Add support for integer64 variable types

[edgararuiz-zz]

Results collected from databases can, including aggregations, could return an integer64 variable type which does not seem to be supported by aes(). The workaround is to coerce as.numeric() prior to plotting.

Code that collects the data:

> by_month <- flights %>%
+     group_by(month) %>%
+     tally() %>%
+     collect() 
> 
> head(by_month)
# A tibble: 6 x 2
  month               n
  <dbl> <S3: integer64>
1    10          556205
2     4          598126
3    12          544958
4     5          606293
5     2          569236
6     7          627931

The plot that is attempted followed by the error's text:

library(ggplot2)

ggplot(by_month) +
  geom_line(aes(x = month, y = n)) 

Don't know how to automatically pick scale for object of type integer64. Defaulting to continuous.

[hadley]

Slightly more minimal reprex:

library(ggplot2)  

df <- data.frame(x = bit64::as.integer64(1:10))
ggplot(df) + geom_histogram(aes(x))
#> Don't know how to automatically pick scale for object of type integer64. Defaulting to continuous.
#> `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
#> Warning: Removed 10 rows containing non-finite values (stat_bin).

^{Created on 2019-06-18 by the reprex package (v0.2.1.9000)}

Fixing this will probably happen naturally when we start using vctrs in ggplot2.

[paleolimbot]

Getting ggplot2 to know that the scale for integer64 objects should be continuous is a quick fix:

scale_type.integer64 <- function(x) "continuous" 

However the current holdup is that scale expansion uses ifelse(), which incorrectly type converts integer64 objects. dplyr::if_else() is too strict, however, as it fails when true and false are integer and numeric (which happens often). If this is still worth fixing, the solution would be to replace these two lines:

ggplot2/R/scale-expansion.r

Line 176 in 7f317d4

limits <- ifelse(is.na(coord_limits), limits, coord_limits)

ggplot2/R/scale-expansion.r

Line 194 in 7f317d4

continuous_range <- ifelse(is.finite(final_scale_limits), final_scale_limits, limits)

...with an improved (probably vctrs-powered) ifelse().

[paleolimbot]

A reprex for future consideration:

library(ggplot2)
scale_type.integer64 <- function(x) "continuous" 

df <- data.frame(x = bit64::as.integer64(1:10))
ggplot(df) + geom_histogram(aes(x))
#> `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
#> Warning: Computation failed in `stat_bin()`:
#> `binwidth` must be positive