Alexander Berntsen <[email protected]> writes:

On 28/06/16 16:01, Dmitry Alexandrov wrote:

...

I guess, you actually want to make ‘C-h’ equal to ‘DEL’ and ‘C-j’ to ‘RET’ *everywhere*, not just in isearch (there is no any evil-mode’s search, as far I know), helm, etc. Nothing is easier:

(define-key key-translation-map (kbd "C-h") (kbd "DEL")) (define-key key-translation-map (kbd "C-j") (kbd "RET"))

I would not reject ‘C-j’ though, it is useful sometimes to have two returns: a ‘smart’ one and a ‘dumb’ one.

I don't want DEL, I want backspace.

Do you have ‘<backspace>’ that is bound somewhere else than ‘DEL’? (See ‘<f1> k <backspace>’.)

So I changed it to (global-set-key (kbd "C-h") (kbd "<backspace>")) instead of delete-backwards-char. But it still doesn't work when searching.

Is there is any reason why you believe that this should work? I did not suggest you to do anything like this.

'/ab^J' will bring up search, and add 'ab' followed by literally the characters '^' and 'J'.

No, it’s followed by literally the character ‘^J’ (this is one character — ‘linefeed’ aka ‘\n’ is C notation). By the way, this is a good illustration on why one should not use ‘^J’ as notation for ‘C-j’, ‘^H’ for ‘C-h’, etc — cases when ‘[(control <something>)]’ inserts itself are extremely rare, mostly keys with control modifier are bound to some command.

'/ab^H' will bring up search, add 'ab' to the query, and then close it.

The problem in helm is different. If I bring up helm using helm-M-x, using ^J or ^H will not work for different reasons. ^J because it is bound to something else, and ^H because it's not bound to anything (i.e. ^H^H will print "C-h C-h is undefined").

Alexander Berntsen <[email protected]> writes:

...

...

...

I guess, you actually want to make ‘C-h’ equal to ‘DEL’ and ‘C-j’ to ‘RET’ *everywhere*, not just in isearch (there is no any evil-mode’s search, as far I know), helm, etc. Nothing is easier:

(define-key key-translation-map (kbd "C-h") (kbd "DEL")) (define-key key-translation-map (kbd "C-j") (kbd "RET"))

I would not reject ‘C-j’ though, it is useful sometimes to have two returns: a ‘smart’ one and a ‘dumb’ one.

I don't want DEL, I want backspace.

Do you have ‘<backspace>’ that is bound somewhere else than ‘DEL’? (See ‘<f1> k <backspace>’.)

...

So I changed it to (global-set-key (kbd "C-h") (kbd "<backspace>")) instead of delete-backwards-char. But it still doesn't work when searching.

Is there is any reason why you believe that this should work? I did not suggest you to do anything like this.

I confused what kbd "DEL" did with what the DEL key does. Using kbd "DEL" achieves the same as kbd "<backspace", as far as I can tell.

...

...

'/ab^J' will bring up search, and add 'ab' followed by literally the characters '^' and 'J'.

No, it’s followed by literally the character ‘^J’ (this is one character — ‘linefeed’ aka ‘\n’ is C notation).

Sorry, I meant visually. It is indeed literally the character '^J'.

In the end though, your suggestion unfortunately doesn't solve anything.

Sorry, I’ve missed *when* did you try this suggestion (to use ‘key-translation-map’). I’ve re-read our conversation (restored in quotes above) — and all you said is that you had merely ingored it and tried something else with ‘global-map’ (‘global-set-key’).

Alexander Berntsen <[email protected]> writes:

On 30/06/16 23:03, Dmitry Alexandrov wrote:

...

Sorry, I’ve missed *when* did you try this suggestion (to use ‘key-translation-map’).

This is what I get for answering email in a hurry. (Twice!)

Your suggestion works perfectly. Thank you. And double thank you for your patience in pointing out that I misread. (Twice!)

You are welcome. :-) Now about flies in the ointment. Being a more low-level thing than ‘global-map’, ‘key-translation-map’ acts not on a whole key sequence, but rather on every single key chord. I. e. it affects not only ‘C-h’ and ‘C-j’ on their own, but also, for instance, ‘C-x C-h’ or ‘C-c C-j’ whatever and wherever they mean. There is should not be any practical problem with ‘C-h’, since it almost everywhere stands for ‘help’ and has ‘<f1>’ as a synonym. But if you’ll stick to ‘C-j’ mapped to ‘C-m’ / ‘RET’, you have to sort it out somehow. The laziest way is just to use ‘C-S-j’ (‘S’ stands for shift) whenever the ‘real’ ‘C-j’ is required. But I have to repeat that I would not consider giving up ‘C-j’ even on its own a good idea — it is very useful to have ‘dumb’ and ‘smart’ return keys, at least for any sort of auto-completion — smart one selects the most relevant completion, while dumb one applies exactly what you’ve typed. (Which chord is dumb and which is smart depends on personal preference — there is no consistency through GNU Emacs packages, for example in ‘icomplete’ ‘C-j’ is smart by default and ‘C-m’ / ‘RET’ is dumb, while in ‘ido’ — exactly the way around.)