it in an XPath expression before /.