AI & Agent Dev Bug Sandbox Back to Radar

XMLOutputParser Streaming Drops Valid Unicode Root Tags

Streaming with XMLOutputParser silently loses data when the root XML element uses a valid Unicode name (e.g., Chinese tags). The parser's start-tag detection regex only matches ASCII letters, colon, and underscore, causing it to skip Unicode roots and either return empty output or omit the root hierarchy.

highConfidence 95%Langchain-Core Affected V1.6.6 Affected Vmaster

Origin Analysis

The _StreamingParser in langchain_core/output_parsers/xml.py uses the regex r"<[a-zA-Z:_]" to locate the first XML element. This regex does not account for Unicode NameStartChar characters defined in XML 1.0, so roots like <答案> are ignored. If the root is Unicode but a child is ASCII, the parser starts at the child, losing the root; if both root and child are Unicode, no start tag is ever matched and streaming yields nothing.

Run the provided Python script with langchain-core==1.6.6. It creates a FakeStreamingListLLM that returns fixed XML strings and pipes them through XMLOutputParser. Compare invoke vs stream/astream for three cases: <答案><选项>蓝色</选项></答案>, <答案><choice>blue</choice></答案>, and <answer><选项>蓝色</选项></answer>. The first returns empty for stream/astream, the second returns only the child without root, the third works correctly.

Fixing Code Block

0x7F 45 4C 46 02 01 01 000x90 12 AB 04 C1 7E 33 08
0xA4 1C 90 3E 77 0B D2 580x11 F6 6A 2C 80 19 DE 44
0x00 00 00 00 03 00 3E 000x01 00 B7 4D 55 90 0C 61
0x6C 69 62 63 2E 73 6F 2E0x36 00 8E 17 42 AA 05 F3
0xDE C0 DE 01 8B 45 FC 890xC7 5D 20 6F 91 03 E4 78
0x13 37 00 4F 9A 2D E1 700x08 B4 5C 22 71 D0 4A 16
0x48 8D 3D 22 01 00 00 E80x91 FF 0A 63 2E 57 B8 04
0x5A 91 C3 0E 44 70 1B 880x02 6F 39 C4 15 80 ED 27
0xFF 15 40 20 60 00 48 890xC3 90 7A 01 4C 28 B1 56
0x0F 1F 44 00 00 55 48 890xE5 C3 18 6D 92 0B 33 A7

Edge Case Audit

This fix broadens the start-tag detection to include all Unicode word characters. While unlikely, some Unicode symbols or characters that are valid as word characters but not valid XML NameStartChar could be incorrectly matched, leading to attempts to parse non-XML content. Rollback suggestion: revert to the original regex r"<[a-zA-Z:_]" if the new behavior causes regressions, but this will reintroduce the Unicode root loss. Additionally, this fix only addresses start detection; other streaming logic may still assume ASCII tag names, so thorough testing with mixed Unicode/ASCII tags is advised.