[ESS] RFC: and .Rhistory -- new default == *no* .Rhistory ??

[Vincent Goulet]

Interesting discussion, here.

I may add on a side note that my Emacs distributions for Windows and OS X have had for four years now

	(setq-default inferior-R-args "--no-restore-history --no-save ")

meaning that, by default, R started from Emacs does not read the .Rhistory file. This is not the same as not creating any history file in the first place, but the "source is real" philosophy is still somewhat enforced.

I thought this might be a useful reminder for users of my distributions.

And sorry if I appear to be plugging my distributions, here!

Vincent

Le 2011-09-03 à 05:58, Martin Maechler a écrit :

> Thank you very much!
> all, who sent in your votes, some offline (for saving list
> bandwidth), almost always providing reasons for your opinion.
> 
> My summary (hopefully mostly unbiased):
> 
> o Most voted *pro* changing the default, with main arguments
>    - we should emphasize working with *.R script files
>    - I'd prefer no .Rhistory - cluttering my file system
>    - emacs users nowadays should be smart to activate it if  they want
> 
> o Those voting *pro* keeping the default argued
>    - beginners may be helped much by its presence
>    - the smart "leet" (~= "elite" - I had to look that up; it's an
>      	  	        interesting Wikipedia reading)
>      users have ~/.emacs customizations anyway and can just add one more
>    - "the principle of least surprise"
>      (changing year long defaults is unpleasant if not frustrating for many)
> 
> There's one thing most correspondents seem to not have been
> aware (even though I tried to get it accross as well):
> My changes (to the svn development version) two days ago have
> finally activated the  ess-history-directory  (customizable) variable
> and if you set that --  very natural to me seems
>        (setq ess-history-directory "~/.R/")
> then there would be one and only one .Rhistory file, namely ~/.R/.Rhistory
> which gives no cluttering and no warnings about missing
> .Rhistory files, but of course looses project, i.e. directory,
> specific history.
> 
> After all these considerations, and notably with the wise
> recommendation from Doug Bates, one of the very earliest authors
> of ESS (then under the name of "S-mode"), I am now proposing to
> keep the default behavior -- for the sake of the unsuspecting
> and the beginners -- but also plan to add a prominent
> recommendation/hint that you easily can turn it off completely by
>        (setq ess-history nil)
> or turn off the cluttering by 
>        (setq ess-history-directory "~/.R/")
> 	;; requires *write* permission to the directory..
> 
> once again with the idea that many of the experienced ESS users
> will do one or the other.
> 
> I know that this may seem as disregarding majority vote, but
> then it was not only a vote (*) rather more a RFC, and I've
> ended up counting those comments more than votes.  Also, this is
> not a decision (as that will rather be made by the full ESS core
> team) but only a strong suggestion...
> 
> Martin
> 
> ---
> (*) the vote can not really be "representative" the way we had to
>   do it on ESS-help, and keeping it up a very limited time, etc etc
> 
> 
>> On Fri, Sep 2, 2011 at 11:03 AM, Douglas Bates <bates at stat.wisc.edu>
>> wrote:
>>> On Fri, Sep 2, 2011 at 7:25 AM, Guy Freeman <gfreeman at hku.hk> wrote:
>>>> I agree with Marianne, including being grateful for your hard work!
>>>> 
>>>> This shouldn't be treated as an election, but as an opportunity to
>>>> explore the consequences for non-"leet" users.
>>>> 
>>>> Keep default -> harmless if messy .Rhistory file created, which can
>>>> be turned off for those who are very obsessive about these things ;)
>>>> Change default -> great for those that expect it, disaster for those
>>>> that didn't
>>> 
>>> I agree that the new capability is very desirable but that there is
>>> probably no need to change the default.  Anyone reading this email
>>> list probably has ess configuration variables set in their ~/.emacs
>>> file anyway so this is just one more to set if you prefer not to have
>>> .Rhistory files created.
>>> 
>>>> On Fri, Sep 2, 2011 at 18:35, Marianne Promberger
>>>> <marianne.promberger at kcl.ac.uk> wrote:
>>>>> Martin Maechler <maechler at stat.math.ethz.ch> 01-Sep-11 16:24:
>>>>>> With this e-mail, we are polling you, the active ESS users.  Or in
>>>>>> other words, we ask you to "vote" on one aspect of this issue, for
>>>>>> which I need to give a bit more context :
>>>>>> 
>>>>>>        nil --  do not attempt to read/write a history file        
>>>>>>  t --  use the default naming scheme for loading the history  
>>>>>>  "fname" -- use this value as the filename.
>>>>>> 
>>>>>>   Note that from above,  't'  corresponds to the current default  
>>>>>> in ESS (which is 'nil' there !).
>>>>> 
>>>>> I would vote for a default of "t"
>>>>> 
>>>>> We are talking about a default. I think it is worse if you expect R
>>>>> to write a .Rhistory and it doesn't than if you get unnecessary
>>>>> clutter (because you don't know about the new option to turn off the
>>>>> Rhistory).
>>>>> 
>>>>> Many ESS users probably don't read this list, or the change notes.
>>>>> 
>>>>> Personally, I would not want to miss .Rhistory files in project
>>>>> folders.
>>>>> 
>>>>> I have a script file for each project. But that is already more
>>>>> cluttered than I'd like it to be (eg, with analyses that I will not
>>>>> ultimately report).
>>>>> 
>>>>> I often work with trial-and-error to achieve a task. I do save the
>>>>> end result to the script file. Later I may have a similar but not
>>>>> identical task, and I vaguely remember that one of the
>>>>> trial-and-error-steps would be helpful now. Going through the
>>>>> .Rhistory is very helpful in that case.
>>>>> 
>>>>> Whichever way you decidee, many thanks for putting together such a
>>>>> tremendously useful tool!
>>>>> 
>>>>> Marianne
>>>>> 
>>>>> --
>>>>> Marianne Promberger http://promberger.info
>>>>> http://participate-in-research.org.uk GnuPG/PGP public key ID
>>>>> 80AD9916
>>>>> 
>>>>> ______________________________________________
>>>> 
>>>> 
>>>> --
>>>> Dr. Guy Freeman, Post-Doctoral Fellow
>>>> School of Public Health
>>>> Li Ka Shing Faculty of Medicine
>>>> The University of Hong Kong
>>>> Tel: +852 - 39062034
>>>> Mob: +852 - 68790289
>>>> Fax: +852 - 35201945
>>>> Add: Unit 624-627, Level 6, Core F, Cyberport 3, 100 Cyberport Road, Hong Kong
>>>> Web: http://sph.hku.hk/faculty_and_staff_detail.php?id=171
>>>> 
> 
> ______________________________________________
> ESS-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/ess-help