libera/#clim - IRC Chatlog

13:11:38 cmack morning! coffee made; ready to look at wayland IME things again.

13:29:06 calsys jackdaniel: I don't know much about input source, but here's my little knowledge: there are 2 type of input sources, one is called keyboard layout, which will remap keys on your keyboard to others

13:30:53 calsys in this situation, C-r will probably being translated into C-す or something. It can be seem as intended, since user have switched the layout. the extra key can probably allowed to be bind with commands

13:32:55 calsys There are japanese (one kind), Vietnamese, Laos and some south-east asia country using this pattern, as the characters in their languages are lesser than total keys on the keyboard

13:34:08 calsys Another type is so-called input method. It need to popup a panel on your screen and compose text inside it

13:34:30 calsys which is adopted by chinese, korean, and now japanese

13:36:34 calsys in this situation, a Ctrl-modified character, like C-r, cannot be recognized by the input method. So it will interrupt the text composing, and being conveyed to the app as-is

13:37:58 jackdaniel calsys: from what I've observed while using SDL2, when I press a key with a modifier and there is a character associated with the key without the modifier

13:38:13 jackdaniel then SDL2 sends two events: text-input event (without setting there the modifier)

13:38:20 jackdaniel and a key event (with key and a modifier)

13:38:32 calsys jackdaniel: as you can see in my GIF, I'm under chinese input source, then I type C-a C-e to successfully navigate

13:39:00 calsys jackdaniel: yes, I found the problem. keydown is sent together

13:39:01 jackdaniel and there is no way to tell whether they correspond to the same user action or not (that's on linux at least)

13:39:49 jackdaniel (this is why I've concluded that SDL2 approach unusable for CLIM that processes both keys and characters)

13:40:38 calsys in my demo, I chosenly ignore keydown events when IME is enabled: only those non-graphic characters (#\Rubout, #\Newline) are processed, others are ignored, as they've all be taken by the IME

13:41:19 calsys it works very fine with chinese input source, probably also all CJK input sources, because they work in very similar way

13:41:28 jackdaniel another observation, and my memory may be slightly off here, is that even when IME window is not opened, there are IME-like events sent by SDL2 for each individual character (even for us layouts)

13:42:17 calsys jackdaniel: You're testing on linux with ibus or fcitx? that can be happened there

13:42:42 jackdaniel no, I had no IME enabled

13:42:54 jackdaniel yet sdl2 was still firing these events for individual characters

13:43:10 calsys have you ever have ibus/fcitx in the background?

13:43:17 jackdaniel no

13:43:22 jackdaniel I don't even have them on my system

13:43:34 calsys oh, i got it.

13:44:17 calsys in my demo i selectively enable text-input of SDL2

13:44:49 calsys only when I probed that user have activated an input method

13:44:56 cmack I haven't been testing with SDL yet but my understanding from reading the docs was that when "Start text input" event was signaled, the other key events should be ignored.

13:44:59 jackdaniel and how do you probe that?

13:45:53 calsys my demo is for macOS, and I write a native interface using Carbon.framework to detect current active IME in real time

13:45:59 jackdaniel cmack: but it sends it even for individual characters, when ime is not enabled; that'd mean that we can't process keys that at the same time have character bindings? (disregarding of whether we have pressed a modifier or not)

13:46:29 jackdaniel calsys: ah, I see; and based on that flag you either handle sdl2 text input events or you ignore them? and if it is "on", then you ignore key presses?

13:46:44 calsys cmack: no, they should not, since non-graphical characters will not be sent through textinput event

13:48:05 calsys jackdaniel: yes, if there is an active input method, then I StartTextInput; otherwise I stop it

13:48:40 cmack jackdaniel: that I don't know, yet... I thought that text-input and text-input-editing wouldn't be sent at all until the "start text input" command was given. that was still on my things to-check list. But I'm focused on the wayland side first :/

13:49:09 jackdaniel I'' answer later, afk; thanks for elaborating

13:49:22 calsys ok, thanks

13:49:56 cmack I want to get some proof-of-concept code together so it'll be easier to be concrete about differences

13:50:28 calsys cmack: I do think you are right, the textinput event will not be sent before StartTextInput. but it will be sent after it started, whether what your current IME is

13:51:19 calsys cmack: I'll try to upload my demo to git service, so you can better examinate it

14:11:03 calsys Now you can inspect my modification here: https://git.sr.ht/~calsys/McCLIM

14:15:27 cmack calsys: thanks! I see some similarities in our approaches but I may continue on my own path for wayland and then compare when I get it working. I have the IME communication but I only have the commit (sdl2 text-input equivalent) event working, not preedit

14:16:05 calsys jackdaniel: also i wich you can checkout here when you have time: the flexi-stream cannot be passed to OPEN in LispWorks, so it will cause platform-specific error in TTF module. The patch is here, you can git am it, or do any change you want: https://paste.sr.ht/~calsys/af01c56880a4a1267ba7bf5cf60ba2e70e4354b4

14:16:43 cmack calsys: out of curiosity, are you @lucky_magick on mastodon?

14:18:23 calsys cmack: no, i don't use mastodon. my id in all platforms are likely calsys, calsys456 or calsys3456, "The Calendrical System". my previous nickname is apr3vau, so-called April & May

14:19:29 calsys cmack: I don't have achieved pre-edit too. Maybe we can get current these things work, then consider preedit after

14:19:49 calsys then we'll not give ourselves too much workload

14:19:51 cmack ok, that person seems to also be macos focused with some CLIM, too. since the community is small, so far, I thought maybe it might be you too.

14:20:36 calsys Yeah, it's small. I'm more active in LispWorks user group actually :D

14:24:07 cmack I see. Ok, I'm going to kinda be afk to focus on code for little bit

14:46:16 calsys jackdaniel: please also check this patch when you have time, the default font path on macOS is not complete. The system library font folder and user library font folder is missing. patch is here: https://paste.sr.ht/~calsys/444fe74bc42cfe9a84be491d851732d2b108734f

14:56:04 jackdaniel I'll look into it, but first I want to make fastgf work. currently, w/o fastgf, I'm slowly approaching ccl speed https://turtleware.eu/static/paste/ff4fdf04-pre-fastgf-faster.png

15:03:46 jackdaniel with it I hope to approach sbcl and lw

15:06:01 calsys wow, i'm really expecting

18:43:31 jackdaniel cmack: re start-text-input, having in mind that I may misremember things (!), -- it was the case that each character was proposed as a separate IME sequence

18:44:29 jackdaniel the patch for search path for osx is trivial, I'll apply it now

18:48:42 jackdaniel calsys: re lispworks refusing to open a stream -- we shouldn't open the stream at all

18:48:55 jackdaniel so I think that this should be fixed in ensure-truetype-font

18:49:23 jackdaniel something like (unless (streamp source) (setf source (open ...)))