Hello,
here is a patch implementing varwidth_bucket (naming is up for
discussion) function which does binning with variable bucket width. The
use-cases are same as for width_bucket (=data analytics, mainly
histograms), the difference is that width_bucket uses buckets of same
width but the varwidth_bucket accepts an sorted array of right-bound
thresholds to define the individual buckets.
Currently applications implement this with long CASE statements which
are quite hard to read/maintain and are much slower than this
implementation which uses binary search.
There are 3 actual functions, one generic and two faster versions for
the int8 and float8 input that take advantage of the static width of
those types.
The research leading to these results has received funding from the
European Union's Seventh Framework Programme (FP7/2007-2013) under grant
agreement n° 318633
--
Petr Jelinek http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services
