roworder: Fast Reordering of Data Frame Rows

In collapse: Advanced and Fast Data Transformation

Fast Reordering of Data Frame Rows

Description

A fast substitute for dplyr::arrange, based on radixorder(v) and inspired by data.table::setorder(v). It returns a sorted copy of the data frame, unless the data is already sorted in which case no copy is made. In addition, rows can be manually re-ordered. roworderv is a programmers version that takes vectors/variables as input.

Use data.table::setorder(v) to sort a data frame without creating a copy.

Usage

roworder(X, ..., na.last = TRUE, verbose = .op[["verbose"]])

roworderv(X, cols = NULL, neworder = NULL, decreasing = FALSE,
          na.last = TRUE, pos = "front", verbose = .op[["verbose"]])

Arguments

X	a data frame or list of equal-length columns.
...	comma-separated columns of X to sort by e.g. var1, var2. Negatives i.e. -var1, var2 can be used to sort in decreasing order of var1. Internally all expressions are turned into strings and startsWith(expr, "-") is used to detect this, thus it does not negate the actual values (which may as well be strings), and you cannot apply any other functions to columns inside roworder() to induce different sorting behavior.
cols	select columns to sort by using a function, column names, indices or a logical vector. The default NULL sorts by all columns in order of occurrence (from left to right).
na.last	logical. If TRUE, missing values in the sorting columns are placed last; if FALSE, they are placed first; if NA they are removed (argument passed to radixorderv).
decreasing	logical. Should the sort order be increasing or decreasing? Can also be a vector of length equal to the number of arguments in cols (argument passed to radixorderv).
neworder	an ordering vector, can be < nrow(X). if pos = "front" or pos = "end", a logical vector can also be supplied. This argument overwrites cols.
pos	integer or character. Different arrangement options if !is.null(neworder) && length(neworder) < nrow(X). Int. String Description 1 "front" move rows in neworder to the front (top) of X (the default). 2 "end" move rows in neworder to the end (bottom) of X. 3 "exchange" just exchange the order of rows in neworder, other rows remain in the same position. 4 "after" place all further selected rows behind the first selected row.
verbose	logical. 1L (default) prints a message when ordering a grouped or indexed frame, indicating that this is not efficient and encouraging reordering the data prior to the grouping/indexing step. Users can also set verbose = 2L to also toggle a message if x is already sorted, implying that no copy was made and the call to roworder(v) is redundant.

Value

A copy of X with rows reordered. If X is already sorted, X is simply returned.

Note

If you don't require a copy of the data, use data.table::setorder (you can also use it in a piped call as it invisibly returns the data).

roworder(v) has internal facilities to deal with grouped and indexed data. This is however inefficient (since in most cases data could be reordered before grouping/indexing), and therefore issues a message if verbose > 0L.

See Also

colorder, Data Frame Manipulation, Fast Grouping and Ordering, Collapse Overview

Examples

head(roworder(airquality, Month, -Ozone))
head(roworder(airquality, Month, -Ozone, na.last = NA))  # Removes the missing values in Ozone

## Same in standard evaluation
head(roworderv(airquality, c("Month", "Ozone"), decreasing = c(FALSE, TRUE)))
head(roworderv(airquality, c("Month", "Ozone"), decreasing = c(FALSE, TRUE), na.last = NA))

## Custom reordering
head(roworderv(mtcars, neworder = 3:4))               # Bring rows 3 and 4 to the front
head(roworderv(mtcars, neworder = 3:4, pos = "end"))  # Bring them to the end
head(roworderv(mtcars, neworder = mtcars$vs == 1))    # Bring rows with vs == 1 to the top

collapse documentation built on Nov. 3, 2024, 9:08 a.m.