If triggering a confluence import / XWiki xar export via the WikiStream converter, the result is not written into a file in xar format, but all pages are written into one big xml file.
There are two "reasons" for this:
- the ConfluenceInputWikiStream never sends beginWiki /endWiki events (makes somewhat sense, as the format seems ont to export spaces, not complete wikis
- the XAROutputWikiStream creates a XARWikiWriter on "beginWiki", however. If it does not have such a writer when starting to create the content, it uses a WikiStreamXMLStreamWriter instead.
I would like to work on a pull request here, but I am not sure at which end this should be fixed (if it should be fixed at all and is not considered a feature). The simplest thing (that "works for me") is to add "beginWiki"/"endWiki" calls in the ConfluenceInputWikiStream.read method, even though one has to invent a name for the wiki (I can see no such data serving as a wiki name in the *.xml.zip file, and I guess Confluence has no need for wiki names either, as it has no multi-wiki feature.