Add s-replace-regexp

[Silex]

Don't merge this yet.

@magnars: I need your guidance for the next points

Interface is quite different:

(s-replace OLD NEW S)
(s-replace-regexp REGEXP REP STRING &optional FIXEDCASE LITERAL SUBEXP START)

Do we want all the additional parameters?

Replace all matches for REGEXP with REP in STRING.

Return a new string containing the replacements.

Optional arguments FIXEDCASE, LITERAL and SUBEXP are like the
arguments with the same names of function ‘replace-match’.  If START
is non-nil, start replacements at that index in STRING.

REP is either a string used as the NEWTEXT arg of ‘replace-match’ or a
function.  If it is a function, it is called with the actual text of each
match, and its value is used as the replacement text.  When REP is called,
the match data are the result of matching REGEXP against a substring
of STRING, the same substring that is the actual text of the match which
is passed to REP as its argument.

To replace only the first match (if any), make REGEXP match up to \'
and replace a sub-expression, e.g.
  (replace-regexp-in-string "\\(foo\\).*\\'" "bar" " foo foo" nil nil 1)
    => " bar foo"

TODO:

• Need to add more examples.
• Need to decide wether we want to limit the additional parameters to s-replace-regexp.
• Need to decide wether to implement s-replace-all-regexp.

[coveralls]

Coverage decreased (-4.06%) to 95.313% when pulling 126b962 on Silex:master into 71f2902 on magnars:master.

[coveralls]

Coverage remained the same at 99.375% when pulling b6cdd53 on Silex:master into 71f2902 on magnars:master.

[magnars]

I think it makes sense to keep the optional parameters. However, the order of the required parameters should match s-replace. In particular, note that all s- functions take the original string as the first argument, for ease of threading with s-with.

[Silex]

In particular, note that all s- functions take the original string as the first argument, for ease of threading with s-with.

Hum, sorry but https://github.com/magnars/s.el#s-replace-old-new-s says otherwise (oldtext, newtext, originalstring)?

I think it makes sense to keep the optional parameters. However, the order of the required parameters should match s-replace

I think it does as shown in my 1st post.

Anyway, that means we are good with (defalias 's-replace-regexp 'replace-regexp-in-string).

I'll add more example, in the meantime what do you think of s-replace-all-regexp? the way s-replace-all works is by building a giant "or" regexp and do one call to replace-regexp-in-string... I believe attempting the same could be problematic, but looping over the cons of replacements and calling N times replace-regexp-in-string could work. That or alternatively we don't implement s-replace-all-regexp at the moment.

[magnars]

Yes, you are right. The original string is supposed to be last. All good then. I think s-replace-all-regexp needs a bit of thought, and can be skipped at the moment.

[coveralls]

Coverage remained the same at 99.375% when pulling 72ab5c0 on Silex:master into 71f2902 on magnars:master.

[Silex]

Alright, review the commit and suggest changes or merge 😉

[salamynder]

Thank you guys for putting so much work into this.

Regarding the necessity of s-replace-all-regexp:

The normal behaviour of replace-regexp-in-string replaces every occurence:

(replace-regexp-in-string "foo" "bar" "foo foo")
"bar bar"

(meaning: we don't need s-replace-regexp-all, but..); if this behaviour is not wished for one shall do subexpression replacement (last argument):

(replace-regexp-in-string "\\(foo\\).*\\'" "bar" " foo foo" nil nil 1)
    => " bar foo"

But this usage would again conflict with (s-with)-threading, because last arg isn't any longer the input-string but as said the subexpression number. (?!)

Overall problem: Cannot use &optional args of (replace-regexp-in-string) with (s-with)

[magnars]

Overall problem: Cannot use &optional args of (replace-regexp-in-string) with (s-with)

Yes, this is the case for all functions with optional args. Unfortunately my early standardisation on having the original string in the last position bites us here.

[magnars]

Looks good to me!

[Silex]

(meaning: we don't need s-replace-regexp-all, but..); if this behaviour is not wished for one shall do subexpression replacement (last argument):

I think you missunderstand s-replace-all-regexp, see https://github.com/magnars/s.el#s-replace-all-replacements-s

That is, s-replace already replaces all occurences.

[salamynder]

replacing all pairs... of course! sorry for the noise. And thanks for this fix!

[michaelmhoffman]

This is what I use as a replacement for s-replace-all-regexp:

(require 'dash)

(defun replace-regexp-cons (s replacement)
  "REPLACEMENT is a cons-cell. Each `car` is replaced with `cdr` in S.
n
`car` is a regexp.
`cdr` can include backreferences."
  (replace-regexp-in-string (car replacement) (cdr replacement) s t))

(defun replace-all-regexp (replacements s)
  "REPLACEMENTS is a list of cons-cells. Each `car` is replaced with `cdr` in S.

`car`s are regexps.
`cdr`s can include backreferences."
  (-reduce-from #'replace-regexp-cons s replacements))

It works well but has a dependency on dash.