emacs-rewriting-pcase

Use pcase to rewrite elisp source.

This elisp library provides a function for rewriting the s-expressions in a buffer according to a predicate. The rewriter performs a simple recursive macro-expansion using the predicate, but instead of prodcuing another sexpr, it replaces the source text corresponding to the S-expression with text for the replacement value.
The purpose of this rewriting is to maintain the relative position of surrounding code elements, including comments.

The current version only performs this function well when the S-expression is being replaced in-toto. This is adequate for use in my unboxed package manager, but eventually I'd like to have a proper rewriting pcase that lets the user move arbitrary subexpressions around and have the relevant comments and formatting be preserved.

The following code will produce a predicate that looks for S-expression patterns commonly used in the source code of packages containing data files. Using this predicate will rewrite the library to use a hard-coded path for locating their data files and allow the library code to be installed in a different location on the file system.

(defun make-replace-library-location (path)
  "Predicate to hard-code library path."
    (lambda (sexpr)
      (pcase sexpr
        (`(file-name-directory load-file-name) `(,path))
        (`(file-name-directory (or load-file-name . ,rest))
         `(,path))
        (`(file-name-directory
           (or . ,(and (pred listp)
               ls
   	           (guard (memq 'load-file-name ls)))))
         `(,path))
        (_ nil))))

Note the path string is in a cons cell to distinguish failure from indicating the S-expression should be replaced by nil. If we put the package's data files in the user's emacs directory under "data/some-package/", the replacement could be performed on the current buffer using

(rewriting-pcase--pcase-replace-sexpr 
  (make-replace-library-location 
    (file-name-concat user-emacs-directory
	                  "data"
					  "some-package/")))

Plan:

1. Get function that properly identifies how each piece of text maps to s-exp and comments correct
2. Convert from function arguments tracking parameters for current sexp to recursive data structure reflecting complete parse, including systematic associate of comments to syntactic units.
3. Derive a version of pcase that operates on the parsed sexp data structure from step 2, instead of on ordinary values. Must support the same patterns, presumably with a few additional operators for working with the enriched information.