updated revdeps #3303
updated revdeps #3303
Conversation
|
@paleolimbot can you begin to go through the list of problems and group them into the same apparent root cause. After that we can begin to tackle the issues one by one |
|
This is what I have so far...I can break these up and file individual issues as well if this is not the best place. geom_ribbon() failureThere is a failure in bayesAB with the use of ggplot(data.frame(thing = c("a", "b", "c"))) +
geom_point(x = c(1, 2, 3), y = 3)
ggplot(data.frame(thing = c("a", "b", "c"))) +
geom_point(aes(x = c(1, 2, 3), y = 3))Theme-related failuresPackages breathteststan, esmisc, fergm, fingertipscharts, jcolors, xpose all fail with error This failure is at https://github.com/tidyverse/ggplot2/blob/master/R/guides-axis.r#L30. This is a result of #2934. All of these packages define complete themes using
rstan does not fail, but defines a default theme in a non-standard way: https://github.com/stan-dev/rstan/blob/develop/rstan/rstan/R/stan_plot_options.R MTLR does not define a complete theme, but has a ggstatsplot tests its custom vdiffr failurevdiffr has two test failures as a result of one figure ( test_that("ggtitle is set correctly", {
p <- ggplot2::ggplot()
expect_doppelganger("Some title", p)
expect_doppelganger("Some other title", p + ggplot2::ggtitle(NULL))
})I think the code is failing because the plot title has a slightly different height. The diff of the failing plot is here: pathGrob failuresPackages ggalt, ggforce, ggpol, ggpolypath, ggsolvencyii, and ggspatial all fail with I assume all of these packages are trying to get holes to show up in grid polygons (for ggspatial and ggpolypath I know this for sure). I don't know enough about grid to know whether this is a grid issue, a ggplot2 issue, or a package developer issue. Still investigating
|
|
Let’s just keep it here until all packages have been gone through. I’m suffering a concussion right now so I’m sadly of less help than I anticipated |
|
No problem, feel better! |
|
@paleolimbot The vdiffr failures are due to this PR: #2996 The package authors will have to fix their vdiffr reference images. However, vdiffr is not run on CRAN, so this should not trigger problems there. However, this should be stated explicitly in the release notes. I'll open an issue to remind us. |
|
The Maybe we should revisit this one more time and accept |
|
Another group of failures: annotate() data with facetsPepsNMR and vidger fail with:
Both are doing something like this: library(ggplot2)
data <- data.frame(x = c(1, 2), y = c(1, 2), facet = c("a", "b"))
p <- ggplot(data, aes(x, y)) + geom_point() + facet_grid(vars(facet))
p + annotate("text", x = 1.5, y = 1.5, label = c("label1", "label2"))
#> Error: Aesthetics must be either length 1 or the same as the data (4): labelIn 3.1.1 it resulted in this: I don't have a good reason for this changing since 3.1.1 except maybe #3242 or #3162. Clearly it isn't the desired behaviour. |
|
In reading over the library(ggplot2)
data <- data.frame(x = c(1, 2), y = c(1, 2), facet = c("a", "b"))
p <- ggplot(data, aes(x, y)) + geom_point() + facet_grid(vars(facet))
p + annotate("text", x = c(1, 2), y = 1.5, label = "label")Created on 2019-05-06 by the reprex package (v0.2.1) Specifically, it only transfers position values into the data frame it creates: Line 57 in 92d2777 Maybe Line 105 in 92d2777 |
|
I made a new issue for the annotation problem (#3305). |
|
Regarding |
|
Another set of errors, possibly related to what broke the annotation code for facets: Something about zero-row dataAll of interflex, HistDAWass, and mlr fail, with but for different reasons. I've simplified the failing code to reprexes: # interflex
library(ggplot2)
ggplot() +
annotate("point", x = numeric(0), y = numeric(0), colour = "red")
#> Error: Elements must equal the number of rows or 1
# histDAWass
library(ggplot2)
df <- data.frame(x = c(7, 8.2, 9.2, 10.8, 15))
ggplot(
df,
aes(x = factor(0),
ymin = df$x[1],
lower = df$x[2],
middle = df$x[3],
upper = df$x[4],
ymax = df$x[5])
) +
geom_boxplot(stat = "identity")
#> Error: Elements must equal the number of rows or 1
# mlr
library(ggplot2)
ggplot(data.frame(y = 1), aes(factor(1), y = y)) + geom_violin()
#> Warning in max(data$density): no non-missing arguments to max; returning -
#> Inf
#> Error: Elements must equal the number of rows or 1In 3.1.1, all of these generated very odd output but did so without erroring. These plots also had other layers in the examples, so it's likely they were intended to fail silently. # interflex
library(ggplot2)
ggplot() +
annotate("point", x = numeric(0), y = numeric(0), colour = "red")# histDAWass
library(ggplot2)
df <- data.frame(x = c(7, 8.2, 9.2, 10.8, 15))
ggplot(
df,
aes(x = factor(0),
ymin = df$x[1],
lower = df$x[2],
middle = df$x[3],
upper = df$x[4],
ymax = df$x[5])
) +
geom_boxplot(stat = "identity")# mlr
library(ggplot2)
ggplot(data.frame(y = 1), aes(factor(1), y = y)) + geom_violin()
#> Warning in max(data$density): no non-missing arguments to max; returning -
#> Inf |
|
Last one for today: different output column ordering between
|
|
This is the rest of the packages: argument order problem in join_keys() is reversed from argument order problem in plyr::join.keys()For incR, there is a plot that facets by date. This was possible in 3.1.1, but is now broken. library(ggplot2)
df <- data.frame(date = as.Date("2019-01-01"), x = 1, y = 1)
p <- ggplot(df, aes(x, y)) + geom_point()
p + facet_grid(vars(date))
#> Error in scale_apply(layer_data, x_vars, "train", SCALE_X, x_scales):
p + facet_wrap(vars(date))
#> Error in scale_apply(layer_data, x_vars, "train", SCALE_X, x_scales):This is a "problem" in df <- data.frame(date = as.Date("2019-01-01"))
df_fct <- data.frame(date = as.factor(df$date))
unlist(ggplot2:::join_keys(df, df_fct, "date"))
#> x y n
#> 1 1 1
unlist(ggplot2:::join_keys(df_fct, df, "date"))
#> x y n
#> 2 1 2
unlist(plyr::join.keys(df, df_fct, "date"))
#> x y n
#> 2 1 2
unlist(plyr::join.keys(df_fct, df, "date"))
#> x y n
#> 1 1 1Installation failuresgastempt failed to install in the revdep check. Testing the package on my machine (MacOS) seemed to not generate any failing tests. I can't find any failures related to ggplot2 from the CMD check. ssMousetrack and trialr failed to install the first time, so all warnings were "newly broken". Package size failuresheatmaply is just barely at the 5 MB limit and may have been pushed over the edge by the update. Connectivity failureslime, OUTRIDER, PathoStat, shiny, stats19. I couldn't install OUTRIDER, and couldn't replicate the failures in problems.md in any of the others. Unknown failuresSCORPIUS doesn't have any errors on my machine using the github version. |
|
@paleolimbot Could you open a separate issue for each remaining group of problems (for which I haven't opened one yet) and tag each with the 3.2.0 milestone? Then we can triage there. Ignore installation failures, package size failures, etc. Cover only the issues that were brought about by some identifiable change in the ggplot2 code. |
|
@clauswilke I think I got them all as separate issues. I think several were caused because there was ggplot2 code that depended on plyr-specific corner case handling. Is the general approach to fix the ggplot2 code or to fix things in |
|
I don't think there's a generic answer to how we address these issues. It depends on how many downstream packages are affected, how difficult it would be for us to restore the previous behavior, and how strongly we believe that the new behavior is more appropriate/correct. Contacting the downstream maintainers and telling them that they need to fix something is always an option, but if it's easy for us to prevent a problem then we should rather do that. |
|
Replaced by #3352 |
|
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/ |
database file
We checked 2622 reverse dependencies (2212 from CRAN + 410 from BioConductor), comparing R CMD check results across CRAN and dev versions of this package.
Issues with CRAN packages are summarised below.
New problems
(This reports the first line of each new failure)
bayesAB
checking examples ... ERROR
breathteststan
checking examples ... ERROR
CSTools
checking tests ...
esmisc
checking examples ... ERROR
fergm
checking examples ... ERROR
fingertipscharts
checking examples ... ERROR
checking tests ...
ggalt
checking whether package ‘ggalt’ can be installed ... WARNING
ggforce
checking whether package ‘ggforce’ can be installed ... WARNING
ggpol
checking whether package ‘ggpol’ can be installed ... WARNING
ggpolypath
checking whether package ‘ggpolypath’ can be installed ... WARNING
ggsolvencyii
checking whether package ‘ggsolvencyii’ can be installed ... WARNING
ggspatial
checking whether package ‘ggspatial’ can be installed ... WARNING
ggstance
checking tests ...
ggstatsplot
checking tests ...
ggtern
checking whether package ‘ggtern’ can be installed ... WARNING
heatmaply
checking installed package size ... NOTE
HistDAWass
checking examples ... ERROR
incR
checking examples ... ERROR
interflex
checking examples ... ERROR
jcolors
checking examples ... ERROR
lime
checking tests ...
mlr
checking examples ... ERROR
MTLR
checking tests ...
SCORPIUS
checking tests ...
shiny
checking tests ...
ssMousetrack
checking installed package size ... NOTE
checking for GNU extensions in Makefiles ... NOTE
stats19
checking tests ...
trialr
checking installed package size ... NOTE
checking dependencies in R code ... NOTE
checking for GNU extensions in Makefiles ... NOTE
vdiffr
checking tests ...
xpose
checking examples ... ERROR
checking tests ...
Failed to check