fix: support struct vector fields in BulkWriter

[zhuwenxing]

What this PR does

• Adds BulkFileType.JSONL and writes one JSON object per line with .jsonl suffix.
• Allows Local/Remote BulkWriter to persist/upload JSONL files instead of skipping them.
• Allows BulkWriter struct-array subfields to use FLOAT_VECTOR, FLOAT16_VECTOR, BFLOAT16_VECTOR, INT8_VECTOR, and BINARY_VECTOR.
• Normalizes struct vector payloads for Parquet output, including FLOAT16/BFLOAT16 byte payloads.
• Uses ml_dtypes for BFLOAT16 JSON/CSV byte-vector decoding and includes it in the bulk_writer extra.
• Adds unit coverage for JSON, JSONL, CSV, and Parquet BulkWriter output across the supported struct vector subfield types.

Tests

• python -m pytest -q tests/unit/test_bulk_writer_buffer.py tests/unit/test_local_bulk_writer.py tests/unit/test_remote_bulk_writer.py
  • 95 passed
• Milvus Python client E2E against local dev-cli deployment:
  • test_import_struct_array_vector_subfield_all_file_types
  • test_import_struct_array_vector_subfield_with_local_bulk_writer
  • 40 passed

[mergify]

Tick the box to add this pull request to the merge queue (same as @mergifyio queue).

• Queue this pull request

[codecov]

Codecov Report

❌ Patch coverage is 92.59259% with 4 lines in your changes missing coverage. Please review.
✅ Project coverage is 93.78%. Comparing base (20a0e42) to head (a8bdef7).
⚠️ Report is 2 commits behind head on master.

Files with missing lines	Patch %	Lines
pymilvus/bulk_writer/buffer.py	93.54%	2 Missing ⚠️
pymilvus/bulk_writer/bulk_writer.py	90.00%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #3631      +/-   ##
==========================================
+ Coverage   93.49%   93.78%   +0.29%     
==========================================
  Files          69       69              
  Lines       15200    15238      +38     
==========================================
+ Hits        14211    14291      +80     
+ Misses        989      947      -42     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[SpadeA-Tang]

/lgtm

[zhuwenxing]

/assign @XuanYang-cn
PTAL

[zhuwenxing]

/hold

[zhuwenxing]

/unhold

[XuanYang-cn]

Found one blocking BulkWriter serialization regression.

nit: pymilvus/bulk_writer/__init__.py still omits ml_dtypes from the existing missing-package guard, so users who have the older bulk_writer extras except ml_dtypes get a raw import failure instead of the package install guidance; add ml_dtypes to expected_pkgs.

[XuanYang-cn]

/lgtm

[sre-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: XuanYang-cn, zhuwenxing

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Details

Needs approval from an approver in each of these files:

• OWNERS [XuanYang-cn]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[pymilvus-bot]

✅ Backport Created
Hi @zhuwenxing, Backport PR for 3.0 has been created: #3637

(cc @XuanYang-cn @SpadeA-Tang)

[pymilvus-bot]

Backport Created: 2.6 -> #3669