QueryParser produces empty BooleanQueries when all clauses are stop words

• Key: LUCENE-25
• Type: Bug
• Status: Resolved
• Resolution: Fixed
• Priority: Major
• Assignee: Unassigned
• Reporter: Puk Witte
• Component/s: QueryParser
• Affects Version/s: 1.0.2

Environment

Operating System: All
Platform: PC

Description

When I want to do the following query (example):
(fieldx : xxxxx OR fieldy : xxxxxxxx)AND(fieldx : stopword OR fieldy : stopword)
it will search (after passing the Analyzer including a StopFilter) for
(fieldx : xxxxx OR fieldy : xxxxxx) AND() and give a wrong searchresult or a
NullPointerException.

Comments

[Otis Gospodnetic]

That's interesting.
Does this happen even if you put spaces in the query 'properly', so that instead
of this:

(fieldx : xxxxx OR fieldy : xxxxxxxx)AND(fieldx : stopword OR fieldy : stopword)

you use:

(fieldx: xxxxx OR fieldy: xxxxxxxx) AND (fieldx: stopword OR fieldy: stopword)

Does it still happen?
If so, please include the stack trace, preferably with line numbers, so that we
know which code to fix.
Thanks.

[Jean-François Halleux]

We will try not to have to organize a party for the second birthday of this one.

Ok, the root cause for this is the QueryParser (once again), more specifically
getBooleanQuery which is returning an empty query even if there are no clauses.
When everything is a stopword, it could be the case that there are no clauses.

There are several possible fixes. Included in the attached is the less
disruptive one I could think of. It is "only" breaking three tests in
TestQueryParser.

However for the long term, I think that we should think of what it means to
parse an empty query ("") and what should be the expected result (null,"",
exception, ...)

Have fun,

Jean-Francois Halleux

[Jean-François Halleux]

Created an attachment (id=10302)
Fix elimination of empty query conjunction - careful this patch is cumulated with my previous regarding escapting character handling

[Christoph Goller]

• • • Bug 9110 has been marked as a duplicate of this bug. ***

[Hoss Man]

revising title/component to better reflect real problem

[Grant Ingersoll]

Attached patch adds a test to QP for this. It now seems like QP is doing reasonable things with stopwords at least for a few cases. Of course, I am not sure what QP should return when all the terms are stopwords (it currently returns a BooleanQuery with zero clauses)

I am not sure when this was fixed.
In other words, I would like to close this.

[Doron Cohen]

I think it is LUCENE-933 that fixed this one.

[Aviran]

This issue also affects version 2.2.0

[Mark Miller]

I don't know when this was fixed, but the example given now works, Grant's tests are already committed, and I think any QueryParser stop word issues that remain should get a new ticket. I don't think this one will help anything anymore.