[python] Preserve declared primary-key order in merge reads

[wangzhigang1999]

Purpose

Fixes #9800

When a composite primary key is declared in a different order from the table columns, PyPaimon can return both the old and updated row. For example, four orders with PRIMARY KEY (order_id, customer_id) produce five rows after one update when customer_id precedes order_id in the table schema.

Build merge-read key fields from the table's ordered trimmed primary-key fields, matching the writer's sort order. This preserves field IDs, types and the storage format. Regression coverage checks complete merged results, value-only projections, partition-key trimming and a primary-key column position change across schema versions.

Tests

Validation: the new regression tests fail before the fix; all 70 tests in reader_primary_key_test, schema_evolution_pk_read_test, projection_predicate_index_test and test_sequence_field_read pass after it. Flake8 on changed files and git diff --check pass. Local validation used Python 3.12.6 and PyArrow 19.0.1.

[Copilot]

🟢 Approval recommended

No unresolved blocking issues were identified.

Pull request overview

Fixes merge-read deduplication by preserving declared composite primary-key order.

Changes:

• Builds merge keys from ordered trimmed primary-key fields.
• Adds coverage for projections, partitions, reordered keys, and schema changes.

File summaries

File	Summary
paimon-python/pypaimon/tests/schema_evolution_pk_read_test.py	Tests primary-key position changes across schemas.
paimon-python/pypaimon/tests/reader_primary_key_test.py	Tests composite-key ordering and projections.
paimon-python/pypaimon/read/split_read.py	Aligns merge-key construction with writer sort order.

Review details

• Files reviewed: 3/3 changed files
• Comments generated: 0
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[JingsongLi]

Requirement fit: SUPPORTED. Implementation: CLEAN.

Reviewed e674f81862d0. Merge keys must follow the declared primary-key order used to sort the data files. Using trimmed_primary_keys_fields fixes duplicate/stale logical rows when schema field order differs, while preserving key IDs and types and excluding partition keys.

Validation: all 70 primary-key reader, schema-evolution, projection/predicate and sequence-field tests passed with PyArrow 19.0.1. Current head CI is green.

No actionable implementation regression found in this review.