Better IME support #658

shibukawa · 2015-12-11T18:53:16Z

Add better IME support (#41) to implement on-the-spot IME conversion. My code is based on #643.

I added the following functions:

glfwSetPreeditCallback() : callback function to notify preedit text
glfwSetPreeditCursorPos() : teach text cursor position for IME windows
glfwGetPreeditCursorPos() : get text cursor position for IME windows

They are not big changes. It keeps backward compatibility about API level.
To use IME feature, UI programmers should implement visualization code for preedit texts.

I implemented platform specific code for Windows/X11/Mac OS X (not Wayland and Mir).

shibukawa · 2015-12-13T14:30:47Z

I start implementing my text box, I think some features are needed in addition to the above functions:

Set/Get current IME status (integer only text box and so on)
IME status change callback (commit text when turn off IME)
Reset preedit text (reset text when widget focus would changed)

I will add these functions to my branch.

elmindreda · 2015-12-13T14:52:46Z

My initial impression is that this looks great! I will review it more closely once it's time for the IME part of 3.2, which should hopefully give you time to land the additional features you mention.

shibukawa · 2015-12-16T16:21:36Z

I implemented all method I mentioned and I added one test application (ime.c, almost as same as event.c, but IME functions are assigned to mouse buttons).

Current status is:

- Windows works almost perfectly except Google Japanese Input doesn't follow cursor position (MS IME works). - Mac OS X's NSTextInputContext doesn't work on my computer (10.9). So resetting preedit text feature doesn't work. - I tested on Ubuntu 15.10, but XIM (X Input Method) implementation is buggy. Test application can receive preedit text and commited text, but my real application can't. I am still debugging.

XIM is very old standard. I think adding Fcitx option (more modern protocol and most popular one now) is better solution (use Fcitx if it is installed, otherwise use XIM). And Fcitx supports Wayland too (I couldn't find Mir information). How about it?

elmindreda · 2016-04-05T10:08:36Z

There is quite a range of programming styles in this patch. Did you write this yourself or are parts of it copied from other sources? If so I need to know what licenses they come under.

elmindreda · 2016-04-05T10:09:46Z

The glfwGetIMEStatus and glfwSetIMEStatus functions should be folded into glfwGetInputMode and glfwSetInputMode.

elmindreda · 2016-04-05T10:14:21Z

The imm32 library should be dynamically loaded the same way as other non-default libraries.

shibukawa · 2016-04-05T17:29:21Z

Maybe 50% is my code and 50% is other's. I referred the following Japanese web sites:

Windows: http://www.kumei.ne.jp/c_lang/sdk3/sdk_278.htm
X11: http://homepage3.nifty.com/rio_i/lab/xlib/012xim.htm

I will ask them about sample code's license.

I couldn't find memo about Mac. Of course, I checked MSDN, Apple's document, X11R6 references.

shibukawa · 2016-04-05T17:38:27Z

I will merge glfw(Get|Set)IMEStatus into glfw(Get|Set)InputMode.

imm32 is provided as static library by Microsoft and it should be bundled with Windows like gdi32.dll. I thinks if(WIN32) is enough to detect imm32.

shibukawa · 2016-04-06T16:06:17Z

I rebased to fix conflict between current master.
I removed glfw(Get|Set)IMEStatus and added new options to glfw(Get|Set)InputMode
Ryo Ichinose (X11 guy) agreed using XIM related code as libpng/zlib license.

Now I am waiting reply from Yasutaka Kumei.

shibukawa · 2016-04-07T16:12:44Z

Yasutaka Kumei agreed using his code with glfw.

elmindreda · 2016-04-08T10:43:02Z

Awesome! Thank you!

elmindreda · 2016-04-08T10:54:31Z

About the imm32 thing, it's not for compatibility with older systems. As you rightly point out, the import library is available everywhere. Rather it's as a convenience for users, just like the current dynamic loading of winmm.

Fixes #41. Closes #658.

hesiod · 2017-02-11T23:21:25Z

@elmindreda It appears that you wanted to close this PR with 779b562, but Github failed to realize that, so you probably have to close this PR manually

dmitshur · 2017-02-12T00:09:15Z

It appears that you wanted to close this PR with 779b562, but Github failed to realize that, so you probably have to close this PR manually

@hesiod That commit is not yet merged into master branch. When it is, due to its commit message, this PR will be closed.

elmindreda · 2017-02-12T04:41:55Z

What @shurcooL said. Expect several more revisions to pop up here before it's ready to merge, as it's a big and multi-platform addition.

This fix is based on shibukawa's fix: glfw#658 shibukawa@d36a164 Some minor coding style changes are made, but not yet follow glfw's one, and some comments doesn't follow recent changes. So further work is needed. Co-authored-by: Yoshiki Shibukawa <[email protected]> Co-authored-by: Takuro Ashie <[email protected]>

This fix is based on shibukawa's fix: glfw#658 shibukawa@d36a164 Co-authored-by: Yoshiki Shibukawa <[email protected]> Co-authored-by: Takuro Ashie <[email protected]>

This fix is based on shibukawa's fix: glfw#658 The differences is the following. * Remove `X_HAVE_UTF8_STRING` branching since the current logic doesn't use it * Replace `XNDestroyCallback` for `XNPreeditAttributes` in `XCreateIC` Co-authored-by: Yoshiki Shibukawa <[email protected]> Co-authored-by: Takuro Ashie <[email protected]>

This fix is from shibukawa's fix: glfw#658 shibukawa@d36a164