R doctests fail for no reason

[sagetrac-felixs]

Whether or not a certain R version is installed, should not affect doctests, if R works as intended.

Component: interfaces

Keywords: r-project

Author: Felix Salfelder

Branch/Commit: u/felixs/r_doctest @ 628d5af

Issue created by migration from https://trac.sagemath.org/ticket/15158

[sagetrac-felixs]

comment:1

This is a defect, because sagelib is sometimes used outside ouf sage (the distribution). It will be a defect within sage (the distribution) as soon as R will not be built unconditionally (e.g. #14796).

[kcrisman]

comment:2

Can you be more elaborate about what you mean by this? The description is remarkably cryptic.

But I think you mean the function tested in the function changed (correctly) in #13686. There is a good reason to test this function; it's a function in Sage, so it needs to be doctested. Perhaps there is a nice clever way to test this without allowing just any old output like ((25,1,3), 'R version -3.2.x') but that would need to be done, so we shouldn't remove the function. If that's not what you're implying, my apologies in advance.

[kcrisman]

Changed keywords from none to r-project

[sagetrac-felixs]

comment:3

Replying to @kcrisman:

Can you be more elaborate about what you mean by this? The description is remarkably cryptic.

OK, I'll try.

But I think you mean the function tested in the function changed (correctly) in #13686.

The functions r_version and version in r.py are affected, also source and _source, this seems unrelated to #13686.

There is a good reason to test this function; it's a function in Sage, so it needs to be doctested.

The return values depend on the R version that is called.
For example newer R has additional L in the source. So in total, four doctests fail.

With this patch
http://trac.sagemath.org/changeset?new=628d5afab70880b7b7a24d3417d7ff996dbb816c&base=728e542453259b59b7e012f382e141652671d425&old=c29a7554ab4cfd9f0d17a8a242558db5b820b12a#file36
they pass (how can I point to a commit from here?).

Perhaps there is a nice clever way to test this without allowing just any old output like ((25,1,3), 'R version -3.2.x')

If R is version -3.2.x, and version returns that, everything works correctly, doesn't it?

but that would need to be done.

It's difficult to tell which version is present without a version call... In which way would the other tests pass, if R is a wrong one?

so we shouldn't remove the function. If that's not what you're implying, my apologies in advance.

No, I dont want to remove the checks, just allow the extra L and be a bit more tolerant about the version number.

[jdemeyer]

comment:4

Needs to be rebased.