I've used noweb references to actually assemble what will be tangled all at
once. See how I did it here
<https://github.com/togakangaroo/daily-programmer/tree/master/sliding-puzzle>
.

The reason why the "incorrect" block is outdented is that tangle
automatically tries to guess indentation level. (Take that as a handwavy
thing - I don't know exactly how it does that nor how configurable it is)
The solution I've found is noweb - basically you *wouldn't* tangle any of
those blocks, but would create another non-evaluated or exportable block
that has noweb references to the above blocks and how you want them
arranged. *This* is the block you tangle.

I haven't yet figured out how, but it should be possible to automatically
configure things to run autopep8 after tangling - that would take care of
indentation issues.

Also I'm pretty sure there's no :weave header arg...at least I haven't seen
it used and can't find it documented anywhere.

On Fri, Nov 29, 2019 at 12:08 PM Norman Walsh <ndw@nwalsh.com> wrote:

> Hi,
>
> I’ve seen a couple of pointers recently to using Org mode and tangle
> to write more literate Emacs configurations. I use Org+babel all the
> time to write “interactive” documents, so I thought I’d try out tangle
> from Org.
>
> I didn’t want to start with something as comlicated as my Emacs
> config :-) so I figured I’d kick the tires with a small python
> program. That did not end well.
>
> Consider:
>
> #+TITLE: Python literate programming
> #+OPTIONS: html-postamble:nil
>
> It starts off as a completely standard Python3 program.
>
> ---%<------------------------------------------------------
> #+BEGIN_SRC python :tangle yes :weave no
> #!/usr/bin/env python3
>
> #+END_SRC
>
> It defines ~a~.
>
> #+BEGIN_SRC python :tangle yes
> def a():
>     print("a")
>
>
> #+END_SRC
>
> And ~b~.
>
> #+BEGIN_SRC python :tangle yes
> def b():
>     print("b")
>
>
> #+END_SRC
>
> Now ~c~ is a little more complicated:
>
> #+BEGIN_SRC python :tangle yes
> def c():
>    print("c")
> #+END_SRC
>
> Not only does ~c~ print “c”, it calls ~a()~ and ~b()~.
>
> #+BEGIN_SRC python :tangle yes
>    b()
>    a()
> #+END_SRC
>
> Finally, make it importable. Not that you’d want to.
>
> #+BEGIN_SRC python :tangle yes
> if __name__ == "__main__":
>     main()
> #+END_SRC
> --->%------------------------------------------------------
>
> That’s the script. It weaves into HTML more-or-less ok (there’s a
> weird black box at the front of indented lines, but I can come back to
> that later).
>
> It’s a complete mess when tangled.
>
> The extra blank lines between functions (to make pylint happy with
> some PEP guideline) have disappeared. I guess I could live with that,
> but the complete failure to preserve indention in the penultimate code
> block is a show stopper:
>
> #!/usr/bin/env python3
>
> def a():
>     print("a")
>
> def b():
>     print("b")
>
> def c():
>    print("c")
>
> b()
> a()
>
> if __name__ == "__main__":
>     main()
>
> (Also, why is there an extra blank line before the incorrectly
> indented block?)
>
> Is this user error on my part somehow? I suppose I could write my own
> version of tangle, though I’m not clear if the whitespace is lost in
> the tangle function or in the Org mode data model.
>
> Thoughts?
>
>                                         Be seeing you,
>                                           norm
>
> --
> Norman Walsh <ndw@nwalsh.com> | We discover in ourselves what others
> http://nwalsh.com/            | hide from us, and we recognize in
>                               | others what we hide from
>                               | ourselves.--Vauvenargues
>