Re: [rkward-devel] R2HTML orphaned on CRAN

[Milan Bouchet-Valat]

Le dimanche 24 août 2014 à 08:27 +0200, Thomas Friedrichsmeier a écrit :
 On Saturday 23 August 2014 23:21:37 Milan Bouchet-Valat wrote:
  Great! I've created a GitHub repository at
  https://github.com/nalimilan/R2HTML
  
  Please apply for commit rights (or give me your username) and I'll give them
  to you.
 
 Username is tfry-git (both tfry and tfried were already taken - this world is 
 so crowded...).
Indeed. ;-)

I've added you.

  I've pushed fixes for the global assignment issue, at least the warnings
  are gone and the tests pass, I've not yet done careful checks. I had to
  remove an obscure feature which overrode the fix() function in order to
  do clever things when fix() was called during an HTMLStart()/HTMLStop()
  session. I don't think this is very useful.
 
 Yes, I've had a glance at that, and it sure looked weird. Regarding the other 
 assignments, I guess .HTML.file is the most touchy one, as this could 
 conceivably be used by some dependent code. I guess it might be useful (not 
 to 
 RKWard, though) to add at least a getter function, so users can mix HTML() 
 and 
 custom writes more easily.
Actually I also need a setter function in my RcmdrPlugin.temis package,
so I'll add both a getter an a setter. The NEWS will have to mention
that this update breaks API anyway.


Regards

--
Slashdot TV.  
Video for Nerds.  Stuff that matters.
http://tv.slashdot.org/
___
RKWard-devel mailing list
RKWard-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/rkward-devel

Re: [rkward-devel] R2HTML orphaned on CRAN

[Thomas Friedrichsmeier]

Hi!

On Sunday 24 August 2014 10:20:56 Milan Bouchet-Valat wrote:
 Actually I also need a setter function in my RcmdrPlugin.temis package,
 so I'll add both a getter an a setter. The NEWS will have to mention
 that this update breaks API anyway.

While we're at breaking the API: I've taken care of the other two notes, 
except for:

 * checking Rd \usage sections ... NOTE
 
 S3 methods shown with full name in documentation object 'HTML.cormat':
   ‘HTML.cormat’
 
 The \usage entries for S3 methods should use the \method markup and not
 their full name.

Now, HTML.cormat.Rd specifically says:

 Although the class \sQuote{cormat} is not defined, the function is called to
 \code{HTML.cormat}, to highlight the fact it should be called only with a
 correlation matrix.

So, using \method markup would seem somewhat wrong, even if it would shut up 
the warning without too bad side-effects. Googling reveals, R2HTML 
is not the first package with that problem(*). There appears to be no 
workaround other than renaming the function. In fact, that may actually be 
reasonable, because HTML.cormat *will* become an S3-method for cormat 
objects, as soon as any package defines that class (not necessarily in a 
compatible way).

I'm uncertain, whether we should
a) use S3-method markup, anyway
b) rename HTML.cormat to HTMLcormat, breaking the API (although of course we 
could keep a deprecated HTML.cormat along with a))
c) Ignore the note, assuming that CRAN maintainers will be ok with it.

I tend to think, b) may be the best option after all.

Regards
Thomas

(*): e.g.:
- http://r.789695.n4.nabble.com/R-CMD-check-non-S3-method-warning-td4576398.html
- 
http://r.789695.n4.nabble.com/S3-methods-with-full-name-in-documentation-td4490309.html
- 
http://r.789695.n4.nabble.com/Need-to-tell-R-CMD-check-that-a-function-qr-R-is-not-a-method-td4642506.html


--
Slashdot TV.  
Video for Nerds.  Stuff that matters.
http://tv.slashdot.org/
___
RKWard-devel mailing list
RKWard-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/rkward-devel

Re: [rkward-devel] R2HTML orphaned on CRAN

[Milan Bouchet-Valat]

Le dimanche 24 août 2014 à 11:46 +0200, Thomas Friedrichsmeier a écrit :
 Hi!
 
 On Sunday 24 August 2014 10:20:56 Milan Bouchet-Valat wrote:
  Actually I also need a setter function in my RcmdrPlugin.temis package,
  so I'll add both a getter an a setter. The NEWS will have to mention
  that this update breaks API anyway.
 
 While we're at breaking the API: I've taken care of the other two notes, 
 except for:
 
  * checking Rd \usage sections ... NOTE
  
  S3 methods shown with full name in documentation object 'HTML.cormat':
‘HTML.cormat’
  
  The \usage entries for S3 methods should use the \method markup and not
  their full name.
 
 Now, HTML.cormat.Rd specifically says:
 
  Although the class \sQuote{cormat} is not defined, the function is called to
  \code{HTML.cormat}, to highlight the fact it should be called only with a
  correlation matrix.
 
 So, using \method markup would seem somewhat wrong, even if it would shut up 
 the warning without too bad side-effects. Googling reveals, R2HTML 
 is not the first package with that problem(*). There appears to be no 
 workaround other than renaming the function. In fact, that may actually be 
 reasonable, because HTML.cormat *will* become an S3-method for cormat 
 objects, as soon as any package defines that class (not necessarily in a 
 compatible way).
 
 I'm uncertain, whether we should
 a) use S3-method markup, anyway
 b) rename HTML.cormat to HTMLcormat, breaking the API (although of course we 
 could keep a deprecated HTML.cormat along with a))
 c) Ignore the note, assuming that CRAN maintainers will be ok with it.
 
 I tend to think, b) may be the best option after all.
Breaking the API because of a silly WARNING does not seem like the best
option to me. Why not a) instead, since it has no effect on the actual
code, just on the docs? The doc page makes it obvious anyway that it's
not a S3 method.


Regards

--
Slashdot TV.  
Video for Nerds.  Stuff that matters.
http://tv.slashdot.org/
___
RKWard-devel mailing list
RKWard-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/rkward-devel

Re: [rkward-devel] R2HTML orphaned on CRAN

[Milan Bouchet-Valat]

Le dimanche 24 août 2014 à 15:29 +0200, Thomas Friedrichsmeier a écrit :
 Hi,
 
 On Sunday 24 August 2014 12:35:16 Milan Bouchet-Valat wrote:
   I'm uncertain, whether we should
   a) use S3-method markup, anyway
   b) rename HTML.cormat to HTMLcormat, breaking the API (although of course
   we could keep a deprecated HTML.cormat along with a))
   c) Ignore the note, assuming that CRAN maintainers will be ok with it.
   
   I tend to think, b) may be the best option after all.
  
  Breaking the API because of a silly WARNING does not seem like the best
  option to me. Why not a) instead, since it has no effect on the actual
  code, just on the docs? The doc page makes it obvious anyway that it's
  not a S3 method.
 
 well, a) is certainly the easiest option, by all means. The (acceptable) 
 downside to this is that the usage-section will then be rendered as
   ## S3 method for class 'cormat'
   HTML(x,[...])
 while you really have to call HTML.cormat(). The example section is not 
 affected, though.
 
 As the package maintainer, it's your call ;-).
What a responsibility! :-)

Really, I don't care about the ## S3 method comment, as long as the
description is explicit, so go ahead with solution a).


Regards

--
Slashdot TV.  
Video for Nerds.  Stuff that matters.
http://tv.slashdot.org/
___
RKWard-devel mailing list
RKWard-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/rkward-devel

Re: [rkward-devel] R2HTML orphaned on CRAN

[Milan Bouchet-Valat]

Le dimanche 24 août 2014 à 16:29 +0200, meik michalke a écrit :
 Am Sonntag, 24. August 2014, 12:35:16 schrieb Milan Bouchet-Valat:
  Breaking the API because of a silly WARNING does not seem like the best
  option to me.
 
 i totally agree. but when it comes to keeping something on CRAN, even a silly 
 WARNING can shut the door. i really admire the quality management they keep 
 up 
 there, but the zero tolerance for even the slightest deviation is sometimes a 
 bit too harsh, for my taste. i had to withdraw roxyPackage from CRAN because 
 of such a stupid problem.[*]
Fully agreed. That's also why I prefer solution a) - solution c) is far
too risky. ;-)

 viele grüße :: m.eik
 
 
 [*] roxyPackage calls tools:::news2Rd() to create NEWS.Rd files -- which is 
 exactly what the manual on news() tells administrators to do. i however was 
 told not to use unexported ::: functions. but to my knowledge there is no 
 other way to do this job. so i asked them to solve this by exporting that 
 particular function in the tools package instead; i mean, if they *want* you 
 to use it, why not export it already? but they did not like the idea, and i 
 was not willing to lose the functionality i was able to implement by using 
 the 
 NEWS.Rd file (especially the RSS feed). apart from that, i find it 
 infuriating 
 that packages are being blocked if the ship a copy of the GPL in their 
 sources.
...and that we were asked to update our BSD-licensed packages to include
a LICENSE file just to mention the copyright holders, which is redundant
with the Copyright field in DESCRIPTION. ;-)


Regards

--
Slashdot TV.  
Video for Nerds.  Stuff that matters.
http://tv.slashdot.org/
___
RKWard-devel mailing list
RKWard-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/rkward-devel