ggvis中的自定义填充颜色（以及其他选项）

Question

我正在尝试使用ggvis为2014赛季创建NFL强度排名图表。数据来自FootballOutsiders.com，稍后我将制作一个Shiny应用程序，在赛季更新时自动从网站上抓取数据。下面的示例非常接近，但我想做一些修改。我想要：

1. 在图表的每个单元格中包含“防守”数值，使可视化效果类似于原始的“df”数据框。

2. 自定义颜色比例，使正值越来越橙色，负值越来越蓝色（即更负=更蓝）。

3. 替代方案2可以是从橙色到蓝色的渐变，并使“防御”的值接近零时不透明度降低至0.5。

4. 能够选择NA的颜色，因为它目前在图表中显示为黑色。

我一直在尝试使用add_scale()和props()进行调整，但迄今为止还没有什么效果。

这是图表： 这是数据：

df <- structure(list(team = c("ARI", "ATL", "BAL", "BUF", "CAR", "CHI", 
"CIN", "CLE", "DAL", "DEN", "DET", "GB", "HOU", "IND", "JAX", 
"KC", "MIA", "MIN", "NE", "NO", "NYG", "NYJ", "OAK", "PHI", "PIT", 
"SD", "SEA", "SF", "STL", "TB", "TEN", "WAS"), w1 = c(17.5, -5.8, 
-12.6, 8.7, -6.8, -13.8, -8.7, 4, -4.6, 0.9, -11.4, -25.9, 4.2, 
-0.2, 4.9, 4.2, 4.2, -5.7, 2.4, 13.5, -0.8, 10.3, -5.6, 10.9, 
8.2, -16.4, 14.4, 13.8, 10.5, -15.7, -6.7, 2.5), w2 = c(-11.4, 
-12.6, 4, 2.4, -0.8, -4.6, 13.5, -5.8, 4.2, -6.7, -15.7, -5.6, 
10.3, 4.9, 4.2, -0.2, -13.8, 4.2, 10.5, 8.2, -16.4, 14.4, 2.5, 
0.9, -8.7, -25.9, 17.5, 8.7, -6.8, -5.7, 13.8, 10.9), w3 = c(-4.6, 
-6.8, 8.2, 17.5, 4, -5.6, 4.2, -8.7, -5.7, -25.9, 14.4, -0.8, 
-11.4, 10.9, 0.9, 2.4, -6.7, -5.8, 10.3, 10.5, 2.5, 8.7, 4.2, 
4.2, -15.7, -13.8, -0.2, -16.4, 13.8, 13.5, -12.6, 4.9), w4 = c(NA, 
10.5, -15.7, 2.5, -8.7, 14.4, NA, NA, -5.8, NA, -5.6, 8.7, -13.8, 
4.2, 17.5, 4.2, 10.3, 13.5, -6.7, 13.8, 4.2, -0.8, 2.4, -4.6, 
-6.8, 10.9, NA, 4.9, NA, 4, 0.9, -11.4), w5 = c(-0.2, -11.4, 
0.9, -0.8, 8.7, -15.7, 4.2, 4.2, 2.5, -16.4, -13.8, 10.5, 13.8, 
-8.7, 4, -4.6, NA, 14.4, -12.6, -6.8, 13.5, 17.5, NA, -5.7, 10.9, 
-5.6, 4.2, -6.7, 4.9, -5.8, 8.2, -25.9), w6 = c(4.2, 8.7, -6.8, 
4.2, -12.6, 13.5, -15.7, 4, -25.9, -5.6, 10.5, 2.4, 0.9, 2.5, 
4.2, NA, 14.4, -0.8, -13.8, NA, 4.9, -0.2, 17.5, -11.4, 8.2, 
10.3, 13.8, -5.7, -4.6, -8.7, 10.9, -16.4), w7 = c(10.3, -8.7, 
13.5, 10.5, 14.4, 2.4, 0.9, 10.9, -11.4, -4.6, -5.8, -15.7, 4, 
-12.6, 8.2, 17.5, 8.7, -13.8, -5.6, -0.8, 13.8, 4.2, -16.4, NA, 
2.5, -6.7, -5.7, -0.2, -25.9, NA, 4.2, 4.2), w8 = c(4.9, -0.8, 
-12.6, -5.6, -25.9, 4.2, -8.7, 10.3, 4.2, 17.5, 13.5, -5.8, 4.2, 
4, 2.4, -5.7, 10.9, -6.8, 8.7, 14.4, NA, -13.8, 8.2, -16.4, 0.9, 
-0.2, -15.7, NA, -6.7, 10.5, 2.5, 13.8), w9 = c(13.8, NA, 4, 
NA, -5.8, NA, 10.9, -6.8, -16.4, 4.2, NA, NA, 4.9, -11.4, -12.6, 
-5.6, 17.5, 4.2, -0.2, -15.7, 0.9, -6.7, -25.9, 2.5, -8.7, 2.4, 
10.3, -5.7, -4.6, 8.2, NA, 10.5), w10 = c(-5.7, -6.8, 4.2, -6.7, 
4.9, 14.4, 8.2, -12.6, 10.9, 10.3, 2.4, 8.7, NA, NA, 13.8, -13.8, 
-0.8, NA, NA, -4.6, -25.9, 4, -0.2, -15.7, -5.6, NA, -11.4, -5.8, 
-16.4, 13.5, -8.7, NA), w11 = c(-0.8, -15.7, NA, 2.4, 13.5, 10.5, 
-5.8, 2.5, NA, -5.7, -16.4, 4.9, 8.2, 4.2, NA, -25.9, -13.8, 
8.7, 0.9, -12.6, -4.6, NA, 17.5, 14.4, 4.2, 10.3, -6.7, -11.4, 
-0.2, 4.2, 4, -6.8), w12 = c(-25.9, 8.2, -5.8, -5.6, NA, -6.8, 
2.5, 13.5, -11.4, 2.4, 4.2, 10.5, -12.6, 10.9, 0.9, 10.3, -0.2, 
14.4, -0.8, -8.7, 13.8, -13.8, -6.7, 4.2, NA, -5.7, -16.4, 4.2, 
17.5, 8.7, 4.9, -4.6), w13 = c(13.5, -16.4, 17.5, 8.2, 10.5, 
-0.8, -6.8, -13.8, 4.9, -6.7, 8.7, 4.2, 4.2, 4.2, -11.4, -0.2, 
-5.6, -15.7, 14.4, 4, 10.9, 2.4, -5.7, 13.8, -5.8, -8.7, -4.6, 
-25.9, 10.3, -12.6, 2.5, 0.9), w14 = c(-6.7, 14.4, 2.4, -0.2, 
-5.8, 13.8, 4, 0.9, 8.7, -13.8, -6.8, 13.5, 10.9, 8.2, 2.5, -16.4, 
-8.7, -5.6, 17.5, -15.7, 4.2, 10.5, -4.6, -25.9, -12.6, 4.2, 
4.9, 10.3, 4.2, -0.8, -11.4, -5.7), w15 = c(-5.7, 4, 10.9, 14.4, 
-6.8, -5.8, 8.2, -12.6, 4.9, 17.5, 10.5, -13.8, 0.9, 2.5, -8.7, 
10.3, 4.2, -0.8, 2.4, 8.7, 4.2, 4.2, -6.7, 13.8, 13.5, -0.2, 
-4.6, -25.9, -16.4, -15.7, -5.6, -11.4), w16 = c(-25.9, -5.8, 
2.5, 10.3, 8.2, -0.8, -0.2, -15.7, 0.9, -12.6, 8.7, -6.8, -8.7, 
13.8, 4.2, 4, 10.5, 2.4, -5.6, 13.5, -5.7, 4.2, -13.8, 4.2, -6.7, 
-4.6, -16.4, 17.5, -11.4, 14.4, 10.9, 4.9), w17 = c(-4.6, -15.7, 
8.2, 4.2, 13.5, 10.5, 4, -8.7, 4.2, 10.3, 14.4, -0.8, 10.9, 4.2, 
2.5, 17.5, -5.6, 8.7, -13.8, -6.8, 4.9, 2.4, -0.2, -11.4, -12.6, 
-6.7, -5.7, -16.4, -25.9, -5.8, 0.9, 13.8)), .Names = c("team", 
"w1", "w2", "w3", "w4", "w5", "w6", "w7", "w8", "w9", "w10", 
"w11", "w12", "w13", "w14", "w15", "w16", "w17"), row.names = c(NA, 
32L), class = "data.frame")

以下是目前的代码：

require(dplyr)
require(ggvis)
require(tidyr) # For the gather function

df2 <- df %>% gather(key, value, w1:w17)
names(df2) <- c("team", "week", "defense")

df2 %>% 
  ggvis(~week, ~team, fill = ~defense) %>%
  layer_rects(width = band(), height = band()) %>%
  scale_nominal("x", padding = 0, points = FALSE) %>%
  scale_nominal("y", padding = 0, points = FALSE)

- rsoren

Answer 1

我通过创建一个新变量def.color来为每个单元格设置颜色，该变量将defense的每个值映射到特定的颜色。在ggplot2中，你可以直接在调用ggplot时使用一行代码设置颜色，例如scale_fill_manual()，而不是向数据框添加颜色变量。我希望在ggvis中也能这样做，但我还没有找到方法。所以，现在我们开始：

# Create a new variable df2$def.color for mapping df2$defense values to colors

# Functions to create color ramps for the blue and orange color ranges
Blue = colorRampPalette(c("darkblue","lightblue"))
Orange = colorRampPalette(c("orange","darkorange3"))

# Negative values of defense get a blue color scale with 10 colors
df2$def.color[!is.na(df2$defense) & df2$defense<0] = 
  as.character(cut(df2$defense[!is.na(df2$defense) & df2$defense<0], 
                   seq(min(df2$defense - 0.1, na.rm=TRUE), 0, length.out=11), 
                   labels=Blue(10)))

# Positive values of defense get an orange color scale with 10 colors
df2$def.color[!is.na(df2$defense) & df2$defense>=0] = 
  as.character(cut(df2$defense[!is.na(df2$defense) & df2$defense>=0], 
                   seq(0, max(df2$defense, na.rm=TRUE)+0.1, length.out=11), 
                   labels=Orange(10)))

# Set NA values in df2$def.color to light gray in df2$def.color
df2$def.color[is.na(df2$defense)] = "#E5E5E5"  

# Set NA values in df2$defense to blanks so that we won't get "NaN" in cells with 
# missing data
df2$defense[is.na(df2$defense)] = ""

现在我们创建图表。为了获取颜色，使用:=将def.color映射到fill。要添加defense的值，请使用layer_text。我对每个单元格内文本的位置不太满意，但目前这是我能想到的最好的方法。

df2 %>% 
  ggvis(~week, ~team, fill:=~def.color) %>% 
  layer_rects(width = band(), height = band()) %>%
  scale_nominal("x", padding = 0, points = FALSE) %>%
  scale_nominal("y", padding = 0, points = FALSE) %>%
  layer_text(text:=~defense, stroke:="white", align:="left", baseline:="top") 

- eipi10

---

似乎有点违反直觉，需要手动将变量映射到颜色上。 ggvis的比例函数还没有实现这个功能吗？ - Peter Ellis

我有一段时间没有尝试使用ggvis了，因为它一直在快速发展。我同意让比例函数来处理这个问题会更好。希望现在或不久的将来会有解决方法。我还想找到一种方法来更好地控制数字放置在单元格中的位置。 - eipi10

我自己只看了几天的ggvis。从我所读的内容来看，轻松实现这一点是未来的一个特性。目前，您可以定义一个填充或描边比例尺，在每个端点处使用任意颜色，但您还不能拥有类似于{scales}的scale_fill_gradientn的等效物——这对于蓝白红或颜色相互混合的情况至关重要。 - Peter Ellis

请问 @eipi10 是否已经成功地为图表添加了图例？ - altimit

Answer 2

我发现一种显示图例的解决方案，利用了scale_ordinal函数。我使用了@eipi10编写的大量代码，非常感谢！

# Functions to create color ramps for the blue and orange color ranges,
# combined in a single palette with 10 colors of each ramp and gray for NAs
Blue <- colorRampPalette(c("darkblue","lightblue"))
Orange <- colorRampPalette(c("orange","darkorange3"))
palette <- c(Blue(10), "#E5E5E5", Orange(10))

# Negative values of defense get a blue color scale with 10 colors, indexes
# from 1 to 10
df2$def.label[!is.na(df2$defense) & df2$defense<0] <- 
    as.character(cut(df2$defense[!is.na(df2$defense) & df2$defense<0], 
                 seq(min(df2$defense - 0.1, na.rm = TRUE), 0, length.out = 11), 
                 labels = palette[1:10]))

# Positive values of defense get an orange color scale with 10 colors,
# indexes from 12 to 21
df2$def.label[!is.na(df2$defense) & df2$defense>=0] <-
    as.character(cut(df2$defense[!is.na(df2$defense) & df2$defense>=0], 
                 seq(0, max(df2$defense, na.rm = TRUE) + 0.1, length.out = 11), 
                 labels = palette[12:21]))

# Set NA values in df2$defense to 11 in def.label, the label for gray color
df2$def.label[is.na(df2$defense)] <- palette[[11]] 

# Define the values to be displayed on the legend
pos.cut.values <- seq(0, max(df2$defense, na.rm = TRUE) + 0.1, length.out = 11)
neg.cut.values <- seq(min(df2$defense - 0.1, na.rm = TRUE), 0, length.out = 11)
legend.values <- c(paste(neg.cut.values[1:10], '..', neg.cut.values[2:11]),
                   'NA', paste(pos.cut.values[1:10], '..', pos.cut.values[2:11]))

# Set NA values in df2$defense to blanks so that we won't get "NaN" in cells
# with missing data
df2$defense[is.na(df2$defense)] <- ""

df2 %>% 
    ggvis(~week, ~team, fill:=~def.label) %>% 
    scale_ordinal('fill', range = palette) %>%
    add_legend(scales = 'fill', values = legend.values) %>%
    layer_rects(width = band(), height = band()) %>%
    scale_nominal("x", padding = 0, points = FALSE) %>%
    scale_nominal("y", padding = 0, points = FALSE) %>%
    layer_text(text := ~defense, stroke := "white", align := "left",
               baseline := "top") 

- skasch