[BUG] Data Misalignment in Excel Plugin - Empty Cell Returns Distant Value from Shopee Report

[gltjk]

Description​

When reading a specific range in a ​Shopee income report template, empty cells return incorrect values from unrelated header locations instead of NULL. This occurs specifically with Shopee's financial report format where metadata headers exist above the target range.

Steps to Reproduce​

1. Use attached Shopee income report template (shopee_income_report.xlsx):

• Original headers at row 1 (卖家帐号, 付款ID, 收款渠道)
• Data section starts at row 6 (编号, 订单编号, 退款ID)
• Target cell C7 is empty

2. Execute DuckDB query:

FROM read_xlsx('shopee_income_report.xlsx', range='A6:C7');

3. Actual Result:

编号 double	订单编号 varchar	退款ID varchar
1.0	yyyyy	卖家账号

4. Expected Result:

编号 double	订单编号 varchar	退款ID double
1.0	yyyyy	NULL

Environment​

• DuckDB v1.3.0 71c5c07cdd
• MacOS 15.5 (24F74) on MacBook Pro M3 Max
• Shopee report template structure (attached anonymized file)

Critical Impact​

This bug poses ​high risk for financial data processing:

• Misinterprets empty refund fields as valid seller account IDs
• Causes silent data corruption in automated report pipelines
• Particularly problematic for e-commerce reports with multi-header layouts
• Undermines trust in data integrity for monetary operations

Additional Context​

• Reproducible with multiple Shopee income reports from different regions
• May relate to shared string table handling in xlsx files
• Test file maintains Shopee's original cell formatting and structure
• If exported reports are opened with Excel and then saved, the results will be as expected