Hello Nicolas, Nicolas Goaziou writes: > Hello, > > Alan Schmitt writes: > >> I've been playing with block chaining to generate some dot file then to >> export then as images. I had a little trouble finding the number of '\' >> I need to put in front of a quote if I want the quote to be quoted. Here >> is a way to make it work: >> >> #+name: foo >> #+begin_src emacs-lisp :exports none >> "bar [label = \"\\\\\"test1\\\\\"\"]\nbaz [label = \"\\\\\"test2\\\\\"\"]" >> #+end_src >> >> #+results: foo >> : bar [label = "\\"test1\\""] >> : baz [label = "\\"test2\\""] >> >> #+begin_src dot :file ~/tmp/test-dot.png :var input=foo :exports results >> graph { >> $input >> } >> #+end_src >> >> My question is: why can't I simply use this: >> >> #+name: foo >> #+begin_src emacs-lisp :exports none >> "bar [label = \"\\\"test1\\\"\"]\nbaz [label = \"\\\"test2\\\"\"]" >> #+end_src >> >> #+results: foo >> : bar [label = "\"test1\""] >> : baz [label = "\"test2\""] >> >> (I guess the answer is in the error in replace-regexp-in-string: >> (error "Invalid use of `\\' in replacement text") >> .) > > Indeed. This function, unless told not to, treats backslashes characters > specially. > >> Would it be problematic to first transform every "\\" into a "\\\\" in >> org-babel-expand-body:dot, before the call to >> replace-regexp-in-string? > > I think `replace-regexp-in-string' should be called with a non-nil > LITERAL argument in this case. Yes. I think it should also not try to mach the case (i.e., FIXEDCASE should be non-nil). Here is a patch to do that.