setkey（）是否追溯工作？ （在R中）

Question

考虑数据帧df1，它被设置为数据表。请注意，未设置密钥。

library(data.table)

df1 <- data.frame(v1 = letters[seq( from = 1, to = 4 )])
setDT(df1)
tables()  #  verify that df1 is now data table. No key.
     NAME NROW NCOL MB COLS KEY
[1,] df1     4    1  1 v1      
Total: 1MB

df2 <- df1  #  create df2 as copy of df1

tables()
     NAME NROW NCOL MB COLS KEY
[1,] df1     4    1  1 v1      
[2,] df2     4    1  1 v1      #  two data tables. No keys.
Total: 2MB

setkey(df2, v1)  #  set key to df2 (only)

tables()
     NAME NROW NCOL MB COLS KEY
[1,] df1     4    1  1 v1   v1  #  both df1 and df2 are assigned keys 
[2,] df2     4    1  1 v1   v1  #  it is as if df1 is assigned key via df2 ('retrospectively')
Total: 2MB

但是，如果我在设置密钥之前对df2进行了更改，比如添加一个列，则setkey（）的行为会有所不同：不会是＆＃34;回顾＆＃34;。 （见下面的代码）。 我想知道为什么会这样。感谢

library(data.table)

df1 <- data.frame(v1 = letters[seq( from = 1, to = 4 )])
setDT(df1)
tables()
     NAME NROW NCOL MB COLS KEY
[1,] df1     4    1  1 v1      
Total: 1MB
df2 <- df1
tables()
     NAME NROW NCOL MB COLS KEY
[1,] df1     4    1  1 v1      
[2,] df2     4    1  1 v1      
Total: 2MB

df2$v2 <- c(1:4) # add column, therefore v2 != v1

setkey(df2, v1)

tables()
     NAME NROW NCOL MB COLS  KEY
[1,] df1     4    1  1 v1         #  in this instance v1 wasn't set a key via df2
[2,] df2     4    2  1 v1,v2 v1 
Total: 2MB