问题

如何在ggplot的geom内使用本机管道和占位符？

当将data.frame传递给ggplot时，本机管道可以正常工作，但在geom内部，当我想要筛选原始data.frame时，我需要使用'magrittr'管道。

所以这样可以正常工作：

mtcars |> 
     ggplot() +
     geom_col(data = . %>% filter(cyl == 4),
              aes(x = gear, y = mpg),
              fill = "grey80") +
     geom_text(data = . %>% filter(cyl == 4 & gear == 3),
               aes(x = gear,
                   y = mpg,
                   label = "ooo"), color = "tomato3") +
     theme_minimal()

通过匿名函数，我可以完全切换到本机管道：

mtcars |> 
     ggplot() +
     geom_col(data = (\(x) filter(x, cyl == 4)),
              aes(x = gear, y = mpg),
              fill = "grey80") +
     geom_text(data = (\(x) filter(x, cyl == 4 & gear == 3)),
               aes(x = gear,
                   y = mpg,
                   label = "ooo"), color = "tomato3") +
     theme_minimal()

但是否有一种优雅的方法可以使用R的本机管道运算符和占位符来执行相同的操作？

这将引发错误'invalid use of pipe placeholder'：

mtcars |> 
     ggplot() +
     geom_col(data = _ |>  filter(cyl == 4),
              aes(x = gear, y = mpg),
              fill = "grey80") +
     geom_text(data = _ |>  filter(cyl == 4 & gear == 3),
               aes(x = gear,
                   y = mpg,
                   label = "ooo"), color = "tomato3") +
     theme_minimal()

mtcars |> 
     ggplot() +
     geom_col(data = . %>% filter(cyl == 4),
              aes(x = gear, y = mpg),
              fill = "grey80") +
     geom_text(data = . %>% filter(cyl == 4 & gear == 3),
               aes(x = gear,
                   y = mpg,
                   label = "ooo"), color = "tomato3") +
     theme_minimal()

mtcars |> 
     ggplot() +
     geom_col(data = (\(x) filter(x, cyl == 4)),
              aes(x = gear, y = mpg),
              fill = "grey80") +
     geom_text(data = (\(x) filter(x, cyl == 4 & gear == 3)),
               aes(x = gear,
                   y = mpg,
                   label = "ooo"), color = "tomato3") +
     theme_minimal()

mtcars |> 
     ggplot() +
     geom_col(data = _ |>  filter(cyl == 4),
              aes(x = gear, y = mpg),
              fill = "grey80") +
     geom_text(data = _ |>  filter(cyl == 4 & gear == 3),
               aes(x = gear,
                   y = mpg,
                   label = "ooo"), color = "tomato3") +
     theme_minimal()

答案1

得分: 3

以下是翻译好的部分：

"其实可以继续使用原生的管道，但可以利用 ggplot2，尽管可能会弃用公式语法，而更喜欢新的匿名函数语法，它仍然有效。

mtcars |> 
  ggplot() +
  geom_col(data = ~filter(., cyl == 4),
           aes(x = gear, y = mpg),
           fill = "grey80") +
  geom_text(data =  ~filter(., cyl == 4 & gear == 3),
            aes(x = gear,
                y = mpg,
                label = "ooo"), color = "tomato3") +
  theme_minimal()

尽管如此，我个人非常相信不要在绘图/图表代码中进行数据操作。
我更愿意提前进行所有数据更改，然后编写纯粹的绘图函数。

d1 <- filter(mtcars, cyl == 4)
d2 <- filter(mtcars, cyl == 4 & gear == 3)

  ggplot() +
  geom_col(data =  d1,
           aes(x = gear, y = mpg),
           fill = "grey80") +
  geom_text(data =  d2,
            aes(x = gear,
                y = mpg,
                label = "ooo"), color = "tomato3") +
  theme_minimal()
```"

<details>
<summary>英文:</summary>

Its possible to stay with the native pipe, but leverage that ggplot2 though perhaps deprecating formula syntax in favour of the new anonymous function syntax, still works.

mtcars |>
ggplot() +
geom_col(data = ~filter(., cyl == 4),
aes(x = gear, y = mpg),
fill = "grey80") +
geom_text(data = ~filter(., cyl == 4 & gear == 3),
aes(x = gear,
y = mpg,
label = "ooo"), color = "tomato3") +
theme_minimal()

That said, I personally am a great believer of not doing data manipulations in my plotting/charting code. 
I would much rather do all my data changes up front, and then write a purely plotting function.

d1 <- filter(mtcars, cyl == 4)
d2 <- filter(mtcars, cyl == 4 & gear == 3)

ggplot() +
geom_col(data = d1,
aes(x = gear, y = mpg),
fill = "grey80") +
geom_text(data = d2,
aes(x = gear,
y = mpg,
label = "ooo"), color = "tomato3") +
theme_minimal()

答案2

得分: 2

1. 基础管道中的占位符 _ |> 指的是随后的函数中的第一个参数。
2. magrittr 管道 %>% 中的占位符 . 指的是正在传递的数据。
3. 在 ggplot2 函数中，数据参数预期是数据框或类似对象（例如 tibble），而不是函数调用。
4. 因此，您不能直接在数据参数中使用基础管道和占位符。
5. 但是，您可以在数据参数中使用匿名函数，该函数以数据框作为输入，并在此函数内部使用基础管道和占位符。
6. 然后，您可以对此数据框应用函数并返回一个数据框，该数据框可以直接供 ggplot2 使用。

因此，您的第二种方法有效！