Compute the variations of the total column

total_variations.py

def get_input_df():

    df = pd.DataFrame({"date": pd.date_range("2017-1-1", "2020-1-1", freq="1D")})
    df["open"] = np.random.choice([1, 0], len(df))
    df["closed"] = np.random.choice([1, 0], len(df))
    df["location"] = np.random.choice(["a", "b", "c"], len(df))
    return df

def compute_total_variations_df(df):
    df = df.copy()
    df["total"] = (df["open"] - df["closed"])
    df = df.set_index("date").groupby("location").resample("1M")["total"].sum().reset_index()
    df["total"] = df.groupby(["date", "location"])["total"].transform("cumsum")

    dfs = []
    for location in df["location"].unique():
        _df = df.loc[lambda df: df["location"] == location].sort_values("date")
        last_year_total_s = _df["total"].shift(12)
        _df["variation_total"] = 100 * ((_df["total"] - last_year_total_s ) / last_year_total_s)
        dfs.append(_df)

    return pd.concat(dfs)

result_df = get_input_df().pipe(compute_total_variations_df)