Add a note on reusing LazyFrames

[avimallu]

Address pola-rs/polars#9786.

[cmdlineluser]

Thanks for this.

Just thought I'd add that maybe it would be helpful to also have a code example demonstrating this?

Copying the code from pola-rs/polars#9786 (comment)

df_a = pl.LazyFrame({"col": ["A", "B", "C", "D"]})
df_b = pl.LazyFrame({"col": ["A", "B", "C", "D"]})

df_b = df_b.groupby("col").all().with_row_count()

df_a.join(df_b, on="col", how="left").collect()

# shape: (4, 2)
# ┌─────┬────────┐
# │ col ┆ row_nr │
# │ --- ┆ ---    │
# │ str ┆ u32    │
# ╞═════╪════════╡
# │ A   ┆ 0      │
# │ B   ┆ 2      │
# │ C   ┆ 3      │
# │ D   ┆ 1      │
# └─────┴────────┘

df_a.join(df_b, on="col", how="left").collect()

# shape: (4, 2)
# ┌─────┬────────┐
# │ col ┆ row_nr │
# │ --- ┆ ---    │
# │ str ┆ u32    │
# ╞═════╪════════╡
# │ A   ┆ 1      │
# │ B   ┆ 3      │
# │ C   ┆ 0      │
# │ D   ┆ 2      │
# └─────┴────────┘

A better example could be made, but seeing this behaviour was what helped me understand what exactly was happening.

[avimallu]

I'd love to @cmdlineluser, but at some point it might just happen that during a compilation of the user guide, the outputs will be the same and will leave new users very thoroughly confused. Any ideas (other than adding a special condition) to make sure that they're different?

I remembered that there is an option of running code but not displaying it on screen. I can just run a while loop to make sure that the results are always different for the sake of illustration, and not show it in the user guide. Any thoughts @ritchie46?

[ritchie46]

I am not familiar with md-book. It seems possible to me?

[ritchie46]

Let's first merge the note and see if we can improve with a code example later. 👍