polars: multiply 2 LazyFrames together by column

Question

I have 2 polars LazyFrames:

df1 = pl.DataFrame(data={
    'foo': np.random.uniform(0,127, size= n).astype(np.float64),
    'bar': np.random.uniform(1e3,32767, size= n).astype(np.float64),
    'baz': np.random.uniform(1e6,2147483, size= n).astype(np.float64)
}).lazy()

df2 = pl.DataFrame(data={
    'foo': np.random.uniform(0,127, size= n).astype(np.float64),
    'bar': np.random.uniform(1e3,32767, size= n).astype(np.float64),
    'baz': np.random.uniform(1e6,2147483, size= n).astype(np.float64)
}).lazy()

I would like to multiply each column in df1 with its respective column in df2.

If I convert these to non-lazy DataFrames I can achieve this:

df1.collect() * df2.collect()

foo         bar         baz
f64         f64         f64
3831.295563 6.4637e6    3.3669e12
164.194271  2.9691e8    2.2696e12
3655.918761 1.9444e7    2.3625e12
7191.48868  3.7044e7    3.1687e12
9559.505277 2.6864e8    2.5426e12

However, if I try to perform the same expression on the LazyFrames, I get an exception

df1 * df2

TypeError: unsupported operand type(s) for *: 'LazyFrame' and 'LazyFrame'

How can I perform column-wise multiplication across 2 LazyFrames?

Answer 1

you'll need to join

(
    df1.with_row_index()
    .join(df2.with_row_index(), on="index")
    .select(pl.col(col) * pl.col(f"{col}_right") for col in df1.columns)
    .collect()
)

shape: (10, 3)
┌─────────────┬──────────┬───────────┐
│ foo         ┆ bar      ┆ baz       │
│ ---         ┆ ---      ┆ ---       │
│ f64         ┆ f64      ┆ f64       │
╞═════════════╪══════════╪═══════════╡
│ 6623.602754 ┆ 2.7173e8 ┆ 3.7654e12 │
│ 2588.499522 ┆ 7.4295e8 ┆ 3.0266e12 │
│ 933.474643  ┆ 3.7090e8 ┆ 4.2794e12 │
│ 7061.625136 ┆ 2.2365e8 ┆ 2.7040e12 │
│ …           ┆ …        ┆ …         │
│ 2717.969236 ┆ 4.9398e7 ┆ 3.0930e12 │
│ 785.760153  ┆ 1.6305e8 ┆ 1.8954e12 │
│ 9534.366291 ┆ 7.3153e8 ┆ 1.9056e12 │
│ 1916.452503 ┆ 1.4976e8 ┆ 3.2704e12 │
└─────────────┴──────────┴───────────┘