Control significant figures with an option about pillar

[hadley]

I've changed my mind - it's causing enough pain for enough people right now that we should have someway to suppress.

Probably easiest to parameterise as option(pillar.sigfigs = 3) allowing option(pillar.sigfigs = Inf)to turn it off.

[hadley]

That is not under consideration right now. It needs more discussion, and we need to get a release out to resolve the biggest problems.

[hadley]

The underlying problem here is #57 - I think if we fix that an option will not be necessary.

I'd prefer to push as far as we can before introducing more options, since options substantially increase the testing complexity, and make the package behave differently for different people. If this new behaviour is too annoying, we can definitely consider options, but I'd like to give it more time (i.e. open again in 6 months time if it's still a problem and I'll be more amenable)

[ilarischeinin]

Unaware of this issue, I was actually already in the middle of implementing this kind of option for my own use. But then I came here to check if there was any discussion in more general terms about the question of left-hand-sides of numbers and significant digits vs. thousands separators.

As a little background, I was surprised at first when I saw that in e.g. 1234 the three digits with a separate color are 123 and not 234. Also, I would've thought that if 123456789 was shown in two colors, 123 and 789 would be shown in one, and 456 in the other color. That is, I thought that the two colors represented thousands separators. I was therefore confused what it actually meant, and only from the documentation learned that it represented significant digits.

Since I didn't want to use it that way, I started to implement an option so I could use options(pillar.sigfigs = Inf).

But then I got to thinking that would the thousands separator idea actually be a more useful use case for the colors than significant digits? Consistency is important, but since sigfig already means different things for the rhs (colors and rounding) and lhs (no rounding, only colors), I'm not sure if that rules out the idea.

I definitely haven't looked into this very deeply, so there might be fairly obvious factors at play that make my suggestion ridiculous. Or maybe it's just plain stupid. But since I have spent a fair amount of time staring at integers on the scale of millions (base pair numbers in the human genome), I for one see a lot of potential in this kind of representation (color formatting) of thousands separators. Use of commas, or extra characters in general, is problematic for copy&paste.

Or maybe keep significant digits behaving as it is, but include a new option for the thousands separator. That one could default to not be shown, but when set, it would override the significant digits on the lhs.

I'm happy to work on this and make a PR if it's deemed useful.

[dpeterson71]

I agree that this has to change. The current implementation makes inspecting tibbles almost impossible for me. I LOVE the alignment along decimal points, but the forced highlight of only three digits is confusing at best. Not only is this unintuitive (as cboettig indicates), it is visually distracting
in many cases.

[krlmlr]

See #78 for a suggestion for visible thousand separators using a very light gray background every other three digits.

[ilarischeinin]

Thank you for the new options in #79!

However, I still think there should either be

• a separation in how significant figures are treated on the left and right hand sides,
• a mode that didn't force the number of significant digits to be set (like with base R), or
• a separate option that would override the formatting of lhs's (#78).

I really like having subtle for things like column headers or how POSIXct's are displayed. So, I would prefer not to turn it off. But if I don't turn it off, I'm forced to set sigfig to some value. If it's small (like 3), numbers (or their lhs's to be exact) look funny. If it's large (like 10), there are many unnecessary zeros on the rhs's (#62 is about this, but I think the functionality discussed on that issue would be a slippery slope to go down).

[krlmlr]

@ilarischeinin: Happy to discuss this in a separate issue or even pull request ;-)

[github-actions]

This old thread has been automatically locked. If you think you have found something related to this, please open a new issue and link to this old issue if necessary.