Additional info regarding the questions raised in your article.

[cmdlineluser]

Hey there, wasn't sure where best to respond but I just thought I'd share some info I've learned as a Polars user which may be of interest:

For to the convert a string to a datetime example:

df.with_columns(dt = pl.col('string_date').str.to_datetime().cast(pl.Datetime))

Perhaps the .cast() is a stray from previous examples, but you already have a datetime so it can be omitted.

df.with_columns(dt = pl.col('string_date').str.to_datetime())

• dt.date()

dt.date() can also be used to get a date from a datetime

df.with_columns(
    pl.col('string_date').str.to_datetime().dt.date()
)

• pl.col("foo").cast(pl.Date)

I’m curious what happens if I simply just .cast('date')

This has been requested, there is a pending PR: pola-rs/polars#10517

[Update]: This was added in pola-rs/polars#12072

• .strptime()

I’ll be buggered if I know the difference between to_datetime() and strptime()

They are essentially aliases, to_date() / to_datetime() were added later and str.strptime() remains for legacy reasons.

• Rename str.strptime/dt.strftime methods pola-rs/polars#8215
• With to_datetime and to_date, is strptime going away? pola-rs/polars#10110

This is why .to_datetime() lives under .str but you do bring up an interesting point as the new name doesn't imply it is tied to strings.

It should be simply pl.col('column').to_datetime() and error if the data types are wrong

• .dt.to_string()

Similar to the strptime case, .dt.to_string() was added as an alias in favour of .dt.strftime()