FIX: Decode Raw UTF-16 data from Conn.getinfo()

[bewithgaurav]

Work Item / Issue Reference

AB#<WORK_ITEM_ID>

GitHub Issue: #318

---

Summary

This pull request introduces improvements to the handling of string encoding in the getinfo method for SQL Server connections, adds support for profiling builds in the Windows build script, and enhances test coverage for string decoding. The most important changes are grouped below:

String Decoding Improvements

• The getinfo method in connection.py now attempts to decode string results from SQL Server using multiple encodings in order: UTF-16LE (Windows default), UTF-8, and Latin-1. This improves robustness when handling driver responses and avoids silent data corruption by returning None if all decoding attempts fail.

Test Coverage

• Added a new test test_getinfo_string_encoding_utf16 in test_003_connection.py to verify that string values returned by getinfo are properly decoded from UTF-16, contain no null bytes, and are non-empty, helping catch encoding mismatches early.

Build Script Cleanup

• Removed redundant logic from build.bat related to copying the msvcp140.dll redistributable, simplifying the post-build process.

[github-actions]

📊 Code Coverage Report

🔥 Diff Coverage 40%

🎯 Overall Coverage 75%

📈 Total Lines Covered: 5092 out of 6756 📁 Project: mssql-python

---

Diff Coverage

Diff: main...HEAD, staged and unstaged changes

• mssql_python/connection.py (40.0%): Missing lines 1193-1194,1202

Summary

• Total: 5 lines
• Missing: 3 lines
• Coverage: 40%

mssql_python/connection.py

Lines 1189-1198

  1189                     # Try encodings in order: UTF-16LE (Windows), UTF-8, Latin-1
  1190                     for encoding in ("utf-16-le", "utf-8", "latin1"):
  1191                         try:
  1192                             return actual_data.decode(encoding).rstrip("\0")
! 1193                         except UnicodeDecodeError:
! 1194                             continue
  1195                     
  1196                     # All decodings failed
  1197                     logger.debug(
  1198                         "error",

---

Lines 1198-1206

  1198                         "error",
  1199                         "Failed to decode string in getinfo with any supported encoding. "
  1200                         "Returning None to avoid silent corruption.",
  1201                     )
! 1202                     return None
  1203                 else:
  1204                     # If it's not bytes, return as is
  1205                     return data
  1206             elif is_yn_type:

---

---

📋 Files Needing Attention

📉 Files with overall lowest coverage (click to expand)

mssql_python.pybind.logger_bridge.cpp: 59.2%
mssql_python.row.py: 66.2%
mssql_python.pybind.ddbc_bindings.cpp: 67.1%
mssql_python.helpers.py: 67.5%
mssql_python.pybind.connection.connection.cpp: 73.6%
mssql_python.pybind.ddbc_bindings.h: 76.9%
mssql_python.ddbc_bindings.py: 79.6%
mssql_python.pybind.connection.connection_pool.cpp: 79.6%
mssql_python.connection.py: 82.2%
mssql_python.cursor.py: 83.8%

---

🔗 Quick Links

⚙️ Build Summary	📋 Coverage Details
View Azure DevOps Build	Browse Full Coverage Report

[saurabh500]

Why latin1 ? :) Asking out of curiosity. IIUC, utf-16-le and utf-8 might be sufficient. If you have other reasons to add latin1, it would be good to know

[bewithgaurav]

You're right, I had included this in code as some kind of fallback strategy - however the expected encoding is UTF-16LE, and UTF-8 is a reasonable fallback for edge cases where data might come pre-decoded
Better to remove it. I'll make this change. Thanks for the suggestion!