带有空 RHS 的 `dcast` [英] `dcast` with empty RHS
问题描述
有没有办法在不首先指定要翻转的变量的情况下翻转我的数据?在我看来，逻辑默认值是组内索引.
Is there any way to flip wide my data without first specifying a variable against which to be flipped? The logical default seems to me to be the in-group index.
例如，
DT <- data.table(id = rep(6:10, each = 3), var = rnorm(15))
DT
# id var
# 1: 6 1.58293930
# 2: 6 0.44234019
# 3: 6 -0.06576521
# 4: 7 -0.65124980
# 5: 7 0.88371933
# 6: 7 -1.94998135
# 7: 8 -1.95746466
# 8: 8 -0.50978195
# 9: 8 -0.40450447
# 10: 9 -0.61097399
# 11: 9 -0.92335213
# 12: 9 -0.19881983
# 13: 10 0.13022635
# 14: 10 -0.30141200
# 15: 10 0.78355188
我想要的基本上是，对于每个 id，var 的每个值在不同的列中(如果有任何 NAscode>id 与较少的 var 值关联)，可以这样做:
What I want is basically, for each id, each value of var in a different column (and NAs if there's any id with fewer var values associated), which can be done like so:
DT[ , I := 1:.N, by = id]
dcast(DT, id ~ I, value.var = "var")
# id 1 2 3
# 1: 6 1.5829393 0.4423402 -0.06576521
# 2: 7 -0.6512498 0.8837193 -1.94998135
# 3: 8 -1.9574647 -0.5097820 -0.40450447
# 4: 9 -0.6109740 -0.9233521 -0.19881983
# 5: 10 0.1302263 -0.3014120 0.78355188
但是，如果我不必先定义 I 会更方便，如下所示:
However, it would be more convenient if I didn't have to define I first, like so:
dcast(DT, id~ ., value.var = "var")
但这不起作用:
缺少聚合函数，默认为 'length'
Aggregate function missing, defaulting to 'length'
# id .
# 1: 6 3
# 2: 7 3
# 3: 8 3
# 4: 9 3
# 5: 10 3
是否有我可以传递的聚合函数来获得所需的效果?
Is there perhaps an aggregating function that I could pass to get the desired effect?
推荐答案
现在可以使用 rowid 函数:
This is now possible using the rowid function:
dcast(DT, id ~ rowid(id), value.var = "var")
# id 1 2 3
# 1: 6 1.1050942 0.1271620 1.3051373
# 2: 7 -0.5441056 -0.6866828 -0.8083762
# 3: 8 -0.6812820 -1.1934716 -1.3913903
# 4: 9 -0.3462497 -0.8229276 -1.0884394
# 5: 10 -0.4600681 0.6173795 -1.0125658
有关更多选项、示例和说明，请参阅 ?rowid.
See ?rowid for more options, examples, and explanation.
这篇关于带有空 RHS 的 `dcast`的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！
