Encounter a ggplot2's problem. pic1 is good, then pic2 is good,but when review pic1,it gets bad

Question

Recently, I encountered a question in ggplot2 field. It's confused for me that everytime I plot first plot with ggplot names "pic1"(the result of running is okay), and then I plotted second one with ggplot2 called "pic2". Of course, the "pic2" is good. But at this moment, I check "pic1", I found the regression line became a vertical line.For example:

1. "pic1"

p <- ggplot()  
p <- p + geom_line(data = MyData, aes(x = otherCrop, y = eta ))
p <- p+ geom_point(data = dat,aes(x =otherCrop,
                                  y = dat$sumEnemies, colour = YEAR ),position = position_jitter(width = .01),size = 1)

 

p <- p+labs(colour = "年份\nYear")    + theme_classic(base_size=18)  +
        theme(axis.title.x=element_text( vjust=0))
p=p + theme(text=element_text(family="Times",   size=18)) 
pic1=p  

2. "pic2"

p <- ggplot()  
p <- p + geom_line(data = MyData, aes(x = SHDI, y = eta ))
p <- p+ geom_point(data = dat,aes(x = dat$SHDI,
                                  y = eta,colour = YEAR ),position = position_jitter(width = .01),size = 1)

 

p <- p+labs(colour = "年份\nYear")    + theme_classic(base_size=18)  +   
        theme(axis.title.x=element_text( vjust=0))
p=p + theme(text=element_text(family="Times",   size=18)) 
pic2=p  

But at this moment, I started to review "pic1", I found it as below: It became a strange short vertical line. This would be difficult because I cannot plot them in a same paper. Does anybody know what's the problem?

Answer 1

I think this is a great example of why using the dataframe$column syntax inside an aes call is discouraged: it makes your plot vulnerable to subsequent changes in your data. Here's a simple example. Start with a data frame with columns x and y:

library(ggplot2)

df <- data.frame(x = 1:10, y = 1:10)

Now make a ggplot, but instead of using aes(x = x, y = y), we make the mistake of doing aes(x = df$x, y = df$y):

vulnerable_plot <- ggplot()
vulnerable_plot <- vulnerable_plot + geom_line(data = df, aes(x = df$x, y = df$y))
pic1 <- vulnerable_plot

Now we review our plot. Sure, ggplot nags us to say we shouldn't use this syntax, but the plot looks fine, so who cares, right?

pic1
#> Warning: Use of `df$x` is discouraged. Use `x` instead.
#> Warning: Use of `df$y` is discouraged. Use `y` instead.

Now, let's make pic2 identical to pic1 except we use the correct syntax:

invulnerable_plot <- ggplot()
invulnerable_plot <- invulnerable_plot + geom_line(data = df, aes(x = x, y = y))
pic2 <- invulnerable_plot

Now we don't get any warning, but the plot looks the same.

pic2

So there's no difference between pic1 and pic2. Or is there? What happens when we change our data frame?

df$y <- 10:1

vulnerable_plot

Oh dear. Our first plot has changed because the plot object has a reference to an external variable that it relies on to build the plot. That's not what we wanted.

However, with the version where we used the correct syntax, a copy of the data was taken and is kept with the plot data, so it remains unaffected by subsequent changes to df:

invulnerable_plot

^{Created on 2020-08-23 by the reprex package (v0.3.0)}