Allow using plot_grid in a pipe

[flying-sheep]

Currently, one needs to do

foo %>% map(ggplot) %>% plot_grid(plotlist = .)

With this change, the following is possible:

foo %>% map(ggplot) %>% plot_grid()

[codecov-io]

Codecov Report

Merging #146 into master will decrease coverage by 1.1%. The diff coverage is 50%.

@@            Coverage Diff             @@
##           master     #146      +/-   ##
==========================================
- Coverage   68.69%   67.58%   -1.11%     
==========================================
  Files          20       20              
  Lines         856      799      -57     
==========================================
- Hits          588      540      -48     
+ Misses        268      259       -9

Impacted Files	Coverage Δ
R/plot_grid.R	89.28% <50%> (-1.46%)	:arrow_down:
R/as_grob.R	30.3% <0%> (-1.13%)	:arrow_down:
R/key_glyph.R	95.38% <0%> (-0.77%)	:arrow_down:
R/themes.R	96.21% <0%> (-0.3%)	:arrow_down:
R/draw.R	46.83% <0%> (-0.23%)	:arrow_down:
R/set_null_device.R	0% <0%> (ø)	:arrow_up:
R/stamp.R	0% <0%> (ø)	:arrow_up:
R/align_plots.R	100% <0%> (ø)	:arrow_up:

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update 597b3ea...759fea9. Read the comment docs.

[clauswilke]

Since plot_grid() is meant to take most anything as an argument, this will lead to problems down the road. This is true regardless of the specific logic you employ to try to detect whether plot_grid() is employed inside a pipe.

[flying-sheep]

I think you misunderstood: The logic in my PR checks if the first and only argument only has the class 'list'. It therefore it breaks nothing, but instead makes the following work:

> # current behavior:
> plot_grid(list(ggdraw()))
Warning message in as_grob.default(plot):
“Cannot convert object of class list into a grob.”
> # behavior after my PR
> plot_grid(list(ggdraw()))
<an empty canvas!>

[clauswilke]

No, I didn't misunderstand. Rewriting function signatures based on the data type of some of the input is fundamentally a bad idea. In your case, if at some point in the future somebody finds a use for as_grob.list(), then that will unexpectedly not work with plot_grid(). And likewise, if somebody gets used to having plot_grid() behave a certain way with input of type list, they may expect the same to happen with input of type gList, and that's also not going to be possible.

[flying-sheep]

Why is it a bad idea? I don’t see how as_grob.list might in any way make sense.