Namespace separation of symbols

Jason Smith — Thu, 08 Sep 2016 05:56:48 -0000

Bug #64 opened. Thank you. I think the obvious fix is likely to succeed, but I wanted a more experienced noodling over it before diving in.

Namespace separation of symbols

Jashua Cloutier — Tue, 06 Sep 2016 01:07:58 -0000

I'll have to think about how to do this one. Please open a bug. I fear the fix may be to store names as "A::X" rather than "X" where the namespace is A as an attribute. But there may be backwards incompatibility for some... something to think about.

Namespace separation of symbols

Jason Smith — Fri, 02 Sep 2016 00:31:16 -0000

First off, fabulous project, thank you.

I am having an issue with symbols defined in different namespaces, clashing. For instance:

namespace A {
class X;
};

namespace B {
class X;
};

A::X is not 'registered' as a fully qualified name in _CppHeader.classes. Hitting B::X in _CppHeader.evaulate_class_stack triggers a name collision error.

Happy to work with you on this, as I've been fixing bugs as I find them in my local 'fork' (I needed to implement a rudimentary preprocessor directive handler for our own needs, that I don't think would be of interest to most), but this one looked to require some more pervasive edits than I am ready to jump into. (I would hope that just fully qualifying the symbol names would be sufficient, but it didn't seem to be quite that simple.)

Macro parsing and nameStack boundaries?

Devon — Thu, 13 Aug 2015 15:53:19 -0000

Hi there,

First off, what a great product. Thanks for all the work you’ve done on this.

I’ve run into an issue parsing a particular style of programming we employ and hope you can offer a suggestion.

In the example header below, we're using macros to inject code snippets at compile time.

The issue I'm having is the token processor does not see a closing bracket (or semi-colon) at the end of the last macro so it attempts to process the first public enum as part of the nameStack created by the macro processing.

E.g:

Evaluating stack ['SCIOBJ_BEGIN_IFC_MAP', '(', 'SCIWizard', ')', 'SCIOBJ_CHAIN_IFC_MAP', '(', 'BASE_CLASS', ')', 'SCIOBJ_END_IFC_MAP', ':', 'typedef', 'enum', '_SomeMode']

Instead of:

Evaluating stack ['SCIOBJ_BEGIN_IFC_MAP', '(', 'SCIWizard', ')', 'SCIOBJ_CHAIN_IFC_MAP', '(', 'BASE_CLASS', ')', 'SCIOBJ_END_IFC_MAP’]

The net result is the _SomeMode enum gets swallowed and doesn’t appear anywhere in the resulting CppHeader class.

Note adding a semi-colon after MACRO_FUNCTION_END below does allow the parser to see the typedef enum, but I am not able to add this semi-colon to all the files that use this programming style.

Is there anything I can do (short of modifying the header files) to get the CppHeaderParser to properly parse these headers?

Thank you,

// devon

Example header file

#define MY_CLASS_H_
#include “macro_header.h"

class MyClass : public MyGenericClass<AnotherClass, MyClass> {
    typedef MyGenericClass<AnotherClass, MyClass> BASE_CLASS;

    MACRO_FUNCTION_BEGIN(AnotherClass)
        MACRO_FUNCTION_MAP(BASE_CLASS)
    MACRO_FUNCTION_END // <— missing ';' creates parsing issue

public:
    typedef enum _SomeMode {
        MODE_ONE,
        MODE_TWO
    } SomeMode;

    enum SomeState {
        STATE_ONE,
        STATE_TWO
    };
};

In the macro_header.h file, we have these macros defined:

#define MACRO_FUNCTION_BEGIN(IfName) public: \
    virtual MyRetPtr<MyObj> queryInterface(const MyStr &sIfID) { \
        typedef IfName __ifNameBaseType; \
        if (sIfID == getIID<IfName>()) \
            return MyRetPtr<MyObj>(static_cast<IfName *>(this));

#define MACRO_FUNCTION_MAP(base) \
        MyRetPtr<MyObj> result = base::queryInterface(sIfID); \
        if (result) \
            return result;

#define MACRO_FUNCTION_END \
        if (sIfID == MYOBJ_INTERFACE) \
            return MyRetPtr<MyObj>(static_cast<MyObj *>(static_cast<__ifNameBaseType *>(this))); \
        \
        return NULL; \
    }

So, basically, after pre-processing we end up with:

#define MY_CLASS_H_
#include "macro_header.h"

class MyClass : public MyGenericClass<AnotherClass, MyClass> {
    typedef MyGenericClass<AnotherClass, MyClass> BASE_CLASS;

    virtual MyRetPtr<MyObj> queryInterface(const MyStr &sIfID) {
        typedef IfName __ifNameBaseType;
        if (sIfID == getIID<IfName>())
            return MyRetPtr<MyObj>(static_cast<IfName *>(this));
        MyRetPtr<MyObj> result = base::queryInterface(sIfID);
        if (result)
           return result;
        if (sIfID == MYOBJ_INTERFACE)
            return MyRetPtr<MyObj>(static_cast<MyObj *>(static_cast<__ifNameBaseType *>(this)));

        return NULL;
    }
public:
    typedef enum _SomeMode {
        MODE_ONE,
        MODE_TWO
    } SomeMode;

    enum SomeState {
        STATE_ONE,
        STATE_TWO
    };
};

Templates support

Jashua Cloutier — Sat, 02 May 2015 18:06:35 -0000

I see, you are after "template<typename T_ObjectType="">" or a parsed version of it. If you look around line 2071 you will see that I actually strip out this part of the code

        # Strip out template declarations
        templateSectionsToSliceOut = []
        try:
            for m in re.finditer("template[\t ]*<[^>]*>", headerFileStr):
                start = m.start()
                # Search for the final '>' which may or may not be caught in the case of nexted <>'s
                for i in range(start, len(headerFileStr)):
                    if headerFileStr[i] == '<':
                        firstBracket = i
                        break
                ltgtStackCount = 1
                #Now look for fianl '>'
                for i in range(firstBracket + 1, len(headerFileStr)):
                    if headerFileStr[i] == '<':
                        ltgtStackCount += 1
                    elif headerFileStr[i] == '>':
                        ltgtStackCount -= 1
                    if ltgtStackCount == 0:
                        end = i
                        break
                templateSectionsToSliceOut.append((start, end))

            # Now strip out all instances of the template
            templateSectionsToSliceOut.reverse()
            for tslice in templateSectionsToSliceOut:
                newlines = headerFileStr[tslice[0]: tslice[1]].count("\n") * "\n" #Keep line numbers the same
                headerFileStr = headerFileStr[:tslice[0]] + newlines  + headerFileStr[tslice[1] + 1:]
        except:
            pass

I do this because dealing with that broke other parts of the code and there wasnt a need for it. I will consider adding this as something that gets parsed in a future release (God help me).

Forward Declarations

arya — Fri, 01 May 2015 18:44:45 -0000

Thank you :)

Templates support

Semenyakin Vladimir — Mon, 27 Apr 2015 07:01:23 -0000

Yeah, thanks for your focusing on static libraries linkage issues, I found this article with pimpl idiom description, that may help to fix such problems.

About API... Let's I show on example:

template<typename T_ObjectType="">
HeapStorage<T_ObjectType> createStorage() {
return HeapStorage<T_ObjectType>(heapAllocator());
}

Return value of this class is specification of HeapStorage template class. I dependes from template argument with name T_ObjectType. So, for correct header convertion, I need some API to get template argument names - to create "template<typename T_ObjectType="">" line (where template argument name is presented). Something like this:

inCppClass["methods"]["public"]["createStorage"]["templateArgs"][0]

Templates support

Jashua Cloutier — Mon, 27 Apr 2015 01:04:59 -0000

First, if I understand correctly, your goal is to publish a .h file that has everything private stripped. If this is the case, remember that if you have a compiled library using the private .h representation of the class, and someone else compiles code with the public version of the .h file an links against your compiled code, the program will have 2 different understandings of the size of your object and things could end badly.

Looking at your code, I still dont understand exactly what you are having trouble with finding. Are you having trouble accessing DVirtualDeviceRAM::HeapStorage? Please specify what data you want to get pragmatically and how then I can help further

Forward Declarations

Jashua Cloutier — Mon, 27 Apr 2015 00:41:15 -0000

Please try 139:231618db8dee. You should see it works

Templates support

Semenyakin Vladimir — Sat, 25 Apr 2015 12:28:43 -0000

Thank you for fast answer!

I attached python covertion code, input and output headers. There are inner classes in output, but, I think, it's not big problem. The problem is that I can't find API to get template argumetns of template classes and methods.

In test cases I found test "class Lizzard_TestCase(unittest.TestCase):", where template key is used. But it's just flag there, not some structure, where detail template data is stored.

P.S.: Sorry for my English.
P.P.S.: Sorry for my Python. It's not my native programing language

Recent posts to Help

Namespace separation of symbols

Namespace separation of symbols

Namespace separation of symbols

Macro parsing and nameStack boundaries?

Templates support

Forward Declarations

Templates support

Templates support

Forward Declarations

Templates support