RDF translator loses creators linked via a symbol [patch]

[fbennett]

The RDF creator attempts to resolve symbol links to creators in other rdf:Description blocks as containers. The identifier is slightly different than a container (there is no _ underscore following the identifier #), so the attempt fails; but the existing list of creator symbols is overwritten, resulting in an empty list.

The attached patch avoids overwriting the list when the link attempt fails. With the patch, more foaf creator data should become available to the translator (I've run across two sites so far that place creator info in an rdf:Description block separate from other metadata ... and I've only looked at two).

[ajlyon]

This patch doesn't apply cleanly to trunk.

[fbennett]

Replacing previous patch. This will apply cleanly to the trunk.

[fbennett]

Sorry about that. New version up now.

[ajlyon]

-	if(typeof(creators[0]) != "string") {	// see if creators are in a container
-		try {
-			var creators = Zotero.RDF.getContainerElements(creators[0]);
-		} catch(e) {}

Do we not need the try-catch? Is there a risk of Zotero.RDF.getContainerElements(..) throwing an error here?

[fbennett]

Not sure, perhaps it's safer to keep it. If it's kept, it should probably log a warning on failure though.

[ajlyon]

patch with exception handling

[ajlyon]

New patch handles exceptions, logs a message. Look good?

[fbennett]

There seems to be a repeated test layer for (typeof(creators[0]) != "string").

[ajlyon]

So there is. Although that was in your patch too. :P New patch attached.

[ajlyon]

Yet another patch. Now with debug logging and no extraneous logic.

[fbennett]

Oops. Looks great now.

[ajlyon]

Fixed in r7913.

[simon]

I reverted this patch in r7947 because it seems to break import of multiple creators. I think the problem is that creators.length == 1, since it's a single reference to the Seq, but newCreators.length > 1. Additionally, I don't know if Zotero.RDF.getContainerElements() throws anymore since the Tabulator transition, but this patch would also throw if it did (since you would reference undefined.length at line 79).

[ajlyon]

fbennett: Do you have sample RDF data that exercises this bit of the code? Before attempting to land this again, I'd like to get some test coverage.