Large intermediate results dureing columns update

Robert Koch gms_xy at gmx.de
Fri Oct 17 11:40:54 CEST 2014


Given a column defined as bigint (int64) with 1 billion rows. Storage size of this column is 8 GB. Values range in space of int (int32).

UPDATE
	sapsr3.dfkkop
SET
	vkont_int32 = vkont_int64
;

This query takes very long (minutes). During execution the data directory grows by 40 GB (!). I would expect the data cells do be converted an written into the new column directly. Even if MonetDB would materialize the whole source column as "intermediate result", I would expect not to exceed 8 GB for "intermediates" and 4 GB for the target column.



More information about the developers-list mailing list