Thanks Hubert. What i tried so far is not collecting the distinct values of the expolded column grouped by key.

An example input is:

key, cities

1, ["milan","paris","new york"]

1, ["London"]

1, ["London","paris"]

1, ["London","paris"]

1, ["London","paris"]

1, ["milan","paris"]

1, ["paris","new york"]

1, ["new york"]

1, ["new york"]

2, ["milan","paris","new york"]

2, ["paris"]

2, ["paris"]

2, ["milan","paris"]

2, ["paris","new york"]

2, ["Tokyo"]

2, ["new york"]

2, ["LA","Tokyo"]

2, ["LA","Tokyo"]

The desired output is:

key, cities

1, ["milan","paris","new york","London"]

2, ["milan","paris","new york","LA","Tokyo"]