Word commands do not respect word boundaries of the current syntax table

[TheBB]

Originally reported by: Mika Vilpas (Bitbucket: sp3ctum, GitHub: sp3ctum)

---

Commands operating on words, such as w and *, don't seem to respect the syntax table of the current mode for me. In this isssue it is described to alter the syntax table to include, for example, dashes (-) as being part of a word: https://bitbucket.org/lyro/evil/issues/222/w-command-is-not-compatibile-with-vims

I have this in my config:

;; http://timothypratley.blogspot.fi/2014/08/clojure-friendly-word-definitions-in.html
;; Treat these characters as word characters. Works great with evil-mode!
(dolist (c (string-to-list ":_-?!#*"))
  (modify-syntax-entry c "w" clojure-mode-syntax-table)
  (modify-syntax-entry c "w" emacs-lisp-mode-syntax-table))

This used to work in evil-20141003.1015, but now I've upgraded to evil-20150818.1313 and the functionality is broken for me. Is there a way to work around this? Have I misconfigured somehow?

---

• Bitbucket: https://bitbucket.org/lyro/evil/issue/565

[TheBB]

Original comment by sykora NA (Bitbucket: sykora, GitHub: sykora):

---

I just upgraded to emacs 25.1, and I can confirm that something's going on with word selection.

Suppose I have the following text (with cursor between the |'s):

#!emacs-lisp
;; This buffer is for te|x|t that is not saved, and for Lisp evaluation.

Then pressing viw selects all the following text (between |s):

#!emacs-lisp
;;| This buffer is for text that is not saved|, and for Lisp evaluation.

After some digging around in the emacs 25.1 release notes, I noticed that there were some changes to subword-mode which looked promising. On enabling subword-mode, I get the correct behavior (only text is selected).

I can't quite figure out how to fix it, but for reference, the link to the relevant changes in the release notes are here.

[TheBB]

Original comment by Mika Vilpas (Bitbucket: sp3ctum, GitHub: sp3ctum):

---

Sorry, I have downgraded to emacs 24.5.1 on Windows and never gave this any more thought. Nothing stuck out in the changes, but I haven't looked into it extensively.

[TheBB]

Original comment by Frank Fischer (Bitbucket: lyro, GitHub: lyro):

---

Anything?

[TheBB]

Original comment by Frank Fischer (Bitbucket: lyro, GitHub: lyro):

---

Interesting. Any idea what is about to change in Emacs 25?

[TheBB]

Original comment by Mika Vilpas (Bitbucket: sp3ctum, GitHub: sp3ctum):

---

I just downgraded to emacs 24.5.1 and the problem is gone.

[TheBB]

Original comment by Mika Vilpas (Bitbucket: sp3ctum, GitHub: sp3ctum):

---

So this manifests like this. Say the current buffer contains this text, and point be at $, without including the $:

$evil-is-the-best-mode-ever

If I input a *, I get a search for evil, instead of evil-is-the-best-mode-ever. In contrast with what I said in the OP, w and b seem to treat the identifier as a single word. However, with point at $, e will jump like this:

evi$l-is-the-best-mode-ever

[egg-at-reify]

@TheBB thanks for all your input on this -- it helped me at least find a workaround. In case it's useful to others: I added a hook to disable subword-mode in clj/s buffers:

(add-hook 'clojure-mode-hook (lambda () (global-subword-mode -1)))

[egg-at-reify]

(or possibly other folks' input now that I look more closely at the comments.)

Note that unlike the comment above mine, which described, enabling subword-mode as giving correct behavior, for me it was the opposite.