The hardware and bandwidth for this mirror is donated by dogado GmbH, the Webhosting and Full Service-Cloud Provider. Check out our Wordpress Tutorial.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]dogado.de.

SciViews HTTP server

Philippe Grosjean (phgrosjean@sciviews.org)

2022-05-10

Principle

The {svHttp} package uses an option of the integrated HTML help system in R to develop a simple inter-process communication using URLs. It is designed to be the simplest as possible. Yet, it allows also to customize the way the server evaluates R code and sends results back to the client. An {svHttp} server with all default arguments is started in an R process using:

library(svHttp)
server_port <- start_http_server()
server_port

The default port is 8888, but it can be changed with the port = argument in start_http_server() (the name of the server can also be changes with name = and modified or retrieved with http_server_name()).

Since there is only one HTML help system in R, starting the {svHttp} server most probably asks for a different port that the one previously chosen by R. In this case the R HTML help system is forced to close and to reopen with the requested port. There is no consequences for help pages displayed later on. A minor annoyance is in R >= 4.2.0 where preexisting HTML help pages may not respond any more. In this case, just close and display the help page again to restore its full features.

On the client side on the same machine, a simple HTTP request can be used to execute R code. It only works locally to avoid security issues because the communication is simple and not encrypted (the HTTPS protocol is not available). The URL is localhost:port/custom/SciViews with an argument being the URL encoded R code to execute.

For instance to execute 2 - 1, you could use the URL http://localhost:8888/custom/SciViews?2-1. With the {svHttp} server started on the default port 8888 in R, if you use this URL in any local web browser, you will get a page with [1] -1. This is the output R produces when the same instruction 2 - 1 is issued at the R console. Within a separate terminal, you can also issue curl 'http://localhost:8888/custom/SciViews?2-1' and obtain the same result. By default, the {svHttp} server produces the same, or very similar outputs to what would be issued at the R console if the same command was issued there. This is useful to implement a remote R console inside another application.

To get the content of an R variable, just use its name as message, as usual at the R prompt. You can get the content of server_port with the URL http://localhost:8888/custom/SciViews?server_port.

Implementing a client

In R you can use, for instance, the {curl} package on the client. Here is a function that captures output of an R command on the {svHttp} server (note that the server sometimes issues a 500 error code while it still processes the command correctly, so make sure to catch that case too).

http_server_run <- function(cmd, port = 8888) {
  # cmd is a string containing the command to process. We have to URLencode it
  cmd <- utils::URLencode(cmd)
  url <- paste0("http://localhost:", port, "/custom/SciViews?", cmd)
  res <- curl::curl_fetch_memory(url)
  if (res$status_code %in% c(200, 500)) {# Should be OK
    rawToChar(res$content)
  } else if (res$status_code > 0) {
    stop("Error while executing ", url, ": error ", res$status_code)
  }
}

Now, you could run arbitrary R commands in the {svHttp} server and get back the console output it produces. Note: do not use this function in the same R process where the {svHttp} server runs. As it is blocking, you would end up in a deadlock situation!

cat(http_server_run("R.version"))
cat(http_server_run("http_server_name()"))
res <- http_server_run("ls()")
# Now you could do whatever you want with res
res

Callback argument

A second argument can be passed to the URL: the name of a callback function. That way, the client gets this information to process the result send by the {svHttp} server. The URL is then http://localhost:port/custom/SciViews?code&callback. Also, there are numerous ways to customize the behavior of the {svHttp} server, see ?par_http_server. A complete client written in JavaScript is implemented here: https://github.com/SciViews/sciviewsk/blob/master/content/js/socket.js. You could get inspiration from there to develop your own client application.

Utilities on the server-side

Several functions are provided to manage the {svHttp} server on the R process where it runs. You can get or change the port or the name of the server with dedicated functions:

http_server_port()
#> [1] 8888
http_server_name()
#> [1] "R"
# Change the name
http_server_name("myHttpServer")
#> [1] "myHttpServer"

At anytime, you can get an idea of the clients that are currently connected to the {svHttp} server (it can serve multiple clients at the same time, but it will , of course, process commands one at a time and the other clients are put on hold). There is a special client named "default" which process all requests by clients that did not identified themselves properly.

http_server_clients()
#> character(0)

Identification of clients is mandatory to allow further customization of the {svHttp} server in a stateful way. Again, examine the code here (https://github.com/SciViews/sciviewsk/blob/master/content/js/socket.js) for an example of complete implementation of all the features the {svHttp} server provides.

Finally, it is possible to close the {svHttp} server, but keep in mind that you also close the internal R HTML help service at the same time. This would only affect currently displayed help pages, since R would reopen the server automatically at the next help request.

stop_http_server()

The stop_http_server() accepts an argument remove.clients, which is FALSE by default. If set to TRUE, the stateful configuration of registered clients is also deleted.

Alternate solutions

There are several other server implementations and inter-process communication protocols implemented for R. See the CRAN task views, in particular, web technologies and services and the high-performance and parallel computing with R.

These binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.
Health stats visible at Monitor.