[ESS-bugs] ?names<-

[Martin Maechler]

>>>>> Vitalie Spinu <spinuvit at gmail.com>
>>>>>     on Tue, 11 Sep 2012 13:51:12 +0200 writes:

    >>> Martin Maechler <maechler at stat.math.ethz.ch>
    >>> on Tue, 11 Sep 2012 12:58:32 +0200 wrote:

    >> In ESS devel
    >> ?names<-

    >> gives an "infinite loop" (well, until C-g and C-c C-c)
    >> whereas

    >> ?"names<-"
    >> or ?'names<-'
    >> work {though with the slightly ugly double quotes in the buffer  name}.

    >> In all previous version of ESS, it did work without the
    >> quotes....
    >> though it never (I think) did work in  'pure R'  {R in the console}.

    >> Note that this is of course related to (but much more important than)
    >> the fact that
    >> ?chol-methods
    >> no longer works {after  require(Matrix) }
    >> but at least the latter quickly gives an error message
    >> (similar to pure-R).


    > Yes, there is no wrapping into help(...) now, everything is passed to
    > the R as it is. So what works in pure R should work in ESS and vice
    > versa.

Ok... we can live with that...
(apart from below)

    > This is simple and consistent. treating every single combination of
    > foo?bar and foo??bar is an overkill.

    >> We have to decide if    ?names<-
    >> should give an error
    >> or if it should work (as previously, but unlike pure-R which
    >> only accepts a quoted version).

    > R thinks that the above is a continuation statement and waits for the
    > next line. A relatively easy workaround would be to wrap everything
    > after ? into quotes. 

well, that's possible (though the ".." or '..' then should not be used
in the buffer name)

    > Or may be it can be solved on the R side?

Well, R's parser acts first, and it parses it as needing
completion.  "The rest of R" does not even look at it before
it's parsed.

ESS should detect R waiting for input in such situations,...
or then we need something like  quoting (above).

Martin