FEAT: Support for streaming large parameters in execute()

[gargsaumya]

Work Item / Issue Reference

AB#33395

GitHub Issue: #<ISSUE_NUMBER>

---

Summary

This pull request adds support for streaming large parameters to SQL Server using ODBC's Data At Execution (DAE) mechanism, particularly for long Unicode strings and binary data. The changes update both the Python and C++ layers to correctly identify large parameters, mark them for DAE, and handle the streaming process during execution. Additional refactoring improves parameter type mapping and memory handling for these cases.

Large parameter streaming (DAE) support:

• Updated the _map_sql_type method in cursor.py to return an is_dae flag for parameters that require streaming (e.g., long Unicode strings, long binary data), and to calculate the correct size for Unicode strings using UTF-16 encoding. [1] [2] [3] [4]
• Modified _create_parameter_types_list in cursor.py to set DAE-related fields (isDAE, strLenOrInd, dataPtr) in the parameter info when streaming is needed.

C++ bindings and execution logic:

• Extended the ParamInfo struct and its Python bindings to include DAE fields (isDAE, strLenOrInd, dataPtr) for use during parameter binding and streaming. [1] [2]
• Added ODBC DAE API function pointers (SQLParamData, SQLPutData) and integrated their loading and usage into the driver handle setup. [1] [2] [3] [4]
• Refactored parameter binding and execution logic in ddbc_bindings.cpp to handle DAE parameters: if a parameter is marked for DAE, the code enters a loop to stream the data in chunks using SQLParamData and SQLPutData. This is done for large Unicode strings and (potentially) binary data. [1] [2]

Build and warning improvements:

• Adjusted MSVC compiler flags to remove the "treat warnings as errors" option, making builds less strict on warnings.

[Copilot]

Pull Request Overview

This pull request adds support for Data At Execution (DAE) streaming in SQL Server for large parameters, particularly NVARCHAR(MAX) fields. The implementation allows streaming of large Unicode strings and binary data through ODBC's DAE mechanism instead of buffering them entirely in memory.

• Introduces DAE support for streaming large parameters (>4000 characters for strings, >8000 bytes for binary)
• Updates parameter type mapping to correctly calculate UTF-16 string sizes and identify streamable parameters
• Refactors C++ bindings to handle SQLParamData/SQLPutData execution flow for large data

Reviewed Changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 6 comments.

File	Description
mssql_python/pybind/ddbc_bindings.h	Adds function pointer declarations for SQLParamData and SQLPutData DAE APIs
mssql_python/pybind/ddbc_bindings.cpp	Implements DAE streaming logic, extends ParamInfo struct, and handles SQLParamData/SQLPutData execution flow
mssql_python/pybind/CMakeLists.txt	Removes /WX compiler flag to make builds less strict on warnings
mssql_python/cursor.py	Updates parameter type mapping to return DAE flags and correctly calculate UTF-16 string sizes

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[Copilot]

Type mismatch: totalBytes is SQLLEN (signed) but offset is size_t (unsigned). This could cause issues when totalBytes is negative or when comparing signed/unsigned values.

Suggested change

	size_t len = std::min(chunkSize, totalBytes - offset);
	size_t totalBytes_sz = static_cast<size_t>(utf16_str.size());
	const size_t chunkSize = 8192;
	for (size_t offset = 0; offset < totalBytes_sz; offset += chunkSize) {
	size_t len = std::min(chunkSize, totalBytes_sz - offset);

[Copilot]

Encoding the string to calculate UTF-16 length is inefficient for large strings. Consider using a more efficient method to estimate the byte length without full encoding.

Suggested change

	utf16_len = len(param.encode("utf-16-le")) // 2
	utf16_len = utf16_codeunits(param)

[Copilot]

Duplicate encoding operation for UTF-16 length calculation. This encoding is performed even for short strings where it may be unnecessary overhead.

Suggested change

	utf16_len = len(param.encode("utf-16-le")) // 2
	utf16_len = len(param)