使用变量时RevoScaleR rxDataStep行选择失败

Question

我试图用rxDataStep对xdf文件执行选择。我使用的是rowSelection，当我使用显式值时，它会工作，但当我使用变量时，它就不会起作用，例如：

tmp <- rxDataStep(alias.Xdf, transforms = list(TT_AMOUNT = DC_AMOUNT * RT_AMOUNT, UNIT_PRICE = RT_VALUE / TT_AMOUNT), varsToKeep = c('DC_AMOUNT', 'RT_AMOUNT', 'RT_VALUE'), 
            rowSelection = (A_ID == 1646041))

但这并不意味着：

x <- 1646041
tmp <- rxDataStep(alias.Xdf, transforms = list(TT_AMOUNT = DC_AMOUNT * RT_AMOUNT, UNIT_PRICE = RT_VALUE / TT_AMOUNT), varsToKeep = c('DC_AMOUNT', 'RT_AMOUNT', 'RT_VALUE'), 
             rowSelection = (A_ID == x))

它失败的消息如下：

ERROR: The sample data set for the analysis has no variables.
Caught exception in file: CxAnalysis.cpp, line: 3848. ThreadID: 31156 Rethrowing.
Caught exception in file: CxAnalysis.cpp, line: 5375. ThreadID: 31156 Rethrowing.

这里怎么了？我已经胡思乱想了好几个小时了，我在网上找到的每一个新税都试过了。谢谢。

Answer 1

我们可能需要在transformObjects上传递它

library(RevoScaleR)
rxDataStep(alias.Xdf, transforms = list(TT_AMOUNT = DC_AMOUNT * RT_AMOUNT, 
       UNIT_PRICE = RT_VALUE / TT_AMOUNT),
       varsToKeep = c('DC_AMOUNT', 'RT_AMOUNT', 'RT_VALUE'), 
         rowSelection = (A_ID == x1), transformObjects = list(x1=x))

使用可复制的示例

set.seed(100)
myData <- data.frame(x = 1:100, y = rep(c("a", "b", "c", "d"), 25),
                 z = rnorm(100), w = runif(100))

z1 <- 2


myDataSubset <- rxDataStep(inData = myData,
                       varsToKeep = c("x", "w", "z"),
                rowSelection = z > zNew,
                        transformObjects = list(zNew=z1))
#Rows Read: 100, Total Rows Processed: 100, Total Chunk Time: 0.007 seconds 
myDataSubset
#   x          w        z
#1 20 0.03609544 2.310297
#2 64 0.79408518 2.581959
#3 96 0.07123327 2.445683

这也可以用dplyr来完成。

library(dplyr)
myData %>%
      select(x, w, z) %>%
      filter(z > z1)
#   x          w        z
#1 20 0.03609544 2.310297
#2 64 0.79408518 2.581959
#3 96 0.07123327 2.445683