Fei Feng created CALCITE-2633: --------------------------------- Summary: match_recognize all rows per match return type bug Key: CALCITE-2633 URL: https://issues.apache.org/jira/browse/CALCITE-2633 Project: Calcite Issue Type: Bug Components: core Affects Versions: 1.16.0 Environment: JDK 8,Calcite 1.6.0 Reporter: Fei Feng Assignee: Julian Hyde
When using sql match recognize with flink cep ,something seems strange。 {color:#6a8759}SELECT * {color}{color:#6a8759}FROM Ticker {color}{color:#6a8759}MATCH_RECOGNIZE ( {color}{color:#6a8759} ORDER BY proctime {color}{color:#6a8759} MEASURES {color}{color:#6a8759} FIRST(DOWN.tstamp) AS top_tstamp, {color}{color:#6a8759} LAST(DOWN.tstamp) AS bottom_tstamp {color}{color:#6a8759} ALL ROWs PER MATCH {color}{color:#6a8759} AFTER MATCH SKIP PAST LAST ROW {color}{color:#6a8759} PATTERN (DOWN\{2,} LAST) {color}{color:#6a8759} DEFINE {color}{color:#6a8759} DOWN AS DOWN.price < PREV(DOWN.price) or PREV(DOWN.price) IS NULL, {color}{color:#6a8759} LAST AS true {color}{color:#6a8759}) AS T{color} {color:#6a8759}With "AlLL ROWS PER MATCH",the result has 4 columns,the proctime is the first column.But with "ONE ROW PER MATCH",the result has 3 columns, and the proctime is not in the result set. {color} {color:#6a8759}Is this a bug?{color} -- This message was sent by Atlassian JIRA (v7.6.3#76005)