Only use data_frame in tests

[thomasp85]

Some tests were added between the new_data_frame implementation and it's merge that used the now masked data.frame constructor. This PR fixes these so tests now passes

[clauswilke]

I see one more problem, here:

ggplot2/tests/testthat/test-geom-polygon.R

Lines 13 to 15 in 4380cb9

	ggplot(faithful, aes(x = eruptions, y = waiting)) +
	stat_density2d(aes(fill = stat(level)), geom = "polygon", colour = "white") +
	xlim(0.5, 6) + ylim(40, 110)

Not immediately sure what the cause is.

library(ggplot2)
ggplot(faithful, aes(x = eruptions, y = waiting)) +
  stat_density2d(aes(fill = stat(level)), geom = "polygon", colour = "white") +
  xlim(0.5, 6) + ylim(40, 110)
#> Warning in validDetails.polygon(x): NAs introduced by coercion

^{Created on 2018-11-15 by the reprex package (v0.2.1)}

[clauswilke]

One other, minor issue that could be fixed as part of this PR: The rep() inside max() seems superfluous here:

ggplot2/R/stat-contour.r

Line 96 in 92d2777

nlevel = rep(levels, lengths) / max(rep(levels, lengths), na.rm = TRUE),

[yutannihilation]

GeomPolygon's error seems because now munched$group is character here, whereas polygonGrob() only accepts numeric vectors. Previously, it was factor. I didn't know the mercy of stringsAsFactors secretly saved us...

ggplot2/R/geom-polygon.r

Lines 96 to 105 in 4380cb9

	polygonGrob(munched$x, munched$y, default.units = "native",
	id = munched$group,
	gp = gpar(
	col = first_rows$colour,
	fill = alpha(first_rows$fill, first_rows$alpha),
	lwd = first_rows$size * .pt,
	lty = first_rows$linetype
	)
	)
	)

[yutannihilation]

The group comes from here. So you can wrap this as.factor() or convert it to integer before passing it to polygonGrob(). I feel the latter approach is right, but I'm not fully sure what it is supposed to be.

ggplot2/R/stat-contour.r

Lines 92 to 101 in 4380cb9

	groups <- paste(data$group[1], sprintf("%03d", pieces), sep = "-")
	new_data_frame(list(
	level = rep(levels, lengths),
	nlevel = rep(levels, lengths) / max(rep(levels, lengths), na.rm = TRUE),
	x = xs,
	y = ys,
	piece = pieces,
	group = groups
	), n = length(xs))

[thomasp85]

Thanks - weirdly this didn't pop up locally... 🤔

[lock]

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/