BUG: Chinese characters are garbled in all import/export functionalities (UTF-8 encoding issue)

[junshiwang]

Affected module
UI / Backend
(All import and export functionalities)

Describe the bug
Chinese characters are corrupted (garbled text) in all import and export functionalities in OpenMetadata.
This issue is not limited to lineage export, but affects every feature that involves exporting or importing metadata files (e.g. CSV ).

Any metadata containing Chinese characters (table names, column names, descriptions, tags, glossary terms, etc.) becomes unreadable after export, or is incorrectly parsed after import.

To Reproduce

Prepare metadata containing Chinese characters (e.g. table name, column name, description, glossary term)

Perform any of the following operations:

Export lineage

Export metadata (tables / columns / glossary / tags, etc.)

Import metadata using CSV

Open the exported file or review the imported result

Observe that Chinese characters appear as garbled text

This issue is consistently reproducible across different import/export features.

Expected behavior
All import and export functionalities in OpenMetadata should fully support UTF-8 encoding.
Chinese characters should remain intact and readable:

After exporting metadata files

After importing metadata files back into OpenMetadata
No manual encoding conversion should be required.

Version:

OpenMetadata version: 1.11.3

Additional context
This appears to be a global encoding issue across OpenMetadata import/export pipelines:

Exported files may not explicitly use UTF-8 encoding

CSV / Excel exports may be missing UTF-8 BOM, causing Excel (Windows) to misinterpret encoding

HTTP response headers may not include charset=utf-8

Import logic may rely on system default encoding instead of enforcing UTF-8

Because the problem affects all import and export features, it likely requires a centralized fix in the export/import framework rather than a single feature-specific patch.

[Darshan3690]

Hi, @harshach maintainers! 👋

I'm working on this issue as part of a hackathon and would love to take
a shot at fixing it. I've analyzed the problem and here's my understanding
and proposed approach:

---

Root Cause Analysis

The garbled Chinese characters issue is likely caused by:

1. CSV export not explicitly enforcing UTF-8 encoding
2. Missing UTF-8 BOM (\uFEFF) in exported files — causes Excel on
   Windows to misread encoding
3. HTTP response headers missing charset=utf-8 in Content-Type
4. Import logic relying on system default encoding instead of
   explicitly specifying UTF-8

---

Proposed Fix

1. CSV Export — Add UTF-8 BOM

// Add BOM at the start of CSV output stream
outputStream.write(new byte[]{(byte)0xEF, (byte)0xBB, (byte)0xBF});

2. Writer/Reader — Enforce UTF-8 explicitly

// Instead of:
new OutputStreamWriter(outputStream)

// Use:
new OutputStreamWriter(outputStream, StandardCharsets.UTF_8)

3. HTTP Response Header

response.setContentType("text/csv; charset=UTF-8");

4. Import — Force UTF-8 on file read

new InputStreamReader(inputStream, StandardCharsets.UTF_8)

---

Files I Plan to Look At

• CSV export/import utility classes
• HTTP response handlers for file downloads
• Any BufferedReader / BufferedWriter / FileWriter usage
  without explicit charset

---

Could you please help me with:

1. ✅ Assigning this issue to me for the hackathon
2. ✅ Pointing me to the relevant files/classes for CSV export and import
3. ✅ Confirming if there's a centralized export utility class I should focus on

Thank you so much! Happy to discuss further. 🙏

— @Darshan3690

[harshach]

@Darshan3690 assigned it to you. make sure to have unit tests and playwrights along with if backend code changed tests there as well

[Darshan3690]

hi @harshach @PubChimps check the pr #27409 . thank you assign me under hackthon.

[Darshan3690]

Hi @harshach

I am participating in the Back to Metadata Hackathon. Could you please add the "good first issue" label to this issue so it becomes eligible for the hackathon?

Thanks!