Reads a dataset from Dataiku's Data Science Studio

dkuReadDataset(
  name,
  partitions = NULL,
  samplingMethod = c("full", "fixed", "head", "ratio"),
  columns = NULL,
  nbRows = NULL,
  ratio = NULL,
  convertEmptyStrings = TRUE,
  colClasses = NA,
  inferColClassesFromData = TRUE,
  na.strings = "NA"
)

Arguments

name: name of dataset
partitions: character vector of partitions to load
samplingMethod: the sampling method to use, if necessary
columns: a character vector of columns to read from dataset
nbRows: An integer. The number of rounds used for sampling
ratio: A numeric. The probability used for sampling each row. 0 < ratio < 1.
convertEmptyStrings: Whether to convert empty strings to NAs
colClasses: Manually-specified column classes. Default is to infer from dataset schema.
inferColClassesFromData: If colClasses is not specified, infer column classes from data instead of dataset schema.
na.strings: Optional list of strings to convert to NAs. Default is "NA".

Value

A data.frame with the requested data

Details

Users can specify which partitions and columns to load, as well as a sampling scheme if the dataset is too large to fit into memory. Possible sampling schemes are fixed sampling, where a set number of rows are randomly chosen from the dataset; head sampling, where the first *n* rows are sampled from the dataset; and ratio sampling, where rows are included randomly with a probability.

Examples

if (FALSE) { # \dontrun{
d = dkuReadDataset("iris")

# read in two columns
d = dkuReadDataset("iris", columns=c("Sepal.Length", "Sepal.Width"))

# explicitly set colClasses
d = dkuReadDataset("iris", colClasses=c("numeric", "numeric", "numeric", "numeric", "character"))

# fixed sampling -- read 100 random rows from the iris dataset
d = dkuReadDataset("iris", samplingMethod="fixed", nbRows=100)

# head sampling -- read the first 100 rows from the iris dataset
d = dkuReadDataset("iris", samplingMethod="head", nbRows=100)

# ratio sampling -- read 30% of the rows (chosen randomly) from the iris dataset
d = dkuReadDataset("iris", samplingMethod="ratio", ratio=0.3)
} # }