Extra ggplot2 lines trip up grade_this_code()

[garrettgman]

Adding an extra call to a ggplot() call results in an obscure code suggestion: I expected ggplot() where you called +. For example, this learnr exercise

### 

__Edit your code from the previous exercise so that `marital` appears in order of increasing frequency.__ Use a forcats function to change the order.

```{r fct_rev, exercise =  TRUE}
gss_cat %>%
  ggplot(aes(marital)) + 
  geom_bar()
```

```{r fct_rev-solution}
gss_cat %>%
  ggplot(aes(fct_rev(fct_infreq(marital)))) + 
  geom_bar()
```

```{r fct_rev-check}
grade_this_code()
```

Returns the message below when a student supplies the code

gss_cat %>%
  ggplot(aes(fct_infreq(marital))) + 
  scale_x_discrete(limits = rev) +
  geom_bar()

Since adding an extra line to a ggplot() call is a common failure mode for ggplot2, we should find a way to provide a better code suggestion.

One idea: perhaps the ggcheck package can come with a ggplot2 specific version of grade_this_code()that can begin by assuming it is working with ggplot2 code, e.g. grade_this_ggplot_code()

[gadenbuie]

I think we should be able to fix this in gradethis. The underlying problem is that we don't do a good job of handling the case when there are more operations than expected:

# Extra addition
code_feedback("a + b + c", "a + c")
#> I didn't expect `+` where you wrote `a + b + c`.

# Extra addition, plus pipe (smaller reprex of the ggplot example)
code_feedback("x %>% a() + b + c", "a(x) + c")
#> In `x %>% a() + b + c`, I expected you to call `a()` where you called `+`.

# Extra pipes
code_feedback("x %>% a() %>% b() %>% c()", "x %>% a() %>% c()")
#> In `x %>% a() %>% b() %>% c()`, I expected you to call `a()` where you called `b()`.

We have an open issue for the extra pipe case: #185