New vector classes designed for better printing?

[hadley]

I wonder if some of the complexities of tibble printing could be pushed off to individual vector formatters. Truncating strings (at least on a local level) could be useful in other places, and could be pushed down to the vector level. Another useful display type would be for logicals, printing TRUEs as "*" and leaving FALSEs blank. Maybe the same principle applies to better formatting for doubles too?

I don't think this changes what tools tibble should provide, but it may influence how we think about providing them.

[krlmlr]

Sounds like a good idea. I suspect this also helps cleaning up the formatting code. I can create a draft the next time I'll work on tibble.

[huftis]

@hadley About printing TRUE as * and FALSE as blank, how would you differentiate between FALSE and NA (or would NA still be displayed as NA, or as N)?

How about T and F instead? Though this might encourage people to begin using T and F in their code, which is bad practice. So perhaps + and −?

[hadley]

NA would still be printed as is

[jankatins]

Regarding the overall approach: if one takes the requirements and algorithm of a overall tibble printing function from #104 (comment), then delegating the printing down to individual vectors requires a two pass approach: one to get the minimum required length of each vector (will probably do something like min(<internal min value fot that vector type>, nchars(name)) and one to do the print in the specific length (bigger or equal than the required length).

So either this becomes two S3 functions (one to get the min length of a vector, one to get the string representation of a vector) or the tibble print still needs to know the min length of each vector type :-/

[krlmlr]

I don't mind two new S3 methods col_width_preference() and format_col(), the former would return a named list that contains minimum and maximum width. In addition, there should be a compound class that also includes the column name and type in the calculations and formatting.

[hadley]

I'd prefer col_width() which could return a named list: min, max, preferred. (That might be a bit too general, but it's what other layout systems do)

[krlmlr]

Closing in favor of #235.

[hadley]

I'd recommend closing in favour of https://github.com/hadley/colformat

[krlmlr]

We can implement the logic in colformat, but let's leave this open as a reminder and for tibble + colformat integration.

[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.