Re: [PATCH 1/3] diff options: Introduce --ignore-submodules

Johannes Schindelin [off-list ref] writes:

quoted

That's exactly the point.  The option reads "ignore submodules", not 
"special option for use only by stash and rebase".

But it also does not read "ignore submodules and those blobs/trees that 
happen to replace (or be replaced by) submodules".

I think "ignore submodules" option, if exists, would mean "ignore changes
that involve submodules", and I think everybody would agree with that.

Because we are talking about an option to "diff", changes involve two
sides (preimage and postimage).  Logically, you can define the class of
changes that involve submodules in two ways:

 * both sides of the change must be a submodule; otherwise the change does
   not involve a submodule.

 * either side of the change is a submodule; such a change involve a
   submodule.

I am saying that the latter makes much more sense (worse yet, what you did
for creation and deletion is inconsistent --- they are void vs submodule
and you treat them as "change that involve submodule").

Didn't you make the option introduction to read "sometimes you are not
interested in submodules..." or something?

Hi,

On Wed, 14 May 2008, Junio C Hamano wrote:

Johannes Schindelin [off-list ref] writes:

quoted

quoted

That's exactly the point.  The option reads "ignore submodules", not 
"special option for use only by stash and rebase".

But it also does not read "ignore submodules and those blobs/trees 
that happen to replace (or be replaced by) submodules".

I think "ignore submodules" option, if exists, would mean "ignore 
changes that involve submodules", and I think everybody would agree with 
that.

Because we are talking about an option to "diff", changes involve two
sides (preimage and postimage).  Logically, you can define the class of
changes that involve submodules in two ways:

 * both sides of the change must be a submodule; otherwise the change does
   not involve a submodule.

 * either side of the change is a submodule; such a change involve a
   submodule.

I am saying that the latter makes much more sense (worse yet, what you 
did for creation and deletion is inconsistent --- they are void vs 
submodule and you treat them as "change that involve submodule").

Okay, maybe the introduction, and the intuitive understanding of "ignore 
submodules" is the latter.

But I maintain that there is no sensible operation you would need it 
for.

So, how about --only-non-submodules?

Ciao,
Dscho