Hive: work around for non equi left join

Question

Hive does not support non equi joins: The common work around is to move the join condition to the where clause, which work fine when you want an inner join. but what about a

Answer 1

Hive 0.10 supports cross joins, so you could handle all your "theta join" (non-equijoin) conditions in the WHERE clause.

Answer 2

Why not use a WHERE clause that allows for NULL cases separately?

SELECT * FROM OrderLineItem li 
LEFT OUTER JOIN  ProductPrice p 
ON p.ProductID=li.ProductID 
WHERE ( StartDate IS NULL OR OrderDate BETWEEN startDate AND EndDate);

That should take care of it - if the left join matches it'll use the date logic, if it doesn't it'll keep the NULL values intact as a left join should.

Answer 3

Not sure if you can avoid using a double join:

SELECT * 
FROM OrderLineItem li 
LEFT OUTER JOIN  (
  SELECT p.*
  FROM ProductPrice p
  JOIN OrderLineItem li 
  ON p.ProductID=li.ProductID 
  WHERE OrderDate BETWEEN StartDate AND EndDate ) p
ON p.ProductId = li.ProductID
WHERE StartDate IS NULL OR 
  OrderDate BETWEEN StartDate AND EndDate;

This way if there is a match and StartDate is not null, there has to be a valid start/end date match.

Answer 4

1. Create a copy of the left table with added serial row numbers:

   CREATE TABLE OrderLineItem_serial AS
   SELECT ROW_NUMBER() OVER() AS serial, * FROM OrderLineItem;
   

   Remark: This may work better for some tables formats (must be WITHOUT COMPRESSION):

   CONCAT(INPUT__FILE__NAME, BLOCK__OFFSET__INSIDE__FILE) AS serial
   

2. Do an inner join:

   CREATE TABLE OrderLineItem_inner AS
   SELECT * FROM OrderLineItem_serial li JOIN ProductPrice p
   on p.ProductID = li.ProductID WHERE OrderDate BETWEEN startDate AND EndDate;
   

3. Left join by serial:

   SELECT * FROM OrderLineItem_serial li
   LEFT OUTER JOIN OrderLineItem_inner i on li.serial = i.serial;