Hi,

Ihor Radchenko writes:
> prin1-to-string is too specific and only solves a single use-case.
prin1-to-string is actually universal in a way, since any other
manipulation can then be achieved with

: (setq var (do-something <<nw>>))

at least as long as you're tangling to a programming language, that
can read lisp strings.
> Consider the following example:
>
> #+BEGIN_SRC emacs-lisp :noweb yes :tangle yes :noweb-prefix no :noweb-trans prin1-to-string
> <<preamble>>
> (setq latex-header <<nw>>)
> #+END_SRC
>
> There are two noweb references here. Setting source block-wide
> :noweb-trans is not helpful because the first reference will be
> incorrectly filtered through prin1-to-string.
Indeed. Originally I had thought of adding a new syntax <<"nw">> to
insert a string representation. I've attached a new patch, that does
this instead of introducing :noweb-trans. Now that I think of the
universality of prin1-to-string, I actually like it slightly better
than :noweb-trans. It would break existing "nw"-like noweb references.

Of course, one can work around this easily enough by using two blocks.
> I'd rather introduce a new syntax to transform the noweb reference
> inline. Something like
>
> #+BEGIN_SRC emacs-lisp :noweb yes :tangle yes :noweb-prefix no
> <<preamble>>
> (setq latex-header <<(prin1-to-string nw)>>)
> #+END_SRC
You'd need to only allow a single function call with only one
argument, or use something like <<(prin1-to-string <<nw>>)>>. The
change would be much more complex than what I propose, for maybe
little benefit.
>> [...]
> This sounds a bit confusing. I would also add an example where it is
> useful to set :noweb-prefix to no.

I've added such an example in the revised patch attached.

Thanks for the feedback.

Regards,

-- 
Sébastien Miquel