Using R package logger with data.frames

Question

I'm using logger for logging. Now I want to log the contents of a data.frame , currently, this yields something like this:

> logger::log_formatter(logger::formatter_paste)
> logger::log_info(data.frame(a = 1:3, b = 4:6))
INFO [2019-05-03 14:36:29] 1:3
INFO [2019-05-03 14:36:29] 4:6

This is not what I want, is it possible to get something like below

INFO [2019-05-03 14:36:29]   a b
INFO [2019-05-03 14:36:29] 1 1 4
INFO [2019-05-03 14:36:29] 2 2 5
INFO [2019-05-03 14:36:29] 3 3 6

as print() would give?

Answer 1

The default formatter in logger is glue (or sprintf if the glue package is not installed), which yields an error on passing data frames:

glue::glue(data.frame(a = 1:3, b = 4:6))
#> Error: All unnamed arguments must be length 1

sprintf(data.frame(a = 1:3, b = 4:6))
#> Error in sprintf(data.frame(a = 1:3, b = 4:6)) : 
#>   'fmt' is not a character vector

So if you want logger to work with data frames, you need a custom formatter, eg

formatter_data_frame <- function(df, ...) {
    pander::pander_return(df, style = 'simple')
}

library(logger)
log_formatter(formatter_data_frame)
log_info(data.frame(a = 1:3, b = 4:6))
#> INFO [2019-05-04 11:33:47] 
#> INFO [2019-05-04 11:33:47]  a   b 
#> INFO [2019-05-04 11:33:47] --- ---
#> INFO [2019-05-04 11:33:47]  1   4 
#> INFO [2019-05-04 11:33:47]  2   5 
#> INFO [2019-05-04 11:33:47]  3   6 
#> INFO [2019-05-04 11:33:47] 

Or with what you tried to achieve with calling paste on each line:

formatter_data_frame <- function(df, ...) {
    apply(df, 1, paste, collapse = ' ')
}

log_formatter(formatter_data_frame)
log_info(data.frame(a = 1:3, b = 4:6))
#> INFO [2019-05-04 11:35:03] 1 4
#> INFO [2019-05-04 11:35:03] 2 5
#> INFO [2019-05-04 11:35:03] 3 6

Of course, you could tweak this further, eg to do this for a data.frame and fall back to glue on other objects etc

Another approach is using an eval helper:

df <- data.frame(a = 1:3, b = 4:6)
log_eval(df)
#> TRACE [2019-05-04 11:37:12] 'df' => 'structure(list(a = 1:3, b = 4:6), class = "data.frame", row.names = c(NA,  -3L))'

If you think there's something missing from the package that would be useful, please open a ticket in the GitHub repo at https://github.com/daroczig/logger