Re: [PATCH] ob-tangle.el: restore :tangle closure nil behavior

[Ihor Radchenko <yantar92@posteo.net>]

Tom Gillespie <tgbugs@gmail.com> writes:

>> What do you mean by "restore"? Were it evaluated in the past?
>> May you please provide a reproducer?
>
> Hrm. I think I may have mixed two commit lines. It is the case that
> :tangle closures used to work, but you are right, the historical behavior
> when tangling closures meant that all parameters were evaluated (tested
> with the block below in 27 and 28).
>
> #+begin_src elisp :var value=(error "oops") :tangle (or "no")
> value
> #+end_src

This example clearly shows that evaluating everything is a bad idea.
:var has no effect during tangling anyway.

> My use case is that I have blocks that I want to tangle that set :var
> from e.g. the library of babel, which isn't always loaded, but which also
> is not required if :tangle is no.

My confusion about you patch comes from the fact that

#+begin_src emacs-lisp :tangle (if (= 1 1) "yes")
2
#+end_src

works just fine on main.

I admit that I don't fully understand your use case.

>> > -(defun org-babel-tangle--unbracketed-link (params)
>> > +(defun org-babel-tangle--unbracketed-link (params &optional info-was-evaled)
>>
>> This is not acceptable. Taking care about evaluating INFO should be done
>> in a single place instead of adding checks across the babel code. If we
>> go the proposed way, I expect a number of bugs appearing when somebody
>> forgets to change the eval check in some place.
>
> I don't like the solution either. I see two potential alternatives.
> 1. change the structure of the info list to indicate whether it has
> already been evaluated
> 2. always call org-babel-read on (cdr (assq :tangle params)) even
> if it may already have been evaluated which can lead to some unexpected
> and potentially nasty results.

I think that instead of repeating (cdr (assq :tangle params)) we need a
common API that will abstract away evaluation and modify PARAMS by side
effect if necessary.

Something like (org-babel-get-heading-arg :tangle info/params)

> I don't think we can consolidate evaluating parameters
> into one place in the general case because there are
> order dependencies where a setting in one param header
> should mask others (as is the case here).

May you please elaborate?

-- 
Ihor Radchenko // yantar92,
Org mode contributor,
Learn more about Org mode at <https://orgmode.org/>.
Support Org development at <https://liberapay.com/org-mode>,
or support my work at <https://liberapay.com/yantar92>