rxOpen-methods: Managing RevoScaleR Data Source Objects

Artigo
07/15/2019

Description

These functions manage RevoScaleR data source objects.

Usage

  rxOpen(src, mode = "r")
  rxClose(src, mode = "r")
  rxIsOpen(src, mode = "r")
  rxReadNext(src)
  rxWriteNext(from, to, ...)

Arguments

`from`

data frame object.

`src`

RxDataSource object.

`to`

RxDataSource object.

`mode`

character string specifying the mode (r or w) to open the file.

`...`

any other arguments to be passed on.

Value

For rxOpen and rxClose, a logical indicating whether the operation was successful. For rxIsOpen, a logical indicating whether or not the RxDataSource is open for the specified mode. For rxReadNext, either a data frame or a list depending upon the value of the returnDataFrame property within src.

Author(s)

Microsoft Corporation Microsoft Technical Support

Examples


 ds <- RxXdfData(file.path(rxGetOption("sampleDataDir"), "claims.xdf"))
 # ds contains only one block of data
 rxOpen(ds) # must open the file before rxReadNext
 rxReadNext(ds) # get the first block
 rxReadNext(ds)
 rxClose(ds)

 # Use a data source to compute means by processing the data in chunks
 # Data processing functions: for each chunk, compute sums of columns and
 # number of rows, then update the results computed from previous chunks
 processData <- function(dframe)
   list(sumCols = colSums(dframe), numRows = nrow(dframe))
 updateResults <- function(x, y)
   list(sumCols = x$sumCols + y$sumCols, numRows = x$numRows + y$numRows)

 # Create data source
 censusWorkers <- file.path(rxGetOption("sampleDataDir"), "CensusWorkers.xdf")
 ds <- RxXdfData(censusWorkers, varsToKeep = c("age", "incwage"),
                 blocksPerRead = 2)

 # Process data and update results
 rxOpen(ds)
 resList <- processData(rxReadNext(ds))
 while(TRUE)
 {
     df <- rxReadNext(ds)
     if (nrow(df) == 0)
         break
     resList <- updateResults(resList, processData(df))
 }
 rxClose(ds)

 # Compute the means of the variables from the accumulated results
 varMeans <- resList$sumCols / resList$numRows
 varMeans

Partilhar via