[R] Summary: Warnings with no INDEX file in a package

[David Kane <David Kane]

Thanks to Henrik Bengtsson, Martin Maechler, Brian Ripley, Jeff Gentry and
David Brahm for taking the time to answer my questions about INDEX files in R
packages.

My key take-aways were:

1) All of this will be changing a lot in R 1.7.0-to-be, so don't sweat the
   details.

2) Many people run R CMD build before they run R CMD check. I found this quite
   surprising. I did not realize that R CMD build actually *changes* the contents
   of your package by, at least, adding an INDEX file if one is not there. I do
   not see this behavior documented in the help for R CMD build. As a simple
   example:

agate|tmp> ls pyri.check
CVS		DESCRIPTION	R		inst		man
agate|tmp> R CMD build pyri.check
* checking for file 'pyri.check/DESCRIPTION' ... OK
* preparing 'pyri.check':
* checking whether 'INDEX' is up-to-date ... NO
* creating new 'INDEX'
* removing junk files
* building 'pyri.check_1.5.tar.gz'

agate|tmp> ls pyri.check
CVS		INDEX		inst
DESCRIPTION	R		man
agate|tmp>

When I first run R CMD check, I get a warning about the lack of an INDEX. But,
since R CMD build creates the index for me, I can now run R CMD check and get
no warnings.

My mistake was to interpret:

"* creating new 'INDEX'"

to mean:

"Making an INDEX file and putting it in your .gz file while leaving your
original package unchanged."


Thanks to all for clarifying this for me.

Dave Kane

-- 
David Kane
Geode Capital Management
617-563-0122
david.d.kane at fmr.com
Please avoid sending me Word or PowerPoint attachments.
See http://www.fsf.org/philosophy/no-word-attachments.html