add jar datasketches-memory-2.0.0.jar;
add jar datasketches-java-3.1.0.jar;
add jar datasketches-hive-1.2.0.jar;

create temporary function data2sketch as 'org.apache.datasketches.hive.quantiles.DataToDoublesSketchUDAF';
create temporary function unionSketches as 'org.apache.datasketches.hive.quantiles.UnionDoublesSketchUDAF';
create temporary function getQuantile as 'org.apache.datasketches.hive.quantiles.GetQuantileFromDoublesSketchUDF';
create temporary function getQuantiles as 'org.apache.datasketches.hive.quantiles.GetQuantilesFromDoublesSketchUDF';

use <your-db-name-here>;

create temporary table quantiles_input (value double, category char(1));
insert into table quantiles_input values
  (1, 'a'), (2, 'a'), (3, 'a'), (4, 'a'), (5, 'a'), (6, 'a'), (7, 'a'), (8, 'a'), (9, 'a'), (10, 'a'),
  (11, 'b'), (12, 'b'), (13, 'b'), (14, 'b'), (15, 'b'), (16, 'b'), (17, 'b'), (18, 'b'), (19, 'b'), (20, 'b');

create temporary table quantiles_intermediate (category char(1), sketch binary);
insert into quantiles_intermediate select category, data2sketch(value) from quantiles_input group by category;

-- get median value per category
select category, getQuantile(sketch, 0.5) from quantiles_intermediate;

Output:
a	6.0
b	16.0

-- union across categories and get median
select getQuantile(unionSketches(sketch), 0.5) from quantiles_intermediate;

Output:
11.0

-- if you want several quantiles it is more efficient to use this function
-- quantile of 0.0 is min value, 0.5 is median, and 1.0 is max value
select getQuantiles(unionSketches(sketch), 0.0, 0.5, 1.0) from quantiles_intermediate;

Output:
[1.0,11.0,20.0]

-- alternative way to get quantiles by specifying a number of evenly spaced fractions
-- this is equivalent to giving fractions of 0.0, 0.5 and 1.0
select getQuantiles(unionSketches(sketch), 3) from quantiles_intermediate;

Output:
[1.0,11.0,20.0]