How to generate samples with no instersection using seed? - snowflake-cloud-data-platform

Question

I am trying to generate unique samples of data using SAMPLE() and SEED().
This means samples should not intersect with each other. Every sample should be unique.
But I am not reaching my target, because Snowflake generates samples with 10% intersection to the previous sample when I am using this approach:
CREATE TABLE employee_sample_10_1 as
SELECT employee_id FROM employee SAMPLE(10) SEED(1);
CREATE TABLE employee_sample_10_2 as
SELECT employee_id FROM employee SAMPLE(10) SEED(2);
CREATE TABLE employee_sample_10_3 as
SELECT employee_id FROM employee SAMPLE(10) SEED(3);
CREATE TABLE employee_sample_10_4 as
SELECT employee_id FROM employee SAMPLE(10) SEED(4);
CREATE TABLE employee_sample_10_5 as
SELECT employee_id FROM employee SAMPLE(10) SEED(5);
CREATE TABLE employee_sample_10_6 as
SELECT employee_id FROM employee SAMPLE(10) SEED(6);
CREATE TABLE employee_sample_10_7 as
SELECT employee_id FROM employee SAMPLE(10) SEED(7);
CREATE TABLE employee_sample_10_8 as
SELECT employee_id FROM employee SAMPLE(10) SEED(8);
CREATE TABLE employee_sample_10_9 as
SELECT employee_id FROM employee SAMPLE(10) SEED(9);
CREATE TABLE employee_sample_10_0 as
SELECT employee_id FROM employee SAMPLE(10) SEED(0);
ideally, all these tables created as samples, should not intersect and their total number of rows should be equal to the number of rows in the original table EMPLOYEE.
In fact, none of these conditions are true. The number of rows won't match, moreover, each individual sample table will intersect with its neighbours by ~10% roughly.
SELECT employee_id FROM employee_sample_10_1 INTERSECT SELECT employee_id FROM employee_sample_10_2; --getting ~10% of data intersect, need 0%
SELECT employee_id FROM employee_sample_10_2 INTERSECT SELECT employee_id FROM employee_sample_10_3; --getting ~10% of data intersect, need 0%
SELECT employee_id FROM employee_sample_10_3 INTERSECT SELECT employee_id FROM employee_sample_10_4; --getting ~10% of data intersect, need 0%
SELECT employee_id FROM employee_sample_10_4 INTERSECT SELECT employee_id FROM employee_sample_10_5; --getting ~10% of data intersect, need 0%
SELECT employee_id FROM employee_sample_10_5 INTERSECT SELECT employee_id FROM employee_sample_10_6; --getting ~10% of data intersect, need 0%
SELECT employee_id FROM employee_sample_10_6 INTERSECT SELECT employee_id FROM employee_sample_10_7; --getting ~10% of data intersect, need 0%
SELECT employee_id FROM employee_sample_10_7 INTERSECT SELECT employee_id FROM employee_sample_10_8; --getting ~10% of data intersect, need 0%
SELECT employee_id FROM employee_sample_10_8 INTERSECT SELECT employee_id FROM employee_sample_10_9; --getting ~10% of data intersect, need 0%
SELECT employee_id FROM employee_sample_10_9 INTERSECT SELECT employee_id FROM employee_sample_10_0; --getting ~10% of data intersect, need 0%
Question: How to make SAMPLE() and SEED() produce only unique sets of values?

Answer 1

You can assign a random employee group to each row; create a temp table off that result set (or use the result_scan method if you prefer) and then insert into the 10 tables accordingly, stripping out the column used to assign the groups if necessary.
Here's a sample using the CUSTOMER table in the Snowflake sample data:
create temp table RANDOM_EMPLOYEE_GROUPS as
select uniform(0, 9, random()) as RANDOM_GROUP, *
from "SNOWFLAKE_SAMPLE_DATA"."TPCH_SF1"."CUSTOMER";
select * from RANDOM_EMPLOYEE_GROUPS limit 100;
-- Insert into the 10 tables from there.