> Unfortunately, this code is not enough. Imagine a mostly identity
> transform with such grouping. What to do with the elements "caption"
> and "table"?
In an identity transform situation, have no-op templates for caption
and table, with suitable predicates eg
<xsl:template match="caption[preceding-sibling::*[1][self::image]"/>
and then the usual way to handle them from the image template is with a mode:
<xsl:template match="image[following-sibling::*[1][self::caption]] .... ">
(output the image markup)
<xsl:apply-templates
select="following-sibling::*[1][self::caption]" mode="image-caption"/>
with
<xsl:template match="caption" mode="image-caption">
...
</
--
Andrew Welch
http://andrewjwelch.com
