httr2 is a ground-up rewrite of httr that provides a pipeable API with an explicit request object that solves more problems felt by packages that wrap APIs (e.g. built-in rate-limiting, retries, OAuth, secure secrets, and more).
You can install httr from CRAN with:
install.packages("httr2")
To use httr2, start by creating a request:
library(httr2)
<- request("https://r-project.org")
req
req#> <httr2_request>
#> GET https://r-project.org
#> Body: empty
You can tailor this request with the req_
family of
functions:
# Add custom headers
%>% req_headers("Accept" = "application/json")
req #> <httr2_request>
#> GET https://r-project.org
#> Headers:
#> • Accept: 'application/json'
#> Body: empty
# Add a body, turning it into a POST
%>% req_body_json(list(x = 1, y = 2))
req #> <httr2_request>
#> POST https://r-project.org
#> Body: json encoded data
# Automatically retry if the request fails
%>% req_retry(max_tries = 5)
req #> <httr2_request>
#> GET https://r-project.org
#> Body: empty
#> Policies:
#> • retry_max_tries: 5
# Change the HTTP method
%>% req_method("PATCH")
req #> <httr2_request>
#> PATCH https://r-project.org
#> Body: empty
And see exactly what httr2 will send to the server with
req_dry_run()
:
%>% req_dry_run()
req #> GET / HTTP/1.1
#> Host: r-project.org
#> User-Agent: httr2/0.1.1 r-curl/4.3.2 libcurl/7.79.1
#> Accept: */*
#> Accept-Encoding: deflate, gzip
Use req_perform()
to perform the request, retrieving a
response:
<- req_perform(req)
resp
resp#> <httr2_response>
#> GET https://www.r-project.org/
#> Status: 200 OK
#> Content-Type: text/html
#> Body: In memory (6158 bytes)
The resp_
functions help you extract various useful
components of the response:
%>% resp_content_type()
resp #> [1] "text/html"
%>% resp_status_desc()
resp #> [1] "OK"
%>% resp_body_html()
resp #> {html_document}
#> <html lang="en">
#> [1] <head>\n<meta http-equiv="Content-Type" content="text/html; charset=UTF-8 ...
#> [2] <body>\n <div class="container page">\n <div class="row">\n ...
You can now create and modify a request without performing it.
This means that there’s now a single function to perform the request and
fetch the result: req_perform()
. (If you want to handle the
response as it streams in, use req_stream()
instead).
req_perform()
replaces httr::GET()
,
httr::POST()
, httr::DELETE()
, and
more.
HTTP errors are automatically converted into R errors. Use
req_error()
to override the defaults (which turn all 4xx
and 5xx responses into errors) or to add additional details to the error
message.
You can automatically retry if the request fails or encounters a
transient HTTP error (e.g. a 429 rate limit request).
req_retry()
defines the maximum number of retries, which
errors are transient, and how long to wait between tries.
OAuth support has been totally overhauled to directly support many more flows and to make it much easier to both customise the built-in flows and to create your own.
You can manage secrets (often needed for testing) with
secret_encrypt()
and friends. You can obfuscate mildly
confidential data with obfuscate()
, preventing it from
being scraped from published code.
You can automatically cache all cacheable results with
req_cache()
. Relatively few API responses are cacheable,
but when they are it typically makes a big difference.
httr2 wouldn’t be possible without curl, openssl, jsonlite, and jose, which are all maintained by Jeroen Ooms. A big thanks also go to Jenny Bryan and Craig Citro who have given me much useful feedback on both the design of the internals and the user facing API.