Jesus Camacho Rodriguez created HIVE-10770:
----------------------------------------------
Summary: Recognize additional common factors in Filter predicates
Key: HIVE-10770
URL: https://issues.apache.org/jira/browse/HIVE-10770
Project: Hive
Issue Type: Bug
Reporter: Jesus Camacho Rodriguez
Assignee: Jesus Camacho Rodriguez
Currently, we canonize predicates at the term level (i.e. "a or b or a" becomes
"a or b" but we do not attempt to recognize terms that are equivalent).
Further, we do not exploit e.g. the symmetry of '=' (i.e. a = b iff b = a).
- A first extension would be to normalize comparisons between field references
and literals so that the lower field reference is always on the left. So, "$6 =
$3" becomes "$3 = $6"; "$6 > $3" becomes "$3< $6". And "literal <= $5" becomes
"$5 >= literal". This would not damage performance, and would improve a few
plans.
- Another possible extension. Given the predicate "(a or b) and ((x and a) or
(y and b))", the first factor can be removed so the expression consists only of
"(x and a) or (y and b)".
One possible way to recognize such cases is to transform the second factor to
CNF i.e. "(x or y) and (x or b) and (a or y) and (a or b)", and as it contains
"(a or b)", we would know that we can discard it. Then we could just use the
original expression i.e. "(x and a) or (y and b)" in the predicate, once we
have done the check.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
