Question

I have three table as follows:

documents (id, content) 
words (id, word) 
word_document (word_id, document_id, count)

Words table have all the words occurred in all documents and word_document relate a word to a document and the count of that word in that document.

I want to write a query to search for two words and returns only the documents which have both word ordered by the sum of the count of both words in documents.

For example

DocA: green apple is not blue
DocB: blue apple is blue
DocC: red apple is red

now a search for apple and blue returns:

DocA, 3
DocB, 2

becaus:

DocA contains both words and 3 of them
DocB contains both words and 2 of them
DocC only contains one word

I used intersect successfully but it doesn’t return the count sum and no order.

Answer 1

For those who want this, this only works:

select wd.document_id, (wd.count + d.count) as tcount from word_document as wd
join words as w on w.id = wd.word_id
join
(select document_id, count from word_document 
join words on words.id = word_document.word_id
where words.word = "apple") d on d.document_id=wd.document_id
where w.word = "blue" order by tcount desc

You can create temp table from the inner query and execute outer on that. It can be done recursively for more words.