lazy_dt: Create a "lazy" data.table for use with dplyr verbs

View source: R/step-first.R

lazy_dtR Documentation

Create a "lazy" data.table for use with dplyr verbs


A lazy data.table lazy captures the intent of dplyr verbs, only actually performing computation when requested (with collect(), pull(),,, or tibble::as_tibble()). This allows dtplyr to convert dplyr verbs into as few data.table expressions as possible, which leads to a high performance translation.

See vignette("translation") for the details of the translation.


lazy_dt(x, name = NULL, immutable = TRUE, key_by = NULL)



A data table (or something can can be coerced to a data table).


Optionally, supply a name to be used in generated expressions. For expert use only.


If TRUE, x is treated as immutable and will never be modified by any code generated by dtplyr. Alternatively, you can set immutable = FALSE to allow dtplyr to modify the input object.


Set keys for data frame, using select() semantics (e.g. key_by = c(key1, key2).

This uses data.table::setkey() to sort the table and build an index. This will considerably improve performance for subsets, summaries, and joins that use the keys.

See vignette("datatable-keys-fast-subset") for more details.


library(dplyr, warn.conflicts = FALSE)

# If you have a data.table, using it with any dplyr generic will
# automatically convert it to a lazy_dt object
dt <- data.table::data.table(x = 1:10, y = 10:1)
dt %>% filter(x == y)
dt %>% mutate(z = x + y)

# Note that dtplyr will avoid mutating the input data.table, so the
# previous translation includes an automatic copy(). You can avoid this
# with a manual call to lazy_dt()
dt %>%
  lazy_dt(immutable = FALSE) %>%
  mutate(z = x + y)

# If you have a data frame, you can use lazy_dt() to convert it to
# a data.table:
mtcars2 <- lazy_dt(mtcars)
mtcars2 %>% select(mpg:cyl)
mtcars2 %>% select(x = mpg, y = cyl)
mtcars2 %>% filter(cyl == 4) %>% select(mpg)
mtcars2 %>% select(mpg, cyl) %>% filter(cyl == 4)
mtcars2 %>% mutate(cyl2 = cyl * 2, cyl4 = cyl2 * 2)
mtcars2 %>% transmute(cyl2 = cyl * 2, vs2 = vs * 2)
mtcars2 %>% filter(cyl == 8) %>% mutate(cyl2 = cyl * 2)

# Learn more about translation in vignette("translation")
by_cyl <- mtcars2 %>% group_by(cyl)
by_cyl %>% summarise(mpg = mean(mpg))
by_cyl %>% mutate(mpg = mean(mpg))
by_cyl %>%
  filter(mpg < mean(mpg)) %>%
  summarise(hp = mean(hp))

dtplyr documentation built on March 31, 2023, 9:13 p.m.