Goden Yao created HAWQ-437: ------------------------------ Summary: Support time series partition filter pushdown Key: HAWQ-437 URL: https://issues.apache.org/jira/browse/HAWQ-437 Project: Apache HAWQ Issue Type: Sub-task Components: PXF Reporter: Goden Yao Assignee: Goden Yao
time series is a very common used folder structure in HDFS usually for logging purpose: e.g. {noformat} For example, let’s say that the data for sales information in a chain is stored in the following directories structure: /data/sales/yyyy/mm/dd/hh/*.data (year/month/day/hour) /data/sales / --- 2012 / --- --- 01 / --- --- --- 01 / --- --- --- --- 00 / --- --- --- --- 01 / ... --- --- --- --- 23 / --- --- --- 02 / ... --- --- --- 31 / ... --- --- 02 / --- --- --- 01 / ... --- --- --- 28 / ... ... --- 2013 / ... --- 2014 / ... {noformat} Time series partitioning will support numeric operation (=, !=, <, >, <=, >=), under the following limitations: * Time template must be of the form year/month/day/hour or any subset of it (e.g. year/month/day or month/day). * The dates must be in numeric values. * Years should be represented by two or four digits (yy or yyyy). * Months should be represented by two digits (mm). The months valid values are 01-12. * Days should be represented by two digits (dd) The days valid values are 01-31. * Hours should be represented by two or four digits (hh or hhhh). Hours valid values are 00-23 or 0000-2359. Any other form should be represented using the regular string partitions. Time series will only regard the supported values - directories with unsupported values will be ignored. -- This message was sent by Atlassian JIRA (v6.3.4#6332)