bind last output to .Last.value

[randy3k]

close #405

Thanks @flying-sheep for pointing out the relevant code. Now .Last.value works as expected (even within a cell).

[randy3k]

I think the test failure is not relevant, but there is a note

Found the following possibly unsafe calls:
File ‘IRkernel/R/utils.r’:
  unlockBinding(".Last.value", .BaseNamespaceEnv)

Basically, it says unlocking .Last.value is not safe, but I don't know there is a way to do it safely. Perhaps assigning the last value to a global variable like .jupyter.last_value rather than base::.Last.value?

[flying-sheep]

well, the docs say that it lives in the base namespace, so you’re doing the right thing.

we’re basically replacing the frontend of the whole R runtime, so it’s expected that we do a few things that are considered “potentially unsafe”.

and of course your changes cause the test failure: we just test for the number of lines the “NOTE” has… we could do this better, but for now we have to update and account for every single line. as you can see it went from 4→9. what lines gets added?

[randy3k]

I have updated the tests, 5 out of 7 of the jobs pass except the R-devel versions on linux. I have no clues how to fix them..

[flying-sheep]

the devel test fails aren’t yours, they’re ours (turns out our tests didn’t actually run for some time)

it’s not “8 lines from the ‘attach’ NOTE”, though. it’s still 4 for that one and 4 for your new one.

[randy3k]

I fixed the comments in .travis.sh

[flying-sheep]

great. i’ll wait for the test results and merge

[flying-sheep]

oh, and of course: thank you 😄

[randy3k]

😀
You are welcome