跳到主要内容

堆叠面积图

堆叠面积图把几个系列叠在一起,于是每个 x 位置上既能看见每个系列贡献了多少,也能看见合起来是多少。要表达「一个 整体沿一条连续轴(通常是时间)由哪些部分构成」,它就是最自然的那张图。

WITH d AS (SELECT * FROM read_csv_auto('{{DFK_BASE_URL}}data/stacked_area.tsv'))
SELECT kuva_render(to_json({
'title': 'Abundance by species',
'x_axis': {'name': 'week', 'tick_format': 'integer'},
'y_axis': {'name': 'abundance'},
'legend': {'position': 'outside_right_top'},
'series': [{
'type': 'stacked_area',
'x': (SELECT list(week ORDER BY week) FROM (SELECT DISTINCT week FROM d)),
'series': (SELECT list({'label': species, 'values': vals} ORDER BY species)
FROM (SELECT species, list(abundance ORDER BY week) AS vals
FROM d GROUP BY species))
}]
})) AS chart;

x 是共享的横轴,每个系列每个 x 一个值。数据是长格式,所以透视交给查询做:一行一个系列, list(… ORDER BY week) 把它的数值收成正确的顺序。

归一化(100 % 堆叠)​

normalized 把每一列重新缩放到各系列加起来等于 100 %,y 轴随之变成 0–100 %。故事是构成而不是量级时用它 —— 否则一个不断增长的总量会把所有小色带压平,把它们之间的此消彼长全藏起来。

WITH d AS (SELECT * FROM read_csv_auto('{{DFK_BASE_URL}}data/stacked_area.tsv'))
SELECT kuva_render(to_json({
'title': 'Relative abundance',
'x_axis': {'name': 'week', 'tick_format': 'integer'},
'y_axis': {'name': 'share of total (%)'},
'legend': {'position': 'outside_right_top'},
'series': [{
'type': 'stacked_area',
'x': (SELECT list(week ORDER BY week) FROM (SELECT DISTINCT week FROM d)),
'series': (SELECT list({'label': species, 'values': vals} ORDER BY species)
FROM (SELECT species, list(abundance ORDER BY week) AS vals
FROM d GROUP BY species)),
'normalized': true
}]
})) AS chart;

顶边描边​

默认会沿每条色带的顶边画一道描边,把色相相近的相邻色带分开。"show_strokes": false 全部去掉,得到更柔和、更平 的观感 —— 颜色本来就够对比时,这样更好看。

WITH d AS (SELECT * FROM read_csv_auto('{{DFK_BASE_URL}}data/stacked_area.tsv'))
SELECT kuva_render(to_json({
'title': 'Without strokes',
'x_axis': {'name': 'week', 'tick_format': 'integer'},
'y_axis': {'name': 'abundance'},
'legend': {'position': 'outside_right_top'},
'series': [{
'type': 'stacked_area',
'x': (SELECT list(week ORDER BY week) FROM (SELECT DISTINCT week FROM d)),
'series': (SELECT list({'label': species, 'values': vals} ORDER BY species)
FROM (SELECT species, list(abundance ORDER BY week) AS vals
FROM d GROUP BY species)),
'show_strokes': false,
'fill_opacity': 0.95
}]
})) AS chart;

图例位置​

legend.position 取图例页上的任意名字。堆叠面积图常用的是默认的右侧,以及图内 四个角 —— 图内图例会遮住数据,所以只有留白足够时才用:

WITH d AS (SELECT * FROM read_csv_auto('{{DFK_BASE_URL}}data/stacked_area.tsv'))
SELECT kuva_render(to_json({
'title': 'Legend inside',
'x_axis': {'name': 'week', 'tick_format': 'integer'},
'y_axis': {'name': 'abundance'},
'legend': {'position': 'inside_top_left'},
'series': [{
'type': 'stacked_area',
'x': (SELECT list(week ORDER BY week) FROM (SELECT DISTINCT week FROM d)),
'series': (SELECT list({'label': species, 'values': vals} ORDER BY species)
FROM (SELECT species, list(abundance ORDER BY week) AS vals
FROM d GROUP BY species)),
'fill_opacity': 0.55
}]
})) AS chart;

样式​

fill_opacity 是每条色带的不透明度(默认 0.7):调低能让网格透出来,1 就是完全不透明。 stroke_width 是顶边描边的宽度(默认 1.5),一旦关掉 show_strokes 它就没有作用了。

WITH d AS (SELECT * FROM read_csv_auto('{{DFK_BASE_URL}}data/stacked_area.tsv'))
SELECT kuva_render(to_json({
'title': 'Nearly opaque bands',
'x_axis': {'name': 'week', 'tick_format': 'integer'},
'y_axis': {'name': 'abundance'},
'legend': {'position': 'outside_right_top'},
'series': [{
'type': 'stacked_area',
'x': (SELECT list(week ORDER BY week) FROM (SELECT DISTINCT week FROM d)),
'series': (SELECT list({'label': species, 'values': vals} ORDER BY species)
FROM (SELECT species, list(abundance ORDER BY week) AS vals
FROM d GROUP BY species)),
'fill_opacity': 0.9,
'stroke_width': 2.5
}]
})) AS chart;

颜色​

系列不给 color 就按调色板顺次取色,超过八个之后循环。分类有约定俗成的颜色时,或者同一个物种会出现在好几张图里、 你应该让它在每张图里都是同一个颜色时,就逐个指定。

WITH d AS (SELECT * FROM read_csv_auto('{{DFK_BASE_URL}}data/stacked_area.tsv')),
s AS (
SELECT species, list(abundance ORDER BY week) AS vals
FROM d GROUP BY species
)
SELECT kuva_render(to_json({
'title': 'Explicit colours',
'x_axis': {'name': 'week', 'tick_format': 'integer'},
'y_axis': {'name': 'abundance'},
'legend': {'position': 'outside_right_top'},
'series': [{
'type': 'stacked_area',
'x': (SELECT list(week ORDER BY week) FROM (SELECT DISTINCT week FROM d)),
'series': (SELECT list({'label': species, 'values': vals,
'color': CASE species
WHEN 'Firmicutes' THEN '#2c7bb6'
WHEN 'Bacteroidetes' THEN '#fdae61'
WHEN 'Proteobacteria' THEN '#d7191c'
ELSE '#abdda4' END}
ORDER BY species) FROM s)
}]
})) AS chart;

字段​

字段类型设置什么
xnumber[]必填。 共享的横轴。
seriesseries[]必填。 每项 {values, label?, color?};每个 x 一个值。
fill_opacitynumber色带不透明度,0–1(默认 0.7)。
stroke_widthnumber顶边描边宽度(默认 1.5)。
show_strokesboolean画顶边描边(默认开)。
normalizedboolean每列重新缩放到 100 %。
legend_positionstring任意图例位置。

说明​

  • x 与每个系列的 values 都是必填;系列比 x 短会按 0 补齐 —— 那会画出一条塌成零的色带而不是报错, 所以要留意长度。
  • normalized 会把 y 轴改成 0–100 %,标签也照这个写。
  • 堆叠顺序就是 series 的顺序,所以聚合时要有意排序 —— label 是分类时,图例顺序与堆叠顺序是同一件事。
  • 某个色带特别大时会把压在上面的色带全部压平;normalized 就是为这种情形存在的。

另见​