I can think of a couple ways to do so that basically rely on an OUTER JOIN With all possible dimension values (0-23h).

To frame the problem I will use a BigQuery public dataset: bigquery-public-data.london_bicycles.cycle_hire where start_date is a TIMESTAMP A field containing the time when each ride started. I will filter out all trips that started before 7 AM with the following query:

SELECT
  *
FROM
  bigquery<span>-</span>public<span>-</span>data.london_bicycles.cycle_hire
WHERE
  EXTRACT(HOUR FROM start_date) > 6

Therefore, I get the same problem when creating a data source with the previous custom query (hours without data are not displayed):

Blending with range data

The first and quick fix would be to create a data source that contains all possible dimension values. This could be done in a spreadsheet or, for example, with an additional 0 B custom query:

SELECT
  *
FROM
  UNNEST(GENERATE_ARRAY(0, 23)) AS hour

and now we blend the data so that the hour array is on the left side (Data Studio uses LEFT OUTER JOIN) and join with HOUR(start_date) from the cycle_hire table:

Then we get the desired histogram (be sure to change the number of bars up to 24 in the STYLE tab and sort by hour ascending in the DATA one):

Aggregate in BigQuery

Another alternative would be to run the calculations for the hourly buckets already in the custom query. For example, in this case, we would generate the hour array, left outer join it with the aggregated data and substitute NULL for 0:

WITH
hours AS (
  SELECT
    *
  FROM
    UNNEST(GENERATE_ARRAY(0, 23)) AS hour),
bike_rides AS (
  SELECT
    EXTRACT(HOUR FROM start_date) AS hour,
    COUNT(*) AS total
  FROM
    bigquery<span>-</span>public<span>-</span>data.london_bicycles.cycle_hire
  WHERE
    EXTRACT(HOUR FROM start_date) > 6
  GROUP BY
    hour )

SELECT
  hour,
  IF(total IS NULL,0,total) AS total
FROM
  bike_rides
RIGHT OUTER JOIN
  hours
USING
  (hour)

Which gives the same result.