ggplot2：将geom_line和geom_point组合在一起的图形中的坐标图例

这是示例数据帧：

dt <- data.frame(x = 1:6, y2017 = 5:10, y2018 = 12:17, y2019 = 15:20)

#  x y2017 y2018 y2019
#  1     5    12    15
#  2     6    13    16
#  3     7    14    17
#  4     8    15    18
#  5     9    16    19
#  6    10    17    20

这是我最初所做的：

ggplot(dt, aes(x = x)) +
  geom_point(aes(y = y2017, color = "2017", fill = "2017", shape = "2017"), size = 2) + 
  geom_line(aes(y = y2017, color = "2017")) +
  geom_point(aes(y = y2018, color = "2018", fill = "2018", shape = "2018"), size = 2) + 
  geom_point(aes(y = y2019, color = "2019", fill = "2019", shape = "2019"), size = 2) + 
  ylab("Y") +
  scale_colour_manual(name = "", 
                      breaks = c("2017", "2018","2019"),
                      values = c("#D16103", "#52854C", "#293352")) +
  scale_shape_manual(name = "", 
                     breaks = c("2017", "2018","2019"),
                     values = c(3, 4, 21)) +
  scale_fill_manual(name = "", 
                    breaks = c("2017", "2018","2019"),
                    values = c("#D16103", "#52854C", "#293352")) + 
  theme_bw()

结果图为：

对于Legend2018和2019，有一条直线贯穿这一点，这不是我想要的。于是，我试着继续color在geom_line中2017出的aes()。

ggplot(dt, aes(x = x)) +
  geom_point(aes(y = y2017, color = "2017", fill = "2017", shape = "2017"), size = 2) + 
# Here is what I change:
  geom_line(aes(y = y2017),  color = "#D16103") +
  geom_point(aes(y = y2018, color = "2018", fill = "2018", shape = "2018"), size = 2) + 
  geom_point(aes(y = y2019, color = "2019", fill = "2019", shape = "2019"), size = 2) + 
  ylab("Y") +
  scale_colour_manual(name = "", 
                      breaks = c("2017", "2018","2019"),
                      values = c("#D16103", "#52854C", "#293352")) +
  scale_shape_manual(name = "", 
                     breaks = c("2017", "2018","2019"),
                     values = c(3, 4, 21)) +
  scale_fill_manual(name = "", 
                    breaks = c("2017", "2018","2019"),
                    values = c("#D16103", "#52854C", "#293352")) + 
  theme_bw()

然后结果变为：

在的图例中2017，与之交叉的线不再存在，而2018和2019是正确的。

我不知道是否有一种方法允许的传奇组合2017在第一张图和2018和2019第二图。

（我知道我可以pivot_longer()用来变换数据集并使其更容易绘制图例，但是对于我的主要数据集来说，这变得更加困难。所以我坚持这种方式）