Empty strings shouldn't be indexed

[mlissner]

If you create an object with attributes that are empty strings, they'll get indexed.

So:

item.attribute = ''

Will result in an item in your index with attribute = ''.

This is very easy to fix since there's already a check to make sure that attributes aren't None. All that needs to happen is for the line to also say, and a != ''.

[tow]

Not obvious to me - why shouldn't an empty string be indexed?

Or rather - why shouldn't an empty string be stored - an attribute on an object may well be stored in Solr without that field being indexed, and it is a perfectly reasonable thing for an object to have an attribute which is an empty string.

[mlissner]

Thanks for the reply.

I guess the way I look at this is that I want one and only one way to know if an attribute has values.

Django itself solves this problem by having the convention that CharFields are never null. So, the only value they can have is "". (As of about Django 1.7, a warning will be issued if your model has a nullable CharField.)

My idea here is along the same lines, but flips it so that blanks are normalized to None. It would also be possible, (though harder, I think?) to make None values into blanks instead.

So, when indexing:

	""	None	Thoughts
current	""	no attr	This is OK, but inconsistent
proposed	no attr	no attr	Good, easy to implement, opposite of Django convention
alt proposal	""	""	Best? Harder to implement, aligns with Django convention

Thanks again for the reply. Glad to see your involvement in the project.