[ESS-bugs] ess-mode 12.09 [rev. 5167 (2012-09-24)]; R-object-tooltip not working (again)

[Vitalie Spinu]

  >> Martin Maechler <maechler at stat.math.ethz.ch>
  >> on Sat, 29 Sep 2012 15:02:08 +0200 wrote:

  >>>>>> Vitalie Spinu <spinuvit at gmail.com>
  >>>>>> on Sat, 29 Sep 2012 13:20:02 +0200 writes:

  >> Hi,

  >> The idea is great. I was actually planning to migrate it
  >> into ESS12.09, but have never done that, because we didn't
  >> get time for our ESSR support package. The package is
  >> needed to replace the pseudo dispatch which Eric is now
  >> using in ess-R-object-tooltip-alist, with R's generic
  >> mechanism.

  >> The function should also be further extended to allow
  >> rotation of meaningfully information, str, summary,
  >> methods arguments, may be even plots. Also it should work
  >> in terminal, and that means to pop an electric buffer.

  >> ---- 
  >> In recent ESS there have been some internal changes and
  >> tooltip-show-at-point was renamed into
  >> ess-tooltip-show-at-point.

  > {"they" knew that;  was already needed for ESS 12.04 }

But the version on the site was still using tooltip-show-at-point.

  > For that reason, I've also decided to use  ess-r-object-tooltip
  > (lower case 'r') as the other functions there.

We are not using -r- or -R- consistently. I made a couple of greps and
it looks like -R- is used much more often, and mainly ess-r-args.el uses
the first one. also the customization group is ess-R. So I would prefer
to stick with ees-R-. 

  >> Here is a modified version
  >> which works with 12.09:

  > HOWEVER, the "error catching" does not work anyway for me (I do
  > use an English locale in R), and so I see the error as a
  > tooltip...
  > ... which is actually ok;
  > so we could also decide to  drop that whole

Yes, that change was deliberate.  It's better to show the exact error.

  >   (unless (re-search-forward "\(object .* not found\)\|unexpected" nil t)

  > check. As I say above, it cannot work in a non-English locale,
  > *and* it is not so bad if it does not work.  The error message
  > "as tooltip" is not a problem.

We cannot currently drop this because there must be a check for a
class. And only after we have the class, we can move further. This is a
pseudo-dispatch that I was talking about. This is why I wasn't in a
hurry to move it into ESS a couple of months ago when I stumbled on
it. It's much better to have an R generic, say essSummary, to retrieve
the corresponding summary, instead of the
ess-R-object-tooltip-alist. The best place for that is ESSR package of
course. But we can implement it even now. Methods are cached in methods
package and there is no interference with user's global environment.

So one way to implement that would be to inject essSummary generic with
a couple of useful methods. Then users can extend/modify the methods for
whatever they are interested in.

essSummary can have an additional argument (or arguments?) that would
indicate the level and type of the summary. I have in mind a cycling
work-flow. C-c C-g shows str, C-c C-g C-g shows summary, C-c C-g C-g C-g
shows plot or table etc. 

The second approach would to try in turn a list of predefined functions
ess-R-summary-functions ("str", "summary", "table", "plot"), a list
which user can customize and add/remove summary functions.

Then we can also implement a shortcut C-c C-g 3 that would jump to 3rd
summary in the list. I hope I am clear enough.

I think the second approach is the best. It is simpler, more powerful
and also easier to implement. We won't need any dispatch whatsoever
because all the summary functions are already generics, and user can add
his own generics if needed.

    Vitalie


  >>>> Martin Maechler <maechler at stat.math.ethz.ch> on Sat, 29
  >>>> Sep 2012 11:52:31 +0200 wrote:

  >>>>>>>> Stefano Conti <s.conti at gmx.co.uk> on Fri, 28 Sep
  >>>>>>>> 2012 18:44:07 +0100 writes:

  >>>> Dear ESS developers team, Since upgrading to the latest
  >>>> Emacs (latest version 24.2.1) ESS package (version
  >>>> 12.09) I have been unable to get R (latest version
  >>>> 2.15.1) object tooltips
  >>>> (u. http://www.emacswiki.org/emacs/ESSToolTips) working
  >>>> in ESS.

  >>>> Specifically, upon invoking the ess-R-object-tooltip
  >>>> command from within Emacs on any R script, I obtain the
  >>>> "No ESS process is associated with this buffer now"
  >>>> message from the mini-buffer.  Note that downgrading ESS
  >>>> to its previous version (12.04-4) restored the
  >>>> functionality of R object tooltips.

  >>>> Hope the information above and below is helpful enough;
  >>>> I am also copying Martin Maechler, whom I have
  >>>> previously received help on a related issue from, in
  >>>> this correspondence.

  >>>> Thank you in advance for any assistance you may provide,
  >>>> and sincere apologies for any cross-posting.  Kind
  >>>> regards,

  MM> Caro Stefano,

  MM> as this extra functionality was never adopted in to ESS
  MM> proper, we of course could not pay attention to its
  MM> continued functioning.

  MM> I think this is quite a nice idea that some, maybe even
  MM> many people may want, and so I consider to add it to ESS
  MM> proper, and then find out what goes wrong (in your
  MM> case).  To adopt it into ESS, I'd also want a
  MM> confirmation by Erik (the author, hence in CC) that he'd
  MM> donate it to ESS as well.  Erik?

  MM> Martin

  MM> _______________________________________________ ESS-bugs
  MM> ESS-bugs at r-project.org
  MM> https://stat.ethz.ch/mailman/listinfo/ess-bugs

  MM> _______________________________________________ ESS-core
  MM> list: https://stat.ethz.ch/mailman/listinfo/ess-core