[Orgmode] Re: A tool for creating source code files from example and src blocks in org files

[Eric H. Neilsen, Jr.]

Chris,

Yes, I am also unhappy with the use of numbering to order chunks, but 
the traditional LP mechanism of using substitution of named chunks has 
some major flaws I have not figured out how to address.

The basic problem is that it can be used as a mechanism for code reuse, 
probably will be if it is present. This is not necessary (code reuse 
like this probably means poor use of the code reuse mechanisms of the 
programming language in any case), hides the fact that the code is 
reused from debugging and performance tools that rely on the source 
code, and makes the task of untangling much harder, and impossible to 
completely automate. Consider a program in which the same chunk appears 
twice in the tangled source code. Using a debugger, you find a bug in 
the chunk. You fix it in the source code, but only in one place. You 
then find another bug, or even the same bug in a different guise, and 
fix it differently in the other appearance. What is the untangler 
supposed to do with the result? Which new version do you want? Do you 
want the changes merged? Do you actually want both versions, and if so, 
how does it edit the org-mode document to include them?

Yes, I know, from an LP purists point of view, the untangler is an 
abomination. Unfortunately, few (if any) code development tools 
(debuggers, performance analyzers, etc.) support debugging or analysis 
of source code embedded in CWEB, noweb, or org-mode text files. This, to 
me, is a deal-breaker. With an untangle command, I can write my code in 
an org-mode file, tangle it and get emacs buffers with all C, headers, 
makefiles, java, or whatever code in it, use emacs's extensive code 
development tools (eg the emacs front end to gdb) to build and debug it, 
and pull the source code back into the org-mode file when the bugs are 
fixed. In fact, the code development tools do not even need to be 
embedded in emacs; anything tools can look at a traditional source file 
becomes useful.

There is an additional problem, although one that can be solved by "just 
doing more work." At present, org-tangle can use the existing org-mode 
export code to do all necessary weaving. org-mode, in turn, takes 
advantage of the many independently supported modes for each language to 
format them properly. If we introduce a new syntax (for substitution) 
inside the literal blocks, it means that we will need an org-weave that 
can properly format each language. I am not sure how practical this is; 
I am certainly not that ambitious.

It may be a little while before my itself makes an appearance. Not only 
do I need to wait for my employer to figure the legal stuff, I have also 
received enough feedback that I want to address issues better before 
anyone else is tempted to use the code.

   -Eric

--
Eric H. Neilsen, Jr.
http://home.fnal.gov/~neilsen