在撰写此答案(OpenCV 3.4.1是最后发布的版本)时,没有办法只过滤默认错误处理程序的输出(我可以想到),也没有办法更改错误处理程序.
但是,您的问题让我开始思考-在highgui模块中,我们已经有了一些函数,可以让我们为鼠标,轨迹栏和按钮事件设置Python回调,因此我们可以从该代码中获得启发并将其添加为新功能.
让我们使Python错误处理程序具有与C++ error handler相匹配的签名：
error_handler([int]status, [str]func_name, [str]err_msg, [str]file_name, [int]line, [obj]userdata) -> None
让我们还添加支持以重置为默认错误处理程序-highgui函数尚未执行的操作.
首先,我们需要一个静态错误处理程序函数,该函数会将参数从C编组到Python,并调用适当的Python函数来处理错误.就像我们从中汲取灵感的功能一样,我们将使用user data参数来保存由功能对象和可选的Python用户数据组成的元组.
static int OnError(int status, const char *func_name, const char *err_msg, const char *file_name, int line, void *userdata)
{
PyGILState_STATE gstate;
gstate = PyGILState_Ensure();
PyObject *o = (PyObject*)userdata;
PyObject *args = Py_BuildValue("isssiO", status, func_name, err_msg, file_name, line, PyTuple_GetItem(o, 1));
PyObject *r = PyObject_Call(PyTuple_GetItem(o, 0), args, NULL);
if (r == NULL) {
PyErr_Print();
} else {
Py_DECREF(r);
}
Py_DECREF(args);
PyGILState_Release(gstate);
return 0; // The return value isn't used
}接下来,我们需要编写函数来实现Python和C之间的绑定.但是,由于我的灵感来自我的suspicions of potential memory leaks函数,因此我们将进行一些补充以解决此问题-我们将跟踪最新的用户数据对象,并在必要时取消引用.
// Keep track of the previous handler parameter, so we can decref it when no longer used
static PyObject* last_on_error_param = NULL;
static PyObject *pycvRedirectError(PyObject*, PyObject *args, PyObject *kw)
{
const char *keywords[] = { "on_error", "param", NULL };
PyObject *on_error;
PyObject *param = NULL;
if (!PyArg_ParseTupleAndKeywords(args, kw, "O|O", (char**)keywords, &on_error, ¶m))
return NULL;
if ((on_error != Py_None) && !PyCallable_Check(on_error)) {
PyErr_SetString(PyExc_TypeError, "on_error must be callable");
return NULL;
}
if (param == NULL) {
param = Py_None;
}
if (last_on_error_param) {
Py_DECREF(last_on_error_param);
last_on_error_param = NULL;
}
if (on_error == Py_None) {
ERRWRAP2(redirectError(NULL));
} else {
last_on_error_param = Py_BuildValue("OO", on_error, param);
ERRWRAP2(redirectError(OnError, last_on_error_param));
}
Py_RETURN_NONE;
}最后,我们必须注册“special method”.与我们从中获得灵感的功能不同,我们不依赖某些GUI,并且希望它始终可用,因此我们将代码修改如下：
static PyMethodDef special_methods[] = {
{"redirectError", (PyCFunction)pycvRedirectError, METH_VARARGS | METH_KEYWORDS, "redirectError(onError [, param]) -> None"},
#ifdef HAVE_OPENCV_HIGHGUI
{"createTrackbar", pycvCreateTrackbar, METH_VARARGS, "createTrackbar(trackbarName, windowName, value, count, onChange) -> None"},
{"createButton", (PyCFunction)pycvCreateButton, METH_VARARGS | METH_KEYWORDS, "createButton(buttonName, onChange [, userData, buttonType, initialButtonState]) -> None"},
{"setMouseCallback", (PyCFunction)pycvSetMouseCallback, METH_VARARGS | METH_KEYWORDS, "setMouseCallback(windowName, onMouse [, param]) -> None"},
#endif
{NULL, NULL},
};现在我们可以重建OpenCV,安装并测试它.我编写了以下脚本来演示功能：
import cv2
def verbose_error_handler(status, func_name, err_msg, file_name, line, userdata):
print "Status = %d" % status
print "Function = %s" % func_name
print "Message = %s" % err_msg
print "Location = %s(%d)" % (file_name, line)
print "User data = %r" % userdata
def silent_error_handler(status, func_name, err_msg, file_name, line, userdata):
pass
print "** Default handler"
try:
cv2.imshow("", None) # This causes an assert
except cv2.error as e:
pass
print "** Using verbose handler"
cv2.redirectError(verbose_error_handler, 42)
try:
cv2.imshow("", None) # This causes an assert
except cv2.error as e:
pass
print "** Using silent handler"
cv2.redirectError(silent_error_handler)
try:
cv2.imshow("", None) # This causes an assert
except cv2.error as e:
pass
print "** Back to default handler"
cv2.redirectError(None)
try:
cv2.imshow("", None) # This causes an assert
except cv2.error as e:
pass使用修补程序的OpenCV版本(基于上述说明),我在控制台上获得以下输出：
** Default handler
OpenCV(3.4.1) Error: Assertion failed (size.width>0 && size.height>0) in cv::imshow, file F:\GitHub\opencv\modules\highgui\src\window.cpp, line 364
** Using verbose handler
Status = -215
Function = cv::imshow
Message = size.width>0 && size.height>0
Location = F:\GitHub\opencv\modules\highgui\src\window.cpp(364)
User data = 42
** Using silent handler
** Back to default handler
OpenCV(3.4.1) Error: Assertion failed (size.width>0 && size.height>0) in cv::imshow, file F:\GitHub\opencv\modules\highgui\src\window.cpp, line 364就在我开始编写此答案时,我的想法之一是,由于default error handler使用以下格式字符串将所有这些消息输出到stderr
"OpenCV(%s) Error: %s (%s) in %s, file %s, line %d"
我们也许可以钩住stderr流,并过滤掉以常量前缀开头的所有行. A,我无法实现这一目标.也许别人可以吗？
编辑：通过一些较小的修改将修补程序合并到opencv：master中(主要是我们摆脱了用户数据参数,因为在Python中是不必要的).
