data.frame 子集化：使用数字日期差异的非常奇怪的行为

Question

在用日期子集数据框时，有一种我不明白的行为。这是示例（末尾的数据）：

> df
   positif1 positif2      date1      date2
1         0        0 2020-05-02 2020-04-30
2         0        0 2020-05-02 2020-04-21
3         0        0 2020-05-02 2020-04-30
.
.

我天真地想把这个子集化：

df[df$positif2 == 0 & df$positif1 == 0 & as.numeric(df$date1 - df$date2) <=2,]

它没有 return 任何东西，尽管显然有满足条件的值：

as.numeric(df[df$positif2 == 0 & df$positif1 == 0,"date1"]-df[df$positif2 == 0 & df$positif1 == 0,"date2"])
 [1]  2 11  2 29 12 18  1 22  5 24  5  6  4 25  9  9 13 16 17 35  5 35 22 51  3 17  8 16 12 15 14 21 14
[34]  4

我意识到正确的方法是：

df[df$positif2 == 0 & df$positif1 == 0 & difftime(df$date1, df$date2,units = "day") <=2,]

而且我的问题是在对数据框进行子集化时单位会发生变化：

> df$date1 - df$date2
Time differences in secs
 [1]  172800  950400  172800 2505600 1036800 1555200 2073600   86400 1900800  432000 2073600  432000
[13]  518400  345600 2160000  777600  777600       0 1123200 1382400 1468800 3024000       0  432000
[25] 3024000 1900800 4406400       0  259200 1468800  691200 1382400 1036800 1296000 1209600 1814400
[37] 1209600  345600


> df[df$positif2 == 0,"date1"] - df[df$positif2 == 0,"date2"]
Time differences in days
 [1]  2 11  2 29 12 18 24  1 22  5 24  5  6  4 25  9  9 13 16 17 35  5 35 22 51  3 17  8 16 12 15 14 21
[34] 14  4

这对我来说毫无意义。我做错了什么吗？这种行为有原因吗？

数据：

df <- structure(list(positif1 = c(0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 
0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 
0, 0, 0, 0, 0, 0), positif2 = c(0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 1L, 0L, 0L, 0L, 0L, 1L, 0L, 
0L, 0L, 0L, 1L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), date1 = structure(c(1588377600, 
1588377600, 1588377600, 1588377600, 1588377600, 1588377600, 1588377600, 
1588377600, 1588377600, 1588377600, 1588377600, 1588377600, 1588377600, 
1588377600, 1588377600, 1588377600, 1588377600, 1584057600, 1588377600, 
1588377600, 1588377600, 1588377600, 1586563200, 1588377600, 1588377600, 
1588377600, 1588377600, 1586908800, 1588377600, 1588377600, 1588377600, 
1588377600, 1588377600, 1588377600, 1588377600, 1588377600, 1588377600, 
1588377600), class = c("POSIXct", "POSIXt"), tzone = "UTC"), 
    date2 = structure(c(1588204800, 1587427200, 1588204800, 1585872000, 
    1587340800, 1586822400, 1586304000, 1588291200, 1586476800, 
    1587945600, 1586304000, 1587945600, 1587859200, 1588032000, 
    1586217600, 1587600000, 1587600000, 1584057600, 1587254400, 
    1586995200, 1586908800, 1585353600, 1586563200, 1587945600, 
    1585353600, 1586476800, 1583971200, 1586908800, 1588118400, 
    1586908800, 1587686400, 1586995200, 1587340800, 1587081600, 
    1587168000, 1586563200, 1587168000, 1588032000), class = c("POSIXct", 
    "POSIXt"), tzone = "UTC")), row.names = c(NA, -38L), class = "data.frame", index = structure(integer(0), "`__positif1__positif2`" = c(1L, 
2L, 3L, 4L, 5L, 6L, 8L, 9L, 10L, 11L, 12L, 13L, 14L, 15L, 16L, 
17L, 19L, 20L, 21L, 22L, 24L, 25L, 26L, 27L, 29L, 30L, 31L, 32L, 
33L, 34L, 35L, 36L, 37L, 38L, 7L, 18L, 23L, 28L)))

Answer 1

来自?difftime

Subtraction of date-time objects gives an object of this class, by calling difftime with units = "auto".

所以我们知道在减去日期时默认 units = "auto"。

其次，

If units = "auto", a suitable set of units is chosen, the largest possible (excluding "weeks") in which all the absolute differences are greater than one.

所以当 units = "auto" 它试图 select 一个最大的单位。所以当我们这样做时

df$date1 - df$date2

在date1和date2中有一些相同的条目使它们的差异为0，所以这里选择"seconds"作为单位。

但是当你在 positif2 (positif2 == 0) 中仅对 0 个条目的日期进行子集化时，最小差异是日期，因此单位 selected 是 "days" 这里。

df$date1[df$positif2 == 0] - df$date2[df$positif2 == 0]

您已经确定的正确方法是使用 difftime 并在其中明确指定 units 参数。

data.frame 子集化：使用数字日期差异的非常奇怪的行为

data.frame subsetting: very strange behavior using numeric date difference

datetime

r

subset

dataframe

posixct