用NA inplace data.table替换"NA"(NA字符串)

Question

我有这个虚拟数据集:

abc <- data.table(a = c("NA", "bc", "x"), b = c(1, 2, 3), c = c("n", "NA", "NA"))

我试图用标准NA替换"NA"; 使用data.table到位.我试过了:

for(i in names(abc)) (abc[which(abc[[i]] == "NA"), i := NA])
for(i in names(abc)) (abc[which(abc[[i]] == "NA"), i := NA_character_])
for(i in names(abc)) (set(abc, which(abc[[i]] == "NA"), i, NA))

但仍然有这个我得到:

abc$a 
"NA" "bc" "x"

我错过了什么？

编辑:我在这个问题中尝试了@frank的答案type.convert().(谢谢坦率;不知道这种模糊但有用的功能)在文档中type.convert()提到:"这主要是read.table的辅助函数." 所以我想彻底测试一下.当您有一个填充"NA"(NA字符串)的完整列时,此功能会产生较小的副作用.在这种情况下type.convert(),将列转换为逻辑列.对于这种情况abc将是:

abc <- data.table(a = c("NA", "bc", "x"), b = c(1, 2, 3), c = c("n", "NA", "NA"), d = c("NA", "NA", "NA"))

EDIT2:总结原始问题中的代码:

for(i in names(abc)) (set(abc, which(abc[[i]] == "NA"), i, NA))

工作正常,但仅限于当前最新版本data.table(> 1.11.4).因此,如果一个人面临这个问题,那么更好地更新data.table并使用这个代码比type.convert()

Answer 1

我会做...

chcols = names(abc)[sapply(abc, is.character)]
abc[, (chcols) := lapply(.SD, type.convert, as.is=TRUE), .SDcols=chcols]

产量

> str(abc)
Classes ‘data.table’ and 'data.frame':  3 obs. of  3 variables:
 $ a: chr  NA "bc" "x"
 $ b: num  1 2 3
 $ c: chr  "n" NA NA
 - attr(*, ".internal.selfref")=<externalptr> 

您的DT[, i :=]代码无效,因为它创建了一个名为"i"的列; 正如@AdamSampson指出的那样,你的set代码已经运行了.(注意:在他们的comp之前,OP从data.table 1.10.4-3升级到1.11.4.)

所以我想彻底测试一下.当您有一个填充"NA"(NA字符串)的完整列时,此功能会产生较小的副作用.在这种情况下type.convert(),将列转换为逻辑列.

啊对.您最初的方法可以更安全地解决此问题:

# op's new example
abc <- data.table(a = c("NA", "bc", "x"), b = c(1, 2, 3), c = c("n", "NA", "NA"), d = c("NA", "NA", "NA"))

# op's original code
for(i in names(abc)) 
  set(abc, which(abc[[i]] == "NA"), i, NA)

附注:NA具有逻辑类型; 并且通常data.table会在将不一致类型的值分配给列时发出警告,但我猜他们在NAs的例外中写道:

DT = data.table(x = 1:2)
DT[1, x := NA]
# no problem, even though x is int and NA is logi

DT = data.table(x = 1:2)
DT[1, x := TRUE]
# Warning message:
# In `[.data.table`(DT, 1, `:=`(x, TRUE)) :
#   Coerced 'logical' RHS to 'integer' to match the column's type. Either change the target column ['x'] to 'logical' first (by creating a new 'logical' vector length 2 (nrows of entire table) and assign that; i.e. 'replace' column), or coerce RHS to 'integer' (e.g. 1L, NA_[real|integer]_, as.*, etc) to make your intent clear and for speed. Or, set the column type correctly up front when you create the table and stick to it, please.