R----data.table包介紹學習

相比dplyr包，data.table包可以更大程度地提升數據的處理速度，這裏就簡單介紹一下data.tale包的使用方法。

data.table：用於快速處理大數據集的哦

數據的讀取

data.table包中數據讀取的函數：fread()

data.table的建立

library(data.table)
DT = data.table(x=rep(c("a","b","c"),each=3), y=c(1,3,6), v=1:9) DT # x y v # 1: a 1 1 # 2: a 3 2 # 3: a 6 3 # 4: b 1 4 # 5: b 3 5 # 6: b 6 6 # 7: c 1 7 # 8: c 3 8 # 9: c 6 9 

基礎操做

行提取

行提取分爲單行提取和多行提取。

單行提取

DT[2] # 2nd row # x y v # 1: a 3 2 DT[2,] # same # x y v # 1: a 3 2 

這裏DT [2]和DT [2]是徹底相同的，這裏的「，」只是說明還有其餘參數可設置，而其餘參數按默認值進行計算。下文全部這樣的最後一個「，」都再也不寫出來。

多行提取

• 數字提取

DT[1:2] # x y v # 1: a 1 1 # 2: a 3 2 DT[c(2,5)] # x y v #1: a 3 2 #2: b 3 5 

• 邏輯提取

DT[c(FALSE,TRUE)] # even rows (usual recycling) # x y v # 1: a 3 2 # 2: b 1 4 # 3: b 6 6 # 4: c 3 8 

此時，C（FALSE，TRUE）會本身重複匹配成與DT的行數相同的向量

列提取

與行提取相同，列的提取也包含單列提取和多列提取。

單列提取

• 數字提取

數字提取時，必定要把心安理得參數設置爲FALSE。

DT[,2,with=FALSE] # 2nd column # y # 1: 1 # 2: 3 # 3: 6 # 4: 1 # 5: 3 # 6: 6 # 7: 1 # 8: 3 # 9: 6 

• 列名提取

DT[,list(v)] # v column (as data.table # v # 1: 1 # 2: 2 # 3: 3 # 4: 4 # 5: 5 # 6: 6 # 7: 7 # 8: 8 # 9: 9 

列名的修改

列名的修改能夠使用setnames（）函數，這個函數好像比對data.frame類型數據名更改的名稱（）和colnames（）函數也要快一些。

dt = data.table(a=1:2,b=3:4,c=5:6) # compare to data.table try(tracemem(dt)) # by reference, no deep or shallow copies setnames(dt,"b","B") # by name, no match() needed (warning if "b" is missing) setnames(dt,3,"C") # by position with warning if 3 > ncol(dt) setnames(dt,2:3,c("D","E")) # multiple setnames(dt,c("a","E"),c("A","F")) # multiple by name (warning if either "a" or "E" is missing) setnames(dt,c("X","Y","Z")) # replace all (length of names must be == ncol(DT)) 

多列提取

• 數字提取

如同上面對按數字對單列的提取，對多列提取也要設置與參數爲FALSE。

DT[,2:3,with=FALSE] # y v # 1: 1 1 # 2: 3 2 # 3: 6 3 # 4: 1 4 # 5: 3 5 # 6: 6 6 # 7: 1 7 # 8: 3 8 # 9: 6 9 DT[,c(1,3),with=FALSE] # x v # 1: a 1 # 2: a 2 # 3: a 3 # 4: b 4 # 5: b 5 # 6: b 6 # 7: c 7 # 8: c 8 # 9: c 9 

• 按列名提取

DT[,list(y, v)] # y v # 1: 1 1 # 2: 3 2 # 3: 6 3 # 4: 1 4 # 5: 3 5 # 6: 6 6 # 7: 1 7 # 8: 3 8 # 9: 6 9 

若是按列名提取時，不使用列表，仍然能對列進行提取，只是結果以向量的形式輸出。

DT[,v] # v column (as vector) # [1] 1 2 3 4 5 6 7 8 9 DT[,c(v)] # same # [1] 1 2 3 4 5 6 7 8 9 DT[, c(y, v)] # [1] 1 3 6 1 3 6 1 3 6 1 2 3 4 5 6 7 8 9 

列的添加與刪除

列的添加

• 單列添加

DT
    # x y v # 1: a 1 1 # 2: a 3 2 # 3: a 6 3 # 4: b 1 4 # 5: b 3 5 # 6: b 6 6 # 7: c 1 7 # 8: c 3 8 # 9: c 6 9 DT[, a := 'k'] DT # x y v a # 1: a 1 1 k # 2: a 3 2 k # 3: a 6 3 k # 4: b 1 4 k # 5: b 3 5 k # 6: b 6 6 k # 7: c 1 7 k # 8: c 3 8 k # 9: c 6 9 k DT[,c:=8] # add a numeric column, 8 for all rows DT # x y v a c # 1: a 1 1 k 8 # 2: a 3 2 k 8 # 3: a 6 3 k 8 # 4: b 1 4 k 8 # 5: b 3 5 k 8 # 6: b 6 6 k 8 # 7: c 1 7 k 8 # 8: c 3 8 k 8 # 9: c 6 9 k 8 DT[,d:=9L] # add an integer column, 9L for all rows DT[2,d:=10L] # subassign by reference to column d DT # x y v a c d # 1: a 1 1 k 8 9 # 2: a 3 2 k 8 10 # 3: a 6 3 k 8 9 # 4: b 1 4 k 8 9 # 5: b 3 5 k 8 9 # 6: b 6 6 k 8 9 # 7: c 1 7 k 8 9 # 8: c 3 8 k 8 9 # 9: c 6 9 k 8 9 DT[, e := d + 2] DT # x y v a c d e # 1: a 1 1 k 8 9 11 # 2: a 3 2 k 8 10 12 # 3: a 6 3 k 8 9 11 # 4: b 1 4 k 8 9 11 # 5: b 3 5 k 8 9 11 # 6: b 6 6 k 8 9 11 # 7: c 1 7 k 8 9 11 # 8: c 3 8 k 8 9 11 # 9: c 6 9 k 8 9 11 

若是添加的列名，數據中已經包含則是對這一列數據的修改。

• 多列的添加

DT[, c('f', 'g') := list( d + 1, c)] DT[, ':='( f = d + 1, g = c)] # same DT # x y v a c d e f g # 1: a 1 1 k 8 9 11 10 8 # 2: a 3 2 k 8 10 12 11 8 # 3: a 6 3 k 8 9 11 10 8 # 4: b 1 4 k 8 9 11 10 8 # 5: b 3 5 k 8 9 11 10 8 # 6: b 6 6 k 8 9 11 10 8 # 7: c 1 7 k 8 9 11 10 8 # 8: c 3 8 k 8 9 11 10 8 # 9: c 6 9 k 8 9 11 10 8 

此處，須要注意的是新建立的列只能依照原有數據列，而不能依照新建立的列。例如這個例子中，G = C是能夠運行，而摹= F則會提示錯誤。

列的刪除

DT[,c:=NULL] # remove column c DT # x y v a d e f g # 1: a 1 1 k 9 11 10 8 # 2: a 3 2 k 10 12 11 8 # 3: a 6 3 k 9 11 10 8 # 4: b 1 4 k 9 11 10 8 # 5: b 3 5 k 9 11 10 8 # 6: b 6 6 k 9 11 10 8 # 7: c 1 7 k 9 11 10 8 # 8: c 3 8 k 9 11 10 8 # 9: c 6 9 k 9 11 10 8 DT[, c('d', 'e', 'f', 'g'):=NULL] DT # x y v a # 1: a 1 1 k # 2: a 3 2 k # 3: a 6 3 k # 4: b 1 4 k # 5: b 3 5 k # 6: b 6 6 k # 7: c 1 7 k # 8: c 3 8 k # 9: c 6 9 k 

列指標的簡單操做

簡單操做主要包括求和，平均值，方差和標準差等。

DT[2:3,sum(v)] # sum(v) over rows 2 and 3 # [1] 5 DT[2:3,mean(v)] # sum(v) over rows 2 and 3 # [1] 2.5 

索引鍵

查看和建立索引

索引是對列而言的，索引建立後，數據將自動按索引值進行從新排序，因此每一個數據最多隻能有一個索引，可是索引能夠由多列組成，這些列能夠是數字，因子，字符串或其餘格式。

單列索引的建立

## methdod first key(DT) # key # NULL setkey(DT,x) # set a 1-column key. No quotes, for convenience. key(DT) [1] "x" DT # x y v a # 1: a 1 1 k # 2: a 3 2 k # 3: a 6 3 k # 4: b 1 4 k # 5: b 3 5 k # 6: b 6 6 k # 7: c 1 7 k # 8: c 3 8 k # 9: c 6 9 k ## method second setkeyv(DT,"y") # same (v in setkeyv stands for vector) key(DT) # [1] "y" 

一旦對數據進行新的索引，原有的索引將消失。

多列索引的建立

## methdod first # key setkey(DT,x,v) # set a 1-column key. No quotes, for convenience. key(DT) # [1] "x" "v" DT # x y v a # 1: a 1 1 k # 2: a 3 2 k # 3: a 6 3 k # 4: b 1 4 k # 5: b 3 5 k # 6: b 6 6 k # 7: c 1 7 k # 8: c 3 8 k # 9: c 6 9 k ## method second setkeyv(DT,c("x", "y")) # same (v in setkeyv stands for vector) key(DT) # [1] "x" "v" DT # x y v a # 1: a 1 1 k # 2: a 3 2 k # 3: a 6 3 k # 4: b 1 4 k # 5: b 3 5 k # 6: b 6 6 k # 7: c 1 7 k # 8: c 3 8 k # 9: c 6 9 k 

經過索引進行數據的提取

按照索引對數據提取，能夠加快提取數據的速度。

單索引

正向提取

setkey(DT, x)
DT["a"] # binary search (fast) # x y v a # 1: a 1 1 k # 2: a 3 2 k # 3: a 6 3 k DT[.(x=="a")] # same; i.e. binary search (fast) # x y v a # 1: a 1 1 k # 2: a 3 2 k # 3: a 6 3 k DT[x=="a"] # same; i.e. binary search (fast) # x y v a # 1: a 1 1 k # 2: a 3 2 k # 3: a 6 3 k 

• 反向提取

DT[!.("a")] # not join # x y v a # 1: b 1 4 k # 2: b 3 5 k # 3: b 6 6 k # 4: c 1 7 k # 5: c 3 8 k # 6: c 6 9 k DT[!"a"] # same # x y v a # 1: b 1 4 k # 2: b 3 5 k # 3: b 6 6 k # 4: c 1 7 k # 5: c 3 8 k # 6: c 6 9 k DT[!2:4] # all rows other than 2:4 # x y v a # 1: a 1 1 k # 2: b 3 5 k # 3: b 6 6 k # 4: c 1 7 k # 5: c 3 8 k # 6: c 6 9 k 

多索引

• 正向提取

setkey(DT, x, y)
# Mehtod First DT["a"] # join to 1st column of key # x y v a # 1: a 1 1 k # 2: a 3 2 k # 3: a 6 3 k DT[.("a")] # same, .() is an alias for list() # x y v a # 1: a 1 1 k # 2: a 3 2 k # 3: a 6 3 k DT[.("a",3)] # join to 2 columns # x y v a # 1: a 3 2 k DT[.("a",3:6)] # join 4 rows (2 missing) # x y v a # 1: a 3 2 k # 2: a 4 NA NA # 3: a 5 NA NA # 4: a 6 3 k DT[.("a",3:6),nomatch=0] # remove missing # x y v a # 1: a 3 2 k # 2: a 6 3 k DT[.("a",3:6),roll=TRUE] # rolling join (locf) # x y v a # 1: a 3 2 k # 2: a 4 2 k # 3: a 5 2 k # 4: a 6 3 k ## Method Second DT[J('a')] # x y v a # 1: a 1 1 k # 2: a 3 2 k # 3: a 6 3 k DT[J("a",3)] # binary search (fast) # x y v a # 1: a 3 2 k DT[J("a",3:6)] # same; i.e. binary search (fast) # x y v a # 1: a 3 2 k # 2: a 4 NA NA # 3: a 5 NA NA # 4: a 6 3 k DT[J("a",3:6), nomatch = 0] # x y v a # 1: a 3 2 k # 2: a 6 3 k DT[J("a",3:6), roll = T] # x y v a # 1: a 3 2 k # 2: a 4 2 k # 3: a 5 2 k # 4: a 6 3 k ## Method Third DT[list("a")] # x y v a # 1: a 1 1 k # 2: a 3 2 k # 3: a 6 3 k DT[list("a",3)] # x y v a # 1: a 3 2 k DT[list("a", 3:6)] # x y v a # 1: a 3 2 k # 2: a 4 NA NA # 3: a 5 NA NA # 4: a 6 3 k DT[list("a", 3:6), nomatch = 0] # x y v a # 1: a 3 2 k # 2: a 6 3 k DT[list("a", 3:6), roll = T] # x y v a # 1: a 3 2 k # 2: a 4 2 k # 3: a 5 2 k # 4: a 6 3 k 

• 反向提取

DT[x!="b" | y!=3] # not yet optimized, currently vector scans # x y v a # 1: a 1 1 k # 2: a 3 2 k # 3: a 6 3 k # 4: b 1 4 k # 5: b 6 6 k # 6: c 1 7 k # 7: c 3 8 k # 8: c 6 9 k DT[!.("b",3)] # same result but much faster # x y v a # 1: a 1 1 k # 2: a 3 2 k # 3: a 6 3 k # 4: b 1 4 k # 5: b 6 6 k # 6: c 1 7 k # 7: c 3 8 k # 8: c 6 9 k 

分類彙總

分類匯老是指按某列的分類指標進行簡單操做，藉助由參數實現。此外，經過參數與索引相互沒有影響這裏。

單指標的分類彙總

• 默認彙總名稱

DT[,sum(v),by=x] # x V1 # 1: a 6 # 2: b 15 # 3: c 24 DT[,sum(v),by=y] # y V1 # 1: 1 12 # 2: 3 15 # 3: 6 18 

• 自定義彙總名稱

DT[,list(sum.v.x = sum(v)),by=x] # x sum.v.x # 1: a 6 # 2: b 15 # 3: c 24 DT[,list(sum.v.y = sum(v)),by=y] # y sum.v.y # 1: 1 12 # 2: 3 15 # 3: 6 18 DT[,sum.v.y := sum(v) ,by=y] # x y v a sum.v.y # 1: a 1 1 k 12 # 2: a 3 2 k 15 # 3: a 6 3 k 18 # 4: b 1 4 k 12 # 5: b 3 5 k 15 # 6: b 6 6 k 18 # 7: c 1 7 k 12 # 8: c 3 8 k 15 # 9: c 6 9 k 18 

• 彙總結果與原始數據進行匹配

DT[,sum.v.y := sum(v) ,by=y] # x y v a sum.v.y # 1: a 1 1 k 12 # 2: a 3 2 k 15 # 3: a 6 3 k 18 # 4: b 1 4 k 12 # 5: b 3 5 k 15 # 6: b 6 6 k 18 # 7: c 1 7 k 12 # 8: c 3 8 k 15 # 9: c 6 9 k 18 

多指標的多個分類彙總

• 默認彙總名稱

DT[,list(mean(v),sum(v)),by=list(x,y)] # keyed by # x y V1 V2 # 1: a 1 1 1 # 2: a 3 2 2 # 3: a 6 3 3 # 4: b 1 4 4 # 5: b 3 5 5 # 6: b 6 6 6 # 7: c 1 7 7 # 8: c 3 8 8 # 9: c 6 9 9 

• 自定義彙總名稱

DT[,list(mean.v = mean(v),sum.v = sum(v)),by=list(x,y)] # keyed by # x y mean.v sum.v #1: a 1 1 1 #2: a 3 2 2 #3: a 6 3 3 #4: b 1 4 4 #5: b 3 5 5 #6: b 6 6 6 #7: c 1 7 7 #8: c 3 8 8 #9: c 6 9 9 

• 彙總結果與原始數據進行匹配

DT[,c("mean.v", "sum.v.y") := list(mean(v),sum(v)) ,by=list(x,y)] # x y v a sum.v.y mean.v # 1: a 1 1 k 1 1 # 2: a 3 2 k 2 2 # 3: a 6 3 k 3 3 # 4: b 1 4 k 4 4 # 5: b 3 5 k 5 5 # 6: b 6 6 k 6 6 # 7: c 1 7 k 7 7 # 8: c 3 8 k 8 8 # 9: c 6 9 k 9 9 

data.table與data.frame的轉化

data.table格式加快了處理速度，而data.frame則更爲基礎。二者的轉化能夠經過data.table（），setDT（）和setDT（）來實現，其中data.table（）和setDT（）函數能夠將數據從data.frame轉化爲data.table，setDF（）函數能夠將數據從data.table轉化爲data.frame。注意使用data.table（），setDT（）和setDT（）時，參數自己的數據類型也會發生變化。

class(DT) # [1] "data.table" "data.frame" class(setDF(DT)) # [1] "data.frame" class(DT) # [1] "data.frame" 

此外，data.table包還能夠與基礎包中的重複的（），惟一的（），子（）函數結合使用。不只如此，data.table包還有一些基礎包的替代函數.rbind（）升級版的rbindlist（），能夠合併列數不一樣和列位置不一樣的數據。比dplyr包中安排（）函數更快的setorder（）排序函數。

來源於：http://xukuang.github.io/blog/2016/04/data-table-in-R/