Refactor options handling

[etiennebacher]

Related to #316. @sorhawell I just made this to show with examples what I’d like to do. I didn’t check tests yet, it’s mostly to discuss about how to handle options.

There are 3 methods related to options:

• pl$options shows a named list with options and their default values,
• pl$set_options() modifies the values but only for the arguments that were specified in the call (and it works with autocomplete),
• pl$reset_options() brings the options back to their default.

I also used this occasion to rename some options (default_maintain_order becomes maintain_order).

I think this is clearer for the user and it also gets rid of a lot of internal code, what do you think? (also curious about your opinion @eitsupi)

library(polars)

pl$options
#> $maintain_order
#> [1] FALSE
#>
#> $do_not_repeat_call
#> [1] FALSE
#>
#> $strictly_immutable
#> [1] TRUE
#>
#> $debug_polars
#> [1] FALSE
#>
#> $no_messages
#> [1] FALSE
original_options <- pl$options

# this modifies options
pl$set_options(strictly_immutable = FALSE, maintain_order = TRUE)
pl$options
#> $maintain_order
#> [1] TRUE
#>
#> $do_not_repeat_call
#> [1] FALSE
#>
#> $strictly_immutable
#> [1] FALSE
#>
#> $debug_polars
#> [1] FALSE
#>
#> $no_messages
#> [1] FALSE

# if an option is not specified, it's not modified (here, maintain_order is
# still TRUE)
pl$set_options(strictly_immutable = TRUE)
pl$options
#> $maintain_order
#> [1] TRUE
#>
#> $do_not_repeat_call
#> [1] FALSE
#>
#> $strictly_immutable
#> [1] TRUE
#>
#> $debug_polars
#> [1] FALSE
#>
#> $no_messages
#> [1] FALSE

# reset to default
pl$reset_options()
identical(pl$options, original_options)
#> [1] TRUE

[eitsupi]

Thanks!

This looks good to me, but I don't know if this should be preferred over the options function that is common in R.
Or can implement the options function way be built on top of this later?

[etiennebacher]

I don't think options() has autocompletion, which is quite nice in pl$set_options():

And I have to think about how annoying it would be to synchronize the options passed to options() and those passed to pl$set_options().

[etiennebacher]

I'll try to find an alternative to the unlockBinding() approach since it raises a warning in R CMD check.

[etiennebacher]

All tests pass now but there are still unresolved comments. @sorhawell @eitsupi if you agree with the changes I'll update the NEWS before merging

[sorhawell]

I will continue review tonight I hope :)

[sorhawell]

Great thanks :)

[eitsupi]

Oops, I didn't expect this to be merged before 0.8.1 since it's a breaking change.

[etiennebacher]

Ah sorry, I can try to revert this