Open kishoreg opened 6 years ago
agrawaldevesh
commented
5 years ago Is there any update on this ? Seems easy enough to do by exposing the DataSchema(cachedDataScheme) from com.linkedin.pinot.core.query.reduce.BrokerReduceService#reduceOnDataTable into BrokerResponseNative. It seems straightforward and backward compatible change that I would be willing to work on it.
It would save me writing some janky client side hacks if this existed.
kishoreg
commented
5 years ago Yes. It’s definitely easy to add. You got the logic right. Feel free to submit a PR. It will be great to enable this via a config flag for now.
snleee
commented
5 years ago @agrawaldevesh We do have some support for preserving types. https://github.com/apache/incubator-pinot/pull/2830
If you add OPTION(preserveType='true'), Pinot will preserve the type for the result. One limitation is that this won't preserve types for grouping keys (Keeping types for grouping key will require considerable refactoring).
Can you play with preserveType option to see if it makes your task easier? However, it would be ideal if we can add the types along with the column names.
today we return the column names but not the data types. Pinot knows about the data type either from the segment metadata or from the schema. Knowing the data types will make it easier to write connectors to Pinot.
{ "selectionResults":{ "columns":[ "Cancelled", "Carrier", "DaysSinceEpoch", "Delayed", "Dest", "DivAirports", "Diverted", "Month", "Origin", "Year" ], "results":[ [ "0", "AA", "16130", "0", "SFO", [], "0", "3", "LAX", "2014" ], [ "0", "AA", "16130", "0", "LAX", [], "0", "3", "SFO", "2014" ], [ "0", "AA", "16130", "0", "SFO", [], "0", "3", "LAX", "2014" ] ] }, "traceInfo":{}, "numDocsScanned":3, "aggregationResults":[], "timeUsedMs":10, "segmentStatistics":[], "exceptions":[], "totalDocs":102 }