Issue 555

[katrinabrock]

I've made an attempt to solve issue #555 , but it's not quite working. The plot I'm getting looks like this:

diff plot

So pane A is almost right and pane B is further off.

I've diagnosed that when usr is set manually, asp is not taken into account.

tinyplot/R/facet.R

Lines 325 to 328 in ba46791

	fusr[[ii]] = c(xext, yext)
	assign(".fusr", fusr, envir = get(".tinyplot_env", envir = parent.env(environment())))
	# Explicitly set (override) the current facet extent
	par(usr = fusr[[ii]])

I've attempted to translate the logic used by plot.window(). However, it's not working. I believe the issue is that grconvertX(1, "npc", "inches") is not actually equivalent to GConvertXUnits(1.0, NPC, INCHES, dd) or maybe I'm calling grconvertX in the wrong context.

[katrinabrock]

Ok, this seems very strange to me. This is totally outside tinyplot context:

print_scale <- function(from = "npc", to = "inches") print(c(
  grconvertX(1, from, to),
  grconvertY(1, from, to)
))

print_scale()

print("set mfrow")
par(mfrow = c(1,2))
print_scale()

print("mfg=c(1,1)")
par(mfg = c(1,1))
print_scale()

print("mfg=c(1,2)")
par(mfg = c(1,2))
print_scale()

Result:

[1] 6.58 6.18
[1] "set mfrow"
[1] 6.58 6.18
[1] "mfg=c(1,1)"
[1] 3.08 6.18
[1] "mfg=c(1,2)"
[1] 6.58 6.18

It seems very odd that setting the first figure vs the second figure to be drawn next would impact npc->inches conversion.

@grantmcdermott or @zeileis I would love some input here. I read #65 which seems related, but I still feel like there's something fundamental I'm missing here.