Lucene queries

We should improve how Lucene queries work.  We discussed this.  Assigning to you for feedback.

I still think that the easiest would be a string query like this:

local query = [[ id = '1' and type = 'A' and (size = '2' or size = '3') ]]

If you really want to enforce the lucene concepts, then this would be possible:

local query = { 
    Query.term_AND{id='1', type='A'}, 
    Query.term_OR{size='2', size='3'} 
}

But I am not sure how we would handle the 'NOT' in the second case. What do you think?

Here is the lst of all query types:

http://lucene.apache.org/core/4_9_0/core/org/apache/lucene/search/Query.html

How is your string approach going to handle all these options?  It requires a whole new syntax, and in fact this has been done in the Lucene QueryParser (which I can't even find).  We shouldn't do this.

Your second solution doesn't make sense to me.  I don't even understand how the 2 queries combine.

The NOT actually isn't a problem because I changed NOT in Luan to only take a boolean value, so having NOT also take a query won't break anything.

fschmidt wrote

The NOT actually isn't a problem because I changed NOT in Luan to only take a boolean value, so having NOT also take a query won't break anything.

If this approach works, then you can do it. Just to confirm, consider this:

local query = Query.term{id='1'} and not Query.term{type='B'} or Query.term{size='2'}

This would work, right?

Yes that would work.

Keeping assigned to me to do.

I worked on this and I am not happy with my code.  Now I just want to force all queries to be strings that would go through the lucene QueryParser described here:

http://lucene.apache.org/core/4_9_0/queryparser/org/apache/lucene/queryparser/classic/package-summary.html#package_description

So a query would look like:

query = "type:cart or cart_user_id:" .. user.id

The user would have to learn this, but at least it is well documented and is similar to Google advanced search.

ok?

That's a relief. You can go ahead.

This didn't work because we map field names.

I went ahead and finished the Query approach in rev 288 and jeans rev 131.  I know you won't like it because it is verbose, but I don't think it is so bad.  I like that the implementation is extremely lightweight and that you can figure out everything by reading the code in Lucene.luan .

Okay. Can you please explain to me how to do ANDs and ORs with that?

Query.all{fld1="x",fld2="y"}
Query.any{fld1="x",fld2="y"}

This is okay. It is not perfect, but I think this could work.

closing thread

I see an issue with queries. We can't search for nulls. For example:

    Query.all{user_id = 12, parent_id = nil}

Of course we could use zero as the nil case and do:

    Query.all{user_id = 12, parent_id = 0}

Do you have another solution?

This is the way Lucene works.  So either use another value like "0" or "NIL", or read all values and filter for nil in code.

Okay, closing thread.

I will try query strings again.  If it works, I will support both styles.

done in rev 313.  You can use Query.parse() or just use a string as a query.

Okay. I added another query example to the lucene query tool (rev 317 luan).

It looks fine.  We discusses quoting and I looked into it.  Lucene uses double-quotes, so you would do something like this:

'type:user AND name:"Joe Smith"'

Okay, it works. Thanks.