Question

I am attempting to write a stored procedure that can take in k words and return back the answers with the k words in them.

word table

• word
• wordId

description word inverted index

• descriptionID
• wordID
• count

So – given k words, how would I efficiently get all the descriptionIDs that contain the ALL the k words?

Answer 1

The easiest way in a single query is to use a combination of the IN and HAVING clauses. Example:

  SELECT dwi.descriptionid
    FROM DESCRIPTION_WORD_INDEX dwi
    JOIN WORD w ON w.wordid = dwi.wordid
   WHERE w.word IN ('a', 'b', 'c')
GROUP BY dwi.descriptionid
  HAVING COUNT(DISTINCT w.word) = 3

The number of values in the IN clause must match the number of distinct values counted in the HAVING clause.

Another means is to JOIN to multiple copies of the WORD table (in this case) for each specific word:

  SELECT dwi.descriptionid
    FROM DESCRIPTION_WORD_INDEX dwi
    JOIN WORD a ON a.wordid = dwi.wordid
               AND a.word = 'a'
    JOIN WORD b ON b.wordid = dwi.wordid
               AND b.word = 'b'
    JOIN WORD c ON c.wordid = dwi.wordid
               AND c.word = 'c'