Re: About the use of PROPERTY "meta lines"...

[Eric Schulte <eric.schulte@gmx.com>]

[-- Attachment #1: Type: text/plain, Size: 1649 bytes --]

cberry@tajo.ucsd.edu writes:

> Torsten Wagner <torsten.wagner@gmail.com> writes:
>
>> Hmm...
>> but this point is really interesting at least worse to write down in
>> the manual.
>> From my understanding a
>> #+PROPERTY: var bar=2
>> sets bar globally to 2
>> somewhere and many lines and headers later
>> #+PROPERTY: var bar=5
>> would change this value to 5 for either the rest of the file or until
>> a new assignment is given...
>
> I think the behavior is trickier than that.
>
> This file:
>
> ,----
> | #+property: var  foo=1
> | #+property: var+ bar=2
> | 
> | #+begin_src emacs-lisp :results value :exports both
> |   (+ foo bar)
> | #+end_src
> | 
> | #+property: var foo=10
> | #+property: var+ bar=20
> | 
> | 
> | #+begin_src emacs-lisp :results value :exports both
> |   (+ foo bar)
> | #+end_src
> `----
>
> Yields '30' after each block upon C-c C-e A, suggesting it is the last
> #+property setting the global property.
>

This makes sense

>
> But this one:
>
> ,----
> | #+property: var  foo=1
> | #+property: var+ bar=2
> | 
> | #+begin_src emacs-lisp :results value :exports both
> |   (+ foo bar)
> | #+end_src
> | 
> | #+property: var foo=10
> | 
> | #+begin_src emacs-lisp :results value :exports both
> |   (+ foo bar)
> | #+end_src
> `----
>
> Yields '3' after each block.
>
> So the global behavior of the second 'var foo' line depends on there
> baing a subsequent 'var+' line.
>
> Is this really the expected behavior?
>

No, the above behavior is not expected.  I've just pushed up a patch
which results in the following behavior, in which the last specification
of a property overwrites any previous specifications.


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: overwriting-properties.org --]
[-- Type: text/x-org, Size: 181 bytes --]

#+property: something  foo=1
#+property: something+ bar=2
#+property: something foo=10

#+begin_src emacs-lisp
  org-file-properties
#+end_src

#+results:
| (something . foo=10) |


[-- Attachment #3: Type: text/plain, Size: 3133 bytes --]


Best,

>
> (Org-mode version 7.8.03)
>
> Chuck
>
>
>> in that way a property line would be an tree-independent global variable
>>
>> in contrast, a property-block is only valid of the given tree (and
>> subtrees?).
>>
>> This brings up the question if there is a need for
>>
>> #+PROPERTY: const bar=2
>>
>> which would behave exactly the same like var but issue an error
>> message if someone tries to set it again somewhere in the file.
>>
>> Torsten
>>
>>
>>
>> On 01/06/2012 04:28 PM, Eric Schulte wrote:
>>> "Sebastien Vauban"<wxhgmqzgwmuf@spammotel.com>  writes:
>>>
>>>> Hi Eric and all,
>>>>
>>>> Eric Schulte wrote:
>>>>> "Sebastien Vauban"<wxhgmqzgwmuf@spammotel.com>  writes:
>>>>>
>>>>>> #+TITLE:     Properties
>>>>>> #+AUTHOR:    Seb Vauban
>>>>>> #+PROPERTY: var  foo=1
>>>>>> #+PROPERTY: var+ bar=2
>>>>>>
>>>>>> * Abstract
>>>>>>
>>>>>> IIUC, properties are set in this way:
>>>>>>
>>>>>> - on a file basis, before any heading, through the =PROPERTY= keyword,
>>>>>> - on a subtree basis, through the =PROPERTIES= block.
>>>>>>
>>>>>> My comprehension is that the =PROPERTY= keyword may not be used inside "trees",
>>>>>> and should be ignored if that would happen.
>>>>>
>>>>> While it is not normal usage, I think that it is legal for #+PROPERTY:
>>>>> lines (or #+Option: lines etc...) to appear inside of subtrees.
>>>>
>>>> I realize this is not especially a Babel question, but more a Org core
>>>> question...
>>>>
>>>> Thanks for your answer -- which generates a new one, though: what is then the
>>>> expected *semantics* of such a construct?
>>>>
>>>> There are at least 3 different views on such a construct: putting a PROPERTY
>>>> line inside a subtree...
>>>>
>>>> - ... resets some values from that point up to the end of the subtree
>>>> - ... resets some values from that point up to the end of the buffer
>>>> - ... defines some values which can have already been by the subtree
>>>>
>>>
>>> I agree this is murky and whatever behavior we want should be clearly
>>> thought out and documented in the manual.  I would argue that you missed
>>> another possible semantics, the simple semantics which are currently
>>> implemented in which a property line *anywhere* in a buffer sets a
>>> global property.
>>>
>>> Cheers,
>>>
>>>>
>>>> Best regards,
>>>>    Seb
>>>>
>>>>>> The following example shows that either:
>>>>>>
>>>>>> - I'm wrong to think so,
>>>>>> - there is a bug.
>>>>>>
>>>>>> What is the right assumption here?
>>>>>>
>>>>>> * Subtree
>>>>>>
>>>>>> Being located in a subtree, the following lines are ill-placed IMHO:
>>>>>>
>>>>>> #+PROPERTY: var  foo="Hello
>>>>>> #+PROPERTY: var+ world"
>>>>>>
>>>>>> Though, they're well taken into account:
>>>>>>
>>>>>> #+begin_src emacs-lisp
>>>>>>    foo
>>>>>> #+end_src
>>>>>>
>>>>>> #+results:
>>>>>> : Hello world
>>>>>>
>>>>>> These lines have even wiped the definition of =bar= (because of the use of =var=
>>>>>> without any =+=):
>>>>>>
>>>>>> #+begin_src emacs-lisp
>>>>>>    (+ foo bar)
>>>>>> #+end_src
>>>>>>
>>>>>> returns the error "Symbol's value as variable is void: bar."
>
>

-- 
Eric Schulte
http://cs.unm.edu/~eschulte/