[ https://issues.apache.org/jira/browse/HIVE-16178?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Zoltan Haindrich updated HIVE-16178: ------------------------------------ Description: h3. corr the standard defines corner cases when it should return null - but the current result is NaN. If N * SUMX2 equals SUMX * SUMX , then the result is the null value. and If N * SUMY2 equals SUMY * SUMY , then the result is the null value. h3. covar_samp returns 0 instead 1 `If N is 1 (one), then the result is the null value.` h3. check (x,y) vs (y,x) args in docs the standard uses (y,x) order; and some of the function names are also contain X and Y...so the order does matter..currently at least corr uses (x,y) order which is okay - because its symmetric; but it would be great to have the same order everywhere (check others) was: h3. corr the standard defines corner cases when it should return null - but the current result is NaN. If N * SUMX2 equals SUMX * SUMX , then the result is the null value. and If N * SUMY2 equals SUMY * SUMY , then the result is the null value. h3. covar_samp returns 0 instead 1 `If N is 1 (one), then the result is the null value.` > corr/covar_samp UDAF standard compliance > ---------------------------------------- > > Key: HIVE-16178 > URL: https://issues.apache.org/jira/browse/HIVE-16178 > Project: Hive > Issue Type: Sub-task > Components: SQL > Reporter: Zoltan Haindrich > Priority: Minor > > h3. corr > the standard defines corner cases when it should return null - but the > current result is NaN. > If N * SUMX2 equals SUMX * SUMX , then the result is the null value. > and > If N * SUMY2 equals SUMY * SUMY , then the result is the null value. > h3. covar_samp > returns 0 instead 1 > `If N is 1 (one), then the result is the null value.` > h3. check (x,y) vs (y,x) args in docs > the standard uses (y,x) order; and some of the function names are also > contain X and Y...so the order does matter..currently at least corr uses > (x,y) order which is okay - because its symmetric; but it would be great to > have the same order everywhere (check others) -- This message was sent by Atlassian JIRA (v6.3.15#6346)