• Stars
    star
    150
  • Rank 241,663 (Top 5 %)
  • Language
    R
  • License
    Other
  • Created over 6 years ago
  • Updated 3 months ago

Reviews

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Repository Details

Coroutines for R

coro

CRAN status R build status

Overview

coro implements coroutines for R, i.e.Β functions that can be suspended and resumed later on. There are two kinds:

  • Async functions, which make it straightforward to program concurrently
  • Generators for iterating over complex sequences

Supported features:

  • Suspending within loops and if/else branches
  • Suspending within tryCatch()
  • on.exit() expressions and stack-based cleanup such as provided by local_ functions in the withr package
  • Step-debugging and browser() within coroutines

Compatibility with:

  • Python iterators from the reticulate package
  • Async operations from the promises package
  • Parallel computations from the future package

Attach the package to follow the examples:

library(coro)

Async/await functions

Concurrent programming is made straightforward by async-await functions. Whenever you are waiting for a result that may take a while (downloading a file, computing a value in an external process), use await(). The argument to await() must return a promise from the promises package.

Concurrent code based on promises can quickly become hard to write and follow. In the following artificial example, we wait for a download to complete, then decide to launch a computation in an external process depending on a property of the downloaded data. We also handle some errors specifically.

my_async <- function() {
  async_download() %>%
    then(function(data) {
      if (ncol(data) > 10) {
        then(future::future(fib(30)), function(fib) {
          data / fib
        })
      } else {
        data
      }
    }, onRejected = function(err) {
      if (inherits(err, "download_error")) {
        NULL
      } else {
        stop(err)
      }
    })
}

Rewriting this function with async/await greatly simplifies the code:

my_async <- async(function() {
  data <- tryCatch(
    await(async_download()),
    download_error = function(err) NULL
  )

  if (is.null(data)) {
    return(NULL)
  }

  if (ncol(data) > 10) {
    fib <- await(future::future(fib(30)))
    data <- data /fib
  }

  data
})

Generators

Generators are based on a simple iteration protocol:

  • Iterators are functions.
  • They can be advanced by calling the function. The new value is returned.
  • An exhausted iterator returns the sentinel symbol exhausted.

The generator() function creates a generator factory which returns generator instances:

# Create a generator factory
generate_abc <- generator(function() {
  for (x in letters[1:3]) {
    yield(x)
  }
})

# Create a generator instance
abc <- generate_abc()

A generator instance is an iterator function which yields values:

abc
#> <generator/instance>
#> function() {
#>   for (x in letters[1:3]) {
#>     yield(x)
#>   }
#> }

abc()
#> [1] "a"

Collect all remaining values from an iterator with collect():

collect(abc)
#> [[1]]
#> [1] "b"
#> 
#> [[2]]
#> [1] "c"

Iterate over an iterator with loop():

loop(for (x in generate_abc()) {
  print(toupper(x))
})
#> [1] "A"
#> [1] "B"
#> [1] "C"

See vignette("generator") for more information.

Compatibility with the reticulate package

Python iterators imported with the reticulate package are compatible with loop() and collect():

suppressMessages(library(reticulate))

py_run_string("
def first_n(n):
    num = 1
    while num <= n:
        yield num
        num += 1
")

loop(for (x in py$first_n(3)) {
  print(x * 2)
})
#> [1] 2
#> [1] 4
#> [1] 6

They can also be composed with coro generators:

times <- generator(function(it, n) for (x in it) yield(x * n))

composed <- times(py$first_n(3), 10)

collect(composed)
#> [[1]]
#> [1] 10
#> 
#> [[2]]
#> [1] 20
#> 
#> [[3]]
#> [1] 30

Limitations

yield() and await() can be used in loops, if/else branches, tryCatch() expressions, or any combinations of these. However they can’t be used as function arguments. These will cause errors:

generator(function() {
  list(yield("foo"))
})

async(function() {
  list(await(foo()))
})

Fortunately it is easy to rewrite the code to work around this limitation:

generator(function() {
  x <- yield("foo")
  list(x)
})

async(function() {
  x <- await(foo())
  list(x)
})

How does it work

Coroutines are an abstraction for state machines in languages that support them. Conversely, you can implement coroutines by rewriting the code source provided by the user as a state machine. Pass internals = TRUE to the print methods of coroutines to reveal the state machine that is running under the hood:

print(generate_abc, internals = TRUE)
#> <generator>
#> function() {
#>   for (x in letters[1:3]) {
#>     yield(x)
#>   }
#> }
#> State machine:
#> {
#>     if (exhausted) {
#>         return(invisible(exhausted()))
#>     }
#>     repeat switch(state[[1L]], `1` = {
#>         iterators[[2L]] <- as_iterator(user(letters[1:3]))
#>         state[[1L]] <- 2L
#>         state[[2L]] <- 1L
#>     }, `2` = {
#>         repeat switch(state[[2L]], `1` = {
#>             if ({
#>                 iterator <- iterators[[2L]]
#>                 if (is_exhausted(elt <- iterator())) {
#>                   FALSE
#>                 } else {
#>                   user_env[["x"]] <- elt
#>                   TRUE
#>                 }
#>             }) {
#>                 state[[2L]] <- 2L
#>             } else {
#>                 break
#>             }
#>         }, `2` = {
#>             user({
#>                 x
#>             })
#>             state[[2L]] <- 3L
#>             suspend()
#>             return(last_value())
#>         }, `3` = {
#>             .last_value <- if (missing(arg)) NULL else arg
#>             state[[2L]] <- 1L
#>         })
#>         iterators[[2L]] <- NULL
#>         length(state) <- 1L
#>         break
#>     })
#>     exhausted <- TRUE
#>     invisible(exhausted())
#> }

Despite this transformation of source code, browser() and step-debugging still work as you would expect. This is because coro keeps track of the source references from the original code.

Acknowledgements

  • The regenerator Javascript package which uses a similar transformation to implement generators and async functions in older versions of Javascript.

  • Gabor Csardi for many interesting discussions about concurrency and the design of coro.

Installation

Install the development version from github with:

# install.packages("devtools")
devtools::install_github("r-lib/coro", build_vignettes = TRUE)

More Repositories

1

devtools

Tools to make an R developer's life easier
R
2,360
star
2

lintr

Static Code Analysis for R
R
1,159
star
3

httr

httr: a friendly http package for R
R
982
star
4

actions

GitHub Actions for the R community
TypeScript
917
star
5

testthat

An R πŸ“¦ to make testing πŸ˜€
R
867
star
6

usethis

Set up commonly used πŸ“¦ components
R
821
star
7

pkgdown

Generate static html documentation for an R package
R
703
star
8

styler

Non-invasive pretty printing of R code
R
692
star
9

pak

A fresh approach to package installation
C
630
star
10

cli

Tools for making beautiful & useful command line interfaces
R
618
star
11

roxygen2

Generate R package documentation from inline R comments
R
579
star
12

rig

The R Installation Manager
Rust
555
star
13

rlang

Low-level API for programming with R
R
484
star
14

progress

Progress bar in your R terminal
R
457
star
15

R6

Encapsulated object-oriented programming for R
R
402
star
16

here

A simpler way to find your files
R
401
star
17

scales

Tools for ggplot2 scales
R
388
star
18

fs

Provide cross platform file operations based on libuv.
C
361
star
19

rex

Friendly regular expressions for R.
R
331
star
20

covr

Test coverage reports for R
R
329
star
21

crayon

πŸ–οΈ R package for colored terminal output β€” now superseded by cli
R
322
star
22

remotes

Install R packages from GitHub, GitLab, Bitbucket, git, svn repositories, URLs
R
321
star
23

memoise

Easy memoisation for R
R
314
star
24

lobstr

Understanding complex R objects with tools similar to str()
R
296
star
25

slider

Sliding Window Functions
R
290
star
26

callr

Call R from R
R
289
star
27

vctrs

Generic programming with typed R vectors
C
280
star
28

waldo

Find differences between R objects
R
275
star
29

zeallot

Variable assignment with zeal! (or multiple, unpacking, and destructuring assignment in R)
R
250
star
30

conflicted

An alternative conflict resolution strategy for R
R
243
star
31

bench

High Precision Timing of R Expressions
R
242
star
32

gmailr

Access the Gmail RESTful API from R.
R
230
star
33

processx

Execute and Control Subprocesses from R
R
227
star
34

httr2

Make HTTP requests and process their responses. A modern reimagining of httr.
R
223
star
35

asciicast

Turn R scripts into terminal screencasts
R
223
star
36

xml2

Bindings to libxml2
R
218
star
37

gh

Minimalistic GitHub API client in R
R
217
star
38

cpp11

cpp11 helps you to interact with R objects using C++ code.
C++
192
star
39

keyring

πŸ” Access the system credential store from R
R
188
star
40

vdiffr

Visual regression testing and graphical diffing with testthat
C++
181
star
41

svglite

A lightweight svg graphics device for R
C++
178
star
42

pillar

Format columns with colour
R
173
star
43

ragg

Graphic Devices Based on AGG
C++
170
star
44

withr

Methods For Temporarily Modifying Global State
R
169
star
45

hugodown

Make websites with hugo and RMarkdown
R
165
star
46

ymlthis

write YAML for R Markdown, bookdown, blogdown, and more
R
163
star
47

rprojroot

Finding files in project subdirectories
R
147
star
48

debugme

Easy and efficient debugging for R packages
R
145
star
49

available

Check if a package name is available to use
R
142
star
50

archive

R bindings to libarchive, supporting a large variety of archive formats
C++
141
star
51

ellipsis

Tools for Working with ...
R
139
star
52

later

Schedule an R function or formula to run after a specified period of time.
C++
136
star
53

gert

Simple git client for R
C
136
star
54

itdepends

R
133
star
55

rray

Simple Arrays
R
130
star
56

isoband

isoband: An R package to generate contour lines and polygons.
C++
130
star
57

prettyunits

Pretty, human readable formatting of quantities
JavaScript
128
star
58

fastmap

Fast map implementation for R
C++
127
star
59

desc

Manipulate DESCRIPTION files
R
121
star
60

tidyselect

A backend for functions taking tidyverse selections
R
121
star
61

gargle

Infrastructure for calling Google APIs from R, including auth
R
113
star
62

rcmdcheck

Run R CMD check from R and collect the results
R
109
star
63

evaluate

A version of eval for R that returns more information about what happened
R
108
star
64

prettycode

Syntax highlight R code in the terminal
R
100
star
65

revdepcheck

R package reverse dependency checking
R
99
star
66

sloop

S language OOP ⛡️
R
99
star
67

mockery

A mocking library for R.
R
98
star
68

tree-sitter-r

C
95
star
69

clock

A Date-Time Library for R
R
95
star
70

pkgdepends

R Package Dependency Resolution
R
93
star
71

systemfonts

System Native Font Handling in R
C++
91
star
72

lifecycle

Manage the life cycle of your exported functions and arguments
R
90
star
73

gtable

The layout packages that powers ggplot2
R
85
star
74

askpass

Password Entry for R, Git, and SSH
R
83
star
75

commonmark

High Performance CommonMark and Github Markdown Rendering in R
C
83
star
76

zip

Platform independent zip compression via miniz
C
82
star
77

rappdirs

Find OS-specific directories to store data, caches, and logs. A port of python's AppDirs
R
81
star
78

downlit

Syntax Highlighting and Automatic Linking
R
81
star
79

clisymbols

Unicode symbols for CLI applications, with fallbacks
R
76
star
80

ps

R package to query, list, manipulate system processes
C
72
star
81

sessioninfo

Print Session Information
R
72
star
82

pkgapi

Create a map of functions for an R package - WORK IN PROGRESS!
R
70
star
83

credentials

Tools for Managing SSH and Git Credentials
R
70
star
84

sodium

R bindings to libsodium
R
68
star
85

roxygen2md

Convert elements of roxygen documentation to markdown
R
67
star
86

backports

Reimplementations of Functions Introduced Since R-3.0.0
R
65
star
87

pkgbuild

Find tools needed to build R packages
R
65
star
88

webfakes

Fake web apps for HTTP testing R packages
C
61
star
89

cliapp

Rich Command Line Applications
R
60
star
90

generics

Common generic methods
R
59
star
91

diffviewer

HTML widget to visually compare files
JavaScript
57
star
92

liteq

Serverless R message queue using SQLite
R
56
star
93

pkgload

Simulate installing and loading a package
R
54
star
94

cachem

Key-value caches for R
R
53
star
95

carrier

Create standalone functions for remote execution
R
50
star
96

brio

Basic R Input Output
R
50
star
97

marquee

Markdown Parser and Renderer for R Graphics
C
49
star
98

jose

Javascript Object Signing and Encryption for R
R
48
star
99

urlchecker

Run CRAN URL checks from older versions of R
R
45
star
100

pkgconfig

Private configuration for R packages
R
41
star