Editar

Compartir por

rxDataStep: Data Step for RevoScaleR data sources

  • Artigo

Description

Transform data from an input data set to an output data set. The rxDataStep function is multi-threaded.

Usage

  rxDataStep(inData = NULL, outFile = NULL, varsToKeep = NULL, varsToDrop = NULL,
                rowSelection = NULL, transforms = NULL, transformObjects = NULL,
                transformFunc = NULL, transformVars = NULL, 
                transformPackages = NULL, transformEnvir = NULL, 
                append = "none", overwrite = FALSE, rowVarName = NULL, 
                removeMissingsOnRead = FALSE, removeMissings = FALSE, 
                computeLowHigh = TRUE, maxRowsByCols = 3000000,
                rowsPerRead = -1, startRow = 1, numRows = -1, 
                returnTransformObjects = FALSE,
                blocksPerRead = rxGetOption("blocksPerRead"),
                reportProgress = rxGetOption("reportProgress"),
                xdfCompressionLevel = rxGetOption("xdfCompressionLevel"), ...)

Arguments

inData

A data source object of these types: RxTextData, RxXdfData, RxSasData, RxSpssData, RxOdbcData, RxSqlServerData, RxTeradata. If not using a distributed compute context such as RxHadoopMR, a data frame, a character string specifying the input .xdf file, or NULL can also be used. If NULL, a data set will be created automatically with a single variable, .rxRowNums, containing row numbers. It will have a total of numRows rows with rowsPerRead rows in each block.

outFile

A character string specifying the output .xdf file or any of these data sources: RxTextData, RxXdfData, RxSasData, RxSpssData, RxOdbcData, RxSqlServerData, RxTeradata, RxHiveData, RxParquetData, RxOrcData. If NULL, a data frame will be returned from rxDataStep unless returnTransformObjects is set to TRUE. Setting outFile to NULL and returnTransformObjects=TRUE allows chunkwise computations on the data without modifying the existing data or creating a new data set. outFile can also be a delimited RxTextData data source if using a native file system and not appending.

varsToKeep

A character vector of variable names to include when reading from the input data file. If NULL, argument is ignored. Cannot be used with varsToDrop or when outFile is the same as the input data file. Variables used in transformations or row selection will be retained even if not specified in varsToKeep. If newName is used in colInfo in a non-xdf data source, the newName should be used in varsToKeep. Not supported for RxTeradata, RxOdbcData, or RxSqlServerData data sources.

varsToDrop

A character vector of variable names to exclude when reading from the input data file. If NULL, argument is ignored. Cannot be used with varsToKeep or when outFile is the same as the input data file. Variables used in transformations or row selection will be retained even if specified in varsToDrop. If newName is used in colInfo in a non-xdf data source, the newName should be used in varsToDrop. Not supported for RxTeradata, RxOdbcData, or RxSqlServerData data sources.

rowSelection

The name of a logical variable in the data set or a logical expression using variables in the data set to specify row selection. For example, rowSelection = old will use only observations in which the value of the variable old is TRUE. rowSelection = (age > 20) & (age < 65) & (log(income) > 10) will use only observations in which the value of the age variable is between 20 and 65 and the value of the log of the income variable is greater than 10. The row selection is performed after processing any data transformations (see the arguments transforms or transformFunc). As with all expressions, rowSelection can be defined outside of the function call using the expression function.

transforms

An expression of the form list(name = expression, ...) representing the first round of variable transformations. As with all expressions, transforms (or rowSelection) can be defined outside of the function call using the expression function. When using function call in the expression, transformObject should be used to pass the function name to remote context. In addition, calling and enclosing environment of the function are very important if the function uses undefined variable.

transformObjects

A named list containing objects that can be referenced by transforms, transformsFunc, and rowSelection.

transformFunc

A variable transformation function. The recommended way to do variable transformation. See rxTransform for details.

transformVars

A character vector of input data set variables needed for the transformation function. See rxTransform for details.

transformPackages

A character vector defining additional R packages (outside of those specified in rxGetOption("transformPackages")) to be made available and preloaded for use in variable transformation functions, e.g., those explicitly defined in RevoScaleR functions via their transforms and transformFunc arguments or those defined implicitly via their formula or rowSelection arguments. The transformPackages argument may also be NULL, indicating that no packages outside rxGetOption("transformPackages") will be preloaded.

transformEnvir

A user-defined environment to serve as a parent to all environments developed internally and used for variable data transformation. If transformEnvir = NULL, a new "hash" environment with parent baseenv() is used instead.

append

Either "none" to create a new files, "rows" to append rows to an existing file, or "cols" to append columns to an existing file. If outFile exists and append is "none", the overwrite argument must be set to TRUE. Ignored for data frames. You cannot append to RxTextData or append columns to RxTeradata data sources, and appending is not supported for composite .xdf files or when using the RxHadoopMR compute context.

overwrite

A logical value. If TRUE, an existing outFile will be overwritten, or if appending columns existing columns with the same name will be overwritten. overwrite is ignored if appending rows. Ignored for data frames.

rowVarName

A character string or NULL. If inData is a data.frame: If NULL, the data frame's row names will be dropped. If a character string, an additional variable of that name will be added to the data set containing the data frame's row names. If a data.frame is being returned, a variable with the name rowVarName will be removed as a column from the data frame and will be used as the row names.

removeMissingsOnRead

A logical value. If TRUE, rows with missing values will be removed on read.

removeMissings

A logical value. If TRUE, rows with missing values will not be included in the output data.

computeLowHigh

A logical value. If FALSE, low and high values will not automatically be computed. This should only be set to FALSE in special circumstances, such as when append is being used repeatedly. Ignored for data frames.

maxRowsByCols

The maximum size of a data frame that will be returned if outFile is set to NULL and inData is an .xdf file , measured by the number of rows times the number of columns. If the number of rows times the number of columns being created from the .xdf file exceeds this, a warning will be reported and the number of rows in the returned data frame will be truncated. If maxRowsByCols is set to be too large, you may experience problems from loading a huge data frame into memory.

rowsPerRead

The number of rows to read for each chunk of data read from the input data source. Use this argument for finer control of the number of rows per block in the output data source. If greater than 0, blocksPerRead is ignored. Cannot be used if inFile is the same as outFile. The default value of -1 specifies that data should be read by blocks according to the blocksPerRead argument.

startRow

The starting row to read from the input data source. Cannot be used if inFile is the same as outFile.

numRows

The number of rows to read from the input data source. If rowSelection or removeMissings are used, the output data set may have fewer rows than specified by numRows. Cannot be used if inFile is the same as outFile.

returnTransformObjects

A logical value. If TRUE, the list of transformObjects will be returned instead of a data frame or data source object. If the input transformObjects have been modified, by using .rxSet or .rxModify in the transformFunc, the updated values will be returned. Any data returned from the transformFunc is ignored. If no transformObjects are used, NULL is returned. This argument allows for user-defined computations within a transformFunc without creating new data. returnTransformObjects is not supported in distributed compute contexts such as RxHadoopMR.

blocksPerRead

The number of blocks to read for each chunk of data read from the data source. Ignored for data frames or if rowsPerRead is positive.

reportProgress

An integer value with options:

  • 0: no progress is reported.
  • 1: the number of processed rows is printed and updated.
  • 2: rows processed and timings are reported.
  • 3: rows processed and all timings are reported.

xdfCompressionLevel

An integer in the range of -1 to 9. The higher the value, the greater the amount of compression - resulting in smaller files but a longer time to create them. If xdfCompressionLevel is set to 0, there will be no compression and files will be compatible with the 6.0 release of Revolution R Enterprise. If set to -1, a default level of compression will be used.

...

Additional arguments to be passed to the input data source. If stringsAsFactorsis specified and the input data source is RxXdfData, strings with be converted to factors when returning a data frame to R.

Value

For rxDataStep, if returnTransformObjects is FALSE)and an outFile is specified, an RxXdfData data source is returned invisibly. If no outFile is specified, a data frame is returned invisibly. Either can be used in subsequent RevoScaleR analysis.
If returnTransformObjects is TRUE, the transformObjects list as modified by the transformation function is returned invisibly. When working with an RxInSqlServer compute context, both the input and output data sources must be RxSqlServerData.

Author(s)

Microsoft Corporation Microsoft Technical Support

See Also

rxImport, RxXdfData, RxTextData, RxSqlServerData, RxTeradata, rxGetInfo, rxGetVarInfo, rxTransform

Examples



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

 # Get a subset rows and columns from the data frame    
 myDataSubset <- rxDataStep(inData = myData,
     varsToKeep = c("x", "w", "z"), rowSelection = z > 0) 
 rxGetInfo(myDataSubset, getVarInfo = TRUE) 

 # Create a simple .xdf file from the data frame (for use below)
 inputFile <- file.path(tempdir(), "testInput.xdf")
 rxDataStep(inData = myData, outFile = inputFile, overwrite = TRUE)

 # Redo, creating a multi-block .xdf file from the data frame
 rxDataStep(inData = myData, outFile = inputFile, rowsPerRead = 50, 
     overwrite = TRUE)
 rxGetInfo( inputFile )

 # Subset rows and columns, creating a new .xdf file
 outputFile <- file.path(tempdir(), "testOutput.xdf")
 rxDataStep(inData = inputFile, outFile = outputFile,
     varsToKeep = c("x", "w", "z"), rowSelection = z > 0,
     overwrite = TRUE)
 rxGetInfo(outputFile)

 # Use transforms list with data frame input and output data
 # Add new columns
 myNewData <- rxDataStep(inData = myData, 
     transforms = list(a = w > 0, b = 100 * z))
 names(myNewData)

 # Use transformFunc to add new columns
 myXformFunc <- function(dataList) {
   dataList$b <- 100 * dataList$z
   return (dataList)
 }
 myNewData <- rxDataStep(inData = myData,
     transformFunc = myXformFunc)
 names(myNewData)

 # Use transforms to remove columns
 rxDataStep(inData = inputFile, outFile = outputFile,
     transforms = list(w = NULL), overwrite = TRUE)
 rxGetInfo(outputFile)

 # use transformFunc to remove columns
 xform <- function(dataList) {
   dataList$w <- NULL
   return (dataList)
 }
 rxDataStep(inData = inputFile, outFile = outputFile,
     transformFunc = xform, overwrite = TRUE)
 rxGetInfo(outputFile)

 # use transform to change data type
 rxDataStep(inData = inputFile, outFile = outputFile, transformVars = c("x", "y"),
     transforms = list(x = as.numeric(x), y = as.numeric(y)), overwrite = TRUE)
 rxGetVarInfo(outputFile)

 # use transformFunc to change data type
 myXform <- function(dataList) {
   dataList <- lapply(dataList, as.numeric)
   return (dataList)
 }
 rxDataStep(inData = inputFile, outFile = outputFile,
     transformFunc = myXform, overwrite = TRUE)
 rxGetVarInfo(outputFile)

 # use transform to create new data
 rxDataStep(inData = inputFile, outFile = outputFile,
     transforms = list(maxZ = max(z), minW = min(w), z = NULL, w = NULL),
     varsToDrop = c("x", "y"), overwrite = TRUE)
 rxGetVarInfo(outputFile)

 # use transformFunc to create new data
 xform <- function(dataList){
   outList <- list()
   outList$maxZ <- max(dataList$z)
   outList$minW <- min(dataList$w)
   return (outList)
 }
 rxDataStep(inData = inputFile, outFile = outputFile,
     transformFunc = xform, overwrite = TRUE)
 rxGetVarInfo(outputFile)

 # add new column by calling function in expression for "transform"
 # "transform" validate the expression at server, so function name should be passed to 
 # remote context. R looking up undefined variable in calling environment, dynamic scoping 
 # should be used to find "const" used in function "myTransform"
 const <- 10
 myTransform <- function(x){
   const <- get("const", parent.frame())
   x * const
 }
 rxDataStep(inData = inputFile, outFile = outputFile,
            transforms = list(x10 = myTransform(x)), transformObjects = list(myTransform = myTransform, const = const),
            overwrite = TRUE)
 rxGetVarInfo(outputFile)

 # using "transform" and "transformEnvir" to add new columns
 env <- new.env()
 env$constant <- 10
 env$myTransform <- function(x){
   x * constant
 }
 environment(env$myTransform) <- env
 data <- rxDataStep(inData=inputFile, outFile = outputFile,
     transforms = list(b = myTransform(x)), transformEnvir = env, overwrite = TRUE)
 rxGetVarInfo(outputFile)

 # Specify which variables to keep in a new data file
 varsToKeep <- c("x", "w", "z")
 rxDataStep(inData = inputFile, outFile = outputFile, varsToKeep = varsToKeep,
     transforms = list(a = w > 0, b = 100 * z), overwrite = TRUE)
 rxGetInfo(outputFile)

 # Alternatively specify which variables to drop
 varsToDrop <- "y"
 rxDataStep(inData = inputFile, outFile = outputFile, varsToDrop = varsToDrop, 
     transforms = list(a = w > 0, b = 100 * z), overwrite = TRUE)
 rxGetInfo(outputFile)

 # Read a specific number of rows of data into a data frame
 myData <- rxDataStep(inData = inputFile, startRow = 5, numRows = 10)
 rxGetInfo(myData)

 # Create a selection variable to take a roughly 25% random sample
 # of each block of data read (in this case 1).
 rxDataStep(inData = inputFile, outFile = outputFile, rowSelection = selVar,
     transforms = list(
         selVar = as.logical(rbinom(.rxNumRows, 1, .25))
     ), overwrite = TRUE)
 rxGetInfo(outputFile)

 # Create a new variable containing row numbers for a multi-block data file
 # Note that .rxStartRow and .rxNumRows are not supported in 
 # a distributed compute context such as RxHadoopMR
 myDataSource <- rxDataStep(inData = inputFile, outFile = outputFile,  
   transforms = list(
             rowNum = .rxStartRow : (.rxStartRow + .rxNumRows - 1)),
   overwrite = TRUE )

 # Use the returned data source object in another call to rxDataStep
 myMiddleData <- rxDataStep(inData = myDataSource, startRow = 55,
   numRows = 5)
 myMiddleData  

 # Create a new factor variable from a continuous variable using the cut function
 rxDataStep(inData = inputFile, outFile = outputFile,
   transforms = list( 
     wFactor = cut(w, breaks = c(0, .33, .66, 1), labels = c("Low", "Med", "High"))
     ), overwrite=TRUE)
 rxGetInfo( outputFile, numRows = 10 )

 # Create a new data set with simulated data.  A row number variable will
 # be automatically created.  It will have 20 rows in each block,
 # with a total of 100 rows
 #(This functionality not supported in distributed compute contexts.)
 outFile <- tempfile(pattern = ".rxTest1", fileext = ".xdf")
 newDS <- rxDataStep( outFile = outFile, numRows = 100, rowsPerRead = 20,
     transforms = list(
         x = (.rxRowNums - 50)/25,
         pnormx = pnorm(x),
         dnormx = dnorm(x))) 
 # Compute summary statistics on the new data file        
 rxSummary(~., newDS)
 file.remove(outFile)

 # Use the data step to chunk through the data and compute
 # the sum of a squared deviation
 inFile <- file.path(rxGetOption("unitTestDataDir"), "test100r5b.xdf")
 myFun <- function( dataList )
 {
     chunkSumDev <- sum((dataList$y - toMyCenter)^2,  na.rm = TRUE)
     toTotalSumDev <<- toTotalSumDev + chunkSumDev
     return(NULL)
 }
 myCenter <- 40
 newTransformVals <- rxDataStep(inData = inFile, 
     transformFunc = myFun,
     transformObjects = list( 
         toMyCenter = myCenter,
         toTotalSumDev = 0),
     transformVars = "y", returnTransformObjects = TRUE)

 newTransformVals[["toTotalSumDev"]]

 ## Not run:

# These examples need to be modified to substitute appropriate paths.
# Convert an xdf file to csv on HDFS
myData <- data.frame(textVar = c("a", "b", "c", "d"), intVar = as.integer(c(1:2, NA, 4)))
# write data frame to an xdf file in local file system
# test1.xdf will be written out to local file system in the current working directory
xdfFile <- "test1.xdf"
xdfDS <- RxXdfData(xdfFile)
rxDataStep(inData = myData, outFile = xdfDS, overwrite=TRUE)
# use RxTextData to write a csv file to HDFS
# /user/RevoShare/myuser is HDFS path that must exist for testCsv.csv to be written successfully.
csvFile <- "/user/RevoShare/myuser/testCsv.csv"
# We are writing a csv with no header (column names), no quotes and empty string for missing values (NA)
hdfsFS <- RxHdfsFileSystem()
myTextDS <- RxTextData(csvFile, missingValueString = "NA", firstRowIsColNames = FALSE, quoteMark = "", fileSystem=hdfsFS)
# This will write out a csv file to HDFS
rxDataStep(inData = xdfDS, outFile = myTextDS, overwrite = TRUE)

# Convert a composite xdf file to csv writing to HDFS using RxTextData.
# Create a test xdf file.
myData <- data.frame(textVar = c("a", "b", "c", "d"), intVar = as.integer(c(1:2, NA, 4)))
hdfsFS <- RxHdfsFileSystem()
# composite xdf file path in HDFS. Path "/user/RevoShare/myuser/test1" must exist.
xdfFile <- "/user/RevoShare/myuser/test1"
xdfDS <- RxXdfData(xdfFile, fileSystem = hdfsFS)
rxDataStep(inData = myData, outFile = xdfDS, overwrite=TRUE)
# use RxTextData to write a csv file to HDFS
# /user/RevoShare/myuser is HDFS path that must exist for testCsv.csv to be written successfully.
csvFile <- "/user/RevoShare/myuser/testCsv.csv"
# We are writing a csv with no header, no quotes and empty string for missing values (NA)
myTextDS <- RxTextData(csvFile, missingValueString = "", firstRowIsColNames = FALSE, quoteMark = "", fileSystem = hdfsFS)
# This will write out a csv file to HDFS
rxDataStep(inData = xdfDS, outFile = myTextDS, overwrite = TRUE)
## End(Not run)