Re: What is a "virtual" group?

[Dave Abrahams <dave@boostpro.com>]

on Tue Dec 25 2012, Lars Ingebrigtsen <larsi-AT-gnus.org> wrote:

> Dave Abrahams <dave@boostpro.com> writes:
>
>> * Is the "has a different group or groups backing it" part essential to
>>   nnvirtual, or is it really more about "has no real backing store of
>>   its own?"
>
> It's more the former.  It's a group that's used to access data from a
> different backend.

OK, thanks.  That's very clear.  In trying to suggest corresponding
documentation edits, though, I arrive at 6.7.1 Virtual Groups in the
manual and find:

  An "nnvirtual group" is really nothing more than a collection of
  other groups.

where I expect to find a definition of "virtual group."  Maybe we need
to insert something like:

   A "virtual group" is a group whose backing store is actually provided
   by a different group or groups.  At this writing there are two kinds
   of virtual groups: nnvirtual groups, which aggregate all articles
   from their backing groups, and nnir search groups, which aggregate
   specific articles found in their backing groups.

I'm not sure about there only being two types of virtual groups.  Is
nnagent a virtual group?  It does use nnvirtual-convert-headers.

Also, should "virtual" in the user interface and docstrings be changed
to "nnvirtual?"  That seems to be what it really means, most of the time.

>> * Suppose I write a function that takes a group as a parameter.  How
>>   does virtual-ness affect what my function should expect to be able to
>>   do with that group?  Are there certain operations that should be
>>   valid or invalid because of the group's virtual-ness?
>
> It's possible.  Grep for "virtual".

Well, I have.  I did try hard to answer this question for myself before
posting, but failed.

>> * Should an nndoc group be considered virtual?
>
> No.

Interesting.  So how do we ever get here:

  http://git.gnus.org/cgit/gnus.git/tree/lisp/gnus-sum.el#n9253

?

>
>> * Does the check for virtual-ness in gnus-warp-to-article actually
>>   belong there?  After all, real groups shouldn't implement the
>>   warp-to-article backend function, should they?
>
> I don't know.

Would you be open to a patch that removes the check?

-- 
Dave Abrahams
BoostPro Computing                  Software Development        Training
http://www.boostpro.com             Clang/LLVM/EDG Compilers  C++  Boost