Using ImmGetCompositionString yields a different result than Microsoft IME.

[changhao6]

Description

Hi, I am using the Windows API ImmGetCompositionString to get the result of IMECompositionEnd. When I call ImmGetCompositionString, Microsoft IME returns "アイウエオ" but Google IME returns "アイ". What is the reason for this result? How can I get the same result as using Microsoft IME? This is very important to me, please take a look at this problem in your busy schedule, thank you!

Expected behavior

Here is MS IME behavior

get 'アイウエオ'

Actual behavior

Here is Google IME behavior

get 'アイ'

Version or commit-id

Google IME 2.28.5050.0

Environment

Windows 10-22H2

Additional context

Here is the demo I used to reproduce this issue, it is based on winforms:
https://github.com/changhao6/IMEDemo

[yukawa]

What's exactly the steps you tried? I actually got アイウエオ with the following steps.

1. Build and install the latest version of Mozc (c048e1f - you can download the installer file from our GitHub Actions page)
2. Build your demo app
3. Select Mozc as the current IME
4. Type aiueo to get あいうえお
5. Hit the Enter key to commit the composing text.

Here is the demo I used to reproduce this issue, it is based on winforms:
https://github.com/changhao6/IMEDemo

Most likely it's unrelated to this particular issue, but the following code would be a bit more accurate on how ImmGetCompositionString should be used from C#. Just fyi.

using System;
using System.Collections.Generic;
using System.Diagnostics;
using System.Drawing;
using System.IO;
using System.Linq;
using System.Runtime.CompilerServices;
using System.Runtime.InteropServices;
using System.Text;
using System.Threading.Tasks;
using System.Windows.Forms;

namespace IME
{
    internal class MyTextBox : TextBox
    {

        const int GCS_RESULTREADSTR = 0x0200;
        public const int WM_IME_COMPOSITION = 0x010F;
        public const int WM_IME_ENDCOMPOSITION = 0x010E;

        protected override void WndProc(ref Message m)
        {
            if (m.Msg == WM_IME_ENDCOMPOSITION)
            {
                MessageBox.Show(GetCompositionString(this, GCS_RESULTREADSTR));
            }
            base.WndProc(ref m);
        }

        static string GetCompositionString(TextBox textBox, uint dwIndex)
        {
            IntPtr hWnd = textBox.Handle;
            IntPtr hIMC = IntPtr.Zero;
            try
            {
                hIMC = ImmGetContext(hWnd);
                int dwSize = ImmGetCompositionString(hIMC, dwIndex, null, 0);
                if (dwSize > 0)
                {
                    byte[] buf = new byte[dwSize];
                    int copiedSize = ImmGetCompositionString(hIMC, dwIndex, buf, dwSize);
                    if (copiedSize >= 0)
                    {
                        return Encoding.Unicode.GetString(buf, 0, dwSize);
                    }
                }
                return string.Empty;
            }
            finally
            {
                if (hIMC != IntPtr.Zero)
                {
                    ImmReleaseContext(hWnd, hIMC);
                }
            }
        }

        [DllImport("imm32.dll", CharSet = CharSet.Unicode)]
        static extern int ImmGetCompositionString(
            IntPtr hIMC,
            uint dwIndex,
            [MarshalAs(UnmanagedType.LPArray, SizeParamIndex = 3)]
            byte[] lpBuf,
            int dwBufLen);
        [DllImport("imm32.dll")]
        public static extern IntPtr ImmGetContext(IntPtr hWnd);
        [DllImport("imm32.dll")]
        public static extern IntPtr ImmReleaseContext(IntPtr hWnd, IntPtr hIMC);
    }
}

[changhao6]

I'm sorry,I didn't make that clear. In fact, I just typed ai, and chose 'あいうえお' in candidate list. When I performed the same operation on Microsoft IME, I can get the whole 'ｱｲｳｴｵ'.

[changhao6]

@yukawa ,I had ran your modified code, but the problem still exists. Could you please take some time to look at it again?

[yukawa]

OK, I can finally reproduce the issue you described.

How can I get the same result as using Microsoft IME? This is very important to me, please take a look at this problem in your busy schedule, thank you!

The short answer to your original question is that unfortunately there is no way to get the same result as MS-IME unless we change the behavior of Mozc / Google Japanese Input. The behavior would be much similar as long as you use the space key to convert text instead of the down arrow key. As long as the user always uses the space key to convert the text, the issue is less likely to happen.

The long answer is that when APIs like ImmGetCompositionString were defined, prefix match suggestion / prediction wasn't popular in Japanese IMEs, and in fact MS-IME always yielded exact match results as if you hit the space key. We released Mozc/Google Japanese Input in the current manner (setting アイ as a reading string for prefix suggestion) in 2009, followed by MS-IME's starting supporting prefix match suggestion / prediction in their way (setting アイウエオ as a reading string based on the result string).

For this particular case, I think it'd make sense to update Mozc's behavior to match the MS-IME's behavior, but we cannot promise whether/when it would happen. I'll update this thread when we update Mozc's behavior to match the MS-IME's behavior.

Hope this helps.

[changhao6]

Thanks for your reply. Your answer has helped me a lot!