Hash Gazetteer and Feature Separator

[greenwoodma]

The hash gazetteer doesn't support feature separators. In itself this isn't an issue except.... if you create an instance of the PR with the default values it uses the ANNIE gazetteer files which contain entries with features (see the state abbreviation list for an example) which mans the entire line becomes the entry in the gazetteer and so will never match the documents.

[johann-petrak]

So I guess the only way to deal with this properly would be to ask for a feature separator and then ignore the features, and also warn about them getting ignored?
On the other hand, what advantage does the HashGazetteer really have over any of the other gazetteers? Why is it still around?

[greenwoodma]

My suggestion would be to just remove the default value for the list URL from the metadata so it doesn't default to using the Annie lists which it doesn't support. I certainly wouldn't spend time updating the code.

[johann-petrak]

OK. If there is no real advantage of the hash gazetteer over the annie/extended ones, maybe we can plan on deprecating/removing it? May help to avoid confusion and cut down on support request emails on the list :)

[greenwoodma]

It's difficult to deprecate given it's in the ANNIE plugin, but I suppose we could make it a hidden PR so that any apps that use it won't break but people won't be able to create new instances. @ianroberts any thoughts on that?