tidypolars provides a polars backend
for the tidyverse. The aim of tidypolars is to enable users to keep
their existing tidyverse code while using polars in the background
to benefit from large performance gains.
See the example below and the “Getting started”
vignette for
a gentle introduction to tidypolars.
tidypolars is built on polars, which is not available on CRAN. This
means that tidypolars also can’t be on CRAN. However, you can install
it from R-universe.
install.packages(
'tidypolars',
repos = c('https://etiennebacher.r-universe.dev', getOption("repos"))
)install.packages(
'tidypolars',
repos = c('https://etiennebacher.r-universe.dev/bin/linux/jammy/4.3', getOption("repos"))
)Suppose that you already have some code that uses dplyr:
library(dplyr, warn.conflicts = FALSE)
iris |>
select(starts_with(c("Sep", "Pet"))) |>
mutate(
petal_type = ifelse((Petal.Length / Petal.Width) > 3, "long", "large")
) |>
filter(between(Sepal.Length, 4.5, 5.5)) |>
head()
#> Sepal.Length Sepal.Width Petal.Length Petal.Width petal_type
#> 1 5.1 3.5 1.4 0.2 long
#> 2 4.9 3.0 1.4 0.2 long
#> 3 4.7 3.2 1.3 0.2 long
#> 4 4.6 3.1 1.5 0.2 long
#> 5 5.0 3.6 1.4 0.2 long
#> 6 5.4 3.9 1.7 0.4 longWith tidypolars, you can provide a Polars DataFrame or LazyFrame
and keep the exact same code:
library(tidypolars)
iris |>
as_polars_df() |>
select(starts_with(c("Sep", "Pet"))) |>
mutate(
petal_type = ifelse((Petal.Length / Petal.Width) > 3, "long", "large")
) |>
filter(between(Sepal.Length, 4.5, 5.5)) |>
head()
#> shape: (6, 5)
#> ┌──────────────┬─────────────┬──────────────┬─────────────┬────────────┐
#> │ Sepal.Length ┆ Sepal.Width ┆ Petal.Length ┆ Petal.Width ┆ petal_type │
#> │ --- ┆ --- ┆ --- ┆ --- ┆ --- │
#> │ f64 ┆ f64 ┆ f64 ┆ f64 ┆ str │
#> ╞══════════════╪═════════════╪══════════════╪═════════════╪════════════╡
#> │ 5.1 ┆ 3.5 ┆ 1.4 ┆ 0.2 ┆ long │
#> │ 4.9 ┆ 3.0 ┆ 1.4 ┆ 0.2 ┆ long │
#> │ 4.7 ┆ 3.2 ┆ 1.3 ┆ 0.2 ┆ long │
#> │ 4.6 ┆ 3.1 ┆ 1.5 ┆ 0.2 ┆ long │
#> │ 5.0 ┆ 3.6 ┆ 1.4 ┆ 0.2 ┆ long │
#> │ 5.4 ┆ 3.9 ┆ 1.7 ┆ 0.4 ┆ long │
#> └──────────────┴─────────────┴──────────────┴─────────────┴────────────┘If you’re used to the tidyverse functions and syntax, this will feel
much easier to read than the pure polars syntax:
library(polars)
# polars syntax
pl$DataFrame(iris)$
select(c("Sepal.Length", "Sepal.Width", "Petal.Length", "Petal.Width"))$
with_columns(
pl$when(
(pl$col("Petal.Length") / pl$col("Petal.Width") > 3)
)$then(pl$lit("long"))$
otherwise(pl$lit("large"))$
alias("petal_type")
)$
filter(pl$col("Sepal.Length")$is_between(4.5, 5.5))$
head(6)
#> shape: (6, 5)
#> ┌──────────────┬─────────────┬──────────────┬─────────────┬────────────┐
#> │ Sepal.Length ┆ Sepal.Width ┆ Petal.Length ┆ Petal.Width ┆ petal_type │
#> │ --- ┆ --- ┆ --- ┆ --- ┆ --- │
#> │ f64 ┆ f64 ┆ f64 ┆ f64 ┆ str │
#> ╞══════════════╪═════════════╪══════════════╪═════════════╪════════════╡
#> │ 5.1 ┆ 3.5 ┆ 1.4 ┆ 0.2 ┆ long │
#> │ 4.9 ┆ 3.0 ┆ 1.4 ┆ 0.2 ┆ long │
#> │ 4.7 ┆ 3.2 ┆ 1.3 ┆ 0.2 ┆ long │
#> │ 4.6 ┆ 3.1 ┆ 1.5 ┆ 0.2 ┆ long │
#> │ 5.0 ┆ 3.6 ┆ 1.4 ┆ 0.2 ┆ long │
#> │ 5.4 ┆ 3.9 ┆ 1.7 ┆ 0.4 ┆ long │
#> └──────────────┴─────────────┴──────────────┴─────────────┴────────────┘Since most of the work is rewriting tidyverse code into polars
syntax, tidypolars and polars have very similar performance.
Click to see a small benchmark
For more serious benchmarks about polars, take a look at DuckDB
benchmarks.
library(collapse, warn.conflicts = FALSE)
#> collapse 2.0.11, see ?`collapse-package` or ?`collapse-documentation`
library(dtplyr)
large_iris <- data.table::rbindlist(rep(list(iris), 100000))
large_iris_pl <- as_polars_lf(large_iris)
large_iris_dt <- lazy_dt(large_iris)
format(nrow(large_iris), big.mark = ",")
#> [1] "15,000,000"
bench::mark(
polars = {
large_iris_pl$
select(c("Sepal.Length", "Sepal.Width", "Petal.Length", "Petal.Width"))$
with_columns(
pl$when(
(pl$col("Petal.Length") / pl$col("Petal.Width") > 3)
)$then(pl$lit("long"))$
otherwise(pl$lit("large"))$
alias("petal_type")
)$
filter(pl$col("Sepal.Length")$is_between(4.5, 5.5))$
collect()
},
tidypolars = {
large_iris_pl |>
select(starts_with(c("Sep", "Pet"))) |>
mutate(
petal_type = ifelse((Petal.Length / Petal.Width) > 3, "long", "large")
) |>
filter(between(Sepal.Length, 4.5, 5.5)) |>
compute()
},
dplyr = {
large_iris |>
select(starts_with(c("Sep", "Pet"))) |>
mutate(
petal_type = ifelse((Petal.Length / Petal.Width) > 3, "long", "large")
) |>
filter(between(Sepal.Length, 4.5, 5.5))
},
dtplyr = {
large_iris_dt |>
select(starts_with(c("Sep", "Pet"))) |>
mutate(
petal_type = ifelse((Petal.Length / Petal.Width) > 3, "long", "large")
) |>
filter(between(Sepal.Length, 4.5, 5.5)) |>
as.data.frame()
},
collapse = {
large_iris |>
fselect(c("Sepal.Length", "Sepal.Width", "Petal.Length", "Petal.Width")) |>
fmutate(
petal_type = data.table::fifelse((Petal.Length / Petal.Width) > 3, "long", "large")
) |>
fsubset(Sepal.Length >= 4.5 & Sepal.Length <= 5.5)
},
check = FALSE,
iterations = 40
)
#> Warning: Some expressions had a GC in every iteration; so filtering is
#> disabled.
#> # A tibble: 5 × 6
#> expression min median `itr/sec` mem_alloc `gc/sec`
#> <bch:expr> <bch:tm> <bch:tm> <dbl> <bch:byt> <dbl>
#> 1 polars 126.4ms 288.63ms 2.89 19.6KB 0.0722
#> 2 tidypolars 152.53ms 202.25ms 4.37 332.28KB 1.09
#> 3 dplyr 5.62s 6.06s 0.164 1.79GB 0.476
#> 4 dtplyr 835.67ms 1.03s 0.957 1.72GB 2.34
#> 5 collapse 487.95ms 653.16ms 1.50 745.96MB 1.09
# NOTE: do NOT take the "mem_alloc" results into account.
# `bench::mark()` doesn't report the accurate memory usage for packages calling
# Rust code.Did you find some bugs or some errors in the documentation? Do you want
tidypolars to support more functions?
Take a look at the contributing guide for instructions on bug report and pull requests.
The website theme was heavily inspired by Matthew Kay’s ggblend
package: https://mjskay.github.io/ggblend/.