Skew Join Salting Key in Python (Demo)

import random


def skew_join_salting_key(left_df, right_df, salt_range=4):
    """
    Demonstrates skew join salting: expand the smaller side with salt keys,
    then attach a salt key to each row on the larger side.
    Returns a list of (left, right, salt) tuples.
    """
    skewed_left = []
    for row in left_df:
        for salt in range(salt_range):
            skewed_left.append((row, salt))

    joined = []
    for l_row, salt in skewed_left:
        for r_row in right_df:
            if r_row["key"] == l_row["key"] and r_row["salt"] == salt:
                joined.append((l_row, r_row["value"], salt))
    return joined


if __name__ == "__main__":
    # Simulate a small 'right' side (dimension table) with prefixed salt keys
    right_df = [
        {"key": "A", "salt": 0, "value": "x"},
        {"key": "A", "salt": 1, "value": "y"},
        {"key": "B", "salt": 0, "value": "z"},
    ]

    # Large left side: each row will pick a random salt in [0, salt_range)
    left_df = [
        {"key": "A", "id": 101},
        {"key": "A", "id": 102},
        {"key": "B", "id": 103},
    ]

    # Assign a random salt to each left row (simulating a distributed join)
    random.seed(42)
    left_df_with_salt = []
    for row in left_df:
        row["salt"] = random.randrange(4)
        left_df_with_salt.append(row)

    result = skew_join_salting_key(left_df_with_salt, right_df)
    for l, r, salt in result:
        print(f"left_id={l['id']} key={l['key']} right_value={r} salt={salt}")