26. 求助！请问你会画dodge, stack 图吗？

最近要复现一个图：

对于练习时常三周半的我来说，这还不简单？

但当我摇摆在fill 指定多个属性以及position 是dodge 还是stack 间时，现实给了我狠狠的一记耳光。正如有个说的：

ggplot2 is powerful but opinionated software- sometimes it is the ggplot way or no way at all. – Sharpie Jul 18 '10 at 19:32

这可真是离谱妈妈给离谱开门，离谱到家了。

于是我退而求其次，按照https://stackoverflow.com/questions/12715635/ggplot2-bar-plot-with-both-stack-and-dodge 建议，选择了facet。

有个参数可以解决这个问题scales = “free”：

(p1 <- ggplot(snv_df) + theme_classic() + 
  geom_bar(aes(Tumor_Sample_Barcode, fill = Variant_Type)) + 
  theme(
    panel.grid = element_blank(),
    axis.text.x = element_blank(), 
    axis.ticks.x = element_blank(), 
    axis.line.x = element_blank()) + 
  labs(x = "") + scale_y_continuous(expand=c(0,0)) +
  scale_x_discrete(expand = expansion(mult = c(0.03,0)))) + 
  facet_grid(.~Patient_Barcode,scales = "free")

free 让每个数据都有自己的数据表现，而space 则可以让每个数据有自己独立的坐标系：

(p1 <- ggplot(snv_df) + theme_classic() + 
  geom_bar(aes(Tumor_Sample_Barcode, fill = Variant_Type)) + 
  theme(
    panel.grid = element_blank(),
    axis.text.x = element_blank(), 
    axis.ticks.x = element_blank(), 
    axis.line.x = element_blank()) + 
  labs(x = "") + scale_y_continuous(expand=c(0,0)) +
  scale_x_discrete(expand = expansion(mult = c(0.03,0)))) + 
  facet_grid(.~Patient_Barcode, scales = "free", space = "free")

然而这样又有新的问题了，分面的注释无法显示完全名称。

这时候我们可以通过strip 修改一下分面注释对象，让其无背景，并且垂直显示：

(p1 <- ggplot(snv_df) + theme_classic() + 
  geom_bar(aes(Tumor_Sample_Barcode, fill = Variant_Type)) + 
  theme(
    panel.grid = element_blank(),
    axis.text.x = element_blank(), 
    axis.ticks.x = element_blank(), 
    axis.line.x = element_blank(),
    axis.title = element_text(size = 18), 
    axis.text = element_text(size = 14), 
    legend.title = element_blank(),
    legend.text = element_text(size = 12)) + 
  theme(strip.background = element_blank(),
        strip.text.x = element_text(angle = 90, size = 12)) + 
  labs(x = "") + scale_y_continuous(expand=c(0,0)) +
  scale_x_discrete(expand = expansion(mult = c(0.03,0)))) + 
  facet_grid(.~Patient_Barcode, scales = "free", space = "free")

接下来就是让facet 的内容从默认的top 变到底部，“假装”它是通过aes 分组出来的内容：

(p1 <- ggplot(snv_df) + theme_classic() + 
  geom_bar(aes(Tumor_Sample_Barcode, fill = Variant_Type)) + 
  theme(
    panel.grid = element_blank(),
    axis.text.x = element_blank(), 
    axis.ticks.x = element_blank(), 
    axis.line.x = element_blank(),
    axis.title = element_text(size = 18), 
    axis.text = element_text(size = 14), 
    legend.title = element_blank(),
    legend.text = element_text(size = 12),
    legend.position = c(.7, .97),
    legend.background = element_rect(fill = "transparent")) + 
  theme(strip.background = element_blank(),
        strip.text.x = element_text(angle = 90, size = 10, face = "bold")) + 
  labs(x = "") + scale_y_continuous(expand=c(0,0)) +
  scale_x_discrete(expand = expansion(mult = c(0.03,0)))) + 
  theme(legend.spacing.x = unit(.5, 'cm')) + 
  guides(fill = guide_legend(direction = "horizontal")) + 
  facet_grid(.~Patient_Barcode, scales = "free", space = "free", switch = "x")

再美化一下：

(p2 <- p1 + scale_fill_paletteer_d("ggsci::default_nejm") + 
    theme(legend.background = element_rect(fill = "white",
                                            color = "gray60"),
          panel.background = element_rect(fill = "#d5e4eb"), 
          plot.background = element_rect(fill = "#d5e4eb"))
)