I have this MWE:
#+TITLE: Org babel tangle and noweb
#+PROPERTY: header-args:shell :shebang "#!/usr/bin/env bash"
* Shell Script
#+BEGIN_SRC shell :noweb yes :tangle code.sh
exec emacs -Q "$@" \
--eval '(progn
<<elisp>>
)' 2>/dev/null </dev/tty
#+END_SRC
* Emacs Configuration
#+BEGIN_SRC emacs-lisp :noweb-ref elisp :noweb-sep "\\n\\n"
(menu-bar-mode -1)
#+END_SRC
Some comment.
#+BEGIN_SRC emacs-lisp :noweb-ref elisp
(line-number-mode 1)
#+END_SRC
When tangled (C-c C-v t), it generates:
Notice that the empty line is created between noweb segments as expected because of ":noweb-sep "\\n\\n"". But what was unexpected to me was the creation of empty line with spaces inserted to match the indentation.
Is this by design?
If it is by design, I would propose to have a new switch called something like ":noweb-notrailingspc" (would like a better suggestion for this switch name).
So the src block where noweb expansion needs to happen will look like this:
#+BEGIN_SRC shell :noweb yes :tangle code.sh :noweb-notrailingspc yes
exec emacs -Q "$@" \
--eval '(progn
<<elisp>>
)' 2>/dev/null </dev/tty
#+END_SRC
Now, tangling will generate
Here is a diff (without whitespace differences) of the change I made to ob-core locally to get this:
diff --git a/lisp/ob-core.el b/lisp/ob-core.el
index cb332ffea92..0f04d8b85e1 100644
--- a/lisp/ob-core.el
+++ b/lisp/ob-core.el
@@ -2730,9 +2730,7 @@ block but are passed literally to the \"example-block\"."
(funcall nb-add (buffer-substring index (point)))
(goto-char (match-end 0))
(setq index (point))
- (funcall
- nb-add
- (with-current-buffer parent-buffer
+ (let ((body-str (with-current-buffer parent-buffer
(save-restriction
(widen)
(mapconcat ;; Interpose PREFIX between every line.
@@ -2794,6 +2792,9 @@ block but are passed literally to the \"example-block\"."
"`org-babel-noweb-error-langs')"))
"")))
"[\n\r]") (concat "\n" prefix))))))
+ (when (string= "yes" (cdr (assq :noweb-notrailingspc (nth 2 info))))
+ (setq body-str (replace-regexp-in-string "^[ \n\r]*\n" "\n" body-str)))
+ (funcall nb-add body-str)))
(funcall nb-add (buffer-substring index (point-max))))
new-body))
Questions:
1. If the above empty lines with prefix number of spaces are not expected, the fix would be simpler.. to just throw in that (setq body-str (replace-regexp-in-string "^[ \n\r]*\n" "\n" body-str)) form in-between.
2a. If the spacey empty lines are as per design, then does this diff look good? Should I work on a proper patch?
2b. What would be a better/more intuitive switch name than ":noweb-notrailingspc"?