Strange output and content after as.data.frame() coercion

[fweber144]

With posterior 1.4.0, I get strange results when printing an object (e.g., a summary) that I applied as.data.frame() to:

library(posterior)
draws_mat <- as_draws_matrix(example_draws())
q_tails <- function(x) quantile(x, probs = c(0.025, 0.975))
smmry <- summarize_draws(draws_mat, "median", "mad", "q_tails")
smmry_dataframe <- as.data.frame(smmry)
print(smmry_dataframe)

gives

   variable median  mad                                                    2.5% 97.5%
1        mu   4.16 3.57  -\033[31m2\033[39m\033[31m.\033[39m\033[31m16\033[39m   10.2
2       tau   3.07 2.89                                                   0.174  14.6
3  theta[1]   5.97 4.87  -\033[31m2\033[39m\033[31m.\033[39m\033[31m81\033[39m   22.7
4  theta[2]   5.13 4.25  -\033[31m3\033[39m\033[31m.\033[39m\033[31m24\033[39m   14.9
5  theta[3]   3.99 4.94 -\033[31m15\033[39m\033[31m.\033[39m\033[31m0\033[39m    13.2
6  theta[4]   4.99 4.51  -\033[31m4\033[39m\033[31m.\033[39m\033[31m64\033[39m   14.8
7  theta[5]   3.72 4.38  -\033[31m8\033[39m\033[31m.\033[39m\033[31m68\033[39m   12.3
8  theta[6]   4.14 4.81  -\033[31m6\033[39m\033[31m.\033[39m\033[31m80\033[39m   12.8
9  theta[7]   5.90 4.54  -\033[31m2\033[39m\033[31m.\033[39m\033[31m57\033[39m   18.1
10 theta[8]   4.64 4.89  -\033[31m6\033[39m\033[31m.\033[39m\033[31m53\033[39m   14.8

I don't know if this is actually posterior-related or if it might be related to a recent update in tidyverse / tibble / pillar or anything else. I just encountered it in the context of posterior because I don't use tibbles anywhere else.

[mjskay]

Looks like control characters from color formatting. My guess is this is because of this:

posterior/R/summarise_draws.R

Lines 234 to 238 in 75ae740

	for (i in seq_cols(out)) {
	if (is.numeric(out[[i]])) {
	out[[i]] <- do.call(tibble::num, c(list(out[[i]]), .num_args))
	}
	}

Probably the "right" thing to do is to instead return a data frame with a subclass here, like c("draws_summary", "data.frame"), and move the formatting step into the print method of that subclass.

[paul-buerkner]

Yes, that could make sense. @mjskay would you have time to do that or shall I?

[mjskay]

I miiigght have time this weekend, but most likely not due to various bits of travel. But if you don't get to it before next week I could do it next week.

[fweber144]

Thanks! I forgot to mention that not only the print() output is strange, but that it's already the content of the data.frame which seems to be broken. This can be seen from typing str(smmry_dataframe$`2.5%`) after the reprex posted above.

[mjskay]

Yeah, that tracks with my reading of the problem - basically the formatting is being applied to the returned data frame instead of being applied at print time.