add support for variable host and port as suggested in #210

[cboettig]

A heck, here's a PR to implement this anyway, that just makes host and port into optional function arguments. Let me know if you'd rather it were done differently. Many thanks

[cboettig]

@hadley ping. I know httr is still in the queue behind lots of other good stuff; but this is a quick fix to an issue that bites me regularly. (I'd just use my branch but seems weird to tell users to do that)

[hadley]

I'll add httr to my list of packages that need updates (it's currently behind devtools and tidyr)

[hadley]

Could you add a bullet point to NEWS please?

[hadley]

@inheritParams ouath_listener would be a bit DRYer

[cboettig]

Thanks for these, will do

On Mon, May 4, 2015, 11:41 AM Hadley Wickham notifications@github.com
wrote:

In R/oauth-init.R
#211 (comment):

@@ -7,10 +7,13 @@
#' \code{\link{oauth_app}}
#' @param permission optional, a string of permissions to ask for.
#' @param is_interactive Is the current environment interactive?
+#' @param host ip address for the listener

@inheritParams ouath_listener would be a bit DRYer

—
Reply to this email directly or view it on GitHub
https://github.com/hadley/httr/pull/211/files#r29610720.

[hadley]

Thanks!

[cboettig]

@hadley Now I realize that it probably would have been better to set the default of the function argument to something like getOption("httr_localhost") since most third-party R packages define their own auth wrappers around httr and aren't exposing this option either.

To my thinking, a better implementation would actually check the environmental variable for the default, since I could then set something like ENV LOCALHOST 0.0.0.0 inside the container and everything would work with no additional intervention. But I'm not sure if calling Sys.getenv("LOCALHOST") is a good idea. Thoughts?

[cboettig]

@hadley More problematically, I'm worried that this PR might have introduced a bug to oauth as well. I currently get the following error with traceback:

 httr::oauth2.0_token(dropbox, dropbox_app)
Error in makeTcpServer(host, port, appWrapper$onHeaders, appWrapper$onBodyData,  : 
  object 'host' not found

Enter a frame number, or 0 to exit   

1: httr::oauth2.0_token(dropbox, dropbox_app, host = "0.0.0.0")
2: Token2.0$new(app = app, endpoint = endpoint, params = params, cache_path = 
3: public_bind_env$initialize(...)
4: self$init_credentials()
5: init_oauth2.0(self$endpoint, self$app, scope = self$params$scope, type = se
6: oauth_listener(authorize_url, is_interactive, host, port)
7: httpuv::startServer(host, port, list(call = listen))
8: makeTcpServer(host, port, appWrapper$onHeaders, appWrapper$onBodyData, appW

If I go into the environment at 6, oauth_listener it says:

Selection: 6
Browse[4]> ls()
[1] "host"           "info"           "is_interactive" "listen"        
[5] "port"           "request_url"   
Browse[4]> host
Error during wrapup: promise already under evaluation: recursive default argument reference or earlier problems?

even though at 1 the value host is correctly identified. I'm not sure if something has changed since the PR, but at that time I could confirm that this set the host correctly without error. Not sure what's up. maybe I should re-open or open a new issue?

[hadley]

An env var seems ok to me, but it should be something like HTTR_LOCALHOST. I'd recommend making a little function to wrap it up.

I can't replicate the bug you're seeing

[cboettig]

@hadley It's definitely a bug. Looks like I should have edited the R6 Token class to include the host variable and define self$host <- host, e.g. https://github.com/hadley/httr/blob/master/R/oauth-token.r#L40-L45

My apologies, I completely overlooked what was going on in the R6 classes when working out this pull request but @richfitz set me straight.

install_github("cboettig/rdrop2")
rdrop2::drop_auth(host="0.0.0.0")

should replicate this issue(?)

It looks like I need to modify https://github.com/cboettig/httr/blob/master/R/oauth-token.r#L40-73 to include host and self as part of the definition, but I'm hesitant to go hacking fundamental class structures without really knowing what I'm doing...

[hadley]

I think you'll need to add a listener field that stores the host and port (and anything else that might come up in the future)