Bump rust-polars to 0.34

[etiennebacher]

Close #404

Changelog: https://github.com/pola-rs/polars/releases/tag/rs-0.34.0

@eitsupi @sorhawell no rush on this, I'm just opening this PR to get the ball rolling. I'm fine with releasing polars 0.9.0 either before or after this is merged.

[etiennebacher]

There shouldn't be remaining test failures on R or Rust side. However, examples using $cumsum() on Series fail because I commented the code on Rust side since I couldn't find the fix

[eitsupi]

Thank you for working on this!
It seems that MSRV is changing to Rust 1.71, so I would like to release 0.9.0 first.

[etiennebacher]

Sure, go ahead with 0.9.0

[etiennebacher]

Looks like the last thing to do is to fix the $cumsum() error

[etiennebacher]

I tried several things and I'm stuck here:

pub fn cumsum(&self, reverse: bool) -> Series {
        pl::cumsum(self, reverse)
                .map(Series)
                .unwrap()
}

The compiler tells me that self is a series::Series while cumsum() expects a polars::prelude::Series. Can you take a look @eitsupi ?

[eitsupi]

Can you take a look @eitsupi ?

I think it could probably be fixed in a74b6af.

[etiennebacher]

@sorhawell if you have time, it would be great if you could review the changes related to the string cache since you discussed it with the rust-polars team

[eitsupi]

Thank you for accomplishing this almost single-handedly!
I am excited to get this merged and have access to the new features.

Obviously there is some testing and documentation missing on the R side, but those can be added later.

[etiennebacher]

Good thing I had a lot of travel by train these last days 😄 many thanks for your help @eitsupi

Just waiting a bit to see if @sorhawell has some remarks, otherwise I'll merge later

[sorhawell]

Not bad :) I think I have something to add pl$with_string_cache ...

[sorhawell]

I was initially concerned for a corner of nested pl$with_string_cache() like

pl$with_string_cache({
  df1 = pl$DataFrame(head(iris, 2))
  pl$with_string_cache("foobar")
  df2 = pl$DataFrame(tail(iris, 2))
  pl$concat(list(df1, df2))
})

as when disabling an inner with_string_cache() that would the affect the outer. You cleverly avoid all that with the if (!pl$using_string_cache()) { ...} guard. So all good. I cannot point out a bad behavior if a user only uses pl$with_string_cache

I commit and suggest we derive pl$with_string_cache() from StringCacheHolder instead, only as that likely is more future proof as py-polars relies on that too and is not derived from enable/disable_string_cache().

Feel free to roll back this commit and we will be fine also with your solution :)

[sorhawell]

@etiennebacher thank you so much, this is very good work

[etiennebacher]

Many thanks @sorhawell, that's definitely something I missed in rust-polars. I intend to use the string cache more in tidypolars so that will be the occasion to test it more extensively.