R 基于来自不同数据帧的分组值更新数据帧

我有一个dataframe，它的列需要根据查找表进行更新。查找表基于设置的分组。如果未找到匹配项，则要更新的值将保留为空

这是我的输入数据：

dput(DF_Generated)
structure(list(PO_ID = c("P1234", "P1234", "P1234", "P1234", 
"P1234", "P1234", "P1234", "P2345", "P2345", "P2345", "P3456", 
"P3456", NA, NA), SO_ID = c("S1", "S1", "S1", "S2", "S2", "S2", 
"S3", "S3", "S4", "S5", "S7", NA, "S10", "S10"), F_Year = c(2012, 
2012, 2012, 2013, 2013, 2013, 2013, 2011, 2011, 2012, 2014, 2014, 
2015, 2015), Product_ID = c("385X", "385X", "385X", "450X", "450X", 
"900X", "N9X", "3700", "3700", "3800", "A11U", "385X", "2700", 
"3700"), Revenue = c(16.6666666666667, 16.6666666666667, 16.6666666666667, 
35, 35, 35, 100, 100, -50, 20, 50, 20, 100, 10), Quantity = c(1, 
1, 1, 10, 10, 20, 20, 20, -10, 20, 20, 5, 40, 20), Location1 = c("MA", 
"NY", "WA", "NY", "WA", "NY", NA, "IL", "IL", NA, "MN", NA, "CA", 
NA), Name = c("N1", "N1", "N1", "N1", "N1", "N1", NA, "N2", "N2", 
NA, "N3", NA, "N4", NA)), class = c("tbl_df", "tbl", "data.frame"
), row.names = c(NA, -14L), .Names = c("PO_ID", "SO_ID", "F_Year", 
"Product_ID", "Revenue", "Quantity", "Location1", "Name"))

dput(DF_Lookup_2)
structure(list(PO_ID = c("P1234", "P1234", "P1234", "P1234", 
"P1234", "P2345", "P2345", "P3456", NA), SO_ID = c("S1", "S1", 
"S1", "S2", "S2", "S3", "S4", "S7", "S10"), F_Year = c(2012, 
2012, 2012, 2013, 2013, 2011, 2011, 2014, 2015), Location1 = c("MA", 
"NY", "WA", "NY", "WA", "IL", "IL", "MN", "CA"), Name = c("N1", 
"N1", "N1", "N1", "N1", "N2", "N2", "N3", "N4")), class = c("tbl_df", 
"tbl", "data.frame"), row.names = c(NA, -9L), .Names = c("PO_ID", 
"SO_ID", "F_Year", "Location1", "Name"))

dput(DFO)
structure(list(PO_ID = c("P1234", "P1234", "P1234", "P1234", 
"P1234", "P1234", "P1234", "P2345", "P2345", "P2345", "P3456", 
"P3456", NA, NA), SO_ID = c("S1", "S1", "S1", "S2", "S2", "S2", 
"S3", "S3", "S4", "S5", "S7", NA, "S10", "S10"), F_Year = c(2012, 
2012, 2012, 2013, 2013, 2013, 2013, 2011, 2011, 2012, 2014, 2014, 
2015, 2015), Product_ID = c("385X", "385X", "385X", "450X", "450X", 
"900X", "N9X", "3700", "3700", "3800", "A11U", "385X", "2700", 
"3700"), Revenue = c(16.6666666666667, 16.6666666666667, 16.6666666666667, 
35, 35, 35, 100, 100, -50, 20, 50, 20, 100, 10), Quantity = c(1, 
1, 1, 10, 10, 20, 20, 20, -10, 20, 20, 5, 40, 20), Location1 = c("MA", 
"NY", "WA", "NY", "WA", "NY", NA, "IL", "IL", NA, "MN", NA, "CA", 
"CA"), Name = c("N1", "N1", "N1", "N1", "N1", "N1", NA, "N2", 
"N2", NA, "N3", NA, "N4", "N4")), .Names = c("PO_ID", "SO_ID", 
"F_Year", "Product_ID", "Revenue", "Quantity", "Location1", "Name"
), row.names = c(NA, 14L), class = "data.frame")

这是我的查找表：

dput(DF_Generated)
structure(list(PO_ID = c("P1234", "P1234", "P1234", "P1234", 
"P1234", "P1234", "P1234", "P2345", "P2345", "P2345", "P3456", 
"P3456", NA, NA), SO_ID = c("S1", "S1", "S1", "S2", "S2", "S2", 
"S3", "S3", "S4", "S5", "S7", NA, "S10", "S10"), F_Year = c(2012, 
2012, 2012, 2013, 2013, 2013, 2013, 2011, 2011, 2012, 2014, 2014, 
2015, 2015), Product_ID = c("385X", "385X", "385X", "450X", "450X", 
"900X", "N9X", "3700", "3700", "3800", "A11U", "385X", "2700", 
"3700"), Revenue = c(16.6666666666667, 16.6666666666667, 16.6666666666667, 
35, 35, 35, 100, 100, -50, 20, 50, 20, 100, 10), Quantity = c(1, 
1, 1, 10, 10, 20, 20, 20, -10, 20, 20, 5, 40, 20), Location1 = c("MA", 
"NY", "WA", "NY", "WA", "NY", NA, "IL", "IL", NA, "MN", NA, "CA", 
NA), Name = c("N1", "N1", "N1", "N1", "N1", "N1", NA, "N2", "N2", 
NA, "N3", NA, "N4", NA)), class = c("tbl_df", "tbl", "data.frame"
), row.names = c(NA, -14L), .Names = c("PO_ID", "SO_ID", "F_Year", 
"Product_ID", "Revenue", "Quantity", "Location1", "Name"))

dput(DF_Lookup_2)
structure(list(PO_ID = c("P1234", "P1234", "P1234", "P1234", 
"P1234", "P2345", "P2345", "P3456", NA), SO_ID = c("S1", "S1", 
"S1", "S2", "S2", "S3", "S4", "S7", "S10"), F_Year = c(2012, 
2012, 2012, 2013, 2013, 2011, 2011, 2014, 2015), Location1 = c("MA", 
"NY", "WA", "NY", "WA", "IL", "IL", "MN", "CA"), Name = c("N1", 
"N1", "N1", "N1", "N1", "N2", "N2", "N3", "N4")), class = c("tbl_df", 
"tbl", "data.frame"), row.names = c(NA, -9L), .Names = c("PO_ID", 
"SO_ID", "F_Year", "Location1", "Name"))

dput(DFO)
structure(list(PO_ID = c("P1234", "P1234", "P1234", "P1234", 
"P1234", "P1234", "P1234", "P2345", "P2345", "P2345", "P3456", 
"P3456", NA, NA), SO_ID = c("S1", "S1", "S1", "S2", "S2", "S2", 
"S3", "S3", "S4", "S5", "S7", NA, "S10", "S10"), F_Year = c(2012, 
2012, 2012, 2013, 2013, 2013, 2013, 2011, 2011, 2012, 2014, 2014, 
2015, 2015), Product_ID = c("385X", "385X", "385X", "450X", "450X", 
"900X", "N9X", "3700", "3700", "3800", "A11U", "385X", "2700", 
"3700"), Revenue = c(16.6666666666667, 16.6666666666667, 16.6666666666667, 
35, 35, 35, 100, 100, -50, 20, 50, 20, 100, 10), Quantity = c(1, 
1, 1, 10, 10, 20, 20, 20, -10, 20, 20, 5, 40, 20), Location1 = c("MA", 
"NY", "WA", "NY", "WA", "NY", NA, "IL", "IL", NA, "MN", NA, "CA", 
"CA"), Name = c("N1", "N1", "N1", "N1", "N1", "N1", NA, "N2", 
"N2", NA, "N3", NA, "N4", "N4")), .Names = c("PO_ID", "SO_ID", 
"F_Year", "Product_ID", "Revenue", "Quantity", "Location1", "Name"
), row.names = c(NA, 14L), class = "data.frame")

预期输出为：

dput(DF_Generated)
structure(list(PO_ID = c("P1234", "P1234", "P1234", "P1234", 
"P1234", "P1234", "P1234", "P2345", "P2345", "P2345", "P3456", 
"P3456", NA, NA), SO_ID = c("S1", "S1", "S1", "S2", "S2", "S2", 
"S3", "S3", "S4", "S5", "S7", NA, "S10", "S10"), F_Year = c(2012, 
2012, 2012, 2013, 2013, 2013, 2013, 2011, 2011, 2012, 2014, 2014, 
2015, 2015), Product_ID = c("385X", "385X", "385X", "450X", "450X", 
"900X", "N9X", "3700", "3700", "3800", "A11U", "385X", "2700", 
"3700"), Revenue = c(16.6666666666667, 16.6666666666667, 16.6666666666667, 
35, 35, 35, 100, 100, -50, 20, 50, 20, 100, 10), Quantity = c(1, 
1, 1, 10, 10, 20, 20, 20, -10, 20, 20, 5, 40, 20), Location1 = c("MA", 
"NY", "WA", "NY", "WA", "NY", NA, "IL", "IL", NA, "MN", NA, "CA", 
NA), Name = c("N1", "N1", "N1", "N1", "N1", "N1", NA, "N2", "N2", 
NA, "N3", NA, "N4", NA)), class = c("tbl_df", "tbl", "data.frame"
), row.names = c(NA, -14L), .Names = c("PO_ID", "SO_ID", "F_Year", 
"Product_ID", "Revenue", "Quantity", "Location1", "Name"))

dput(DF_Lookup_2)
structure(list(PO_ID = c("P1234", "P1234", "P1234", "P1234", 
"P1234", "P2345", "P2345", "P3456", NA), SO_ID = c("S1", "S1", 
"S1", "S2", "S2", "S3", "S4", "S7", "S10"), F_Year = c(2012, 
2012, 2012, 2013, 2013, 2011, 2011, 2014, 2015), Location1 = c("MA", 
"NY", "WA", "NY", "WA", "IL", "IL", "MN", "CA"), Name = c("N1", 
"N1", "N1", "N1", "N1", "N2", "N2", "N3", "N4")), class = c("tbl_df", 
"tbl", "data.frame"), row.names = c(NA, -9L), .Names = c("PO_ID", 
"SO_ID", "F_Year", "Location1", "Name"))

dput(DFO)
structure(list(PO_ID = c("P1234", "P1234", "P1234", "P1234", 
"P1234", "P1234", "P1234", "P2345", "P2345", "P2345", "P3456", 
"P3456", NA, NA), SO_ID = c("S1", "S1", "S1", "S2", "S2", "S2", 
"S3", "S3", "S4", "S5", "S7", NA, "S10", "S10"), F_Year = c(2012, 
2012, 2012, 2013, 2013, 2013, 2013, 2011, 2011, 2012, 2014, 2014, 
2015, 2015), Product_ID = c("385X", "385X", "385X", "450X", "450X", 
"900X", "N9X", "3700", "3700", "3800", "A11U", "385X", "2700", 
"3700"), Revenue = c(16.6666666666667, 16.6666666666667, 16.6666666666667, 
35, 35, 35, 100, 100, -50, 20, 50, 20, 100, 10), Quantity = c(1, 
1, 1, 10, 10, 20, 20, 20, -10, 20, 20, 5, 40, 20), Location1 = c("MA", 
"NY", "WA", "NY", "WA", "NY", NA, "IL", "IL", NA, "MN", NA, "CA", 
"CA"), Name = c("N1", "N1", "N1", "N1", "N1", "N1", NA, "N2", 
"N2", NA, "N3", NA, "N4", "N4")), .Names = c("PO_ID", "SO_ID", 
"F_Year", "Product_ID", "Revenue", "Quantity", "Location1", "Name"
), row.names = c(NA, 14L), class = "data.frame")

逻辑：

查找基于三列：

PO\u ID

，

SO\u ID

，

F\u Year

。如果找到匹配项，则仅当数据帧需要更新时才会覆盖条目。例如，列

Location1

和

Name

的行不应被覆盖，因为它们的值已经存在于查找表中。但是，

Location1

和

Name

列中的行条目，例如

PO_ID

=

NA

，

SO_ID

=

S10

和

F_Year

=

2015

需要使用查找表中的有效值进行更新，分别为

CA

和

N4

。我尝试使用

数据。表

，但我下面的代码覆盖了所有条目，这是不正确的

我读了线程，并试图这样做，但代码覆盖了我不需要查找的现有条目

这是我的密码：

data.table::setDT(DF_Generated)
data.table::setDT(DF_Lookup_2)
data.table::setkey(DF_Generated,PO_ID,SO_ID,F_Year)
data.table::setkey(DF_Lookup_2,PO_ID,SO_ID,F_Year)

DF_Generated[DF_Lookup_2,on=c("PO_ID","SO_ID","F_Year"),c("Location1","Name"):=list(i.Location1,i.Name)]

我有两个问题：

问题：1）我使用的是

数据表

，因为我的实际数据很大。因此，我正在寻找基于

data.table

的解决方案。如何修复我的

数据表
代码

问题2）如果建议这样做，我也愿意接受
dplyr

然而，对于我的学习，如果你能在这两个解决方案上帮助我，我将不胜感激。我是一名初学者，仍在学习这两个软件包。
我们可以在加入
上的
时创建两个新列，分别为'PO_ID'、'SO_ID'和'F_Year'，然后使用
set

setDT(DF_Generated)[setDT(DF_Lookup_2), c("Location1N", "NameN") := list(i.Location1, 
                               i.Name),on = .(PO_ID, SO_ID, F_Year)]
nm1 <- c("Location1", "Name")
nm2 <- paste0(nm1, "N")
for(j in seq_along(nm1)){
 set(DF_Generated, i = which(is.na(DF_Generated[[nm1[j]]])), j=nm1[j], 
                     value = DF_Generated[[nm2[j]]][is.na(df1[[nm1[j]]])])
}
DF_Generated[, (nm2) := NULL][]

identical(setDT(DFO), DF_Generated)
#[1] TRUE

setDT（DF_生成）[setDT（DF_查找_2），c（“位置1n”，“名称”）：=list（i.Location1，
i、 名称），on=（采购订单ID，销售订单ID，F年）]
nm1这样做的另一个方法是（根据条件）连接到
DF_Lookup_2
，然后分配回
DF_Generated
。这样做的另一个原因是，在和
X[Y]
join中，生成的join的长度为
Y
，因此，类型为
DF\u Lookup\u 2[DF\u Generated]
的join将为我们提供所需的长度。然后，我们可以按原样将其放回
DF_生成的

DF_Generated[is.na(Location1) | is.na(Name), c("Location1", "Name"):= 
               DF_Lookup_2[DF_Generated[is.na(Location1) | is.na(Name)], 
                           .(Location1, Name), on = .(PO_ID, SO_ID, F_Year)]]


identical(DF_Generated, setDT(DFO))
## [1] TRUE

@阿克伦-谢谢你的帮助。如果运行上述代码，您将看到
Location1
将全部
WA
（与原始数据集
DF_生成的
不同。请在
Location1
和
Name
列中查看预期输出。这有帮助吗？可能我没有听你的问题。非常感谢你的帮助。如果你不介意的话，我有两个后续问题：a）你认为
dplyr
会更简单吗你认为这段代码的性能会好吗？我这样问是因为在我的原始数据集中，我有大约100万行，28列。我还没有在原始数据集上运行您的代码，因为定制上述代码需要时间。我是一个初学者，没有太多的线索。因此，我很感激您的想法。@watchtower在
tidyverse
中有一个
coalesce
函数，但是如果有重复的，那么
左连接将不起作用，谢谢。非常恭敬地说，我不知道为什么会出错<代码>“data.table:：set中的错误（生成的DF_，i=is.na（生成的DF_[[nm1[j]]）），：分配给现有列“Location1”的RHS的长度为零，但不为NULL。如果要删除该列，请使用NULL。否则，RHS的长度必须大于0；例如，NA_integer_。如果试图将列类型更改为空列表列，则与所有列类型更改一样，提供完整长度的RHS向量，如vector（'list'，nrow（DT））；即新专栏中的'plonk'。
知道为什么会发生这种情况吗？@watchtower这是基于您文章中的示例还是原始数据集。这是由于软件包版本吗？我使用的是data.table_1.10。0@watchtower如果在新版本中出现问题，请继续使用Davidernburg的代码