bug(bq,sf,rs|clustering):ST_CLUSTERKMEANS remove duplicated coords

[vdelacruzb]

Description

Shortcut

• Story: https://app.shortcut.com/cartoteam/story/397702/internal-st-cluster-k-means-component-result-has-less-clusters-than-number-of-clusters-value
• Autolink: [sc-397702]

Some tables containing duplicates points might result in less clusters done expected. As a solution for this we are removing duplicates before doing the processing.

Type of change

• Fix

Acceptance

### Bigquery

with a as (
SELECT `carto-un`.carto.ST_CLUSTERKMEANS(ARRAY_AGG(geom ORDER BY ST_ASBINARY(geom)), 40) arr
from `cartodb-data-engineering-team`.vdelacruz_carto.clustering_table_copy
)
select count(distinct(element.cluster)) from a, UNNEST(arr) element;
-- 36

with a as (
SELECT `cartodb-data-engineering-team`.vdelacruz_carto.ST_CLUSTERKMEANS(ARRAY_AGG(geom ORDER BY ST_ASBINARY(geom)), 40) arr
from `cartodb-data-engineering-team`.vdelacruz_carto.clustering_table_copy
)
select count(distinct(element.cluster)) from a, UNNEST(arr) element;
-- 40

### Snowflake

CREATE table  CARTO_DATA_ENGINEERING_TEAM.vdelacruz_carto.clustering_table_duplicateds AS
SELECT ST_GEOGFROMTEXT('POINT(-72.3539 -37.47262)') as geom
UNION ALL SELECT ST_GEOGFROMTEXT('POINT(-72.3539 -37.47262)')
UNION ALL SELECT ST_GEOGFROMTEXT('POINT(-71.61442 -35.39392)')
UNION ALL SELECT ST_GEOGFROMTEXT('POINT(-71.61442 -35.39392)')
UNION ALL SELECT ST_GEOGFROMTEXT('POINT(-71.61442 -35.39392)')
UNION ALL SELECT ST_GEOGFROMTEXT('POINT(-71.33815 -29.9541)');

with a as (
SELECT CARTO_DEV_DATA.carto.ST_CLUSTERKMEANS(ARRAY_AGG(ST_ASGEOJSON(geom)::STRING), 6) arr
from CARTO_DATA_ENGINEERING_TEAM.vdelacruz_carto.clustering_table_duplicateds
)
select count(*) FROM a, LATERAL FLATTEN(input=>arr);
-- 6 rows

with a as (
SELECT CARTO_DATA_ENGINEERING_TEAM.vdelacruz_carto.ST_CLUSTERKMEANS(ARRAY_AGG(ST_ASGEOJSON(geom)::STRING), 6) arr
from CARTO_DATA_ENGINEERING_TEAM.vdelacruz_carto.clustering_table_duplicateds
)
select count(*) FROM a, LATERAL FLATTEN(input=>arr);
-- 3 rows. Looked at the results and it's keeping the distinct points

### Redshift

CREATE table  vdelacruz_carto.clustering_table_duplicateds AS
SELECT ST_GEOMFROMTEXT('POINT(-72.3539 -37.47262)') as geom
UNION ALL SELECT ST_GEOMFROMTEXT('POINT(-72.3539 -37.47262)')
UNION ALL SELECT ST_GEOMFROMTEXT('POINT(-71.61442 -35.39392)')
UNION ALL SELECT ST_GEOMFROMTEXT('POINT(-71.61442 -35.39392)')
UNION ALL SELECT ST_GEOMFROMTEXT('POINT(-71.61442 -35.39392)')
UNION ALL SELECT ST_GEOMFROMTEXT('POINT(-71.33815 -29.9541)');

SELECT get_array_length( carto.ST_CLUSTERKMEANS(ST_GEOMFROMTEXT('MULTIPOINT ((-72.3539 -37.47262), (-72.3539 -37.47262), (-71.61442 -35.39392), (-71.61442 -35.39392), (-71.61442 -35.39392), (-71.33815 -29.9541))', 9)));
-- 6

SELECT get_array_length( vdelacruz_carto.ST_CLUSTERKMEANS(ST_GEOMFROMTEXT('MULTIPOINT ((-72.3539 -37.47262), (-72.3539 -37.47262), (-71.61442 -35.39392), (-71.61442 -35.39392), (-71.61442 -35.39392), (-71.33815 -29.9541))', 6)));
-- 3. Looked to the results and is producing disting values

[shortcut-integration]

This pull request has been linked to Shortcut Story #397702: [internal] ST Cluster K-Means component result has less clusters than Number of clusters value.