Re: [O] [babel] Trying to add ERT test cases

[Eric Schulte]

> * Test
>
> #+begin_src emacs-lisp
> (ert-deftest ob-tangle/speed-command-r ()
>   "Test that speed command `r' does demote the headline."
>   (org-test-at-id "4ee368b8-cf7c-4269-98c0-b28dcf94ff2b"
>     (goto-char (point-at-bol))
>     (org-self-insert-command ?r)
>     (goto-char (point-at-bol))
>     (should (looking-at "\\*\\* Speed command"))
>     (delete-char 1)))
> #+end_src
>
> When running the test _once_, I get my level-1 headline preceded by 114
> occurrences of ^M (yes, 114 for 1 test run!):
>
> ^M^M^M...^M^M^M* Speed command (this must be at level-1 headline)
>
> and no demotion of my headline.
>
> Do you understand such?  The above should have been working, if I read you
> correctly.
>

So it looks like these self-insert-command functions are special cases.
They don't look to their arguments to see what key-press invoked them,
but rather they call the `this-command-keys' function for this purpose.
We can force the behavior we want by overriding the definition of this
function locally, taking this approach the following test case worked
for me

#+begin_src emacs-lisp
  (ert-deftest ob-tangle/speed-command-r ()
    "Test that speed command `r' does demote the headline."
    (org-test-with-temp-text "* Speed command"
      (flet ((this-command-keys () "r")) (org-self-insert-command ?r))
      (goto-char (point-min))
      (should (looking-at "\\*\\* Speed command"))))
#+end_src

>
>
>>> - when `should' is failing, the `delete-char' does not take place. This is
>>>   still mysterious to me, at this point in time.
>>
>> I suppose ERT aborts a test when the first should form fails.
>
> I now do think you're right: an error is an error, hence the test is aborting.
> My cleanup is then useless in such a case.
>
>> Many testing frameworks have a way of defining "fixtures" which serve as
>> test wrappers...
>>
>> Hoping to find an ERT tutorial I googled "ert tutorial emacs test" and
>> the first page was [1], which we should probably update to reflect the
>> actual test framework.  The info page on ERT does look to be informative
>> and may specify how to ensure that "cleanup" code is run -- although in
>> general it may be a better idea to simply run tests in a temporary
>> buffer `with-temp-buffer' so no cleanup is required.
>
> ... which means, I indeed should use a temp buffer. Thanks for the hint.
>
> I read thru the page you link to. Just a minor thing: link to "See
> ert-publish-test.el for the implementation" is broken.
>
> Do you have a solution for checking against internal broken links?

Nothing comes immediately to mind.

> I still have my idea of letting DOT draw an graph of the
> interconnection between pages, so that we see missing pages, and
> unreachable ones (files on the file system, but never linked). I
> should (find the time to) extend it enough to serve that purpose.
>

This could be useful, like a site map for worg.  As an intermediate step
I could see it being useful to simply print out all internal links whose
target does not exist on the file system.

Best -- Eric

>
> Best regards,
>   Seb

-- 
Eric Schulte
http://cs.unm.edu/~eschulte/