7.26.314.1 Issues with character encoding, prone to garbled text

[kidzgy]

Version：7.26.314.1

Steps to reproduce:

1. Create a new text document, paste the following content into it, and save it in ANSI encoding.

新的
旧的

2. Drag the text file onto notepad3.exe to open it; you will notice that the text appears as garbled characters.
3. Close the text file; when you drag it onto notepad3.exe a second time, the text will appear normally again.
4. Edit Notepad3.INI, clear the history under [Recent Files], then drag the text file into notepad3.exe again—it will display as garbled text once more.

Video:
https://github.com/user-attachments/assets/ba39e2ca-3af3-4e8d-823a-7a8f3f0e93d7

[RaiKoHoff]

Yes, there are to few characters available, that the Machine-Learning model of UCHARDET is able to detect the correct ANSI encoding. Possibilities:

• Use UTF-8 encoding everywhere (recommended) (https://utf8everywhere.org/)
• Use menu File -> Encoding -> Revert from File as -> GB18030 (Ctrl-Shift-F10)
• Use menu File -> Encoding -> Revert from File as -> More Encodings... (select correct encoding)
• Enable Parse encoding file tags and put as first/last line "encoding:<ENC>", where <ENC> is your encoding, e.g. GB18030
• Type more different characters into document (best a normal text), so the UCHARDET detector is able to identify the correct encoding, cause UCHARDET is trained on normal text.

If you set [Settings2] DevDebugMode=true in your Notepad3.ini, you will see some more information about encoding detection in title bar.

[kidzgy]

Personally, I don't think the encoding format should be determined by the number of characters. After all, the encoding is set at the time each document is created, and Notepad should simply prioritize using the specified encoding. Arbitrarily changing the encoding will only make things worse.
For example, my CSS and HTML files are usually in UTF-8, but my batch files are in ANSI, so different files have different encoding formats.
This issue is just one example I've cited; I've also encountered cases where files still display garbled text after opening, and the encoding is arbitrarily changed from UTF-8 to GB18030.

[RaiKoHoff]

There are more ANSI Encodings in the world than only yours. Notepad3 tries to support them all. If there is no other hint for the right encoding, the best method is to train a Machine-Learning model to predict what is the right encoding by learning from language the corresponding ANSI language (https://www.freedesktop.org/wiki/Software/uchardet).

Personally, I don't think the encoding format should be determined by the number of characters.

Of cause this prediction gets better the more character are in the document.

After all, the encoding is set at the time each document is created, and Notepad should simply prioritize using the specified encoding.

Wrong, after writing the file in any ANSI Encoding, there is no hint, which encoding the byte-sequence in this file has.
So you have to find a way to interpret the byte-sequence and predict an encoding.

For example, my CSS and HTML files are usually in UTF-8

In lots of older files you will find an encoding hint hin the first line of these files, meanwhile UTF-8 is accepted as the golden standard - rtf https://utf8everywhere.org/zh-cn.

This issue is just one example I've cited; I've also encountered cases where files still display garbled text after opening, and the encoding is arbitrarily changed from UTF-8 to GB18030.

Please provide examples for analysis (only, if the contain more than 12 chars).

[RaiKoHoff]

Maybe in your case the best settings would be:

Upcoming PR has a bugfix for the pipeline, where UCHARDET prediction is skipped.

[kidzgy]

Maybe in your case the best settings would be:

Upcoming PR has a bugfix for the pipeline, where UCHARDET prediction is skipped.

Video_2026-03-27_193210.mp4

Following your advice, I recorded a video.
The following issues have occurred:
(1) When opening CSS and JS files, the encoding is automatically changed from UTF-8 to ANSI.
(2) When opening batch files, Chinese characters appear as garbled text. Although the encoding hasn’t been altered, as I mentioned yesterday, the text is garbled the first time I open the file, unless I open it a second time.

I've packaged a few test files into test.zip.

What I mean is, I don’t need Notepad3 to determine whether the encoding is ANSI or UTF-8. I just want the encoding to be the same when I open the file in the system’s built-in Notepad as when I open it in Notepad3. Currently, this isn’t possible in version 7.26.314.
If the encoding is changed without my knowledge and I press CTRL+S, it could lead to potential issues.

test.zip

[kidzgy]

Video_2026-03-27_195155.mp4

Now let's compare it with 7.26.306: the file was originally encoded in a certain format, and when opened in Notepad3, it remains in that same format—no changes were made.