winapi - How might I wrap the FindXFile-style APIs to the STL-style Iterator Pattern in C++?

Question

I'm working on wrapping up the ugly innards of the FindFirstFile/FindNextFile loop (though my question applies to other similar APIs, such as RegEnumKeyEx or RegEnumValue, etc.) inside iterators that work in a manner similar to the Standard Template Library's istream_iterators.

I have two problems here. The first is with the termination condition of most "foreach" style loops. STL style iterators typically use operator!= inside the exit condition of the for, i.e.

std::vector<int> test;
for(std::vector<int>::iterator it = test.begin(); it != test.end(); it++) {
 //Do stuff
}

My problem is I'm unsure how to implement operator!= with such a directory enumeration, because I do not know when the enumeration is complete until I've actually finished with it. I have sort of a hack together solution in place now that enumerates the entire directory at once, where each iterator simply tracks a reference counted vector, but this seems like a kludge which can be done a better way.

The second problem I have is that there are multiple pieces of data returned by the FindXFile APIs. For that reason, there's no obvious way to overload operator* as required for iterator semantics. When I overload that item, do I return the file name? The size? The modified date? How might I convey the multiple pieces of data to which such an iterator must refer to later in an ideomatic way? I've tried ripping off the C# style MoveNext design but I'm concerned about not following the standard idioms here.

class SomeIterator {
public:
 bool next(); //Advances the iterator and returns true if successful, false if the iterator is at the end.
 std::wstring fileName() const;
 //other kinds of data....
};

EDIT: And the caller would look like:

SomeIterator x = ??; //Construct somehow
while(x.next()) {
    //Do stuff
}

Thanks!

Billy3

EDIT2: I have fixed some bugs and written some tests.

Implementation:

#pragma once
#include <queue>
#include <string>
#include <boost/noncopyable.hpp>
#include <boost/make_shared.hpp>
#include <boost/iterator/iterator_facade.hpp>
#include <Windows.h>
#include <Shlwapi.h>
#pragma comment(lib, "shlwapi.lib")
#include "../Exception.hpp"

namespace WindowsAPI { namespace FileSystem {

template <typename Filter_T = AllResults, typename Recurse_T = NonRecursiveEnumeration>
class DirectoryIterator;

//For unit testing
struct RealFindXFileFunctions
{
    static HANDLE FindFirst(LPCWSTR lpFileName, LPWIN32_FIND_DATAW lpFindFileData) {
        return FindFirstFile(lpFileName, lpFindFileData);
    };
    static BOOL FindNext(HANDLE hFindFile, LPWIN32_FIND_DATAW lpFindFileData) {
        return FindNextFile(hFindFile, lpFindFileData);
    };
    static BOOL Close(HANDLE hFindFile) {
        return FindClose(hFindFile);
    };
};

inline std::wstring::const_iterator GetLastSlash(std::wstring const&pathSpec) {
    return std::find(pathSpec.rbegin(), pathSpec.rend(), L'\').base();
}

class Win32FindData {
    WIN32_FIND_DATA internalData;
    std::wstring rootPath;
public:
    Win32FindData(const std::wstring& root, const WIN32_FIND_DATA& data) :
        rootPath(root), internalData(data) {};
    DWORD GetAttributes() const {
        return internalData.dwFileAttributes;
    };
    bool IsDirectory() const {
        return (internalData.dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY) != 0;
    };
    bool IsFile() const {
        return !IsDirectory();
    };
    unsigned __int64 GetSize() const {
        ULARGE_INTEGER intValue;
        intValue.LowPart = internalData.nFileSizeLow;
        intValue.HighPart = internalData.nFileSizeHigh;
        return intValue.QuadPart;
    };
    std::wstring GetFolderPath() const {
        return rootPath;
    };
    std::wstring GetFileName() const {
        return internalData.cFileName;
    };
    std::wstring GetFullFileName() const {
        return rootPath + L"" + internalData.cFileName;
    };
    std::wstring GetShortFileName() const {
        return internalData.cAlternateFileName;
    };
    FILETIME GetCreationTime() const {
        return internalData.ftCreationTime;
    };
    FILETIME GetLastAccessTime() const {
        return internalData.ftLastAccessTime;
    };
    FILETIME GetLastWriteTime() const {
        return internalData.ftLastWriteTime;
    };
};

template <typename FindXFileFunctions_T>
class BasicNonRecursiveEnumeration : public boost::noncopyable
{
    WIN32_FIND_DATAW currentData;
    HANDLE hFind;
    std::wstring currentDirectory;
    void IncrementCurrentDirectory() {
        if (hFind == INVALID_HANDLE_VALUE) return;
        BOOL success =
            FindXFileFunctions_T::FindNext(hFind, &currentData);
        if (success)
            return;
        DWORD error = GetLastError();
        if (error == ERROR_NO_MORE_FILES) {
            FindXFileFunctions_T::Close(hFind);
            hFind = INVALID_HANDLE_VALUE;
        } else {
            WindowsApiException::Throw(error);
        }
    };
    bool IsValidDotDirectory()
    {
        return !Valid() &&
            (!wcscmp(currentData.cFileName, L".") || !wcscmp(currentData.cFileName, L".."));
    };
    void IncrementPastDotDirectories() {
        while (IsValidDotDirectory()) {
            IncrementCurrentDirectory();
        }
    };
    void PerformFindFirstFile(std::wstring const&pathSpec)
    {
        hFind = FindXFileFunctions_T::FindFirst(pathSpec.c_str(), &currentData);
        if (Valid()
            && GetLastError() != ERROR_PATH_NOT_FOUND
            && GetLastError() != ERROR_FILE_NOT_FOUND)
            WindowsApiException::ThrowFromLastError();
    };
public:
    BasicNonRecursiveEnumeration() : hFind(INVALID_HANDLE_VALUE) {};
    BasicNonRecursiveEnumeration(const std::wstring& pathSpec) :
        hFind(INVALID_HANDLE_VALUE) {
        std::wstring::const_iterator lastSlash = GetLastSlash(pathSpec);
        if (lastSlash != pathSpec.begin())
            currentDirectory.assign(pathSpec.begin(), lastSlash-1);
        PerformFindFirstFile(pathSpec);
        IncrementPastDotDirectories();
    };
    bool equal(const BasicNonRecursiveEnumeration<FindXFileFunctions_T>& other) const {
        if (this == &other)
            return true;
        return hFind == other.hFind;
    };
    Win32FindData dereference() {
        return Win32FindData(currentDirectory, currentData);
    };
    void increment() {
        IncrementCurrentDirectory();
    };
    bool Valid() {
        return hFind == INVALID_HANDLE_VALUE;
    };
    virtual ~BasicNonRecursiveEnumeration() {
        if (!Valid())
            FindXFileFunctions_T::Close(hFind);
    };
};

typedef BasicNonRecursiveEnumeration<RealFindXFileFunctions> NonRecursiveEnumeration;

template <typename FindXFileFunctions_T>
class BasicRecursiveEnumeration : public boost::noncopyable
{
    std::wstring fileSpec;
    std::deque<std::deque<Win32FindData> > enumeratedData;
    void EnumerateDirectory(const std::wstring& nextPathSpec) {
        std::deque<Win32FindData> newDeck;
        BasicNonRecursiveEnumeration<FindXFileFunctions_T> begin(nextPathSpec), end;
        for(; !begin.equal(end); begin.increment()) {
            newDeck.push_back(begin.dereference()); 
        }
        if (!newDeck.empty()) {
            enumeratedData.push_back(std::deque<Win32FindData>()); //Swaptimization
            enumeratedData.back().swap(newDeck);
        }
    };
    void PerformIncrement() {
        if (enumeratedData.empty()) return;
        if (enumeratedData.back().front().IsDirectory()) {
            std::wstring nextSpec(enumeratedData.back().front().GetFullFileName());
            nextSpec.append(L"\*");
            enumeratedData.back().pop_front();
            EnumerateDirectory(nextSpec);
        } else {
            enumeratedData.back().pop_front();
        }
        while (Valid() && enumeratedData.back().empty())
            enumeratedData.pop_back();
    }
    bool CurrentPositionNoMatchFileSpec() const
    {
        return !enumeratedData.empty() && !PathMatchSpecW(enumeratedData.back().front().GetFileName().c_str(), fileSpec.c_str());
    }
public:
    BasicRecursiveEnumeration() {};
    BasicRecursiveEnumeration(const std::wstring& pathSpec) {
        std::wstring::const_iterator lastSlash = GetLastSlash(pathSpec);
        if (lastSlash == pathSpec.begin()) {
            fileSpec = pathSpec;
            EnumerateDirectory(L"*");
        } else {
            fileSpec.assign(lastSlash, pathSpec.end());
            std::wstring firstQuery(pathSpec.begin(), lastSlash);
            firstQuery.push_back(L'*');
            EnumerateDirectory(firstQuery);
            while (CurrentPositionNoMatchFileSpec())
                PerformIncrement();
        }
    };
    void increment() {
        do
        {
            PerformIncrement();
        } while (CurrentPositionNoMatchFileSpec());
    };
    bool equal(const BasicRecursiveEnumeration<FindXFileFunctions_T>& other) const {
        if (!Valid())
            return !other.Valid();
        if (!other.Valid())
            return false;
        return this == &other;
    };
    Win32FindData dereference() const {
        return enumeratedData.back().front();
    };
    bool Valid() const {
        return !enumeratedData.empty();
    };
};

typedef BasicRecursiveEnumeration<RealFindXFileFunctions> RecursiveEnumeration;

struct AllResults
{
    bool operator()(const Win32FindData&) {
        return true;
    };
}; 

struct FilesOnly
{
    bool operator()(const Win32FindData& arg) {
        return arg.IsFile();
    };
};

template <typename Filter_T, typename Recurse_T>
class DirectoryIterator : 
    public boost::iterator_facade<
        DirectoryIterator<Filter_T, Recurse_T>,
        Win32FindData,
        std::input_iterator_tag,
        Win32FindData
    >
{
    friend class boost::iterator_core_access;
    boost::shared_ptr<Recurse_T> impl;
    Filter_T filter;
    void increment() {
        do {
            impl->increment();
        } while (impl->Valid() && !filter(impl->dereference()));
    };
    bool equal(const DirectoryIterator& other) const {
        return impl->equal(*other.impl);
    };
    Win32FindData dereference() const {
        return impl->dereference();
    };
public:
    DirectoryIterator(Filter_T functor = Filter_T()) :
        impl(boost::make_shared<Recurse_T>()),
        filter(functor) {
    };
    explicit DirectoryIterator(const std::wstring& pathSpec, Filter_T functor = Filter_T()) :
        impl(boost::make_shared<Recurse_T>(pathSpec)),
        filter(functor) {
    };
};

}}

Tests:

#include <queue>
#include "../WideCharacterOutput.hpp"
#include <boost/test/unit_test.hpp>
#include "../../WindowsAPI++/FileSystem/Enumerator.hpp"
using namespace WindowsAPI::FileSystem;


struct SimpleFakeFindXFileFunctions
{
    static std::deque<WIN32_FIND_DATAW> fakeData;
    static std::wstring insertedFileName;

    static HANDLE FindFirst(LPCWSTR lpFileName, LPWIN32_FIND_DATAW lpFindFileData) {
        insertedFileName.assign(lpFileName);
        if (fakeData.empty()) {
            SetLastError(ERROR_PATH_NOT_FOUN

Answer 1

To solve the first problem, you can have end() return some sentinel value, then in your iterator's increment function, set the iterator equal to that sentinel value when it reaches the end of the container. This is effectively what the directory iterator in Boost.Filesystem does.

For the second problem, I'm not entirely familiar with the FindXFile APIs, but one option would be to return some struct that contains all the data you need or that has member functions to get each of the pieces of data you might want.