Re: evaluation context in call statements

[Rick Frankel <rick@rickster.com>]

On 2013-06-26 02:29, Achim Gratz wrote:
> Eric Schulte writes:
> In defense of the existing behavior, I don't see the benefit of calling
> a code block with the same arguments from multiple locations and
> subsequently littering a file with multiple identical results blocks.
> 
> I agree that this didn't make all that much sense in the past, but with
> property evaluation and elisp argument evaluation now anchored to the
> point of call, the hierarchical position of the call could and (as the
> test case from Rick) will be used to distinguish between invocations
> with the same arguments.  Since the current way to find the results
> doesn't know anything about this, it will generally not do the right
> thing anymore.  Note that calls using a session had that property all
> the time: multiple calls with the same arguments into the same session
> are useful, but Babel would only keep the last result.

Agreed. The only way to know that the arguments are the "same" is to
evaluated them :).

> My only thought about a :target header argument is that it would need 
> to
> be implemented for other types of code blocks as well, which could lead
> to very confusing behavior if we have a named code block with a :target
> header argument which differs from the name.
> 
> Oh yes, the specification of that would be interesting.  I'll try to 
> see
> how this "beam the result anywhere" functionality sprang into existence
> and what the intended use case was (I expect something to do with
> sessions).

I believe the ability to replace named results anywhere was added by
Nicolas in commit 2f2a80fe (quick look at ob-core w/ vc-annotate).

> My current suggestion is however to limit the results block search to
> the same subtree and stop searching at later #+CALL and #+BEGIN_SRC
> line.  We could make this conditional on a :[no]clobber argument to 
> keep
> compatibility with the current behaviour (clobbering the first result
> would be the current and perhaps default behaviour).

These search bounds make sense, but i think this should be the
default behavior. I don't see the current behavior as making
sense---at least to me. At the time (late 2012) I found Nicolases
changes (named results blocks, attributes and captions on the results
block and not the source, etc) confusing. I still find it odd that you
need to evaluate a source block before you can e.g, add a caption or
attributes to the results (previous behavior was that header arguments
on the source block were used for the results in exporting.)

Also, i think a new value for :replace ("original"?) would make more
sense than a new :clobber option.

> My vote is for adding #+name support to call lines, and then handling
> their results in the same manner as code block results.
> 
> I'm not sure what this would entail other than replacing the call with
> its arguments with the name of the call in the results line.  But yes,
> that'd be a step forward, although you'd have to be careful when 
> copying
> calls.

It seems inconsistent to add #+name support to call lines but not the 
other
block modifiers (#+header :var ..., etc). I think call lines are a
special case, so would be ok with the new :target option.

rick