[ https://issues.apache.org/jira/browse/CALCITE-1909?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16107996#comment-16107996 ]
Julian Hyde edited comment on CALCITE-1909 at 7/31/17 9:43 PM: --------------------------------------------------------------- [~ransom], What do you think? was (Author: julianhyde): [~zhiqiang,he] What do you think? > The output rowType should also include partition by and order by key columns > ---------------------------------------------------------------------------- > > Key: CALCITE-1909 > URL: https://issues.apache.org/jira/browse/CALCITE-1909 > Project: Calcite > Issue Type: Bug > Reporter: Dian Fu > Assignee: Julian Hyde > Labels: match > > According to the documentation, the output rowType should also include > partition by key columns and order by key columns. > {quote} > MATCH_RECOGNIZE is called the row pattern output table. The shape (row type) > of the row pattern output table depends on the choice of ONE ROW PER MATCH or > ALL ROWS PER MATCH: > * If ONE ROW PER MATCH is specied or implied, then the columns of the row > pattern output table are the row pattern partitioning columns in their order > of declaration, followed by the row pattern measure columns in their order of > declaration. Since a table must have at least one column, this implies that > there must be at least one row pattern partitioning column or one row pattern > measure column. > * If ALL ROWS PER MATCH is speci ed, then the columns of the row pattern > output table are the row pattern partitioning columns in their order of > declaration, the ordering columns in their order of declaration, the row > pattern measure columns in their order of declaration, and nally any > remaining columns of the row pattern input table, in the order they occur in > the row pattern input table. > {quote} > Currently, partition key columns and order by key columns are not included in > the output row type. -- This message was sent by Atlassian JIRA (v6.4.14#64029)