如何在非等连接的结果中确定订单？

Question

我试图理解如何在-variable的每个级别内data.table排序非equi连接的结果的基本逻辑.on

只是从头开始说清楚:我对订单本身没有任何问题,或者在连接后以所需的方式订购输出.但是,因为我发现所有其他data.table操作的输出高度一致,我怀疑在非equi连接中也有一个排序模式.

我将举两个例子,其中两个不同的"大"数据集以较小的方式连接.我试图描述在输出中最明显的图案内各加入,以及当所不同图案之间的两个数据集的联接.

library(data.table)
# the first 'large' data set
d1 <- data.table(x = c(rep(c("b", "a", "c"), each = 3), c("a", "b")),
                 y = c(rep(c(1, 3, 6), 3), 6, 6),
                 id = 1:11) # to make it easier to track the original order in the output    
#     x y  id
# 1:  b 1   1
# 2:  b 3   2
# 3:  b 6   3
# 4:  a 1   4
# 5:  a 3   5
# 6:  a 6   6
# 7:  c 1   7
# 8:  c 3   8
# 9:  c 6   9
# 10: a 6  10
# 11: b 6  11

# the small data set
d2 <- data.table(id = 1:2, val = c(4, 2))   
#     id val
# 1:   1   4
# 2:   2   2

第一个大数据集和小数据集之间的非等连接on = .(y >= val).

d1[d2, on = .(y >= val)]
#     x y  id  i.id
# 1:  b 4   3     1 # Row 1-5, first match: y >= val[1]; y >= 4
# 2:  a 4   6     1 # The rows within this match have the same order as the original data
# 3:  c 4   9     1 # and runs consecutively from first to last match
# 4:  a 4  10     1
# 5:  b 4  11     1

# 6:  b 2   2     2 # Row 6-13, second match: y >= val[2]; y >= 2 
# 7:  a 2   5     2 # The rows within this match do not have the same order as the original data
# 8:  c 2   8     2 # Rather, they seem to be come in chunks (6-8, 9-11, 12-13) 
                    # First chunk starts with the match with lowest index, y[2] 
# 9:  b 2   3     2  
# 10: a 2   6     2 
# 11: c 2   9     2 

# 12: a 2  10     2
# 13: b 2  11     2

---

第二个"大"数据集:

d3 <- data.table(x = rep(c("a", "b", "c"), each = 3),
                 y = c(6, 1, 3),
                 id = 1:9)
#    x y id
# 1: a 6  1
# 2: a 1  2
# 3: a 3  3
# 4: b 6  4
# 5: b 1  5
# 6: b 3  6
# 7: c 6  7
# 8: c 1  8
# 9: c 3  9

第二个大数据集与小的相同的非equi连接:

d3[d2, on = .(y >= val)]

#    x y   id i.id
# 1: a 4   1     1 # Row 1-3, first match (y >= 4), similar to output above
# 2: b 4   4     1
# 3: c 4   7     1

# 4: a 2   3     2 # Row 4-9, second match (y >= 2).  
# 5: b 2   6     2 # Again, rows not consecutive.
# 6: c 2   9     2 # However, now the first chunk does not start with the match with lowest index,
                   # y[3] instead of y[1]

# 7: a 2   1     2 # y[1] appears after y[3]
# 8: b 2   4     2 # ditto
# 9: c 2   7     2

---

任何人都可以解释(1)-variable的每个级别内的顺序的逻辑on,特别是在第二个匹配中,其中数据的原始顺序不保留在结果中.和(2)为何顺序之间的块内的匹配时使用有所不同的两个不同的数据集？

Answer 1

感谢您抓住这个并在此报告SO,并在Github上归档.现在应该在当前的开发版本(编写本文时为v1.10.5)中修复此问题.

它应该很快就可以在CRAN上作为v1.10.6使用.

来自新闻条目:

22. 在# i1991中报告的某些情况下,非equi连接中返回的行顺序不正确.这已经修复了.感谢@ Henrik-P的报道.