The following aims to help you
decide how to use crul
in different scenarios.
crul
is aimed a bit more at developers than at the
casual user doing HTTP requests. That is, crul
is probably
a better fit for an R package developer, mainly because it heavily uses
R6
- an interface that’s very unlike the interface in
httr
but very similar to interacting with classes in
Ruby/Python.
Most likely you’ll want to do a GET
request - so let’s
start with that - though the details are not much different for other
HTTP verbs.
And in most cases you’ll likely not want to do asynchronous requests - though see below if you do.
You’ll probably want to write a small function, like so (annotated for clarity)
make_request <- function(url) {
# create a HttpClient object, defining the url
cli <- crul::HttpClient$new(url = url)
# do a GET request
res <- cli$get()
# check to see if request failed or succeeded
# - if succeeds this will return nothing and proceeds to next step
res$raise_for_status()
# parse response to plain text (JSON in this case) - most likely you'll
# want UTF-8 encoding
txt <- res$parse("UTF-8")
# parse the JSON to an R list
jsonlite::fromJSON(txt)
}
Use the function
make_request("https://hb.opencpu.org/get")
#> $args
#> named list()
#>
#> $headers
#> $headers$Accept
#> [1] "application/json, text/xml, application/xml, */*"
#>
#> $headers$`Accept-Encoding`
#> [1] "gzip, deflate"
#>
#> $headers$Connection
#> [1] "close"
#>
#> $headers$Host
#> [1] "httpbin:8080"
#>
#> $headers$`User-Agent`
#> [1] "libcurl/8.6.0 r-curl/5.2.1 crul/1.5.0"
#>
#>
#> $origin
#> [1] "172.18.0.2"
#>
#> $url
#> [1] "http://httpbin:8080/get"
Now you can use the make_request
function in your script
or package.
Once you get more familiar (or if you’re already familiar with HTTP) you may want to have more control, toggle more switches.
In the next function, we’ll allow for users to pass in curl options, use a custom HTTP status checker, and xxx.
make_request2 <- function(url, ...) {
# create a HttpClient object, defining the url
cli <- crul::HttpClient$new(url = url)
# do a GET request, allow curl options to be passed in
res <- cli$get(...)
# check to see if request failed or succeeded
# - a custom approach this time combining status code,
# explanation of the code, and message from the server
if (res$status_code > 201) {
mssg <- jsonlite::fromJSON(res$parse("UTF-8"))$message$message
x <- res$status_http()
stop(
sprintf("HTTP (%s) - %s\n %s", x$status_code, x$explanation, mssg),
call. = FALSE
)
}
# parse response
txt <- res$parse("UTF-8")
# parse the JSON to an R list
jsonlite::fromJSON(txt)
}
Use the function
make_request2("https://api.crossref.org/works?rows=0")
#> $status
#> [1] "ok"
#>
#> $`message-type`
#> [1] "work-list"
#>
#> $`message-version`
#> [1] "1.0.0"
#>
#> $message
#> $message$facets
#> named list()
#>
#> $message$`total-results`
#> [1] 160424746
#>
#> $message$items
#> list()
#>
#> $message$`items-per-page`
#> [1] 0
#>
#> $message$query
#> $message$query$`start-index`
#> [1] 0
#>
#> $message$query$`search-terms`
#> NULL
No different from the first function (besides the URL). However, now we can pass in curl options:
make_request2("https://api.crossref.org/works?rows=0", verbose = TRUE)
make_request2("https://api.crossref.org/works?rows=0", timeout_ms = 1)
We can also pass named parameters supported in the get
method, including query
, disk
, and
stream
.
make_request2("https://api.crossref.org/works", query = list(rows = 0))
#> $status
#> [1] "ok"
#>
#> $`message-type`
#> [1] "work-list"
#>
#> $`message-version`
#> [1] "1.0.0"
#>
#> $message
#> $message$facets
#> named list()
#>
#> $message$`total-results`
#> [1] 160424746
#>
#> $message$items
#> list()
#>
#> $message$`items-per-page`
#> [1] 0
#>
#> $message$query
#> $message$query$`start-index`
#> [1] 0
#>
#> $message$query$`search-terms`
#> NULL
In addition, the failure behavior is different, and customized to the specific web resource we are working with
You may want to use asynchronous HTTP requests when any one HTTP request takes “too long”. This is of course all relative. You may be dealing with a server that responds very slowly, or other circumstances.
See the async with crul vignette for more details on asynchronous requests.