All zero values with `scale_y_log10()` gives an error

[billdenney]

When generating a figure with all zero values, scale_y_log10() gives an error instead of an empty plot. I would have hoped for a figure with all the points along the bottom (ideally as it is similar to the typical result for log scales) or otherwise an empty plot (sub-optimal since it removes some information).

library(ggplot2)
p <-
  ggplot(data.frame(x=0, y=0), aes(x=x, y=y)) +
  geom_point()
p

p + scale_y_log10()
#> Warning: Transformation introduced infinite values in continuous y-axis
#> Error in seq.default(min, max, by = by): 'from' must be a finite number

Created on 2018-09-03 by the reprex package (v0.2.0).

[billdenney]

I think that this would be best fixed around here:

ggplot2/R/coord-cartesian-.r

Line 121 in 01155ba

range <- range(scale$transform(limits))

I think that the change would be to check if the transformation yielded all infinite values prior to passing it to range, and if so add some nominal value to the range (like scale$transform(1)).

The alternate solution seems like it would be a check in scales::log_breaks() at the top if all(x %in% c(Inf, 0, NA_real_)) then rng <- c(0, 0).

[smouksassi]

What I always found strange is that when you have facets the facet with all illegal points are still rendered as empty the others with points show some points but when no facets and all points invalid the plot aborts.

require(ggplot2)
#> Loading required package: ggplot2
databug<- data.frame(
  x=c(0,1,-1,0,-1,1),
  y=c(0,1,-1,0,-1,1),
  pnl=c("A","B","C","D","D","D")
  
)

ggplot(databug, aes(x=x, y=y))+geom_point()+scale_y_log10()+facet_grid(~pnl)
#> Warning in self$trans$transform(x): NaNs produced
#> Warning: Transformation introduced infinite values in continuous y-axis
#> Warning: Removed 2 rows containing missing values (geom_point).

Created on 2018-09-12 by the reprex package (v0.2.0).

[billdenney]

@smouksassi, I think that I have an answer for that one: With the facets as you have specified them, the scales are determined based on all of the data, regardless of what is plotted within one facet. If you make the facet: facet_grid(~pnl, scales="free"), it fails (or "free_x" or "free_y", in this case).

[paleolimbot]

I think this is an issue with scales::log_breaks(), which can't compute breaks when there are no finite values:

library(scales)

log_breaks()(c(0, Inf))
#> Error in seq.default(min, max, by = by): 'from' must be a finite number
extended_breaks()(c(0, Inf))
#> [1] 0 0 0 0 0

log_breaks()(c(0, NA))
#> [1] 0
log_breaks()(c(NA, NA))
#> Warning in min(x, na.rm = na.rm): no non-missing arguments to min;
#> returning Inf
#> Warning in max(x, na.rm = na.rm): no non-missing arguments to max;
#> returning -Inf
#> Warning in log_breaks()(c(NA, NA)): NaNs produced
#> Error in if (max == min) return(base^min): missing value where TRUE/FALSE needed
log_breaks()(c(-Inf, Inf))
#> Warning in log_breaks()(c(-Inf, Inf)): NaNs produced
#> Error in if (max == min) return(base^min): missing value where TRUE/FALSE needed
log_breaks()(numeric(0))
#> Warning in min(x): no non-missing arguments to min; returning Inf
#> Warning in max(x): no non-missing arguments to max; returning -Inf
#> Warning in log_breaks()(numeric(0)): NaNs produced
#> Error in if (max == min) return(base^min): missing value where TRUE/FALSE needed

^{Created on 2019-05-26 by the reprex package (v0.2.1)}

[yutannihilation]

This seems already fixed on scales' side.

library(scales)

log_breaks()(c(NA, NA))
#> numeric(0)
log_breaks()(numeric(0))
#> numeric(0)

^{Created on 2021-03-25 by the reprex package (v1.0.0)}

library(ggplot2)

ggplot(data.frame(x=0, y=0), aes(x=x, y=y)) +
  geom_point() + 
  scale_y_log10()
#> Warning: Transformation introduced infinite values in continuous y-axis

^{Created on 2021-03-25 by the reprex package (v1.0.0)}