This message was deleted.

what would your

key

and

value

be in this case?

Key would be string and values would be some computes quantilesDoublesSketch - A result of

groupBy

query is going to be used in a lookup table for quicker access.

group by query will not give you a sketch. It will get you a number

{
  "queryType": "groupBy",
  "dataSource": "table_name",
  "granularity": "all",
  "dimensions": [
    "column_name"
  ],
  "aggregations": [
    {
      "type": "quantilesDoublesSketch",
      "fieldName": "double_sketch",
      "k": 256,
      "name": "baseline_quantile_sketch"
    }
  ],
  "postAggregations": [
    {
      "type": "fieldAccess",
      "fieldName": "baseline_quantile_sketch",
      "name": "baseline_quantile_sketch_raw"
    }
  ],
  "filter": {
    "type": "and",
    "fields": [
      {
        "type": "selector",
        "dimension": "__model_id__",
        "value": "079c6aa9-2027-4c0e-8fbf-c3c0817514d1"
      },
      {
        "type": "in",
        "dimension": "column_name",
        "values": [
          "area",
          "baths",
          "beds",
          "datePriceChanged"
        ]
      }
    ]
  },
  "intervals": "2015-08-28/2025-10-20"
}

This is my query, intension is to store

column_name

as key and

baseline_quantile_sketch_raw

as value. I’m able to store

baseline_quantile_sketch_raw

as the value but in string datatype - wanted to see if this can be stored as sketch type?

if you just use an aggregation then you can get the sketch. The post aggregation will result in a number. I think what yo are trying to do is us a rilled table and join with another table. As far as I know this can't be done using a lookup. You need a join.

FYI: @Apoorv Gupta

@Kumar Saket Is there a quick way to verify?

SELECT
  TIME_FLOOR(CASE WHEN CAST("t_hour" AS BIGINT) > 0 THEN MILLIS_TO_TIMESTAMP(CAST("t_hour" AS BIGINT)) ELSE TIME_PARSE("t_hour") END, 'PT1H') AS __time,
  "channel",
  SUM("__count") AS "__count",
  DS_THETA(COMPLEX_DECODE_BASE64('thetaSketch', "theta_user")) AS "theta_user",
  DS_HLL(COMPLEX_DECODE_BASE64('HLLSketch', "hll_user")) AS "hll_user"
FROM "source"
GROUP BY 1, 2
PARTITIONED BY DAY

May be using

COMPLEX_DECODE_BASE64

with

DS_THETA

or something similar should do the Job. Not sure though..

As far as I know you cannot go from string to sketch (sketch to string is possible as the sketches output string in the post agg for. Debugging