bug: CodeEditor IME input on Windows resets caret and inserts both Pinyin and Chinese

[Qye-Leisure]

Duplicate Check

• I have searched the opened issues and there are no duplicates

Describe the bug

Just like the screen recording I attached, only English data works fine. Once switched to Chinese data, the cursor position automatically jumps to the beginning, and it inputs both Pinyin and Chinese characters together. Python supports Chinese variables.

Code sample

Code

[import flet_code_editor as fce

import flet as ft

CODE = """# 1
# 2
# 3
import json
import textwrap

print("Folding demo")
"""


def main(page: ft.Page):
    editor = fce.CodeEditor(
        language=fce.CodeLanguage.PYTHON,
        value=CODE,
        selection=ft.TextSelection(base_offset=41, extent_offset=62),
        autofocus=True,
        
        expand=True,
        on_selection_change=lambda e: print("Selection:", e),
    )

    async def fold_imports():
        await editor.fold_imports()

    async def fold_comment():
        await editor.fold_comment_at_line_zero()

    page.add(
        ft.Row(
            [
                ft.Button("Fold imports", on_click=fold_imports),
                ft.Button("Fold comment", on_click=fold_comment),
            ]
        ),
        editor,
    )


if __name__ == "__main__":
    ft.run(main)]

To reproduce

Just like the screen recording I attached, only English data works fine. Once switched to Chinese data, the cursor position automatically jumps to the beginning, and it inputs both Pinyin and Chinese characters together. Python supports Chinese variables.

Expected behavior

No response

Screenshots / Videos

Captures

QQ2026225-171352.mp4

Operating System

Windows

Operating system details

11

Flet version

0.81.0

Regression

No, it isn't

Suggestions

No response

Logs

Logs

[Paste your logs here]

Additional details

No response

[FeodorFitsner]

Thanks for the report and video, we were able to reproduce the behavior.

What we’re seeing looks like an IME composition sync issue on Windows (Chinese input): during composition, the editor state appears to be re-applied too aggressively (value / selection), which can reset the caret to the beginning and cause both pinyin + committed Chinese text to appear.

From code review, the likely hotspot is the extension-side sync flow in CodeEditor where selection/value updates are echoed in both directions during active editing.

We can try the following:

1. Skip applying external value / selection while IME composition is active.
2. Only apply external selection when it is an explicit app-side override (not a reflected frontend update).
3. Reduce selection round-trips unless needed (for example when on_selection_change is enabled).
4. Add a regression test for CJK IME composition in the code editor extension.

If this sounds good, we can put together a focused patch around those guards.

[Qye-Leisure]

感谢您提供的报告和视频，我们已经能够重现该问题。

我们看到的似乎是 Windows 上的 IME 编辑同步问题（中文输入法）：在编辑过程中，编辑器状态似乎被过于激进地重新应用（ value / selection ），这可能会将光标重置到开头，导致拼音和已提交的中文文本同时出现。

从代码审查来看，可能的热点是 CodeEditor 中的扩展端同步流，其中在活动编辑期间，选择/值更新会在两个方向上回显。

我们可以尝试以下方法：

1. 在 IME 合成激活期间，跳过应用外部 value / selection 。
2. 仅当存在明确的应用程序端覆盖（而不是反映的前端更新）时才应用外部选择。
3. 除非必要（例如启用 on_selection_change 时），否则减少选择往返次数。
4. 在代码编辑器扩展中添加针对中日韩输入法输入法的回归测试。

如果这个方案可行，我们可以围绕这些守卫制定一个重点补丁。

Thank you for your reply. I am a Python beginner, and Flet is the first framework I have learned. I really like this framework and would like to offer some useful suggestions. However, I have a limited understanding of the underlying technical details. Therefore, I can only contribute by identifying obvious bugs or proposing new features while I learn and use Flet.