Polars is a lightning fast Data Frame library. Its embarrassingly parallel execution, cache efficient algorithms and expressive API makes it perfect for efficient data wrangling, data pipelines, snappy APIs, and much more besides. Polars also supports “streaming mode” for out-of-memory operations. This allows users to analyze datasets many times larger than RAM.
The underlying computation engine is written in Rust and is built on the Apache Arrow columnar memory format. It can be used in Rust or via Python bindings. The polars R-package provides equivalent bindings from R. To help distinguish the different language implementations, we typically use a convention of referring to them with prefixes: rust-polars, py-polars, r-polars, nodejs-polars, etc. But within each language, the relevant library is always just called polars.
polars users can expect orders of magnitude(s) improvement compared to dplyr for simple transformations on datasets >500Mb. The automatic Polars optimization framework means that that this speed boost can be even greater for complex queries that chain together many operations. Performance is similar to that of data.table, although polars supports additional functionality via its relationship to the Apache Arrow memory model. For example, it can scan multiple Parquet files and datasets and selectively import random subsets without having to read all of the data.
Polars syntax is similar to that of Spark, but the workflow is column-oriented rather than row-oriented. Since R is itself a column-oriented language, this should immediately feel familiar to most R users. Like Spark and modern SQL variants, Polars optimizes queries for memory consumption and speed, so you don’t have to. However, unlike Spark, Polars is natively multithreaded instead of multinoded. This makes (r)polars much simpler to install and can be used as one would any other R package.
This R port relies on the excellent extendr package, which is the R equivalent to pyo3+maturin. extendr is very convenient for calling Rust from R, and vice versa, and is what we use to build the polars package. Once built, however, polars has no other dependencies other than R itself. This makes it very fast and lightweight to install, and so polars can immediately be used to tackle your big (or small!) data wrangling tasks.
Users can find detailed documentation for all objects, functions, and
methods on the Reference page of this website. This
documentation can also be accessed from the R console using the typical
? syntax. For example, we will later use the
DataFrame() constructor function and apply the
group_by() method to a DataFrame object. The
documentation for these can be accessed by typing these commands:
The Polars book
offers a great introduction to the Polars data frame library, with a
very large number of examples in Python and Rust. The syntax and
expressions in the polars package for R are (deliberately)
as close to the Python implementation as possible, so you can always
refer to the polars book
for more ideas. Just remember to switch out any “.” (Python) for a “$”
(R) when chaining methods. For example, here are two equivalent lines of
code for some hypothetical dataset.
Series and DataFramesIn polars, objects of class Series are
analogous to R vectors. Objects of class DataFrame are
analogous to R data frames. Notice that to avoid collision with classes
provided by other packages, the class name of all objects created by
polars starts with “RPolars”. For example, a
polars DataFrame has the class
“RPolarsDataFrame”.
To create Polars Series and DataFrames
objects, we load the library and use constructor functions with the
pl$ prefix. This prefix is very important, as most of the
polars functions are made available via
pl$:
library(polars)
pl$Series(name = "a", values = (1:5) * 5)
#> polars Series: shape: (5,)
#> Series: 'a' [f64]
#> [
#> 5.0
#> 10.0
#> 15.0
#> 20.0
#> 25.0
#> ]
pl$DataFrame(a = 1:5, b = letters[1:5])
#> shape: (5, 2)
#> ┌─────┬─────┐
#> │ a ┆ b │
#> │ --- ┆ --- │
#> │ i32 ┆ str │
#> ╞═════╪═════╡
#> │ 1 ┆ a │
#> │ 2 ┆ b │
#> │ 3 ┆ c │
#> │ 4 ┆ d │
#> │ 5 ┆ e │
#> └─────┴─────┘Or, to convert existing R objects to Polars objects, we can use the
as_polars_series() and as_polars_df() generic
functions.
ser = as_polars_series((1:5) * 5)
ser
#> polars Series: shape: (5,)
#> Series: '' [f64]
#> [
#> 5.0
#> 10.0
#> 15.0
#> 20.0
#> 25.0
#> ]
dat = as_polars_df(mtcars)
dat
#> shape: (32, 11)
#> ┌──────┬─────┬───────┬───────┬───┬─────┬─────┬──────┬──────┐
#> │ mpg ┆ cyl ┆ disp ┆ hp ┆ … ┆ vs ┆ am ┆ gear ┆ carb │
#> │ --- ┆ --- ┆ --- ┆ --- ┆ ┆ --- ┆ --- ┆ --- ┆ --- │
#> │ f64 ┆ f64 ┆ f64 ┆ f64 ┆ ┆ f64 ┆ f64 ┆ f64 ┆ f64 │
#> ╞══════╪═════╪═══════╪═══════╪═══╪═════╪═════╪══════╪══════╡
#> │ 21.0 ┆ 6.0 ┆ 160.0 ┆ 110.0 ┆ … ┆ 0.0 ┆ 1.0 ┆ 4.0 ┆ 4.0 │
#> │ 21.0 ┆ 6.0 ┆ 160.0 ┆ 110.0 ┆ … ┆ 0.0 ┆ 1.0 ┆ 4.0 ┆ 4.0 │
#> │ 22.8 ┆ 4.0 ┆ 108.0 ┆ 93.0 ┆ … ┆ 1.0 ┆ 1.0 ┆ 4.0 ┆ 1.0 │
#> │ 21.4 ┆ 6.0 ┆ 258.0 ┆ 110.0 ┆ … ┆ 1.0 ┆ 0.0 ┆ 3.0 ┆ 1.0 │
#> │ 18.7 ┆ 8.0 ┆ 360.0 ┆ 175.0 ┆ … ┆ 0.0 ┆ 0.0 ┆ 3.0 ┆ 2.0 │
#> │ … ┆ … ┆ … ┆ … ┆ … ┆ … ┆ … ┆ … ┆ … │
#> │ 30.4 ┆ 4.0 ┆ 95.1 ┆ 113.0 ┆ … ┆ 1.0 ┆ 1.0 ┆ 5.0 ┆ 2.0 │
#> │ 15.8 ┆ 8.0 ┆ 351.0 ┆ 264.0 ┆ … ┆ 0.0 ┆ 1.0 ┆ 5.0 ┆ 4.0 │
#> │ 19.7 ┆ 6.0 ┆ 145.0 ┆ 175.0 ┆ … ┆ 0.0 ┆ 1.0 ┆ 5.0 ┆ 6.0 │
#> │ 15.0 ┆ 8.0 ┆ 301.0 ┆ 335.0 ┆ … ┆ 0.0 ┆ 1.0 ┆ 5.0 ┆ 8.0 │
#> │ 21.4 ┆ 4.0 ┆ 121.0 ┆ 109.0 ┆ … ┆ 1.0 ┆ 1.0 ┆ 4.0 ┆ 2.0 │
#> └──────┴─────┴───────┴───────┴───┴─────┴─────┴──────┴──────┘pl$DataFrame() is similar to data.frame(),
and as_polars_df() is similar to
as.data.frame().
Both Polars and R are column-orientated. So you can think of
DataFrames (data.frames) as being made up of a collection
of Series (vectors). In fact, you can create a new Polars
DataFrame as a mix of Series and/or regular R
vectors.
pl$DataFrame(
a = as_polars_series((1:5) * 5),
b = as_polars_series(letters[1:5]),
c = as_polars_series(c(1, 2, 3, 4, 5)),
d = c(15, 14, 13, 12, 11),
c(5, 4, 3, 2, 1),
1:5
)
#> shape: (5, 6)
#> ┌──────┬─────┬─────┬──────┬────────────┬──────────────┐
#> │ a ┆ b ┆ c ┆ d ┆ new_column ┆ new_column_1 │
#> │ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- │
#> │ f64 ┆ str ┆ f64 ┆ f64 ┆ f64 ┆ i32 │
#> ╞══════╪═════╪═════╪══════╪════════════╪══════════════╡
#> │ 5.0 ┆ a ┆ 1.0 ┆ 15.0 ┆ 5.0 ┆ 1 │
#> │ 10.0 ┆ b ┆ 2.0 ┆ 14.0 ┆ 4.0 ┆ 2 │
#> │ 15.0 ┆ c ┆ 3.0 ┆ 13.0 ┆ 3.0 ┆ 3 │
#> │ 20.0 ┆ d ┆ 4.0 ┆ 12.0 ┆ 2.0 ┆ 4 │
#> │ 25.0 ┆ e ┆ 5.0 ┆ 11.0 ┆ 1.0 ┆ 5 │
#> └──────┴─────┴─────┴──────┴────────────┴──────────────┘Series and DataFrame can be operated on
using many standard R functions. For example:
# Series
length(ser)
#> [1] 5
max(ser)
#> [1] 25
# DataFrame
dat[c(1:3, 12), c("mpg", "hp")]
#> shape: (4, 2)
#> ┌──────┬───────┐
#> │ mpg ┆ hp │
#> │ --- ┆ --- │
#> │ f64 ┆ f64 │
#> ╞══════╪═══════╡
#> │ 21.0 ┆ 110.0 │
#> │ 21.0 ┆ 110.0 │
#> │ 22.8 ┆ 93.0 │
#> │ 16.4 ┆ 180.0 │
#> └──────┴───────┘
names(dat)
#> [1] "mpg" "cyl" "disp" "hp" "drat" "wt" "qsec" "vs" "am" "gear"
#> [11] "carb"
dim(dat)
#> [1] 32 11
head(dat, n = 2)
#> shape: (2, 11)
#> ┌──────┬─────┬───────┬───────┬───┬─────┬─────┬──────┬──────┐
#> │ mpg ┆ cyl ┆ disp ┆ hp ┆ … ┆ vs ┆ am ┆ gear ┆ carb │
#> │ --- ┆ --- ┆ --- ┆ --- ┆ ┆ --- ┆ --- ┆ --- ┆ --- │
#> │ f64 ┆ f64 ┆ f64 ┆ f64 ┆ ┆ f64 ┆ f64 ┆ f64 ┆ f64 │
#> ╞══════╪═════╪═══════╪═══════╪═══╪═════╪═════╪══════╪══════╡
#> │ 21.0 ┆ 6.0 ┆ 160.0 ┆ 110.0 ┆ … ┆ 0.0 ┆ 1.0 ┆ 4.0 ┆ 4.0 │
#> │ 21.0 ┆ 6.0 ┆ 160.0 ┆ 110.0 ┆ … ┆ 0.0 ┆ 1.0 ┆ 4.0 ┆ 4.0 │
#> └──────┴─────┴───────┴───────┴───┴─────┴─────┴──────┴──────┘Although some simple R functions work out of the box on
polars objects, the full power of Polars is realized
via methods. Polars methods are accessed using the
$ syntax. For example, to sort a Series
object, we use the $sort() method:
ser$sort()
#> polars Series: shape: (5,)
#> Series: '' [f64]
#> [
#> 5.0
#> 10.0
#> 15.0
#> 20.0
#> 25.0
#> ]There are numerous methods designed to accomplish various tasks:
ser$max()
#> [1] 25
dat$slice(offset = 2, length = 3)
#> shape: (3, 11)
#> ┌──────┬─────┬───────┬───────┬───┬─────┬─────┬──────┬──────┐
#> │ mpg ┆ cyl ┆ disp ┆ hp ┆ … ┆ vs ┆ am ┆ gear ┆ carb │
#> │ --- ┆ --- ┆ --- ┆ --- ┆ ┆ --- ┆ --- ┆ --- ┆ --- │
#> │ f64 ┆ f64 ┆ f64 ┆ f64 ┆ ┆ f64 ┆ f64 ┆ f64 ┆ f64 │
#> ╞══════╪═════╪═══════╪═══════╪═══╪═════╪═════╪══════╪══════╡
#> │ 22.8 ┆ 4.0 ┆ 108.0 ┆ 93.0 ┆ … ┆ 1.0 ┆ 1.0 ┆ 4.0 ┆ 1.0 │
#> │ 21.4 ┆ 6.0 ┆ 258.0 ┆ 110.0 ┆ … ┆ 1.0 ┆ 0.0 ┆ 3.0 ┆ 1.0 │
#> │ 18.7 ┆ 8.0 ┆ 360.0 ┆ 175.0 ┆ … ┆ 0.0 ┆ 0.0 ┆ 3.0 ┆ 2.0 │
#> └──────┴─────┴───────┴───────┴───┴─────┴─────┴──────┴──────┘One advantage of using methods is that many more operations are possible on Polars objects using methods than through base R functions.
A second advantage is Methods Chaining, a core part of the Polars workflow. If you are coming from one of the other popular data wrangling libraries in R, then you probably already have an innate sense of what this means. For instance,
dat |> filter(...) |> select(...)DT[i, j, by][...]In polars our method chaining syntax takes the form
object$m1()$m2(), where object is our data
object, and $m1() and $m2() are appropriate
methods, like subsetting or aggregation expressions.
This might all seem a little abstract, so let’s walk through some
quick examples to help make things concrete. We continue with the
mtcars dataset that we coerced to a DataFrame
in the introduction.1
To start, say we compute the maximum value in each column. We can use
the $max() method:
dat$max()
#> shape: (1, 11)
#> ┌──────┬─────┬───────┬───────┬───┬─────┬─────┬──────┬──────┐
#> │ mpg ┆ cyl ┆ disp ┆ hp ┆ … ┆ vs ┆ am ┆ gear ┆ carb │
#> │ --- ┆ --- ┆ --- ┆ --- ┆ ┆ --- ┆ --- ┆ --- ┆ --- │
#> │ f64 ┆ f64 ┆ f64 ┆ f64 ┆ ┆ f64 ┆ f64 ┆ f64 ┆ f64 │
#> ╞══════╪═════╪═══════╪═══════╪═══╪═════╪═════╪══════╪══════╡
#> │ 33.9 ┆ 8.0 ┆ 472.0 ┆ 335.0 ┆ … ┆ 1.0 ┆ 1.0 ┆ 5.0 ┆ 8.0 │
#> └──────┴─────┴───────┴───────┴───┴─────┴─────┴──────┴──────┘Now, we first use the $tail() method to select the last
10 rows of the dataset, and then use the $max() method to
compute the maximums in those 10 rows:
dat$tail(10)$max()
#> shape: (1, 11)
#> ┌──────┬─────┬───────┬───────┬───┬─────┬─────┬──────┬──────┐
#> │ mpg ┆ cyl ┆ disp ┆ hp ┆ … ┆ vs ┆ am ┆ gear ┆ carb │
#> │ --- ┆ --- ┆ --- ┆ --- ┆ ┆ --- ┆ --- ┆ --- ┆ --- │
#> │ f64 ┆ f64 ┆ f64 ┆ f64 ┆ ┆ f64 ┆ f64 ┆ f64 ┆ f64 │
#> ╞══════╪═════╪═══════╪═══════╪═══╪═════╪═════╪══════╪══════╡
#> │ 30.4 ┆ 8.0 ┆ 400.0 ┆ 335.0 ┆ … ┆ 1.0 ┆ 1.0 ┆ 5.0 ┆ 8.0 │
#> └──────┴─────┴───────┴───────┴───┴─────┴─────┴──────┴──────┘Finally, we convert the result to a standard R data frame:
dat$tail(10)$max() |>
as.data.frame()
#> mpg cyl disp hp drat wt qsec vs am gear carb
#> 1 30.4 8 400 335 4.43 3.845 18.9 1 1 5 8Below, we will introduce several other methods, including
$select(), $filter(), and
$group_by() which allow us to do powerful data
manipulations easily. To give you a small taste, we now take group-wise
means:
dat$group_by("cyl")$mean()
#> shape: (3, 11)
#> ┌─────┬───────────┬────────────┬────────────┬───┬──────────┬──────────┬──────────┬──────────┐
#> │ cyl ┆ mpg ┆ disp ┆ hp ┆ … ┆ vs ┆ am ┆ gear ┆ carb │
#> │ --- ┆ --- ┆ --- ┆ --- ┆ ┆ --- ┆ --- ┆ --- ┆ --- │
#> │ f64 ┆ f64 ┆ f64 ┆ f64 ┆ ┆ f64 ┆ f64 ┆ f64 ┆ f64 │
#> ╞═════╪═══════════╪════════════╪════════════╪═══╪══════════╪══════════╪══════════╪══════════╡
#> │ 6.0 ┆ 19.742857 ┆ 183.314286 ┆ 122.285714 ┆ … ┆ 0.571429 ┆ 0.428571 ┆ 3.857143 ┆ 3.428571 │
#> │ 8.0 ┆ 15.1 ┆ 353.1 ┆ 209.214286 ┆ … ┆ 0.0 ┆ 0.142857 ┆ 3.285714 ┆ 3.5 │
#> │ 4.0 ┆ 26.663636 ┆ 105.136364 ┆ 82.636364 ┆ … ┆ 0.909091 ┆ 0.727273 ┆ 4.090909 ┆ 1.545455 │
#> └─────┴───────────┴────────────┴────────────┴───┴──────────┴──────────┴──────────┴──────────┘We can now start chaining together various methods (expressions) to
manipulate it in different ways. For example, we can subset the data by
rows ($filter())
and also columns ($select()):
dat$filter(pl$col("cyl") == 6)
#> shape: (7, 11)
#> ┌──────┬─────┬───────┬───────┬───┬─────┬─────┬──────┬──────┐
#> │ mpg ┆ cyl ┆ disp ┆ hp ┆ … ┆ vs ┆ am ┆ gear ┆ carb │
#> │ --- ┆ --- ┆ --- ┆ --- ┆ ┆ --- ┆ --- ┆ --- ┆ --- │
#> │ f64 ┆ f64 ┆ f64 ┆ f64 ┆ ┆ f64 ┆ f64 ┆ f64 ┆ f64 │
#> ╞══════╪═════╪═══════╪═══════╪═══╪═════╪═════╪══════╪══════╡
#> │ 21.0 ┆ 6.0 ┆ 160.0 ┆ 110.0 ┆ … ┆ 0.0 ┆ 1.0 ┆ 4.0 ┆ 4.0 │
#> │ 21.0 ┆ 6.0 ┆ 160.0 ┆ 110.0 ┆ … ┆ 0.0 ┆ 1.0 ┆ 4.0 ┆ 4.0 │
#> │ 21.4 ┆ 6.0 ┆ 258.0 ┆ 110.0 ┆ … ┆ 1.0 ┆ 0.0 ┆ 3.0 ┆ 1.0 │
#> │ 18.1 ┆ 6.0 ┆ 225.0 ┆ 105.0 ┆ … ┆ 1.0 ┆ 0.0 ┆ 3.0 ┆ 1.0 │
#> │ 19.2 ┆ 6.0 ┆ 167.6 ┆ 123.0 ┆ … ┆ 1.0 ┆ 0.0 ┆ 4.0 ┆ 4.0 │
#> │ 17.8 ┆ 6.0 ┆ 167.6 ┆ 123.0 ┆ … ┆ 1.0 ┆ 0.0 ┆ 4.0 ┆ 4.0 │
#> │ 19.7 ┆ 6.0 ┆ 145.0 ┆ 175.0 ┆ … ┆ 0.0 ┆ 1.0 ┆ 5.0 ┆ 6.0 │
#> └──────┴─────┴───────┴───────┴───┴─────┴─────┴──────┴──────┘
dat$filter(pl$col("cyl") == 6 & pl$col("am") == 1)
#> shape: (3, 11)
#> ┌──────┬─────┬───────┬───────┬───┬─────┬─────┬──────┬──────┐
#> │ mpg ┆ cyl ┆ disp ┆ hp ┆ … ┆ vs ┆ am ┆ gear ┆ carb │
#> │ --- ┆ --- ┆ --- ┆ --- ┆ ┆ --- ┆ --- ┆ --- ┆ --- │
#> │ f64 ┆ f64 ┆ f64 ┆ f64 ┆ ┆ f64 ┆ f64 ┆ f64 ┆ f64 │
#> ╞══════╪═════╪═══════╪═══════╪═══╪═════╪═════╪══════╪══════╡
#> │ 21.0 ┆ 6.0 ┆ 160.0 ┆ 110.0 ┆ … ┆ 0.0 ┆ 1.0 ┆ 4.0 ┆ 4.0 │
#> │ 21.0 ┆ 6.0 ┆ 160.0 ┆ 110.0 ┆ … ┆ 0.0 ┆ 1.0 ┆ 4.0 ┆ 4.0 │
#> │ 19.7 ┆ 6.0 ┆ 145.0 ┆ 175.0 ┆ … ┆ 0.0 ┆ 1.0 ┆ 5.0 ┆ 6.0 │
#> └──────┴─────┴───────┴───────┴───┴─────┴─────┴──────┴──────┘
dat$select(pl$col(c("mpg", "hp")))
#> shape: (32, 2)
#> ┌──────┬───────┐
#> │ mpg ┆ hp │
#> │ --- ┆ --- │
#> │ f64 ┆ f64 │
#> ╞══════╪═══════╡
#> │ 21.0 ┆ 110.0 │
#> │ 21.0 ┆ 110.0 │
#> │ 22.8 ┆ 93.0 │
#> │ 21.4 ┆ 110.0 │
#> │ 18.7 ┆ 175.0 │
#> │ … ┆ … │
#> │ 30.4 ┆ 113.0 │
#> │ 15.8 ┆ 264.0 │
#> │ 19.7 ┆ 175.0 │
#> │ 15.0 ┆ 335.0 │
#> │ 21.4 ┆ 109.0 │
#> └──────┴───────┘Of course, we can chain those methods to create a pipeline:
dat$filter(
pl$col("cyl") == 6
)$select(
pl$col(c("mpg", "hp", "cyl"))
)
#> shape: (7, 3)
#> ┌──────┬───────┬─────┐
#> │ mpg ┆ hp ┆ cyl │
#> │ --- ┆ --- ┆ --- │
#> │ f64 ┆ f64 ┆ f64 │
#> ╞══════╪═══════╪═════╡
#> │ 21.0 ┆ 110.0 ┆ 6.0 │
#> │ 21.0 ┆ 110.0 ┆ 6.0 │
#> │ 21.4 ┆ 110.0 ┆ 6.0 │
#> │ 18.1 ┆ 105.0 ┆ 6.0 │
#> │ 19.2 ┆ 123.0 ┆ 6.0 │
#> │ 17.8 ┆ 123.0 ┆ 6.0 │
#> │ 19.7 ┆ 175.0 ┆ 6.0 │
#> └──────┴───────┴─────┘The $select() method that we introduced above also
supports data modification, so you can simultaneously transform it while
you are subsetting. However, the result will exclude any columns that
weren’t specified as part of the expression. To modify or add some
columns—whilst preserving all others in the dataset—it is therefore
better to use the $with_columns()
method. This next code chunk is equivalent to
mtcars |> dplyr::mutate(sum_mpg = sum(mpg), sum_hp = sum(hp), .by = cyl).
# Add the grouped sums of some selected columns.
dat$with_columns(
pl$col("mpg")$sum()$over("cyl")$alias("sum_mpg"),
pl$col("hp")$sum()$over("cyl")$alias("sum_hp")
)
#> shape: (32, 13)
#> ┌──────┬─────┬───────┬───────┬───┬──────┬──────┬─────────┬────────┐
#> │ mpg ┆ cyl ┆ disp ┆ hp ┆ … ┆ gear ┆ carb ┆ sum_mpg ┆ sum_hp │
#> │ --- ┆ --- ┆ --- ┆ --- ┆ ┆ --- ┆ --- ┆ --- ┆ --- │
#> │ f64 ┆ f64 ┆ f64 ┆ f64 ┆ ┆ f64 ┆ f64 ┆ f64 ┆ f64 │
#> ╞══════╪═════╪═══════╪═══════╪═══╪══════╪══════╪═════════╪════════╡
#> │ 21.0 ┆ 6.0 ┆ 160.0 ┆ 110.0 ┆ … ┆ 4.0 ┆ 4.0 ┆ 138.2 ┆ 856.0 │
#> │ 21.0 ┆ 6.0 ┆ 160.0 ┆ 110.0 ┆ … ┆ 4.0 ┆ 4.0 ┆ 138.2 ┆ 856.0 │
#> │ 22.8 ┆ 4.0 ┆ 108.0 ┆ 93.0 ┆ … ┆ 4.0 ┆ 1.0 ┆ 293.3 ┆ 909.0 │
#> │ 21.4 ┆ 6.0 ┆ 258.0 ┆ 110.0 ┆ … ┆ 3.0 ┆ 1.0 ┆ 138.2 ┆ 856.0 │
#> │ 18.7 ┆ 8.0 ┆ 360.0 ┆ 175.0 ┆ … ┆ 3.0 ┆ 2.0 ┆ 211.4 ┆ 2929.0 │
#> │ … ┆ … ┆ … ┆ … ┆ … ┆ … ┆ … ┆ … ┆ … │
#> │ 30.4 ┆ 4.0 ┆ 95.1 ┆ 113.0 ┆ … ┆ 5.0 ┆ 2.0 ┆ 293.3 ┆ 909.0 │
#> │ 15.8 ┆ 8.0 ┆ 351.0 ┆ 264.0 ┆ … ┆ 5.0 ┆ 4.0 ┆ 211.4 ┆ 2929.0 │
#> │ 19.7 ┆ 6.0 ┆ 145.0 ┆ 175.0 ┆ … ┆ 5.0 ┆ 6.0 ┆ 138.2 ┆ 856.0 │
#> │ 15.0 ┆ 8.0 ┆ 301.0 ┆ 335.0 ┆ … ┆ 5.0 ┆ 8.0 ┆ 211.4 ┆ 2929.0 │
#> │ 21.4 ┆ 4.0 ┆ 121.0 ┆ 109.0 ┆ … ┆ 4.0 ┆ 2.0 ┆ 293.3 ┆ 909.0 │
#> └──────┴─────┴───────┴───────┴───┴──────┴──────┴─────────┴────────┘For what it’s worth, the previous query could have been written more concisely as:
dat$with_columns(
pl$col(c("mpg", "hp"))$sum()$over("cyl")$name$prefix("sum_")
)
#> shape: (32, 13)
#> ┌──────┬─────┬───────┬───────┬───┬──────┬──────┬─────────┬────────┐
#> │ mpg ┆ cyl ┆ disp ┆ hp ┆ … ┆ gear ┆ carb ┆ sum_mpg ┆ sum_hp │
#> │ --- ┆ --- ┆ --- ┆ --- ┆ ┆ --- ┆ --- ┆ --- ┆ --- │
#> │ f64 ┆ f64 ┆ f64 ┆ f64 ┆ ┆ f64 ┆ f64 ┆ f64 ┆ f64 │
#> ╞══════╪═════╪═══════╪═══════╪═══╪══════╪══════╪═════════╪════════╡
#> │ 21.0 ┆ 6.0 ┆ 160.0 ┆ 110.0 ┆ … ┆ 4.0 ┆ 4.0 ┆ 138.2 ┆ 856.0 │
#> │ 21.0 ┆ 6.0 ┆ 160.0 ┆ 110.0 ┆ … ┆ 4.0 ┆ 4.0 ┆ 138.2 ┆ 856.0 │
#> │ 22.8 ┆ 4.0 ┆ 108.0 ┆ 93.0 ┆ … ┆ 4.0 ┆ 1.0 ┆ 293.3 ┆ 909.0 │
#> │ 21.4 ┆ 6.0 ┆ 258.0 ┆ 110.0 ┆ … ┆ 3.0 ┆ 1.0 ┆ 138.2 ┆ 856.0 │
#> │ 18.7 ┆ 8.0 ┆ 360.0 ┆ 175.0 ┆ … ┆ 3.0 ┆ 2.0 ┆ 211.4 ┆ 2929.0 │
#> │ … ┆ … ┆ … ┆ … ┆ … ┆ … ┆ … ┆ … ┆ … │
#> │ 30.4 ┆ 4.0 ┆ 95.1 ┆ 113.0 ┆ … ┆ 5.0 ┆ 2.0 ┆ 293.3 ┆ 909.0 │
#> │ 15.8 ┆ 8.0 ┆ 351.0 ┆ 264.0 ┆ … ┆ 5.0 ┆ 4.0 ┆ 211.4 ┆ 2929.0 │
#> │ 19.7 ┆ 6.0 ┆ 145.0 ┆ 175.0 ┆ … ┆ 5.0 ┆ 6.0 ┆ 138.2 ┆ 856.0 │
#> │ 15.0 ┆ 8.0 ┆ 301.0 ┆ 335.0 ┆ … ┆ 5.0 ┆ 8.0 ┆ 211.4 ┆ 2929.0 │
#> │ 21.4 ┆ 4.0 ┆ 121.0 ┆ 109.0 ┆ … ┆ 4.0 ┆ 2.0 ┆ 293.3 ┆ 909.0 │
#> └──────┴─────┴───────┴───────┴───┴──────┴──────┴─────────┴────────┘Similarly, here’s how we could have aggregated (i.e., collapsed) the
dataset by groups instead of modifying them. We need simply invoke the
$group_by() and $agg()
methods.
dat$group_by(
"cyl",
maintain_order = TRUE
)$agg(
pl$col(c("mpg", "hp"))$sum()
)
#> shape: (3, 3)
#> ┌─────┬───────┬────────┐
#> │ cyl ┆ mpg ┆ hp │
#> │ --- ┆ --- ┆ --- │
#> │ f64 ┆ f64 ┆ f64 │
#> ╞═════╪═══════╪════════╡
#> │ 6.0 ┆ 138.2 ┆ 856.0 │
#> │ 4.0 ┆ 293.3 ┆ 909.0 │
#> │ 8.0 ┆ 211.4 ┆ 2929.0 │
#> └─────┴───────┴────────┘(arg maintain_order = TRUE is optional, since
polars doesn’t sort the results of grouped operations
by default. This is similar to what data.table does and
is also true for newer versions of dplyr.)
The same principles of method chaining can be combined very flexibly to group by multiple variables and aggregation types.
dat$group_by(
"cyl",
manual = pl$col("am")$cast(pl$Boolean)
)$agg(
mean_mpg = pl$col("mpg")$mean(),
med_hp = pl$col("hp")$median()
)
#> shape: (6, 4)
#> ┌─────┬────────┬───────────┬────────┐
#> │ cyl ┆ manual ┆ mean_mpg ┆ med_hp │
#> │ --- ┆ --- ┆ --- ┆ --- │
#> │ f64 ┆ bool ┆ f64 ┆ f64 │
#> ╞═════╪════════╪═══════════╪════════╡
#> │ 8.0 ┆ false ┆ 15.05 ┆ 180.0 │
#> │ 4.0 ┆ true ┆ 28.075 ┆ 78.5 │
#> │ 6.0 ┆ false ┆ 19.125 ┆ 116.5 │
#> │ 6.0 ┆ true ┆ 20.566667 ┆ 110.0 │
#> │ 4.0 ┆ false ┆ 22.9 ┆ 95.0 │
#> │ 8.0 ┆ true ┆ 15.4 ┆ 299.5 │
#> └─────┴────────┴───────────┴────────┘Note that we used the $cast() method to convert the data
type of the am column. See the section below for more
details on data types.
Polars supports data reshaping, going from both long to wide (a.k.a.
“pivotting”, or pivot_wider() in tidyr), and
from wide to long (a.k.a. “unpivotting”, “melting”, or
pivot_longer() in tidyr). Let’s switch to the
Indometh dataset to demonstrate some basic examples. Note
that the data are currently in long format.
indo = as_polars_df(Indometh)
indo
#> shape: (66, 3)
#> ┌─────────┬──────┬──────┐
#> │ Subject ┆ time ┆ conc │
#> │ --- ┆ --- ┆ --- │
#> │ cat ┆ f64 ┆ f64 │
#> ╞═════════╪══════╪══════╡
#> │ 1 ┆ 0.25 ┆ 1.5 │
#> │ 1 ┆ 0.5 ┆ 0.94 │
#> │ 1 ┆ 0.75 ┆ 0.78 │
#> │ 1 ┆ 1.0 ┆ 0.48 │
#> │ 1 ┆ 1.25 ┆ 0.37 │
#> │ … ┆ … ┆ … │
#> │ 6 ┆ 3.0 ┆ 0.24 │
#> │ 6 ┆ 4.0 ┆ 0.17 │
#> │ 6 ┆ 5.0 ┆ 0.13 │
#> │ 6 ┆ 6.0 ┆ 0.1 │
#> │ 6 ┆ 8.0 ┆ 0.09 │
#> └─────────┴──────┴──────┘To go from long to wide, we use the $pivot() method.
Here we pivot the data so that every subject takes its own column.
indo_wide = indo$pivot(values = "conc", index = "time", on = "Subject")
indo_wide
#> shape: (11, 7)
#> ┌──────┬──────┬──────┬──────┬──────┬──────┬──────┐
#> │ time ┆ 1 ┆ 2 ┆ 3 ┆ 4 ┆ 5 ┆ 6 │
#> │ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- │
#> │ f64 ┆ f64 ┆ f64 ┆ f64 ┆ f64 ┆ f64 ┆ f64 │
#> ╞══════╪══════╪══════╪══════╪══════╪══════╪══════╡
#> │ 0.25 ┆ 1.5 ┆ 2.03 ┆ 2.72 ┆ 1.85 ┆ 2.05 ┆ 2.31 │
#> │ 0.5 ┆ 0.94 ┆ 1.63 ┆ 1.49 ┆ 1.39 ┆ 1.04 ┆ 1.44 │
#> │ 0.75 ┆ 0.78 ┆ 0.71 ┆ 1.16 ┆ 1.02 ┆ 0.81 ┆ 1.03 │
#> │ 1.0 ┆ 0.48 ┆ 0.7 ┆ 0.8 ┆ 0.89 ┆ 0.39 ┆ 0.84 │
#> │ 1.25 ┆ 0.37 ┆ 0.64 ┆ 0.8 ┆ 0.59 ┆ 0.3 ┆ 0.64 │
#> │ … ┆ … ┆ … ┆ … ┆ … ┆ … ┆ … │
#> │ 3.0 ┆ 0.12 ┆ 0.32 ┆ 0.22 ┆ 0.16 ┆ 0.13 ┆ 0.24 │
#> │ 4.0 ┆ 0.11 ┆ 0.2 ┆ 0.12 ┆ 0.11 ┆ 0.11 ┆ 0.17 │
#> │ 5.0 ┆ 0.08 ┆ 0.25 ┆ 0.11 ┆ 0.1 ┆ 0.08 ┆ 0.13 │
#> │ 6.0 ┆ 0.07 ┆ 0.12 ┆ 0.08 ┆ 0.07 ┆ 0.1 ┆ 0.1 │
#> │ 8.0 ┆ 0.05 ┆ 0.08 ┆ 0.08 ┆ 0.07 ┆ 0.06 ┆ 0.09 │
#> └──────┴──────┴──────┴──────┴──────┴──────┴──────┘To go from wide to long, we use the $melt() method.
# indo_wide$melt(id_vars = "time") # default column names are "variable" and "value"
indo_wide$unpivot(index = "time", variable_name = "subject", value_name = "conc")
#> shape: (66, 3)
#> ┌──────┬─────────┬──────┐
#> │ time ┆ subject ┆ conc │
#> │ --- ┆ --- ┆ --- │
#> │ f64 ┆ str ┆ f64 │
#> ╞══════╪═════════╪══════╡
#> │ 0.25 ┆ 1 ┆ 1.5 │
#> │ 0.5 ┆ 1 ┆ 0.94 │
#> │ 0.75 ┆ 1 ┆ 0.78 │
#> │ 1.0 ┆ 1 ┆ 0.48 │
#> │ 1.25 ┆ 1 ┆ 0.37 │
#> │ … ┆ … ┆ … │
#> │ 3.0 ┆ 6 ┆ 0.24 │
#> │ 4.0 ┆ 6 ┆ 0.17 │
#> │ 5.0 ┆ 6 ┆ 0.13 │
#> │ 6.0 ┆ 6 ┆ 0.1 │
#> │ 8.0 ┆ 6 ┆ 0.09 │
#> └──────┴─────────┴──────┘Basic functionality aside, it should be noted that
$pivot() can perform aggregations as part of the reshaping
operation. This is useful when you have multiple observations per ID
variable that need to be collapsed into unique values. The aggregating
functions can be arbitrarily complex, but let’s consider a relatively
simple example using our dat (“mtcars”) DataFrame from
earlier: what is the median MPG value (mpg) across
cylinders (cyl), cut by different combinations of
transmission type (am) and engine shape
(vs)?
dat$pivot(
values = "mpg",
index = c("am", "vs"),
on = "cyl",
aggregate_function = "median" # aggregating function
)
#> shape: (4, 5)
#> ┌─────┬─────┬───────┬──────┬──────┐
#> │ am ┆ vs ┆ 6.0 ┆ 4.0 ┆ 8.0 │
#> │ --- ┆ --- ┆ --- ┆ --- ┆ --- │
#> │ f64 ┆ f64 ┆ f64 ┆ f64 ┆ f64 │
#> ╞═════╪═════╪═══════╪══════╪══════╡
#> │ 1.0 ┆ 0.0 ┆ 21.0 ┆ 26.0 ┆ 15.4 │
#> │ 1.0 ┆ 1.0 ┆ null ┆ 30.4 ┆ null │
#> │ 0.0 ┆ 1.0 ┆ 18.65 ┆ 22.8 ┆ null │
#> │ 0.0 ┆ 0.0 ┆ null ┆ null ┆ 15.2 │
#> └─────┴─────┴───────┴──────┴──────┘Here, "median" is a convenience string that is
equivalent to the more verbose pl$element()$median(). Other
convenience strings include "first", "last",
"min", "max", "mean",
"sum", and "count".
As a final example of how polars can be used for standard data wrangling tasks, let’s implement a (left) join. For this example, we’ll borrow some datasets from the nycflights13 package.
data("flights", "planes", package = "nycflights13")
flights = as_polars_df(flights)
planes = as_polars_df(planes)
flights$join(
planes,
on = "tailnum",
how = "left"
)
#> shape: (336_776, 27)
#> ┌──────┬───────┬─────┬──────────┬───┬─────────┬───────┬───────┬───────────┐
#> │ year ┆ month ┆ day ┆ dep_time ┆ … ┆ engines ┆ seats ┆ speed ┆ engine │
#> │ --- ┆ --- ┆ --- ┆ --- ┆ ┆ --- ┆ --- ┆ --- ┆ --- │
#> │ i32 ┆ i32 ┆ i32 ┆ i32 ┆ ┆ i32 ┆ i32 ┆ i32 ┆ str │
#> ╞══════╪═══════╪═════╪══════════╪═══╪═════════╪═══════╪═══════╪═══════════╡
#> │ 2013 ┆ 1 ┆ 1 ┆ 517 ┆ … ┆ 2 ┆ 149 ┆ null ┆ Turbo-fan │
#> │ 2013 ┆ 1 ┆ 1 ┆ 533 ┆ … ┆ 2 ┆ 149 ┆ null ┆ Turbo-fan │
#> │ 2013 ┆ 1 ┆ 1 ┆ 542 ┆ … ┆ 2 ┆ 178 ┆ null ┆ Turbo-fan │
#> │ 2013 ┆ 1 ┆ 1 ┆ 544 ┆ … ┆ 2 ┆ 200 ┆ null ┆ Turbo-fan │
#> │ 2013 ┆ 1 ┆ 1 ┆ 554 ┆ … ┆ 2 ┆ 178 ┆ null ┆ Turbo-fan │
#> │ … ┆ … ┆ … ┆ … ┆ … ┆ … ┆ … ┆ … ┆ … │
#> │ 2013 ┆ 9 ┆ 30 ┆ null ┆ … ┆ null ┆ null ┆ null ┆ null │
#> │ 2013 ┆ 9 ┆ 30 ┆ null ┆ … ┆ null ┆ null ┆ null ┆ null │
#> │ 2013 ┆ 9 ┆ 30 ┆ null ┆ … ┆ null ┆ null ┆ null ┆ null │
#> │ 2013 ┆ 9 ┆ 30 ┆ null ┆ … ┆ null ┆ null ┆ null ┆ null │
#> │ 2013 ┆ 9 ┆ 30 ┆ null ┆ … ┆ null ┆ null ┆ null ┆ null │
#> └──────┴───────┴─────┴──────────┴───┴─────────┴───────┴───────┴───────────┘More information on the polars joining method can be found in the reference manual.
The package supports many other data manipulation operations, which we won’t cover here. Hopefully, you will already have a sense of the key syntax features. We now turn to another core idea of the Polars ecosystem: lazy execution.
While the “eager” execution engine of polars works perfectly well—as evidenced by all of the previous examples—to get the most out of the package you need to go lazy. Lazy execution enables several benefits, but the most important is that it improves performance. Delaying execution until the last possible moment allows Polars to apply automatic optimization to every query. Let’s take a quick look.
To create a so-called “LazyFrame”
from an existing object in memory, we can invoke the
$lazy() method.
ldat = dat$lazy()
ldat
#> polars LazyFrame
#> $explain(): Show the optimized query plan.
#>
#> Naive plan:
#> DF ["mpg", "cyl", "disp", "hp"]; PROJECT */11 COLUMNS; SELECTION: NoneOr, use the as_polars_lf() generic function.
as_polars_lf(dat)
#> polars LazyFrame
#> $explain(): Show the optimized query plan.
#>
#> Naive plan:
#> DF ["mpg", "cyl", "disp", "hp"]; PROJECT */11 COLUMNS; SELECTION: NoneNow consider what happens when we run our subsetting query from earlier on this LazyFrame.
subset_query = ldat$filter(
pl$col("cyl") == 6
)$select(
pl$col(c("mpg", "hp", "cyl"))
)
subset_query
#> polars LazyFrame
#> $explain(): Show the optimized query plan.
#>
#> Naive plan:
#> SELECT [col("mpg"), col("hp"), col("cyl")] FROM
#> FILTER [(col("cyl")) == (6.0)] FROM
#> DF ["mpg", "cyl", "disp", "hp"]; PROJECT */11 COLUMNS; SELECTION: NoneRight now we only have a tree of instructions. But underneath the hood, Polars has already worked out a more optimized version of the query. We can view this optimized plan this by requesting it.
cat(subset_query$explain())
#> DF ["mpg", "cyl", "disp", "hp"]; PROJECT 3/11 COLUMNS; SELECTION: [(col("cyl")) == (6.0)]Here we see a simple, but surprisingly effective component in query optimization: projection. Changing the order in which our subsetting operations occurs—in this case, subsetting on columns first—reduces the memory overhead of the overall query and leads to a downstream speedup. Of course, you would hardly notice a difference for this small dataset. But the same principles carry over to much bigger datasets and more complex queries.
To actually execute the plan, we just need to invoke the
$collect() method. This should feel very familiar if you
have previously used other lazy execution engines like those provided by
arrow or dbplyr.
subset_query$collect()
#> shape: (7, 3)
#> ┌──────┬───────┬─────┐
#> │ mpg ┆ hp ┆ cyl │
#> │ --- ┆ --- ┆ --- │
#> │ f64 ┆ f64 ┆ f64 │
#> ╞══════╪═══════╪═════╡
#> │ 21.0 ┆ 110.0 ┆ 6.0 │
#> │ 21.0 ┆ 110.0 ┆ 6.0 │
#> │ 21.4 ┆ 110.0 ┆ 6.0 │
#> │ 18.1 ┆ 105.0 ┆ 6.0 │
#> │ 19.2 ┆ 123.0 ┆ 6.0 │
#> │ 17.8 ┆ 123.0 ┆ 6.0 │
#> │ 19.7 ┆ 175.0 ┆ 6.0 │
#> └──────┴───────┴─────┘polars supports data import of both CSV and Parquet
files formats. Here we demonstrate using the airquality
dataset that also comes bundled with base R.
write.csv(airquality, "airquality.csv", row.names = FALSE)
pl$read_csv("airquality.csv")
#> shape: (153, 6)
#> ┌───────┬─────────┬──────┬──────┬───────┬─────┐
#> │ Ozone ┆ Solar.R ┆ Wind ┆ Temp ┆ Month ┆ Day │
#> │ --- ┆ --- ┆ --- ┆ --- ┆ --- ┆ --- │
#> │ str ┆ str ┆ f64 ┆ i64 ┆ i64 ┆ i64 │
#> ╞═══════╪═════════╪══════╪══════╪═══════╪═════╡
#> │ 41 ┆ 190 ┆ 7.4 ┆ 67 ┆ 5 ┆ 1 │
#> │ 36 ┆ 118 ┆ 8.0 ┆ 72 ┆ 5 ┆ 2 │
#> │ 12 ┆ 149 ┆ 12.6 ┆ 74 ┆ 5 ┆ 3 │
#> │ 18 ┆ 313 ┆ 11.5 ┆ 62 ┆ 5 ┆ 4 │
#> │ NA ┆ NA ┆ 14.3 ┆ 56 ┆ 5 ┆ 5 │
#> │ … ┆ … ┆ … ┆ … ┆ … ┆ … │
#> │ 30 ┆ 193 ┆ 6.9 ┆ 70 ┆ 9 ┆ 26 │
#> │ NA ┆ 145 ┆ 13.2 ┆ 77 ┆ 9 ┆ 27 │
#> │ 14 ┆ 191 ┆ 14.3 ┆ 75 ┆ 9 ┆ 28 │
#> │ 18 ┆ 131 ┆ 8.0 ┆ 76 ┆ 9 ┆ 29 │
#> │ 20 ┆ 223 ┆ 11.5 ┆ 68 ┆ 9 ┆ 30 │
#> └───────┴─────────┴──────┴──────┴───────┴─────┘Again, however, the package works best if we take the lazy approach.
pl$scan_csv("airquality.csv")
#> polars LazyFrame
#> $explain(): Show the optimized query plan.
#>
#> Naive plan:
#> Csv SCAN [airquality.csv]
#> PROJECT */6 COLUMNSWe could obviously append a set of query operators to the above LazyFrame and then collect the results. However, this workflow is even better suited to Parquet files, since we can leverage their efficient storage format on disk. Let’s see an example.
as_polars_df(airquality)$write_parquet("airquality.parquet")
# eager version (okay)
aq_collected = pl$read_parquet("airquality.parquet")
# lazy version (better)
aq = pl$scan_parquet("airquality.parquet")
aq$filter(
pl$col("Month") <= 6
)$group_by(
"Month"
)$agg(
pl$col(c("Ozone", "Temp"))$mean()
)$collect()
#> shape: (2, 3)
#> ┌───────┬───────────┬───────────┐
#> │ Month ┆ Ozone ┆ Temp │
#> │ --- ┆ --- ┆ --- │
#> │ i32 ┆ f64 ┆ f64 │
#> ╞═══════╪═══════════╪═══════════╡
#> │ 6 ┆ 29.444444 ┆ 79.1 │
#> │ 5 ┆ 23.615385 ┆ 65.548387 │
#> └───────┴───────────┴───────────┘Finally, we can read/scan multiple files in the same directory through pattern globbing.
dir.create("airquality-ds")
# Create a hive-partitioned dataset with the function from the arrow package
arrow::write_dataset(airquality, "airquality-ds", partitioning = "Month")
# Use pattern globbing to scan all parquet files in the folder
aq2 = pl$scan_parquet("airquality-ds/**/*.parquet")
# Scan the first two rows
aq2$fetch(2)
#> shape: (2, 5)
#> ┌───────┬─────────┬──────┬──────┬─────┐
#> │ Ozone ┆ Solar.R ┆ Wind ┆ Temp ┆ Day │
#> │ --- ┆ --- ┆ --- ┆ --- ┆ --- │
#> │ i32 ┆ i32 ┆ f64 ┆ i32 ┆ i32 │
#> ╞═══════╪═════════╪══════╪══════╪═════╡
#> │ 41 ┆ 190 ┆ 7.4 ┆ 67 ┆ 1 │
#> │ 36 ┆ 118 ┆ 8.0 ┆ 72 ┆ 2 │
#> └───────┴─────────┴──────┴──────┴─────┘Before continuing, don’t forget to clean up by removing the newly created temp files and directory on disk.
It is possible to mix R code with Polars by passing R functions to polars. This can unlock a lot of flexibility, but note that it can inhibit performance. R functions are typically slower, so we recommend using native Polars functions and expressions wherever possible.
as_polars_df(iris)$select(
pl$col("Sepal.Length")$map_batches(\(s) { # map with a R function
x = as.vector(s) # convert from Polars Series to a native R vector
x[x >= 5] = 10
x[1:10] # if return is R vector, it will automatically be converted to Polars Series again
})
) |>
as.data.frame()
#> Sepal.Length
#> 1 10.0
#> 2 4.9
#> 3 4.7
#> 4 4.6
#> 5 10.0
#> 6 10.0
#> 7 4.6
#> 8 10.0
#> 9 4.4
#> 10 4.9Polars is strongly
typed and new types can be created with the dtypes
constructor. For example:
The full list of valid Polars types can be found by typing
pl$dtypes into your R console. These include
Boolean, Float32(64), Int32(64),
Utf8, Categorical, Date, etc. Note that some
type names differ from what they are called in R (e.g., Boolean
in Polars is equivalent to logical() in R). This might
occasionally require you to look up a specific type. But the good news
is that polars generally does a good job of inferring
types automatically.
Similar to how (most) data.table
operations are limited to objects of class data.table, we
can only perform polars operations on objects that have been converted
to an appropriate polars class. Later on, we’ll see how
to read data from disk directly in Polars format.↩︎