wf_request() doesn't expose the url anymore

[khufkens]

@eliocamp by default the wf_request() doesn't expose the url anymore, the task is not submitted.

in wf_request() this

  if (transfer) {
    request$transfer(time_out = time_out)
    if (request$is_success()) {
      return(request$get_file())
    }
    message("Transfer was not successfull.")
  }

  return(request)

should be rewritten along these lines to be able to just submit a job and query on a later date (or through the web interface)

   # submit request
   request$transfer(time_out = time_out) 

    if (request$is_success() & transfer) {
      return(request$get_file())
    } else {
     message("Transfer was not successfull - please check your request later at:")
     message(request$url)
    }

  }
  return(request)

Would this make sense? Don't have time to figure out the R6 setup, but I think it does not break anything.

New branch with fix here:
https://github.com/bluegreen-labs/ecmwfr/tree/expose_url

[khufkens]

Just to clarify, what I want to put back in as it is missing now is to just submit a request and return the API endpoint and that's it. With this you can collect data on a later point using wf_transfer() or on the data request pages (online).

[eliocamp]

I think that would break request_batch, since transfer() locks up the R session waiting for the request to finish, which is what transfer = FALSE should prevent.

How about this?

  # Submit the request
  request$submit()
  
  # Only wait for request to finish if transfer == TRUE
  if (transfer) {
    request$transfer(time_out = time_out)
    if (request$is_success()) {
      return(request$get_file())
    }
     message("Transfer was not successfull - please check your request later at:")
     message(request$url)
  }

  return(request)

(I'm not on my PC so I didn't test it)

[khufkens]

This works, but it seems the url by default is private and not obviously returned in the print statement.

This can be accessed via:

 request$.__enclos_env__$private$url

But this isn't really intuitive. Having a request$get_url() would be helpful and basically cover this. Will do some digging later as I'm not familiar with the R6 classes etc. (have a rather full plate)

[eliocamp]

Yes, you're right. I added a get_url() method.

[khufkens]

Thanks merged.

Still have to address the webapi issue on mac, but I'll have to dig into this some more. Closing this with:

#96