ggsave in lapply - erroneously renaming plots on save

[MichelineCampbell]

Hi all,

I have found a weird problem when saving lists of plots with ggsave and lapply.

I have a large dataframe of elemental data for different samples which I typically plot in batches using lapply, with plots titled by element. I noticed that the saved versions of the Fe and Sr plots get mis-titled, even when appearing as they should in the plot viewer. I have encountered both being titled "Sr" and both being titled "Fe". The behaviour seems to resolve itself if you use something like lab(plot = paste("Sample", el)), and of the 15 elements I am plotting it only seems to be Sr and Fe that get mis-titled.

This error reproduces, I have made an example below.

I am using R version 4.4.2 and ggplot2 version 3.5.2.

library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
library(ggplot2)
Sr <- data.frame(x = 1:100,
           y = rnorm(100),
           sample = "Sr")

Fe <- data.frame(x = 1:100,
                 y = rnorm(100) +100,
                 sample = "Fe")

all <- rbind(Sr, Fe)

vars <- unique(all$sample)

lapply(vars, function(i){
  
  el <- i
  
plot <-  all %>% filter(sample == !!i) %>% 
    ggplot(aes(x = x, y = y)) +
    geom_line() +
    labs(title = el,
         y = "Concentration (ppm)",
         x = "Year (CE)") +
    theme_bw() +
    theme(strip.background = element_blank(),
          legend.position = "bottom") +
    guides(colour=guide_legend(nrow=2,byrow=TRUE))


  
  ggsave(width = 15, height = 10, units= "cm", filename = paste0("test2", el, ".png"))
  print(plot)
  dev.off()
  
  print(plot)
})
#> [[1]]
#> 
#> [[2]]

Expected output (copied from the plot viewer)

Saved output

[joranE]

I initially assumed this was just some problematic lazy evaluation, but unless I'm missing something obvious, this pared down example seems to demonstrate clearly that the behavior does in fact depends on the specific text in the title, which seems very odd. Is labs (or something downstream) doing some sophisticated parsing to try to work around lazy evaluation issues?

I ran this in R 4.5.0 and ggplot2 3.5.2.

library(dplyr)
library(ggplot2)

vals1 <- c("S","F")
mt_comb1 <- bind_rows(
  mtcars |> mutate(grp = vals1[1]),
  mtcars[1:10,] |> mutate(grp = vals1[2])
)

# Plot titles are "F" & "S"
lapply(
  X = vals1, 
  FUN = function(nm){
    p <- 
      ggplot(
        data = mt_comb1 |> filter(grp == nm),
        aes(x = hp, y = mpg)
      ) +
      geom_point() +
      labs(title = nm)
    
    ggsave(
      plot = p, 
      filename = paste0("figs/test1_", nm, ".png")
    )
  })


vals2 <- c("Sr","Fe")
mt_comb2 <- bind_rows(
  mtcars |> mutate(grp = vals2[1]),
  mtcars[1:10,] |> mutate(grp = vals2[2])
)

# Plot titles are both "Sr"
lapply(
  X = vals2,
  FUN = function(nm){
    p <- 
      ggplot(
        data = mt_comb2 |> filter(grp == nm),
        aes(x = hp,y = mpg)
      ) + 
      geom_point() + 
      labs(title = nm)
    
    ggsave(
      plot = p,
      filename = paste0("figs/test2_", nm, ".png")
    )
  }
)

[teunbrand]

No, labs() should evaluate all its arguments immediately, so there should in theory be no promise shenanigans.
Inspecting the output of the test2_* series by digging into the grid grobs, it seems that the labels land in the right places.

[aitap]

plot <-  all %>% filter(sample == !!i) %>% 
    ggplot(aes(x = x, y = y)) +
    geom_line() +
    labs(title = el,
         y = "Concentration (ppm)",
         x = "Year (CE)") +
    theme_bw() +
    theme(strip.background = element_blank(),
          legend.position = "bottom") +
    guides(colour=guide_legend(nrow=2,byrow=TRUE))


  
  ggsave(width = 15, height = 10, units= "cm", filename = paste0("test2", el, ".png"))
  print(plot)
  dev.off()
  
  print(plot)

First you produce the plot but do not print it, so it's not getting displayed. Next you call ggsave(), which tries to access last_plot(). What's the last_plot() when plot has not been printed yet? Try explicitly giving plot to ggsave().

[teunbrand]

@joranE was providing the ggsave(plot) argument, so that is not the issue.
I've boiled it down to a grid problem rather than a ggsave() problem. With ragg png we get the following:

library(grid)

# Render as text grobs
vals <- c("Sr", "Fe")
grobs <- lapply(vals, textGrob)

# Confirm label is correct
grobs[[1]]$label
#> [1] "Sr"
grobs[[2]]$label
#> [1] "Fe"

# Create device for temp file
tmp <- tempfile(fileext = ".png")
ragg::agg_png(tmp)

# Draw the grobs
grid.newpage()
grid.draw(grobs[[1]])
grid.draw(grobs[[2]])

dev.off()
#> png 
#>   2
knitr::include_graphics(tmp)

^{Created on 2025-07-02 with reprex v2.1.1}

Note that the file should superimpose Sr and Fe, but it just displays Sr. For comparison, the same thing but with png(type = "cairo"):

library(grid)

vals <- c("Sr", "Fe")

grobs <- lapply(vals, textGrob)

grobs[[1]]$label
#> [1] "Sr"
grobs[[2]]$label
#> [1] "Fe"

tmp <- tempfile(fileext = ".png")
png(tmp, type = "cairo")

grid.newpage()
grid.draw(grobs[[1]])
grid.draw(grobs[[2]])

dev.off()
#> png 
#>   2
knitr::include_graphics(tmp)

^{Created on 2025-07-02 with reprex v2.1.1}

So maybe this is just a ragg png problem?

[joranE]

I've only done a few minutes of experimenting with your grid/ragg example @teunbrand but I can't seem to find an obvious pattern regarding the specific text used or the order in which they are drawn. 🤷

Edit: It seems specific to the exact combination of "Sr" and "Fe" as far as I can tell?!

[MichelineCampbell]

plot <- all %>% filter(sample == !!i) %>%
ggplot(aes(x = x, y = y)) +
geom_line() +
labs(title = el,
y = "Concentration (ppm)",
x = "Year (CE)") +
theme_bw() +
theme(strip.background = element_blank(),
legend.position = "bottom") +
guides(colour=guide_legend(nrow=2,byrow=TRUE))

ggsave(width = 15, height = 10, units= "cm", filename = paste0("test2", el, ".png"))
print(plot)
dev.off()

print(plot)
First you produce the plot but do not print it, so it's not getting displayed. Next you call ggsave(), which tries to access last_plot(). What's the last_plot() when plot has not been printed yet? Try explicitly giving plot to ggsave().

@aitap I get the same behaviour if I print plots to a list and then loop through to save the plots from the list, rather than relying on last_plot().

[teunbrand]

Thanks everyone for help on narrowing down the cause of this bug. I'm convinced this is not a ggplot specific bug, so I've put up an issue in r-lib/ragg#193 to replace this one.