Wenchen Fan created SPARK-61978:
-----------------------------------

             Summary: Fix CHAR padding of IN lists that mix CHAR columns with 
other elements
                 Key: SPARK-61978
                 URL: https://issues.apache.org/jira/browse/SPARK-61978
             Project: Spark
          Issue Type: Bug
          Components: SQL
    Affects Versions: 5.0.0
            Reporter: Wenchen Fan


`value IN (e1, e2, ...)` should return what `value = e1 OR value = e2 OR ...` 
returns. When the value is a CHAR column and the IN list mixes CHAR columns 
with STRING/VARCHAR columns or string literals, ApplyCharTypePadding pads 
nothing, so the CHAR-vs-CHAR pairs are compared unpadded. IN then misses rows 
that the equality matches, and NOT IN keeps rows it should drop.

{code:sql}
CREATE TABLE t(c CHAR(2), c5 CHAR(5), s STRING, v VARCHAR(5)) USING parquet;
INSERT INTO t VALUES ('a', 'a', 'a', 'a');

SELECT c = c5, c IN (c5, s), c NOT IN (c5, s), c IN (c5, v), c IN (c5, 'x'), c 
IN (s, 'a') FROM t;
-- actual:   true, false, true, false, false, false
-- expected: true, true, false, true, true, true
{code}




--
This message was sent by Atlassian Jira
(v10.3.26#10030026)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to