multi: Async Concurrent Requests

multiR Documentation

Async Concurrent Requests


AJAX style concurrent requests, possibly using HTTP/2 multiplexing. Results are only available via callback functions. Advanced use only! For downloading many files in parallel use multi_download instead.


multi_add(handle, done = NULL, fail = NULL, data = NULL, pool = NULL)

multi_run(timeout = Inf, poll = FALSE, pool = NULL)

multi_set(total_con = 50, host_con = 6, multiplex = TRUE, pool = NULL)

multi_list(pool = NULL)


new_pool(total_con = 100, host_con = 6, multiplex = TRUE)

multi_fdset(pool = NULL)



a curl handle with preconfigured url option.


callback function for completed request. Single argument with response data in same structure as curl_fetch_memory.


callback function called on failed request. Argument contains error message.


(advanced) callback function, file path, or connection object for writing incoming data. This callback should only be used for streaming applications, where small pieces of incoming data get written before the request has completed. The signature for the callback function is write(data, final = FALSE). If set to NULL the entire response gets buffered internally and returned by in the done callback (which is usually what you want).


a multi handle created by new_pool. Default uses a global pool.


max time in seconds to wait for results. Use 0 to poll for results without waiting at all.


If TRUE then return immediately after any of the requests has completed. May also be an integer in which case it returns after n requests have completed.


max total concurrent connections.


max concurrent connections per host.


enable HTTP/2 multiplexing if supported by host and client.


Requests are created in the usual way using a curl handle and added to the scheduler with multi_add. This function returns immediately and does not perform the request yet. The user needs to call multi_run which performs all scheduled requests concurrently. It returns when all requests have completed, or case of a timeout or SIGINT (e.g. if the user presses ESC or CTRL+C in the console). In case of the latter, simply call multi_run again to resume pending requests.

When the request succeeded, the done callback gets triggered with the response data. The structure if this data is identical to curl_fetch_memory. When the request fails, the fail callback is triggered with an error message. Note that failure here means something went wrong in performing the request such as a connection failure, it does not check the http status code. Just like curl_fetch_memory, the user has to implement application logic.

Raising an error within a callback function stops execution of that function but does not affect other requests.

A single handle cannot be used for multiple simultaneous requests. However it is possible to add new requests to a pool while it is running, so you can re-use a handle within the callback of a request from that same handle. It is up to the user to make sure the same handle is not used in concurrent requests.

The multi_cancel function can be used to cancel a pending request. It has no effect if the request was already completed or canceled.

The multi_fdset function returns the file descriptors curl is polling currently, and also a timeout parameter, the number of milliseconds an application should wait (at most) before proceeding. It is equivalent to the curl_multi_fdset and curl_multi_timeout calls. It is handy for applications that is expecting input (or writing output) through both curl, and other file descriptors.

See Also

Advanced download interface: multi_download


results <- list()
success <- function(x){
  results <<- append(results, list(x))
failure <- function(str){
  cat(paste("Failed request:", str), file = stderr())
# This handle will take longest (3sec)
h1 <- new_handle(url = "")
multi_add(h1, done = success, fail = failure)

# This handle writes data to a file
con <- file("output.txt")
h2 <- new_handle(url = "", postfields = "bla bla")
multi_add(h2, done = success, fail = failure, data = con)

# This handle raises an error
h3 <- new_handle(url = "")
multi_add(h3, done = success, fail = failure)

# Actually perform the requests
multi_run(timeout = 2)

# Check the file

curl documentation built on Oct. 2, 2023, 5:06 p.m.