Basically, I want to split a multiline string at the regular expression that, in Perl, would be specified with \n(?=[^\W\d]\w*=).

Note that the Perl regexp features a "zero-width lookahead match" ((?=...)), meaning that the substrings that match this part of the regexp will not be "consumed"; they will appear in their corresponding entries in the resulting list.

For example, the string whose printed representation is


would be split into the following three-element list:

("A=1" "B_C_D=\nfoo\nbar\nbaz\n  =whatever=" "X_Y_Z=quux\n")

Even though Elisp regexps do not support lookahead/lookbehind expressions, this sort of splitting problem (in which part or all of each matching substring is not consumed) is common enough that I hope there are some other standard ways to solve it in Elisp.

Moreover, the ultimate goal here is to split the output of the shell command

/usr/bin/zsh -ilc 'printenv' 2>/dev/null

so that each element of the result would correspond to one "setting" of an environment variable1. So I also figure that Elisp may possibly have functions that facilitate the handling of this special case2.

1 The desired result is conceptually similar to value of the variable process-environment, but in general it is not be identical to it. Also, the side-effects associated with setting and updating process-environment do not enter into my problem.

2 Unfortunately, it appears that process-environment is initialized in C-code, so whatever functions do that are not available in Elisp.


This might not be quite the answer you want, but use printenv -0 instead. This puts a NUL between each "line" of output, which in this case will be between each var=value pair:


Parsing that will be much easier.


(split-string (shell-command-to-string "printenv -0") "\0" t) 
| improve this answer | |

I think db48x has the right answer for your underlying problem (without the -0 argument, printenv's output is ambiguous in case one of your env vars contains ...\nfoo=bar).

But as for how to do it in Elisp, I'd do it as follows:

(defun parse-foo (str)
    (insert str)
    (goto-char (point-min))
    (let ((start (point))
          (strs ()))
    (while (re-search-forward "\n\\w+=" nil 'move)
      (push (buffer-substring start (match-beginning 0)) strs)
      (setq start (1+ (match-beginning 0))))
    (cons (buffer-substring start (point)) strs)))
| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.