Boost.Container flat_map and std::string_view - dictionary

For some time I’ve used Boost’s flat_map as my go-to associative collection, for the reasons explained cited on their documentation intro, and (originally) the fact that it gave newer features before the compiler’s std implementation, and it was the same across platforms.
Now, I wanted to start using string_view to prevent copying strings, when these are taken from substrings of a larger input. string_view points to a range of characters within the larger string, without having to copy them into a new std::string instance.
In reaching for a map to use, I recalled that another progressive feature of Boost.Container that I’ve enjoyed in the past is conformal keys, where you could use anything that compared correctly against the stored key, rather than converting to the actual type of key.
But now I can’t find any mention of that in the documentation. I know the std::map can do that now (since C++14) but I’d rather use the flat_map for tiny collections.
What could I have seen that allowed this flexibility, years ago, if it’s not apparent in boost::flat_map::insert etc.? What are good flat collections to use now with up-to-date compilers?

Support for polymorphic lookup functions has been added only recently to Boost.Container. If everything is good, it should be released with Boost 1.68.
In the meantime you can emulate flat associative containers with an ordered std::vector and std::lower_bound.
typedef std::pair< std::string, int > element_type;
std::vector< element_type > map;
struct element_order
{
bool operator()(element_type const& left, element_type const& right) const
{
return left.first < right.first;
}
bool operator()(std::string_view const& left, element_type const& right) const
{
return left < right.first;
}
bool operator()(element_type const& left, std::string_view const& right) const
{
return left.first < right;
}
};
auto find_element(std::string_view const& key)
{
auto it = std::lower_bound(map.begin(), map.end(), key, element_order());
if (it != map.end() && it->first == key)
return it;
return map.end();
}

Perhaps this is not what you are referring to, but if you use std::string_view as the key type, all operations already work via the implicit conversion to std::string_view:
Live On Coliru
#include <boost/container/flat_map.hpp>
#include <string_view>
int main() {
boost::container::flat_map<std::string_view, int> m {
{ "one", 1 },
{ "two", 2 },
{ "three", 3 },
{ "four", 4 },
};
std::string key = "one";
auto one = m.at(key);
auto range = m.equal_range(key);
auto it = m.find(key);
m[key] = 1;
}
The Inverse
Here you'd actually need to use a container that supports compatible-key lookup indeed. It doesn't need to be overly complicated to roll one:
Here's one:
Live On Coliru
#include <initializer_list>
#include <algorithm>
#include <utility>
#include <stdexcept>
#include <boost/container/small_vector.hpp>
template <typename K, typename V, typename Cmp = std::less<K>, typename Storage = boost::container::small_vector<std::pair<K, V>, 10> >
struct flat_map {
using key_type = K;
using mapped_type = V;
using key_compare = Cmp;
using storage = Storage;
using value_type = typename storage::value_type;
using iterator = typename Storage::iterator;
using const_iterator = typename Storage::const_iterator;
struct value_compare {
key_compare _cmp;
template <typename A, typename B>
bool operator()(A const& a, B const& b) const { return _cmp(access(a), access(b)); }
private:
static auto& access(value_type const& v) { return v.first; }
template <typename Other>
static auto& access(Other const& v) { return v; }
} _cmp;
storage _data;
flat_map(std::initializer_list<value_type> i) : _data(i) {}
iterator begin() { return _data.begin(); }
iterator end() { return _data.end(); }
const_iterator begin() const { return _data.begin(); }
const_iterator end() const { return _data.end(); }
template <typename Key>
mapped_type& operator[](Key&& key) { return find(std::forward<Key>(key))->second; }
template <typename Key>
mapped_type const& operator[](Key&& key) const { return find(std::forward<Key>(key))->second; }
template <typename Key>
iterator find(Key&& key) {
auto r = equal_range(std::forward<Key>(key));
return (r.first == r.second)? end() : r.first;
}
template <typename Key>
const_iterator find(Key&& key) const {
auto r = equal_range(std::forward<Key>(key));
return (r.first == r.second)? end() : r.first;
}
template <typename Key>
mapped_type& at(Key&& key) {
auto r = equal_range(std::forward<Key>(key));
if (r.first == r.second) throw std::out_of_range("key");
return r.first->second;
}
template <typename Key>
mapped_type const& at(Key&& key) const {
auto r = equal_range(std::forward<Key>(key));
if (r.first == r.second) throw std::out_of_range("key");
return r.first->second;
}
template <typename Key>
auto equal_range(Key&& key) { return std::equal_range(begin(), end(), std::forward<Key>(key), _cmp); }
template <typename Key>
auto equal_range(Key&& key) const { return std::equal_range(begin(), end(), std::forward<Key>(key), _cmp); }
};
It supports precisely the inverse of the first scenario (given the comparator of std::less<>):
#include <string_view>
#include <string>
int main() {
flat_map<std::string, int, std::less<> > m {
{ "one", 1 },
{ "two", 2 },
{ "three", 3 },
{ "four", 4 },
};
std::string_view key = "one";
auto one = m.at(key);
auto range = m.equal_range(key);
auto it = m.find(key);
m[key] = 1;
}

Related

can coroutine be used to solve stack overflow due to too deep recursive function call?

Has anyone tried to use coroutine to solve stack overflow caused by too deep recursive function call? according to the document on coroutines, the coroutine state will be saved on heap instead of on stack, which could have the potential to avoid the limitation imposed by the limited stack size and thus provide a way to solve the stack overflow issue in a generic way. i have tried with some code but it looks like the stack over flow issue persists. anyone has any tips/advice to share? or point me to some tutorial? thanks in advance.
// file main
#include "RecursiveCall.h"
// coroutine
static ReturnObject DoIntegration(Context& ctx, ReturnObject::promise_type* parent, double x_n)
{
double* dummyA = new double[(int)((x_n + 1) * 2)]; // an effort to prevent heap allocation from "optimized out"
co_await AwaitableBase(ctx, parent, x_n);
ctx._dummyVec.push_back(dummyA); // part of the effort to prevent heap allocation from "optimized out"
}
// caller
static double Invoke(Context& ctx, ReturnObject::promise_type* parent, double x_n)
{
auto ret = DoIntegration(ctx, parent, x_n);
std::coroutine_handle<ReturnObject::promise_type> h = ret._coroH;
auto p = h.promise();
while (!h.done())
{
if (p.AreChildrenReady())
{
h();
break;
}
}
return p._area;
}
bool AwaitableBase::await_suspend(std::coroutine_handle<PromiseType> h)
{
_promise = &h.promise();
if (_parent)
{
_parent->RegisterChild(h);
}
if (_x_n <= _ctx._b)
{
_promise->_x_n = 0.0;
_promise->_area = 0.0;
return false;
}
_promise->_area = GetArea(_x_n, _ctx._incr);
double newX = _x_n - _ctx._incr;
_promise->_x_n = newX;
double area = Invoke(_ctx, &h.promise(), newX);
//post calculation
_promise->_area += area;
return true;
}
double CallRecursive(double x0, double x_n, double incr)
{
Context ctx{ x0, incr };
return Invoke(ctx, nullptr, x_n);
}
int main()
{
double x0 = 0.0;
double x_n = 4.5;
double incr = 0.5; // no stackoveflow
//double incr = 0.0015; // stack oveflow
auto area = CallRecursive(x0, x_n, incr);
std::cout << "integrated result: " << area << "\n";
}
// file RecrusiveCall.h
#include <coroutine>
#include <exception>
#include <map>
#include <iostream>
#include <vector>
/* integration certainly can and should be done in a sequencial way in real world. but here is just use it as a simple example of recursive call, so the integration is implemented as a recursive function call and is done from high limit of x to the lower limit */
static double GetY(double x)
{
using CurvePoint = std::pair<double, double>;
constexpr CurvePoint curve[10] = { {0.0, 1.0}, {0.5, 1.2}, {1.0, 1.0}, {1.5, 1.2}, {2.0, 1.0},
{2.5, 1.2}, {3.0, 1.0}, {3.5, 1.2}, {4.0, 1.0}, {4.5, 1.2} };
if (x < curve[0].first || x > curve[9].first)
return 0.0;
CurvePoint newPoint;
const auto p1 = std::lower_bound(&curve[0], &curve[10], x, [](const auto& a, const auto& b) constexpr { return a.first < b; });
// check for special cases: first
const auto p0 = p1 - 1;
return (p1->second - p0->second) * (x - p0->first) / (p1->first - p0->first) + p0->second;
}
static double GetArea(double end, double incr)
{
return (GetY(end) + GetY(end - incr)) * 0.5 * incr;
}
struct Context
{
double _b; // lower limit of the integration range
double _incr; // increment steplength
std::vector<double*> _dummyVec; // effort to prevent heap allocation from being optimzed out
~Context()
{
for (auto p : _dummyVec)
delete p;
}
};
struct ReturnObject
{
struct promise_type
{
using Handle = std::coroutine_handle<promise_type>;
ReturnObject get_return_object() {
return { std::coroutine_handle<promise_type>::from_promise(*this) };
}
std::suspend_never initial_suspend() { return {}; }
std::suspend_always final_suspend() noexcept { return {}; }
void unhandled_exception() {}
void return_void() {}
void RegisterChild(Handle& childH)
{
_children.push_back(childH);
}
bool AreChildrenReady()
{
for (auto c : _children)
{
if (!c.done())
return false;
}
return true;
}
double GetValue() const { return _area; }
std::vector<Handle> _children;
double _area{ 0 };
double _x_n{ 0 };
};
ReturnObject(promise_type::Handle coro) : _coroH(coro)
{
}
operator std::coroutine_handle<promise_type>() const { return _coroH; }
// A coroutine_handle<promise_type> converts to coroutine_handle<>
operator std::coroutine_handle<>() const { return _coroH; }
std::coroutine_handle<promise_type> _coroH;
};
struct AwaitableBase
{
typedef Context Ctx;
using PromiseType = ReturnObject::promise_type; // todo: remove
bool await_ready()
{
return false;
}
bool await_suspend(std::coroutine_handle<PromiseType> h);
PromiseType* await_resume()
{
return _promise;
}
AwaitableBase(Ctx& ctx, PromiseType* parent, double x_n) : _ctx(ctx), _x_n(x_n), _parent(parent)
{
}
~AwaitableBase()
{
}
Ctx& _ctx;
PromiseType* _parent{ nullptr };
PromiseType* _promise{ nullptr };
double _x_n{ 0.0 };
};
no.
the coroutine's stack frame remains allocated. what's pushed to heap (as custom handle struct) is minimal register state plus struct data, to later resume the stack frame.
that is how you can access all local variables after resuming where you left off.

Is there a map-like tool in QT that can be iterated over inserted index?

From the Qt documentation about QMap::iterator :
Unlike QHash, which stores its items in an arbitrary order, QMap
stores its items ordered by key. Items that share the same key
(because they were inserted using QMap::insertMulti(), or due to a
unite()) will appear consecutively, from the most recently to the
least recently inserted value.
What I want is to interate a map by inserted index. For example this map.
const static QMap<QString, int> MEASUREMENT_COLUMNS{{"ID", MY_SQL_BIGINT}, {"logging_id", MY_SQL_INT}, {"calibration_id", MY_SQL_INT}, {"logging_comment", MY_SQL_VARCHAR255}, {"measurement_date_time", MY_SQL_DATETIME}, {"ADC0", MY_SQL_FLOAT},
{"ADC0", MY_SQL_FLOAT},
{"ADC1", MY_SQL_FLOAT},
{"ADC2", MY_SQL_FLOAT},
But the problem is as the documentation says above about QMap and QHashmap. They will not work for be if I want to iterate a map by inserted index.
For example, first ID, then logging_id, then calibration_id etc.
So I need to select something else than QMap and QHash.
Question:
Is there a map-like tool in QT that can be iterated over inserted index?
You can use two QVector, or use QVector<QPair<QString, int> > instead.
Here's the start of a QHash derivative which provides this functionality. DISCLAIMER: This is not entirely perfected! Not every function / feature of QHash has yet been accounted for. As long as you only use the functions / operator overloads provided here, you'll be fine for sure. If someone wants to keep developing this and repost a truly "finished" class, that would be great!
Note that performance will of course be degraded a bit, and memory consumption will increase, using this vs the natural QHash, but for small data sets that should be negligible.
OrderedHash.h
#ifndef ORDEREDHASH_H
#define ORDEREDHASH_H
#include <QHash>
#include <QVector>
#include <QDataStream>
#include <QDebug>
template<class K, class V>
class OrderedHash : public QHash<K,V>
{
public:
using QHash<K,V>::QHash;
#ifdef Q_COMPILER_INITIALIZER_LISTS
OrderedHash( std::initializer_list<std::pair<K, V>> list )
: QHash<K,V>::QHash()
{ foreach( auto p, list ) insert( std::get<0>(p), std::get<1>(p) ); }
#endif
// Returns the keys in the order they were inserted.
// If the ordered keys vector is blatantly out of sync with the hash
// (as may occur via the use of QHash functions not accounted for
// by this override!), this returns UNordered keys, since those are at
// least accurate.
QList<K> orderedKeys() const {
if( QHash<K,V>::size() != orderedKeys_.size() )
{
qWarning() << "OrderedHash keys are out of sync!";
return QHash<K,V>::keys();
}
return orderedKeys_.toList();
}
// This insert override "appends" to the "end" of the hash. If the key is
// already present, the entry is "moved" to the new end.
typename QHash<K,V>::iterator insert( const K &key, const V &value )
{
//qDebug() << "OrderedHash insert: " << key << ":" << value;
orderedKeys_.removeAll( key );
orderedKeys_.push_back( key );
return QHash<K,V>::insert( key, value );
}
// This additional update function perseveres the "key order" while
// modifying the value. If the key is not yet present, the entry is
// appended to the "end" of the hash.
typename QHash<K,V>::iterator update( const K &key, const V &value )
{
if( !QHash<K,V>::contains( key ) ) return insert( key, value );
return QHash<K,V>::insert( key, value );
}
int remove( const K &key )
{
orderedKeys_.removeAll( key );
return QHash<K,V>::remove( key );
}
void clear()
{
orderedKeys_.clear();
QHash<K,V>::clear();
}
private:
QVector<K> orderedKeys_;
};
// COPIED AND TWEAKED QT SOURCE FOR THESE STREAM OPERATOR OVERLOADS
template <class Key, class T>
Q_OUTOFLINE_TEMPLATE QDataStream &operator>>(QDataStream &in, OrderedHash<Key, T> &hash)
{
QDataStream::Status oldStatus = in.status();
in.resetStatus();
hash.clear();
quint32 n;
in >> n;
for (quint32 i = 0; i < n; ++i) {
if (in.status() != QDataStream::Ok)
break;
Key k;
T t;
in >> k >> t;
/* ORGINAL QT SOURCE
hash.insertMulti(k, t);
*/
//---------------------------------
hash.insert(k, t);
//---------------------------------
}
if (in.status() != QDataStream::Ok)
hash.clear();
if (oldStatus != QDataStream::Ok)
in.setStatus(oldStatus);
return in;
}
template <class Key, class T>
Q_OUTOFLINE_TEMPLATE QDataStream &operator<<(QDataStream &out, const OrderedHash<Key, T>& hash)
{
out << quint32(hash.size());
/* ORGINAL QT SOURCE
typename QHash<Key, T>::ConstIterator it = hash.end();
typename QHash<Key, T>::ConstIterator begin = hash.begin();
while (it != begin) {
--it;
out << it.key() << it.value();
}
*/
//---------------------------------
const QList<Key> keys( hash.orderedKeys() );
foreach( auto key, keys ) out << key << hash.value(key);
//---------------------------------
return out;
}
#endif // ORDEREDHASH_H
Not in QT (to my knowledge, at least).
Can you use Boost, e.g. boost::multiindex? Another option is to combine map with vector in a class +- like this (this is likely to contain errors; it's supposed to illustrate the general idea, not to be a fully working piece of code):
template<typename K, typename V>
class indexed_map
{
map<K, V> m_map;
vector<K> m_insertionOrder;
public:
void insert(const K& k, const V& v)
{
m_map.insert(k,v);
m_insertionOrder.push_back(k);
}
V byKey(const K& k) const {return m_map.at(k)};
V byOrder(size_t n) const {return m_map.at(m_insertionOrder.at(n));}
};
Of course you'll have to write some boilerplate (ok, lots of it in fact), iterators might be also tricky.

How to create a QVariant-based generic model?

Quite often I find myself in need of some custom scheme model, mandating the implementation of more and more models, made even more tedious by the inability of QObject derived classes to be templates.
Qt has the QStandardItemModel but that seems a bit verbose and inconvenient to use, especially from the qml side, and total overkill for a basic list model.
There is also the basic qml ListModel, but that is limited and not elegant to use on the C++ side, and I do suspect a tad more bloated than it needs to be.
Qt has QVariant, which is what its model/view architecture uses internally, so it is surprising that the framework doesn't provide something as simple as:
// qml code
VarMod {
roles: ["name", "age", "weight"]
Component.onCompleted: {
insert(["Jack", 34, 88.5], -1) // qml doesn't support
insert(["Mary", 26, 55.3], -1) // default arg values
}
}
// cpp code
VarMod vm { "name", "age", "weight" }; // member declaration
vm.insert({ "Jack", 34, 88.5 });
vm.insert({ "Mary", 26, 55.3 });
And here it is.
Note that you do have to be responsible with the parameters, as there is no type safety, in fact it has implicit analog to ListModel's dynamicRoles - that is, it will accept and work with any QVariant compatible value on every role slot.
As for memory efficiency, consider that QVariant has 8 bytes for data, plus 4 bytes for type id, plus another 4 bytes of padding, for a total of 16 bytes. That is not insignificant if you are using it for small data types, like say bool, so in case you have a data scheme that has a lot of small (1 - 4 bytes) fields and a scores of items, implementing a full model will still be the better option. It is still a lot better than the generic object model I am using, which has to carry the bloat of QObject, and even more significant in the case of qml objects.
Additionally, QVariant being 16 bytes, I opted to not use the convenience of QVariantList for data storage, which has an underlying QList, making the situation worse than it needs to be. Although that is fixed in Qt 6, which gets rid of QList as it is, and replaces it with an alias of QVector. Still, std::vector helps to avoid that in any case, plus it might actually be a tad faster, since it doesn't have to deal with COW and atomic ref counters. There are several auxiliary methods to help with pre-allocation and release of memory as well.
The model has a safeguard against the change the roles for obvious reasons, the latter is primarily intended to be initialized just once, but there is reset() that is intended to be used in a more dynamic qml context, making it possible to redefine the model schema on the fly and provide a compatible delegate. For the sake of certainty, the roles can only be redefined after the model has been explicitly reset.
There is a minute difference in inserting, on the c++ side, the parameter pack is passed wrapped in {}, in qml it is wrapped in [], both leveraging implicit conversion in the context specific way. Also, note that qml currently doesn't support omitting parameters with default values provided on the c++ side, so for appending you do have to provide an invalid index. Naturally, it would be trivial to add convenience methods for appending and prepending if needed.
In addition to the syntax example of the question, it is also possible to add multiple items at once, from "declarative-y" qml structure such as:
let v = [["Jack", 34, 88.5],
["Mary", 26, 55.3],
["Sue", 22, 69.6]]
vm.insertList(v, -1)
Finally, type safety is possible to implement if the scenario really calls for it, then each role can be specified with the expected type to go with it, such as:
VarMod vm {{"name", QMetaType::QString},
{"age", QMetaType::Int},
{"weight", QMetaType::QReal}};
and then iterating and making the necessary checks to ensure type safety when inserting.
Update: I also added serialization, and save/load from disk features, note that this will serialize the data together with the mode schema.
class VarMod : public QAbstractListModel {
Q_OBJECT
Q_PROPERTY(QVariantList roles READ roles WRITE setRoles NOTIFY rolesChanged)
QVariantList vroles;
QVariantList roles() const { return vroles; }
QHash<int, QByteArray> _roles;
std::vector<std::vector<QVariant>> _data;
inline bool checkArgs(int rc) const {
if (rc == _roles.size()) return true;
qWarning() << "arg size mismatch, got / expected" << rc << _roles.size();
return false;
}
inline bool inBounds(int i, bool ok = false) const {
if (i > -1 && i < (int)_data.size()) return true;
if (!ok) qWarning() << "out of bounds" << i; // do not warn if intentionally appending
return false;
}
inline bool validRole(int r) const { return (r > -1 && r < _roles.size()); }
protected:
QHash<int, QByteArray> roleNames() const override { return _roles; }
int rowCount(const QModelIndex &) const override { return _data.size(); }
QVariant data(const QModelIndex &index, int r) const override {
r = r - Qt::UserRole - 1;
if (inBounds(index.row()) && validRole(r)) return _data[index.row()][r];
return QVariant();
}
public:
VarMod() {} // for qml
VarMod(std::initializer_list<QByteArray> r) {
int rc = Qt::UserRole + 1;
for (const auto & ri : r) {
_roles.insert(rc++, ri);
vroles << QString::fromLatin1(ri);
}
rolesChanged();
}
inline void insert(std::initializer_list<QVariant> s, int i = -1) {
if (!checkArgs(s.size())) return;
insert(QVariantList(s), i);
}
inline bool setItem(int i, std::initializer_list<QVariant> s) {
if (checkArgs(s.size())) return setItem(i, QVariantList(s));
return false;
}
void setRoles(QVariantList r) {
if (_roles.empty()) {
int rc = Qt::UserRole + 1;
for (const auto & vi : r) _roles.insert(rc++, vi.toByteArray());
vroles = r;
rolesChanged();
} else qWarning() << "roles are already initialized";
}
void read(QDataStream & d) {
reset();
QVariantList vr;
d >> vr;
quint32 s;
d >> s;
_data.resize(s);
for (uint i = 0; i < s; ++i) {
_data[i].reserve(vr.size());
for (int c = 0; c < vr.size(); ++c) {
QVariant var;
d >> var;
_data[i].push_back(std::move(var));
}
}
setRoles(vr);
beginResetModel();
endResetModel();
}
void write(QDataStream & d) const {
d << vroles;
d << (quint32)_data.size();
for (const auto & v : _data) {
for (const auto & i : v) d << i;
}
}
public slots:
void insert(QVariantList s, int i) {
if (!inBounds(i, true)) i = _data.size();
if (!checkArgs(s.size())) return;
beginInsertRows(QModelIndex(), i, i);
_data.insert(_data.begin() + i, { s.cbegin(), s.cend() });
endInsertRows();
}
void insertList(QVariantList s, int i) {
if (!inBounds(i, true)) i = _data.size();
int added = 0;
for (const auto & il : s) {
QVariantList ll = il.value<QVariantList>();
if (checkArgs(ll.size())) {
_data.insert(_data.begin() + i + added++, { ll.cbegin(), ll.cend() });
}
}
if (added) {
beginInsertRows(QModelIndex(), i, i + added - 1);
endInsertRows();
}
}
bool setData(int i, int r, QVariant d) {
if (!inBounds(i) || !validRole(r)) return false;
_data[i][r] = d;
dataChanged(index(i), index(i));
return true;
}
bool setDataStr(int i, QString rs, QVariant d) { // a tad slower
int r = _roles.key(rs.toLatin1()); // role is resolved in linear time
if (r) return setData(i, r - Qt::UserRole - 1, d);
qWarning() << "invalid role" << rs;
return false;
}
bool setItem(int i, QVariantList d) {
if (!inBounds(i) || !checkArgs(d.size())) return false;
_data[i] = { d.cbegin(), d.cend() };
dataChanged(index(i), index(i));
return true;
}
QVariantList item(int i) const {
if (!inBounds(i)) return QVariantList();
const auto & v = _data[i];
return { v.begin(), v.end() };
}
QVariant getData(int i, int r) const {
if (inBounds(i) && validRole(r)) return _data[i][r];
return QVariant();
}
QVariant getDataStr(int i, QString rs) const {
int r = _roles.key(rs.toLatin1()); // role is resolved in linear time
if (r) return getData(i, r);
qWarning() << "invalid role" << rs;
return QVariant();
}
QVariantList take(int i) {
QVariantList res = item(i);
if (res.size()) remove(i);
return res;
}
bool swap(int i1, int i2) {
if (!inBounds(i1) || !inBounds(i2)) return false;
std::iter_swap(_data.begin() + i1, _data.begin() + i2);
dataChanged(index(i1), index(i1));
dataChanged(index(i2), index(i2));
return true;
}
bool remove(int i) {
if (!inBounds(i)) return false;
beginRemoveRows(QModelIndex(), i, i);
_data.erase(_data.begin() + i);
endRemoveRows();
return true;
}
void clear() {
beginResetModel();
_data.clear();
_data.shrink_to_fit();
endResetModel();
}
void reset() {
clear();
_roles.clear();
vroles.clear();
rolesChanged();
}
void reserve(int c) { _data.reserve(c); }
int size() const { return _data.size(); }
int capacity() const { return _data.capacity(); }
void squeeze() { _data.shrink_to_fit(); }
bool fromFile(QString path) {
QFile f(path);
if (!f.open(QIODevice::ReadOnly)) return false;
QDataStream d(&f);
read(d); // assumes correct data
return true;
}
bool toFile(QString path) const {
QFile f(path);
if (!f.open(QIODevice::WriteOnly)) return false;
QDataStream d(&f);
write(d);
return true;
}
signals:
void rolesChanged();
};
I also created this sorting/filtering view to supplement the model:
class View : public QSortFilterProxyModel {
Q_OBJECT
Q_PROPERTY(QJSValue filter READ filter WRITE set_filter NOTIFY filterChanged)
Q_PROPERTY(bool reverse READ reverse WRITE setReverse NOTIFY reverseChanged)
bool reverse() const { return _reverse; }
void setReverse(bool v) {
if (v == _reverse) return;
_reverse = v;
reverseChanged();
sort(0, (Qt::SortOrder)_reverse);
}
bool _reverse = false;
mutable QJSValue m_filter;
QJSValue & filter() const { return m_filter; }
void set_filter(QJSValue & f) {
if (!m_filter.equals(f))
m_filter = f;
filterChanged();
invalidateFilter();
}
}
public:
View(QObject *parent = 0) : QSortFilterProxyModel(parent) { sort(0, (Qt::SortOrder)_reverse); }
signals:
void filterChanged();
void reverseChanged();
protected:
bool filterAcceptsRow(int sourceRow, const QModelIndex &) const override {
if (!m_filter.isCallable()) return true;
VarMod * vm = qobject_cast<VarMod *>(sourceModel());
if (!vm) {
qWarning() << "model is not varmod";
return true;
}
return m_filter.call({_engine->toScriptValue(vm->item(sourceRow))}).toBool();
}
bool lessThan(const QModelIndex &left, const QModelIndex &right) const override {
VarMod * vm = qobject_cast<VarMod *>(sourceModel());
if (!vm) {
qWarning() << "model is not varmod";
return false;
}
return vm->getData(left.row(), sortRole()) < vm->getData(right.row(), sortRole());
}
};
For sorting, you just have to specify the sorting role, note that it is the index of the "column" rather than the int value from the roles hash. For filtering it works via a qml functor that receives the model item as a JS array, and expects to return a bool, a c++ functor can be easily added via std::function if needed. Also note that it needs a pointer to the actual qml engine.
View {
id: vv
sourceModel: vm
sortRole: sr.value
reverse: rev.checked
filter: { sa.value; o => o[1] < sa.value } // "capturing" sa.value to react to value changes
}

C++11 range-based for-loop: how to use a container with objects and pointers-to-object

I have two containers:
std::vector< ObjectClass > vecD; // container of objects
std::vector< ObjectClass* > vecP; // container of pointers
in my code, I want to loop over all elements. As far as I know, I need to write distinct for loops, that means
// container of objects
for ( const auto& elem : vecD )
elem.doStuff();
// container of pointers
for ( const auto& elem : vecP )
elem->doStuff(); // the "->" is needed instead of "."
Is there a way to tell the loop "if the elemets are objects, use them directly. Otherwise, dereference them first"?
update
Here is a more elaborate example to clarify:
I have those containers. These are each used in a templated function:
template< typename ContainerT >
void myfunc( const ContainerT& container)
{
for ( const auto& elem : container )
{
if ( elem_is_a_pointer ) //how can this work?
elem->doStuff(); // member function
else
elem.doStuff();
}
}
You could write little helpers, which perform the dereferencing if neccessary
template<typename T>
T& deref(T* p) { return *p; }
template<typename T>
T& deref(T& p) { return p; }
then the code for the loops is equivalent
// container of objects
for ( const auto& elem : vecD )
deref(elem).doStuff();
// container of pointers
for ( const auto& elem : vecP )
deref(elem).doStuff();
You could write a non-member function to do the dispatch:
void doStuff (const ObjectClass& obj) {
obj.doStuff();
}
void doStuff (const ObjectClass* obj) {
obj->doStuff();
}
Then you can write both loops in the same way:
for ( const auto& elem : vecD )
doStuff(elem);
for ( const auto& elem : vecP )
doStuff(elem);
For the code in your edit, you could write a simple helper function to strip off pointers, then call on that:
template <typename T>
T& strip_pointers (T& obj) {
return obj;
}
template <typename T>
T& strip_pointers (T* obj) {
return *obj;
}
template< typename ContainerT >
void myfunc( const ContainerT& container)
{
for ( const auto& elem : container )
{
strip_pointers(elem).doStuff();
}
}
Nope, but you could use next approach:
for ( const auto* pElem : vecP )
{
const auto& elem = *pElem;
elem.doStuff();
}

find_if in MFC container with iterator derived from std::iterator

I have a working iterator for MFC CObList - BaseMFCIter. It works for iterating in loop but i still didn't managed to make ListIter to work properly with STL algorithm find_if.
Code
#include < iterator >
#include "afxwin.h"
#include "afxtempl.h"
#include <iostream>
#include <algorithm>
#include <cstdlib>
class myCObject : public CObject
{
public:
myCObject( std::string val )
{
x = val;
}
std::string x;
};
template < typename Item, class Cont, class Key = POSITION >
class BaseMFCIter : public std::iterator < std::input_iterator_tag, Item >
{
public:
// Define types for the 2 member functions to be used:
typedef Key (Cont::*GetFirstFunctionPtr) () const;
typedef Item (Cont::*GetNextFunctionPtr) (Key&) const;
// Default constructor, makes a null iterator, equal to BaseMFCIter::end()
BaseMFCIter() : m_pCont(0), m_Pos(0), m_GetFirstFunc(0), m_GetNextFunc(0), m_End(true) {}
// Constructor taking pointer to container and the iteration functions
BaseMFCIter(Cont* pCont, GetFirstFunctionPtr pFF, GetNextFunctionPtr pNF)
: m_pCont(pCont), m_Pos(0), m_GetFirstFunc(pFF), m_GetNextFunc(pNF)
{ init(); }
// Copy constructor, initialises iterator to first element
BaseMFCIter(const BaseMFCIter& vi) : m_pCont(vi.m_pCont), m_Pos(0),
m_GetFirstFunc(vi.m_GetFirstFunc), m_GetNextFunc(vi.m_GetNextFunc)
{ init(); }
// Assignment operator, initialises iterator to first element
BaseMFCIter& operator=(const BaseMFCIter& vi)
{
m_pCont = vi.m_pCont;
m_GetFirstFunc = vi.m_GetFirstFunc;
m_GetNextFunc = vi.m_GetNextFunc;
init();
return *this;
}
bool operator == (const BaseMFCIter& rhs) const
{ return (m_Pos == rhs.m_Pos && m_End == rhs.m_End); }
bool operator != (const BaseMFCIter& rhs) const
{ return !operator==(rhs); }
BaseMFCIter& operator ++ () { advance(); return *this; }
BaseMFCIter& operator ++ (int) { BaseMFCIter ret(*this); advance(); return ret; }
Item operator * () { return m_Item; }
Item operator -> () { return m_Item; }
static BaseMFCIter end () { return BaseMFCIter(); } // end() returns default null iterator
private:
Item m_Item; // Current item from container
Cont* m_pCont; // Pointer to container
Key m_Pos; // Key to item in container
bool m_End; // Flag to indicate end of container reached
// Pointers to container iteration functions
GetFirstFunctionPtr m_GetFirstFunc;
GetNextFunctionPtr m_GetNextFunc;
// Use container GetFirst & GetNext functions to set to first element, or end() if not found
void init()
{
m_Pos = 0;
m_End = true;
if (m_pCont && m_GetFirstFunc != 0)
{
m_Pos = (m_pCont->*m_GetFirstFunc)();
advance();
}
}
// Use container GetNext function to find next element in container
void advance()
{
m_End = m_Pos ? false : true;
m_Item = (m_Pos && m_pCont && m_GetNextFunc != 0) ?
(m_pCont->*m_GetNextFunc)(m_Pos) : Item();
}
};
struct Container : public CObList
{
myCObject* GetNext(POSITION& rPosition)
{
return dynamic_cast<myCObject*>(CObList::GetNext(rPosition));
}
myCObject const* GetNext(POSITION& rPosition) const
{
return dynamic_cast<const myCObject*>(CObList::GetNext(rPosition));
}
};
class ListIter : public BaseMFCIter < const myCObject*, Container, POSITION >
{
public:
ListIter( Container* pObj = 0)
: BaseMFCIter< const myCObject*, Container, POSITION >
(pObj, &CObList::GetHeadPosition, &Container::GetNext)
{
}
};
struct Comparator
{
std::string stringToCompare;
bool operator() ( const myCObject* lhs )
{
return (bool) lhs->x.compare( stringToCompare );
}
};
void main( )
{
myCObject* m = new myCObject( "one" );
myCObject* n = new myCObject( "two" );
myCObject* p = new myCObject( "three" );
myCObject* q = new myCObject( "four" );
Container cont;
cont.AddHead( m );
cont.AddHead( n );
cont.AddHead( p );
cont.AddHead( q );
Comparator pred;
pred.stringToCompare = "1";
ListIter iter = ListIter( &cont );
ListIter endIter = ListIter( );
ListIter foundIter = std::find_if( iter, endIter, pred );
std::cout << "foundIter x is: " << foundIter->x.c_str() << std::endl;
}
gives me foundIter x is: four. This propably happens because of the way the end position is defined so
_InIt _Find_if(_InIt _First, _InIt _Last, _Pr _Pred)
{ // find first satisfying _Pred
_DEBUG_RANGE(_First, _Last);
_DEBUG_POINTER(_Pred);
for (; _First != _Last; ++_First)
if (_Pred(*_First))
break;
return (_First);
}
doesn't iterate properly but i can't figure out how to fix it.
A number of issues fixed:
(bool) lhs->x.compare( stringToCompare ) returns true _whenever the string don't match** (see string::compare)
you were searching for "1", which doesn't exist
since the predicate was wrong, you received the first match, which was the first element, also inserted the last, and the name was "four" :)
you didn't check whether a valid match was found (dereferencing the end-iterator is illegal and may crash your program or do worse things: undefined behaviour)
you had a superflous x.c_str() in the output statement
I changed the Compare predicate around to be more idiomatic:
initialize stringToCompare from the constructor
make the field const
make the operator() a const method
This should do the trick (untested code, I'm not near a compiler the coming hours)
Update
After arriving home, I finally broke out the debugger to track that strange behaviour (see comments).
To my dismay, I found out that the BaseMFCIter was designed by someone with very limited understanding of what an iterator is: the copy constructor and assignment operator were completely wrong: they had the effect of creating a new begin iterator - for the same collection. This however, means that an iterator could never be returned from a function.
Therefore, I fixed it (first by implementing it right, later by removing the now-redundant constructor and operator= in favour of the compiler-generated default implementations).
See the full history of my and your edits:
git clone git://gist.github.com/1353471.git
sehe 11 minutes ago rely on default generated copy constructor and assignment instead
sehe 12 minutes ago fixed broken copy constructor and assignment
sehe 65 minutes ago tentative
Dmitry 73 minutes ago Attempt at find_if with predicate
sehe Heeren 25 hours ago Fixed and Tested (VS2010)
sehe 25 hours ago ( STL iterator for MFC container CObList )
#include "afxwin.h"
#include "afxtempl.h"
#include <iostream>
#include <algorithm>
#include <cstdlib>
#include <string>
#include <iterator>
class myCObject : public CObject
{
public:
myCObject( const std::string& val ) { x = val; }
std::string x;
};
template < typename Item, class Cont, class Key = POSITION >
class BaseMFCIter : public std::iterator < std::input_iterator_tag, Item >
{
public:
// Define types for the 2 member functions to be used:
typedef Key (Cont::*GetFirstFunctionPtr) () const;
typedef Item (Cont::*GetNextFunctionPtr) (Key&) const;
// Default constructor, makes a null iterator, equal to BaseMFCIter::end()
BaseMFCIter() : m_pCont(0), m_Pos(0), m_GetFirstFunc(0), m_GetNextFunc(0), m_End(true) {}
// Constructor taking pointer to container and the iteration functions
BaseMFCIter(Cont* pCont, GetFirstFunctionPtr pFF, GetNextFunctionPtr pNF)
: m_pCont(pCont), m_Pos(0), m_GetFirstFunc(pFF), m_GetNextFunc(pNF)
{ init(); }
bool operator == (const BaseMFCIter& rhs) const
{ return (m_Pos == rhs.m_Pos && m_End == rhs.m_End); }
bool operator != (const BaseMFCIter& rhs) const
{ return !operator==(rhs); }
BaseMFCIter& operator ++ () { advance(); return *this; }
BaseMFCIter& operator ++ (int) { BaseMFCIter ret(*this); advance(); return ret; }
Item operator * () { return m_Item; }
Item operator -> () { return m_Item; }
static BaseMFCIter end () { return BaseMFCIter(); } // end() returns default null iterator
private:
Item m_Item; // Current item from container
Cont* m_pCont; // Pointer to container
Key m_Pos; // Key to item in container
bool m_End; // Flag to indicate end of container reached
// Pointers to container iteration functions
GetFirstFunctionPtr m_GetFirstFunc;
GetNextFunctionPtr m_GetNextFunc;
// Use container GetFirst & GetNext functions to set to first element, or end() if not found
void init()
{
m_Pos = 0;
m_End = true;
if (m_pCont && m_GetFirstFunc != 0)
{
m_Pos = (m_pCont->*m_GetFirstFunc)();
advance();
}
}
// Use container GetNext function to find next element in container
void advance()
{
m_End = m_Pos ? false : true;
m_Item = (m_Pos && m_pCont && m_GetNextFunc != 0) ?
(m_pCont->*m_GetNextFunc)(m_Pos) : Item();
}
};
struct Container : public CObList
{
myCObject* GetNext(POSITION& rPosition)
{
return dynamic_cast<myCObject*>(CObList::GetNext(rPosition));
}
myCObject const* GetNext(POSITION& rPosition) const
{
return dynamic_cast<const myCObject*>(CObList::GetNext(rPosition));
}
};
class ListIter : public BaseMFCIter < const myCObject*, Container, POSITION >
{
public:
ListIter( Container* pObj = 0)
: BaseMFCIter< const myCObject*, Container, POSITION >
(pObj, &CObList::GetHeadPosition, &Container::GetNext)
{
}
};
struct Comparator
{
Comparator(const std::string& compareTo) : stringToCompare(compareTo) {}
bool operator() ( const myCObject* lhs ) const
{
return 0 == lhs->x.compare( stringToCompare );
}
private:
const std::string stringToCompare;
};
void main( )
{
myCObject* m = new myCObject( "one" );
myCObject* n = new myCObject( "two" );
myCObject* p = new myCObject( "three" );
myCObject* q = new myCObject( "four" );
Container cont;
cont.AddHead( m );
cont.AddHead( n );
cont.AddHead( p );
cont.AddHead( q );
Comparator pred("three");
ListIter iter = ListIter(&cont),
endIter = ListIter( );
ListIter foundIter = std::find_if( iter, endIter, pred );
if (endIter != foundIter)
{
std::cout << "foundIter x is: " << foundIter->x << std::endl;
}
else
{
std::cout << "not found" << std::endl;
}
}

Resources