I assume it's a bug. In xmlAddPrevSibling, the check is done as follow:
if (elem->type == XML_TEXT_NODE) {
if (cur->type == XML_TEXT_NODE) {
xmlChar *tmp;
tmp = xmlStrdup(elem->content);
tmp = xmlStrcat(tmp, cur->content);
xmlNodeSetContent(cur, tmp);
xmlFree(tmp);
xmlFreeNode(elem);
return(cur);
}
if ((cur->prev != NULL) && (cur->prev->type == XML_TEXT_NODE) &&
(cur->name == cur->prev->name)) {
xmlNodeAddContent(cur->prev, elem->content);
xmlFreeNode(elem);
return(cur->prev);
}
which means that if you have something like
<a/>prev<cur/>
adding a text node before <cur/> does not merge the text. I don't understand
why it's checking for cur->name == cur->prev->name instead of elem->name ==
cur->prev->name.
I also wonder how much of problem it is to keep text nodes unmerged (as I
believe it's possible when handling multiple cdata sections or entities, etc
like in:
<![CDATA[a]]>b<![CDATA[c]]>
Regards,
Jerome
_______________________________________________
xml mailing list, project page http://xmlsoft.org/
[email protected]
http://mail.gnome.org/mailman/listinfo/xml
