dtplyr Documentation

To use dtplyr, load dtplyr, dplyr, and optionally data.table. Use lazy_dt() to wrap a data frame into a "lazy" data table. This object tracks all subsequent dplyr operations without executing them immediately. To execute the operations and retrieve the results, use as.data.table(), as.data.frame(), or as_tibble().

library(data.table)
library(dtplyr)
library(dplyr, warn.conflicts = FALSE)

# Create a lazy data table
mtcars2 <- lazy_dt(mtcars)

# Perform transformations and execute to get a tibble
result <- mtcars2 %>%
  filter(wt < 5) %>%
  mutate(l100k = 235.21 / mpg) %>%
  group_by(cyl) %>%
  summarise(l100k = mean(l100k)) %>%
  as_tibble()

You can install the stable version of dtplyr from CRAN or the development version from GitHub using pak.

# From CRAN
install.packages("dtplyr")

# From GitHub (development version)
# install.packages("pak")
pak::pak("tidyverse/dtplyr")

You can inspect the data.table code that dtplyr has generated for your dplyr pipeline by simply printing the lazy object. This is useful for debugging and verifying the translation.

mtcars2 %>%
  filter(wt < 5) %>%
  mutate(l100k = 235.21 / mpg) %>%
  group_by(cyl) %>%
  summarise(l100k = mean(l100k))
# This will print the Source, Call (the data.table syntax), and the result

By default, dtplyr follows dplyr semantics where mutate() does not modify the object in place, which may require making copies of the data. To allow in-place modification (matching data.table behavior) and potentially improve performance, you can set immutable = FALSE when calling lazy_dt().

# Example of opting out of immutability
mtcars2 <- lazy_dt(mtcars, immutable = FALSE)