[
https://issues.apache.org/jira/browse/IMPALA-7310?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16621126#comment-16621126
]
Paul Rogers commented on IMPALA-7310:
-------------------------------------
As noted above, the code uses -1 as an "undefined" selectivity marker. We saw
that, if NDV=0, we leave the value at "undefined". A quick read suggests we do
special handling further up the expression tree. See
{{PlanNode.coputeCombinedSelectify}}.
1. Remove conjuncts (expressions) with undefined selectivity.
2. If no expressions remain, add one with {{DEFAULT_SELECTIVITY}}.
This means that if we have no metadata, we treat the following identically:
{noformat}
WHERE x = 10
WHERE x = 10 AND y = 20 AND z = 30
{noformat}
Both are assumed to have a 0.1 selectivity, when one might argue that the
second should have 0.1^3 = 0.001 selectivity. (Actual code uses an exponential
back-off, ignored here.)
Note also the somewhat over-generalize use of {{DEFAULT_SELECTIVITY}}. If x =
10 has 0.1 selectivity, then does it make sense to assume that x < 10 has the
same? Some DB's assume a larger number (0.5, say) for inequalities. This
subtlety is lost in the 0.1 for everything approach.
Is this all this a bug or a feature?
> Compute Stats not computing NULLs as a distinct value causing wrong estimates
> -----------------------------------------------------------------------------
>
> Key: IMPALA-7310
> URL: https://issues.apache.org/jira/browse/IMPALA-7310
> Project: IMPALA
> Issue Type: Bug
> Components: Frontend
> Affects Versions: Impala 2.7.0, Impala 2.8.0, Impala 2.9.0, Impala 2.10.0,
> Impala 2.11.0, Impala 3.0, Impala 2.12.0
> Reporter: Zsombor Fedor
> Assignee: Paul Rogers
> Priority: Major
>
> As seen in other DBMSs
> {code:java}
> NDV(col){code}
> not counting NULL as a distinct value. The same also applies to
> {code:java}
> COUNT(DISTINCT col){code}
> This is working as intended, but when computing column statistics it can
> cause some anomalies (i.g. bad join order) as compute stats uses NDV() to
> determine columns NDVs.
>
> For example when aggregating more columns, the estimated cardinality is
> [counted as the product of the columns' number of distinct
> values.|https://github.com/cloudera/Impala/blob/64cd0bb0c3529efa0ab5452c4e9e2a04fd815b4f/fe/src/main/java/org/apache/impala/analysis/Expr.java#L669]
> If there is a column full of NULLs the whole product will be 0.
>
> There are two possible fix for this.
> Either we should count NULLs as a distinct value when Computing Stats in the
> query:
> {code:java}
> SELECT NDV(a) + COUNT(DISTINCT CASE WHEN a IS NULL THEN 1 END) AS a, CAST(-1
> as BIGINT), 4, CAST(4 as DOUBLE) FROM test;{code}
> instead of
> {code:java}
> SELECT NDV(a) AS a, CAST(-1 as BIGINT), 4, CAST(4 as DOUBLE) FROM test;{code}
>
>
> Or we should change the planner
> [function|https://github.com/cloudera/Impala/blob/2d2579cb31edda24457d33ff5176d79b7c0432c5/fe/src/main/java/org/apache/impala/planner/AggregationNode.java#L169]
> to take care of this bug.
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
