Merge pull request #2346 from chriseth/rangeMask

Documentation and tests for RangeMask.

libdevcore/RangeMask.h

@@ -24,6 +24,7 @@
 #include <map>
 #include <utility>
 #include <vector>
+#include <iterator>
 #include <iostream>
 #include <assert.h>
 
@@ -35,6 +36,12 @@ class RLPStream;
 using UnsignedRange = std::pair<unsigned, unsigned>;
 using UnsignedRanges = std::vector<UnsignedRange>;
 
+/**
+ * Set of elements of a certain "ground range" representable by unions of ranges inside this
+ * ground range.
+ * Ranges are given as pairs (begin, end), denoting the interval [begin, end), i.e. end is excluded.
+ * Supports set-theoretic operators, size and iteration.
+ */
 template <class T>
 class RangeMask
 {
@@ -44,14 +51,19 @@ public:
 	using Range = std::pair<T, T>;
 	using Ranges = std::vector<Range>;
 
+	/// Constructs an empty range mask with empty ground range.
 	RangeMask(): m_all(0, 0) {}
+	/// Constructs an empty range mask with ground range [_begin, _end).
 	RangeMask(T _begin, T _end): m_all(_begin, _end) {}
+	/// Constructs an empty range mask with ground range _c.
 	RangeMask(Range const& _c): m_all(_c) {}
 
+	/// @returns the union with the range mask _m, taking also the union of the ground ranges.
 	RangeMask unionedWith(RangeMask const& _m) const { return operator+(_m); }
 	RangeMask operator+(RangeMask const& _m) const { return RangeMask(*this) += _m; }
 
-	RangeMask lowest(T _items) const
+	/// @returns a new range mask containing the smallest _items elements (not ranges).
+	RangeMask lowest(decltype(T{} - T{}) _items) const
 	{
 		RangeMask ret(m_all);
 		for (auto i = m_ranges.begin(); i != m_ranges.end() && _items; ++i)
@@ -59,8 +71,10 @@ public:
 		return ret;
 	}
 
+	/// @returns the complement of the range mask relative to the ground range.
 	RangeMask operator~() const { return inverted(); }
 
+	/// @returns a copy of this range mask representing the complement relative to the ground range.
 	RangeMask inverted() const
 	{
 		RangeMask ret(m_all);
@@ -76,6 +90,8 @@ public:
 		return ret;
 	}
 
+	/// Changes the range mask to its complement relative to the ground range and returns a
+	/// reference to itself.
 	RangeMask& invert() { return *this = inverted(); }
 
 	template <class S> RangeMask operator-(S const& _m) const { auto ret = *this; return ret -= _m; }
@@ -92,61 +108,13 @@ public:
 		return *this;
 	}
 	RangeMask& operator+=(Range const& _m) { return unionWith(_m); }
-	RangeMask& unionWith(Range const& _m)
-	{
-		for (auto i = _m.first; i < _m.second;)
-		{
-			assert(i >= m_all.first);
-			assert(i < m_all.second);
-			// for each number, we find the element equal or next lower. this, if any, must contain the value.
-			auto uit = m_ranges.upper_bound(i);
-			auto it = uit == m_ranges.begin() ? m_ranges.end() : std::prev(uit);
-			if (it == m_ranges.end() || it->second < i)
-				// lower range is too low to merge.
-				// if the next higher range is too high.
-				if (uit == m_ranges.end() || uit->first > _m.second)
-				{
-					// just create a new range
-					m_ranges[i] = _m.second;
-					break;
-				}
-				else
-				{
-					if (uit->first == i)
-						// move i to end of range
-						i = uit->second;
-					else
-					{
-						// merge with the next higher range
-						// move i to end of range
-						i = m_ranges[i] = uit->second;
-						i = uit->second;
-						m_ranges.erase(uit);
-					}
-				}
-			else if (it->second == i)
-			{
-				// if the next higher range is too high.
-				if (uit == m_ranges.end() || uit->first > _m.second)
-				{
-					// merge with the next lower range
-					m_ranges[it->first] = _m.second;
-					break;
-				}
-				else
-				{
-					// merge with both next lower & next higher.
-					i = m_ranges[it->first] = uit->second;
-					m_ranges.erase(uit);
-				}
-			}
-			else
-				i = it->second;
-		}
-		return *this;
-	}
+	/// Modifies this range mask to also include the range _m, which has to be a subset of
+	/// the ground range.
+	RangeMask& unionWith(Range const& _m);
 
+	/// Adds the single element _i to the range mask.
 	RangeMask& operator+=(T _m) { return unionWith(_m); }
+	/// Adds the single element _i to the range mask.
 	RangeMask& unionWith(T _i)
 	{
 		return operator+=(Range(_i, _i + 1));
@@ -181,10 +149,12 @@ public:
 		m_all = std::make_pair(0, 0);
 	}
 
+	/// @returns the ground range.
 	std::pair<T, T> const& all() const { return m_all; }
+	/// Extends the ground range to include _i.
 	void extendAll(T _i) { m_all = std::make_pair(std::min(m_all.first, _i), std::max(m_all.second, _i + 1)); }
 
-	class const_iterator
+	class const_iterator: public std::iterator<std::forward_iterator_tag, T>
 	{
 		friend class RangeMask;
 
@@ -208,6 +178,8 @@ public:
 
 	const_iterator begin() const { return const_iterator(*this, false); }
 	const_iterator end() const { return const_iterator(*this, true); }
+	/// @returns the smallest element in the range mask that is larger than _t or the end of the
+	/// base range if such an element does not exist.
 	T next(T _t) const
 	{
 		_t++;
@@ -219,6 +191,7 @@ public:
 		return uit == m_ranges.end() ? m_all.second : uit->first;
 	}
 
+	/// @returns the number of elements (not ranges) in the range mask.
 	size_t size() const
 	{
 		size_t c = 0;
@@ -228,7 +201,9 @@ public:
 	}
 
 private:
+	/// The ground range.
 	UnsignedRange m_all;
+	/// Mapping begin -> end containing the ranges.
 	std::map<T, T> m_ranges;
 };
 
@@ -241,4 +216,65 @@ template <class T> inline std::ostream& operator<<(std::ostream& _out, RangeMask
 	return _out;
 }
 
+template <class T>
+RangeMask<T>& RangeMask<T>::unionWith(typename RangeMask<T>::Range const& _m)
+{
+	for (auto i = _m.first; i < _m.second;)
+	{
+		assert(i >= m_all.first);
+		assert(i < m_all.second);
+		// For each number, we find the element equal or next lower. this, if any, must contain the value.
+		// First range that starts after i.
+		auto rangeAfter = m_ranges.upper_bound(i);
+		// Range before rangeAfter or "end" if the rangeAfter is the first ever...
+		auto it = rangeAfter == m_ranges.begin() ? m_ranges.end() : std::prev(rangeAfter);
+		if (it == m_ranges.end() || it->second < i)
+		{
+			// i is either before the first range or between two ranges (with some distance
+			// so that we cannot merge it onto "it").
+			// lower range is too low to merge.
+			// if the next higher range is too high.
+			if (rangeAfter == m_ranges.end() || rangeAfter->first > _m.second)
+			{
+				// just create a new range
+				m_ranges[i] = _m.second;
+				break;
+			}
+			else
+			{
+				if (rangeAfter->first == i)
+					// move i to end of range
+					i = rangeAfter->second;
+				else
+				{
+					// merge with the next higher range
+					// move i to end of range
+					i = m_ranges[i] = rangeAfter->second;
+					m_ranges.erase(rangeAfter);
+				}
+			}
+		}
+		else if (it->second == i)
+		{
+			// The range before i ends with i.
+			// if the next higher range is too high.
+			if (rangeAfter == m_ranges.end() || rangeAfter->first > _m.second)
+			{
+				// merge with the next lower range
+				m_ranges[it->first] = _m.second;
+				break;
+			}
+			else
+			{
+				// merge with both next lower & next higher.
+				i = m_ranges[it->first] = rangeAfter->second;
+				m_ranges.erase(rangeAfter);
+			}
+		}
+		else
+			i = it->second;
+	}
+	return *this;
+}
+
 }

libethereum/BlockChainSync.h

@@ -24,7 +24,6 @@
 #include <mutex>
 
 #include <libdevcore/Guards.h>
-#include <libdevcore/RangeMask.h>
 #include <libethcore/Common.h>
 #include <libp2p/Common.h>
 #include "CommonNet.h"

libethereum/EthereumHost.h

@@ -31,7 +31,6 @@
 
 #include <libdevcore/Guards.h>
 #include <libdevcore/Worker.h>
-#include <libdevcore/RangeMask.h>
 #include <libethcore/Common.h>
 #include <libp2p/Common.h>
 #include "CommonNet.h"

libethereum/EthereumPeer.h

@@ -31,7 +31,6 @@
 
 #include <libdevcore/RLP.h>
 #include <libdevcore/Guards.h>
-#include <libdevcore/RangeMask.h>
 #include <libethcore/Common.h>
 #include <libp2p/Capability.h>
 #include "CommonNet.h"

libp2p/Host.h

@@ -33,7 +33,6 @@
 
 #include <libdevcore/Guards.h>
 #include <libdevcore/Worker.h>
-#include <libdevcore/RangeMask.h>
 #include <libdevcrypto/Common.h>
 #include <libdevcrypto/ECDHE.h>
 #include "NodeTable.h"

libp2p/Session.h

@@ -31,7 +31,6 @@
 
 #include <libdevcore/Common.h>
 #include <libdevcore/RLP.h>
-#include <libdevcore/RangeMask.h>
 #include <libdevcore/Guards.h>
 #include "RLPXFrameCoder.h"
 #include "RLPXSocket.h"

test/libdevcore/RangeMask.cpp

@@ -0,0 +1,148 @@
+/*
+	This file is part of cpp-ethereum.
+
+	cpp-ethereum is free software: you can redistribute it and/or modify
+	it under the terms of the GNU General Public License as published by
+	the Free Software Foundation, either version 3 of the License, or
+	(at your option) any later version.
+
+	cpp-ethereum is distributed in the hope that it will be useful,
+	but WITHOUT ANY WARRANTY; without even the implied warranty of
+	MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+	GNU General Public License for more details.
+
+	You should have received a copy of the GNU General Public License
+	along with cpp-ethereum.  If not, see <http://www.gnu.org/licenses/>.
+*/
+/** @file RangeMask.cpp
+ * @author Christian <c@ethdev.com>
+ * @date 2015
+ */
+
+#include <libdevcore/RangeMask.h>
+#include "../TestHelper.h"
+
+using namespace std;
+using namespace dev;
+
+namespace dev
+{
+namespace test
+{
+
+BOOST_AUTO_TEST_SUITE(RangeMaskTest)
+
+BOOST_AUTO_TEST_CASE(constructor)
+{
+	using RM = RangeMask<unsigned>;
+	using Range = pair<unsigned, unsigned>;
+	for (RM r: {RM(), RM(1, 10), RM(Range(2, 10))})
+	{
+		BOOST_CHECK(r.empty());
+		BOOST_CHECK(!r.contains(0));
+		BOOST_CHECK(!r.contains(1));
+		BOOST_CHECK_EQUAL(0, r.size());
+	}
+	BOOST_CHECK(RM().full());
+	BOOST_CHECK(!RM(1, 10).full());
+	BOOST_CHECK(!RM(Range(2, 10)).full());
+}
+
+BOOST_AUTO_TEST_CASE(simple_unions)
+{
+	using RM = RangeMask<unsigned>;
+	using Range = pair<unsigned, unsigned>;
+	RM m(Range(0, 2000));
+	m.unionWith(Range(1, 2));
+	BOOST_CHECK_EQUAL(m.size(), 1);
+	m.unionWith(Range(50, 250));
+	BOOST_CHECK_EQUAL(m.size(), 201);
+	m.unionWith(Range(10, 16));
+	BOOST_CHECK_EQUAL(m.size(), 207);
+	BOOST_CHECK(m.contains(1));
+	BOOST_CHECK(m.contains(11));
+	BOOST_CHECK(m.contains(51));
+	BOOST_CHECK(m.contains(200));
+	BOOST_CHECK(!m.contains(2));
+	BOOST_CHECK(!m.contains(7));
+	BOOST_CHECK(!m.contains(17));
+	BOOST_CHECK(!m.contains(258));
+}
+
+BOOST_AUTO_TEST_CASE(empty_union)
+{
+	using RM = RangeMask<unsigned>;
+	using Range = pair<unsigned, unsigned>;
+	RM m(Range(0, 2000));
+	m.unionWith(Range(3, 6));
+	BOOST_CHECK_EQUAL(m.size(), 3);
+	m.unionWith(Range(50, 50));
+	BOOST_CHECK_EQUAL(m.size(), 3);
+	m.unionWith(Range(0, 0));
+	BOOST_CHECK_EQUAL(m.size(), 3);
+	m.unionWith(Range(1, 1));
+	BOOST_CHECK_EQUAL(m.size(), 3);
+	m.unionWith(Range(2, 2));
+	BOOST_CHECK_EQUAL(m.size(), 3);
+	m.unionWith(Range(3, 3));
+	BOOST_CHECK_EQUAL(m.size(), 3);
+}
+
+BOOST_AUTO_TEST_CASE(overlapping_unions)
+{
+	using RM = RangeMask<unsigned>;
+	using Range = pair<unsigned, unsigned>;
+	RM m(Range(0, 2000));
+	m.unionWith(Range(10, 20));
+	BOOST_CHECK_EQUAL(10, m.size());
+	m.unionWith(Range(30, 40));
+	BOOST_CHECK_EQUAL(20, m.size());
+	m.unionWith(Range(15, 30));
+	BOOST_CHECK_EQUAL(40 - 10, m.size());
+	m.unionWith(Range(50, 60));
+	m.unionWith(Range(45, 55));
+	// [40, 45) still missing here
+	BOOST_CHECK_EQUAL(60 - 10 - 5, m.size());
+	m.unionWith(Range(15, 56));
+	BOOST_CHECK_EQUAL(60 - 10, m.size());
+	m.unionWith(Range(15, 65));
+	BOOST_CHECK_EQUAL(65 - 10, m.size());
+	m.unionWith(Range(5, 70));
+	BOOST_CHECK_EQUAL(70 - 5, m.size());
+}
+
+BOOST_AUTO_TEST_CASE(complement)
+{
+	using RM = RangeMask<unsigned>;
+	using Range = pair<unsigned, unsigned>;
+	RM m(Range(0, 2000));
+	m.unionWith(7).unionWith(9);
+	m = ~m;
+	m.unionWith(7).unionWith(9);
+	m = ~m;
+	BOOST_CHECK(m.empty());
+
+	m += Range(0, 10);
+	m += Range(1000, 2000);
+	m.invert();
+	BOOST_CHECK_EQUAL(m.size(), 1000 - 10);
+}
+
+BOOST_AUTO_TEST_CASE(iterator)
+{
+	using RM = RangeMask<unsigned>;
+	using Range = pair<unsigned, unsigned>;
+	RM m(Range(0, 2000));
+	m.unionWith(Range(7, 9));
+	m.unionWith(11);
+	m.unionWith(Range(200, 205));
+
+	vector<unsigned> elements;
+	copy(m.begin(), m.end(), back_inserter(elements));
+	BOOST_CHECK(elements == (vector<unsigned>{7, 8, 11, 200, 201, 202, 203, 204}));
+}
+
+BOOST_AUTO_TEST_SUITE_END()
+
+}
+}