Suggested new functions

[restonslacker]

Public versus private indicate by leading dot, though any of these could go either way.

• .isValidKey(key) - key is a string
• .isValidDevice(device, key) - device string is associated with key
• pbValidateConf(conf=NULL) - if null, check env variables, otherwise a path to json file, or a json string.
• .isValidChannel(chan) - chan is a string

Thoughts? Additional functions?

[eddelbuettel]

I think these are all good. Maybe split into two PRs?

I also realized that my conf is probably different from your as I have a test-channel in there. Happy to share that one if you want to see if you can blast messages at it. I can twitter-DM you the name. Or maybe send a pushbullet message to you :)

[restonslacker]

WRT test-channel, i actually set one up after looking over simpleTests.R after starting on this. So i think that's okay unless we need to coordinate some testing effort in the future.

[restonslacker]

code for public functions is now set to go:

• pbValidateConf(conf=NULL)
• pbGetChannelInfo(channel, no_recent_pushes=FALSE) - S3 like pdGetDevices
• pbGetUser(apikey=.getKey()) - S3 like pdGetDevices

I wrote the 2nd and 3rd functions because once i looked up the API for the .isValid*() functions, they were a trivial addition.

[eddelbuettel]

If you feel super-adventurous, I think some folks (Jenny Bryan and others) figured out how to test token based web apis from Travis without spilling the token. But I can't remember where that was. :-/ It would be sweet if we could actually hit some tests as the code base it getting better and bigger. Not urgent, obviously.

[restonslacker]

I did look into this some. The general approach seems like we encrypt the info with our public keys and travis's public key. something like:

conf <- readLines("~/.rpushbullet.json")
save(list(dirk=openssl::rsa_encrypt(conf, dirks_key),
           seth=openssl::rsa_encrypt(conf, seths_key),
           travis=openssl::rsa_encrypt(conf, traviss_key)),
      file="conf.RData")

conf.RData is then added to the repo and somehow we tell travis to use it.

We'd still have to setup a throw-away gmail account but maybe that's ok. I don't have my notes in front of me, but i actually started to do that previously. Also, we'd have to convert simpleTests.R to a proper test framework which checks return codes. Not sure how to test for the issue we had in #39, however.

[eddelbuettel]

Well put, that's the way to do it. I actually do have a throw-away gmail account for the work I do with Philippe on his/our gtrendsR package.

That is definitely worth playing with, and as you say once we have it we may as well switch to proper tests. I know RUnit pretty well (include setup/teardown) and testthat not at all :)