mlr3db (version 0.6.0)

DataBackendPolars: DataBackend for Polars

Description

A mlr3::DataBackend using RPolarsLazyFrame from package polars. Can be easily constructed with as_polars_backend(). mlr3::Tasks can interface out-of-memory files if the polars::RPolarsLazyFrame was imported using a polars::scan_x function. Streaming, a polars alpha feature, is always enabled, but only used when applicable. A connector is not required but can be useful e.g. for scanning larger than memory files

Super class

mlr3::DataBackend -> DataBackendPolars

Public fields

levels

(named list())
List (named with column names) of factor levels as character(). Used to auto-convert character columns to factor variables.

connector

(function())
Function which is called to re-connect in case the connection became invalid.

Active bindings

rownames

(integer())
Returns vector of all distinct row identifiers, i.e. the contents of the primary key column.

colnames

(character())
Returns vector of all column names, including the primary key column.

nrow

(integer(1))
Number of rows (observations).

ncol

(integer(1))
Number of columns (variables), including the primary key column.

Methods

Public methods

• DataBackendPolars$new()

• DataBackendPolars$data()

• DataBackendPolars$head()

• DataBackendPolars$distinct()

• DataBackendPolars$missings()

Inherited methods

• mlr3::DataBackend$format()
• mlr3::DataBackend$print()

---

Method new()

Creates a backend for a polars::RPolarsDataFrame object.

Usage

DataBackendPolars$new(
  data,
  primary_key,
  strings_as_factors = TRUE,
  connector = NULL
)

Arguments

data

(polars::RPolarsLazyFrame)
The data object.

Instead of calling the constructor itself, please call mlr3::as_data_backend() on a polars::RPolarsLazyFrame or polars::RPolarsDataFrame. Note that only polars::RPolarsLazyFrames will be converted to a DataBackendPolars. polars::RPolarsDataFrame objects without lazy execution will be converted to a mlr3::DataBackendDataTable.

primary_key

(character(1))
Name of the primary key column. Because polars does not natively support primary keys, uniqueness of the primary key column is expected but not enforced.

strings_as_factors

(logical(1) || character())
Either a character vector of column names to convert to factors, or a single logical flag: if FALSE, no column will be converted, if TRUE all string columns (except the primary key). For conversion, the backend is queried for distinct values of the respective columns on construction and their levels are stored in $levels.

connector

(function())
Optional function which is called to re-connect to e.g. a source file in case the connection became invalid.

---

Method data()

Returns a slice of the data.

The rows must be addressed as vector of primary key values, columns must be referred to via column names. Queries for rows with no matching row id and queries for columns with no matching column name are silently ignored.

Usage

DataBackendPolars$data(rows, cols)

Arguments

rows

(integer())
Row indices.

cols

(character())
Column names.

---

Method head()

Retrieve the first n rows.

Usage

DataBackendPolars$head(n = 6L)

Arguments

n

(integer(1))
Number of rows.

Returns

data.table::data.table() of the first n rows.

---

Method distinct()

Returns a named list of vectors of distinct values for each column specified. If na_rm is TRUE, missing values are removed from the returned vectors of distinct values. Non-existing rows and columns are silently ignored.

Usage

DataBackendPolars$distinct(rows, cols, na_rm = TRUE)

Arguments

rows

(integer())
Row indices.

cols

(character())
Column names.

na_rm

(logical(1))
Whether to remove NAs or not.

Returns

Named list() of distinct values.

---

Method missings()

Returns the number of missing values per column in the specified slice of data. Non-existing rows and columns are silently ignored.

Usage

DataBackendPolars$missings(rows, cols)

Arguments

rows

(integer())
Row indices.

cols

(character())
Column names.

Returns

Total of missing values per column (named numeric()).

See Also

https://pola-rs.github.io/r-polars/

Examples

if (mlr3misc::require_namespaces("polars", quietly = TRUE)) {
  # Backend using a in-memory data set
  data = iris
  data$Sepal.Length[1:30] = NA
  data$row_id = 1:150
  data = polars::as_polars_lf(data)
  b = DataBackendPolars$new(data, primary_key = "row_id")

  # Object supports all accessors of DataBackend
  print(b)
  b$nrow
  b$ncol
  b$colnames
  b$data(rows = 100:101, cols = "Species")
  b$distinct(b$rownames, "Species")

  # Classification task using this backend
  task = mlr3::TaskClassif$new(id = "iris_polars", backend = b, target = "Species")
  print(task)
  head(task)

  # Write a parquet file to scan
  data$collect()$write_parquet("iris.parquet")
  data = polars::pl$scan_parquet("iris.parquet")

  # Backend that re-reads the parquet file if the connection fails
  b = DataBackendPolars$new(data, "row_id",
    connector = function() polars::pl$scan_parquet("iris.parquet"))
  print(b)

  # Define a backend on a subset of the database: do not use column "Sepal.Width"
  data = data$select(
    polars::pl$col(setdiff(colnames(data), "Sepal.Width"))
  )$filter(
    polars::pl$col("row_id")$is_in(1:120) # Use only first 120 rows
  )

  # Backend with only scanned data
  b = DataBackendPolars$new(data, "row_id", strings_as_factors = TRUE)
  print(b)

  # Query disinct values
  b$distinct(b$rownames, "Species")

  # Query number of missing values
  b$missings(b$rownames, b$colnames)

  # Cleanup
  if (file.exists("iris.parquet")) {
    file.remove("iris.parquet")
  }
}