Lazy Polars backend functionality

[cigrainger]

This issue tracks the implementation for full backend functionality for lazy Polars. The intent is to enable splitting things up more easily.

IO read

• from_csv (Add IO operations "from/load" to the LazyFrame implementation #451)
• load_csv (Add IO operations "from/load" to the LazyFrame implementation #451)
• from_parquet (Add IO operations "from/load" to the LazyFrame implementation #451)
• load_parquet (Add IO operations "from/load" to the LazyFrame implementation #451)
• from_ipc (Add IO operations "from/load" to the LazyFrame implementation #451)
• load_ipc (Add IO operations "from/load" to the LazyFrame implementation #451)
• from_ndjson (Add IO operations "from/load" to the LazyFrame implementation #451)
• load_ndjson (Add IO operations "from/load" to the LazyFrame implementation #451)
• load_ipc_stream (b326afd)

PS: from_ipc_stream for the Lazy backend is not possible because is not available in Polars.

IO write

• Certain IO write operations can be done lazily for streaming: https://genserver.social/notice/AQn5LNxxRIfvNvvkNE - we need to investigate which and break it down
• to_parquet (Add LazyFrame version of to_parquet/3 #496)
• to_ipc (Raise for unavailable LazyFrame functions explicitly #497) - available only in main branch of Polars
• to_ipc_stream (Raise for unavailable LazyFrame functions explicitly #497) - It's not streamable, even so the name contains "stream". collect() |> write()
• to_csv (Raise for unavailable LazyFrame functions explicitly #497) - It's not streamable. collect() |> write()
• to_ndjson (Raise for unavailable LazyFrame functions explicitly #497) - It's not streamable. collect() |> write()

Conversion

• lazy (Add lazy backend #202)
• to_lazy (Add lazy backend #202)
• collect (Add lazy backend #202)

Introspection

• names (Add lazy backend #202)
• dtypes (Add lazy backend #202)
• shape (Raise for unavailable LazyFrame functions explicitly #497)
• n_rows (Raise for unavailable LazyFrame functions explicitly #497)
• n_columns (Add lazy backend #202)

Single table verbs

• head (Add lazy backend #202)
• tail (Add lazy backend #202)
• select (Add lazy backend #202)
• filter (Add filter_with/2 for Polars LazyFrame backend #476)
• mutate
• arrange
• distinct
• rename (Add some LazyFrame functions - part III #488)
• dummies (Raise for unavailable LazyFrame functions explicitly #497)
• sample (Raise for unavailable LazyFrame functions explicitly #497) - this one may be possible to implement, but not sure how.
• pull (Raise for unavailable LazyFrame functions explicitly #497)
• slice (Enables the slicing of LazyFrames #450)
• drop_nil (Add some LazyFrame functions - part III #488)
• pivot_wider (Raise for unavailable LazyFrame functions explicitly #497)
• pivot_longer (Add some LazyFrame functions - part III #488)
• group_by (Add lazy backend #202)
• ungroup (Add lazy backend #202)
• summarise

Two or more table verbs

• join (Add LazyFrame version of "join" #491)
• concat_rows (Add LazyFrame concat of rows and columns #492)
• concat_columns (Add LazyFrame concat of rows and columns #492)

[josevalim]

All functions not listed here should raise when invoked saying the lazy operation is not allowed.

[philss]

FYI I'm working on the IO functions of this issue :)

[josevalim]

FWIW, we decided to focus on the non-grouped versions. We will support grouped ones later.

[philss]

I created #498 to track the groups operations.

I'm closing this issue, since most of the functions were implemented or are explicitly raising error now.