如何使用cppyy 1.7.0读取C++函数返回的MlpiAxisInformation结构体数组中的char16_t类型字符串

I’ve tackled similar cppyy string-handling hurdles with older versions, so let’s break down how to fix your issue step by step. The core challenges are properly allocating the struct array for the MLPI function call and converting the char16_t (WCHAR16) arrays to valid Python strings—something your ctypes code struggled with.

Step 1: Load MLPI Types and Functions in cppyy #

First, make cppyy aware of the C++ definitions. You can either include the official MLPI header file directly (if available in your environment) or define the types inline:

import cppyy

# Define MLPI types inline (replace with cppyy.include("<your-mlpi-header.h>") if possible)
cppyy.cppdef("""
typedef unsigned int ULONG;
typedef char16_t WCHAR16;

struct MlpiAxisRef {
    int controlNo;
    int axisNo;
};

enum MlpiAxisType {
    MLPI_AXIS_TYPE_REAL = 0,
    MLPI_AXIS_TYPE_VIRTUAL = 1
    // Add other enum values as per your MLPI version
};

#define MLPI_MOTION_MAX_AXIS_NAME_LEN 100

struct MlpiAxisInformation {
    MlpiAxisRef axis;
    ULONG deviceAddress;
    MlpiAxisType axisType;
    WCHAR16 name[MLPI_MOTION_MAX_AXIS_NAME_LEN];
};

MLPI_API long mlpiMotionGetConfiguredAxes(unsigned long long connection, MlpiAxisInformation* configAxes, const ULONG numElements, ULONG *numElementsRet);
""")

# Alias types for cleaner code
MlpiAxisRef = cppyy.gbl.MlpiAxisRef
MlpiAxisInformation = cppyy.gbl.MlpiAxisInformation
mlpiMotionGetConfiguredAxes = cppyy.gbl.mlpiMotionGetConfiguredAxes

Step 2: Call the MLPI Function and Process Results #

cppyy handles pointer conversions more intuitively than ctypes, but you still need to allocate a struct array for the function to populate. Here’s the full implementation, including the critical char16_t string conversion:

from typing import List

def MLPIGetConfiguredAxes(self) -> List[dict]:
    max_axes = 100
    num_elements = cppyy.gbl.ULONG(max_axes)
    num_elements_ret = cppyy.gbl.ULONG(0)

    # Allocate an array of MlpiAxisInformation structs
    axes_array = (MlpiAxisInformation * max_axes)()

    # Execute the MLPI function
    result = mlpiMotionGetConfiguredAxes(self.con, axes_array, num_elements, cppyy.addressof(num_elements_ret))

    # Handle errors (adjust based on your MLPI error code conventions)
    if result != 0:
        raise RuntimeError(f"mlpiMotionGetConfiguredAxes failed with code {result}")

    # Parse results and convert char16_t strings
    configured_axes = []
    for i in range(num_elements_ret.value):
        axis_data = axes_array[i]
        
        # Manually convert char16_t array to Python string (cppyy 1.7.0 doesn't auto-convert this)
        name_chars = axis_data.name
        # Find the null terminator to avoid trailing garbage
        end_idx = 0
        while end_idx < MLPI_MOTION_MAX_AXIS_NAME_LEN and name_chars[end_idx] != 0:
            end_idx += 1
        # Convert UTF-16 bytes to Python string (adjust endianness to 'utf-16-be' if needed for your system)
        axis_name = bytes(name_chars[:end_idx]).decode('utf-16-le')

        # Package data into a dict for easy use (or a custom class if preferred)
        configured_axes.append({
            "control_no": axis_data.axis.controlNo,
            "axis_no": axis_data.axis.axisNo,
            "device_address": axis_data.deviceAddress,
            "axis_type": axis_data.axisType,
            "name": axis_name
        })

    return configured_axes

Critical Notes for cppyy 1.7.0 #

• Char16_t Conversion: Unlike newer cppyy versions, 1.7.0 doesn’t auto-convert char16_t arrays to Python strings. The manual byte conversion ensures you handle the UTF-16 encoding correctly—double-check endianness for your target system if the string appears garbled.
• Array Allocation: Using (MlpiAxisInformation * max_axes)() creates a static array that cppyy can pass directly as a pointer to the C++ function, mirroring native C behavior.
• Pointer Handling: cppyy.addressof(num_elements_ret) safely retrieves the pointer to the output parameter, just like byref in ctypes but with more native C++ compatibility.

Why This Fixes Your ctypes Issue #

Your ctypes code used ctypes.c_wchar*100, which assumes UTF-16LE but can fail with null terminators or system-specific endianness quirks. The cppyy implementation gives you explicit control over the string conversion process, ensuring you get valid, readable axis names every time.

内容的提问来源于stack exchange，提问作者araisch