Replacing string in a region and conditional replace

Question

My goal is to write a simple function that can do the following:

Staring with an expression that looks like

$$ x = y $$
$$ y = z $$
$$ z = x $$

(There can more lines of such equations)

I want to write a elisp function that when I run it after selecting this region, I get the following

\begin{align}
x = y \\
y = z \\
z = x
\end{align}

I need to do a conditional replace on $$ depending on whether it is near the beginning of the line or the end of the line. Moreover, I also need to do not replace the last $$ with \\ but with \end{align}. I am new to elisp and I could not figure out how to do a conditional replace and also how to restrict the replace to the region I selected.

Any help is greatly appreciated.

Answer 1

If we can make some simple assumption the task is relatively simple and can even be done interactively with two consecutive normal query-replace-regexp commands.

Assumptions:

1. There are no empty equations. I.e., all strings between equation delimiters $$ that consist of space syntax characters and newlines only belong to separators between equations.
2. If one removes all equation delimiters $$ from the first assumption the remaining delimiters only belong to those ones that need to be replaced by \begin{align*} and \end{align*}. For an example there are no strange comments where a single equation delimiter $$ in them.

In the first run use assumption 1 and replace equation delimiters $$ that just separate equations:

M-C-% $$[[:space:]^J]+$\$ RET \\\\^J RET

The emacs manual describes regular expressions in full extent. The most important details for our use case are listed in the following:

• The dollar sign $ stands for itself except at the end of the regular expression where it matches the empty string at end of line. That is the reason why the last dollar sign is escaped with a backslash.
• The outer brackets [...] stand for a single character from a character set.
• The inner brackets within the character set [:space:] stand for characters with the space syntax class.
• You should not copy paste the character ^J. This character stands for the newline which you get if you type C-q C-j. The quoted-insert command bound to C-q causes that the following character is not looked-up in the active keymaps but inserted literally and C-j stands for the newline character #x0a. Note: C-a == 1, C-b == 2, ..., C-i == 9, C-j == 10 == #x0a.
• The + sign behind the character set indicates that one or more characters from the character can match (we eat one or more spaces or newlines).

In the second run replace the even equation delimiters $$ with \begin{align*} and the odd ones with \end{align*}. You can use the lisp evaluation form \,(...) to differentiate between odd and even in the replacement string.

M-C-% $\$ RET \\\,(if (evenp \#) "begin" "end"){align*} RET

Answer 2

This might suffice:

(defun my-command (beginning end)
  "Convert \"$$ ... $$\" lines in the region to an {align} block."
  (interactive "*r")
  (setq end (copy-marker end))
  (save-match-data
    (save-excursion
      (goto-char beginning)
      (insert "\\begin{align}\n")
      (while (re-search-forward "^\\(\\$\\$ \\).*\\( \\$\\$ *\\)$" end t)
        (replace-match " \\\\\\\\" nil nil nil 2) ;; replace the suffix
        (delete-region (match-beginning 1) (match-end 1))) ;; delete the prefix
      (delete-region (- (point) 3) (point)) ;; delete the final " \\"
      (insert "\n\\end{align}")
      (set-marker end nil))))