My issues with dplyr failing intermittently (esp when examples are allowed to run).

dplyr_issues.md

CRAN maintainers now require that examples not be boxed in by the \dontrun{ tags. Since most of rOpenSci's packages retrieve data from the web, this can be problematic, especially if the API requires a key. But in cases where a key isn't required, I'm fine with allowing the examples to run.

When I use dplyr::rbind_all(list_name) over do.call(rbind, list_name) to combine data.frames, I get this issue when running a devtools::check().

> ### ** Examples
> 
> data_by_code <- aw_code(code = "casent0104669")
Warning in rbind_all(image_data) :
  Unequal factor levels: coercing to character
Warning in rbind_all(image_data) :
  Unequal factor levels: coercing to character
Warning in rbind_all(image_data) :
  Unequal factor levels: coercing to character
Warning in rbind_all(image_data) :
  Unequal factor levels: coercing to character
Error: incompatible type (data index: 3, column: 'img1', was collecting: character (dplyr::Collecter_Impl<16>), incompatible with data of type: INTSXP
Execution halted
Error: Command failed (1)

How to replicate the problem

git clone -b dplyr_issues https://github.com/ropensci/AntWeb.git
cd AntWeb
R
library(devtools)
check()

However, when the examples in master are set to not run, check runs fine.

git clone -b master https://github.com/ropensci/AntWeb.git
cd AntWeb
R
library(devtools)
check()

Exposing examples without using dplyr::rbind_all also works fine and passes all checks.

git clone -b no-dplyr https://github.com/ropensci/AntWeb.git
cd AntWeb
R
library(devtools)
check()

• Repo in question
• Lines where the first error occurred

Wat? CRAN is removing \dontrun?

This was my last complaint about CRAN. Brian Ripley wont accept my packages unless I allow the examples to run.

Can you send me the email you get from BDR? And does this error occur when you run outside of check?

I'll forward that email thread to you shortly. The error almost never happen outside of check (see ecoengine/#36). On occasion, the dplyr::rbind_all call segfaults but I have not been able to reproduce it consistently.

Ugh...so what does he want us to do with API clients?