Add configurable scipen bias setting for number formatting in Data Explorer

[kv9898]

Release Notes

New Features

• Add a new setting for scientific notation penalty bias (mimicking that in R) for number formatting in Data Explorer
• Resolves Data Explorer - scientific notation - does not adhere to options(scipen = xxx) #4818.

Bug Fixes

• N/A

QA Notes

[juliasilge]

Thank you once again for contributing to Positron, @kv9898! 🙌 It has been so exciting to see the ways in which you have been able to improve the experience of working in the IDE.

Unfortunately, we are not wanting to address #4818 with the approach in this PR, i.e. formatting the values from the front end using a user-defined IDE setting. We want to go in a direction where we respect the formatting options set inside the kernel (e.g. R's digits and scipen options). I know it's disappointing to hear something like this, and I'm sorry that we're not able to accept his PR. 😔

If you are interested in working on this problem from a different approach, I can point you to some PRs in ark that work on some related ideas:

• Data Explorer: Add support for formatting options ark#382 is about formatting options for the summary stats, and I think would be helpful to look at
• Fix bug when formatting dates containing NAs for the data explorer ark#421 is about how we format dates
• Data Explorer: Add support for truncating outputs at specified length ark#446 is about how we truncate information over the Data Explorer comm

To get the value of an option from inside ark, check out how we use harp::get_option().

[kv9898]

@juliasilge Thanks for the explanation! Nevertheless, I don't think they are contradictory, though. There are cases where no 'kernal' setting is available at all. For example, when the user opens a csv/xlsx file directly in the data explorer.

I personally think it would be nice if we could respect the kernal setting while keeping a fall-back option for when no kernal is involved, like what this PR does. After all, this PR does not make implementing the kernal-based logic more difficult. :)

Also, from my understanding, accommodating scipen for kernals is not in a near milestone, and it will likely be done on a kernal-by-kernal basis. This implies a long wait, during which the user may use thr setting provided by this PR as a workaround.