implement n log(n) compression algorigthm

This commit is contained in:
Mikhail Surin 2018-06-16 21:40:43 +03:00
parent b09323c547
commit 984fe8e4fd
2 changed files with 105 additions and 59 deletions

View File

@ -7,27 +7,36 @@
namespace DB {
namespace ErrorCodes
{
extern const int NUMBER_OF_ARGUMENTS_DOESNT_MATCH;
extern const int ILLEGAL_TYPE_OF_ARGUMENT;
extern const int BAD_ARGUMENTS;
}
namespace {
AggregateFunctionPtr createAggregateFunctionHistogram(const std::string & name, const DataTypes & arguments, const Array & params)
{
if (params.size() != 1)
{
throw Exception("Function " + name + " requires only bins count");
}
throw Exception("Function " + name + " requires bins count", ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH);
UInt32 bins_count = applyVisitor(FieldVisitorConvertToNumber<UInt32>(), params[0]);
if (bins_count == 0)
throw Exception("Bin count should be positive", ErrorCodes::BAD_ARGUMENTS);
assertUnary(name, arguments);
UInt32 bins_count;
#define READ(VAL, PARAM) \
VAL = applyVisitor(FieldVisitorConvertToNumber<decltype(VAL)>(), PARAM);
READ(bins_count, params[0]);
#undef READ
AggregateFunctionPtr res(createWithNumericType<AggregateFunctionHistogram>(*arguments[0], bins_count));
if (!res)
throw Exception("Illegal type " + arguments[0]->getName() + " of argument for aggregate function " + name, ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
return res;
}
}
void registerAggregateFunctionHistogram(AggregateFunctionFactory & factory)
{
factory.registerFunction("histogram", createAggregateFunctionHistogram);

View File

@ -2,7 +2,8 @@
#include <Common/Arena.h>
#include <Common/PODArray.h>
#include <Common/RadixSort.h>
#include <Common/AutoArray.h>
#include <Columns/ColumnVector.h>
#include <Columns/ColumnTuple.h>
#include <Columns/ColumnArray.h>
@ -17,6 +18,8 @@
#include <AggregateFunctions/IAggregateFunction.h>
#include <queue>
namespace DB {
namespace ErrorCodes
@ -31,8 +34,8 @@ namespace ErrorCodes
class AggregateFunctionHistogramData
{
public:
using Mean = Float32;
using Weight = Float32;
using Mean = Float64;
using Weight = Float64;
private:
struct WeightedValue {
@ -41,83 +44,113 @@ private:
WeightedValue operator + (const WeightedValue& other)
{
return {(other.mean * other.weight + mean * weight) / (other.weight + weight), other.weight + weight};
return {mean + other.weight * (other.mean - mean) / (other.weight + weight), other.weight + weight};
}
};
struct RadixSortTraits
{
using Element = WeightedValue;
using Key = Mean;
using CountType = UInt32;
using KeyBits = UInt32;
static constexpr size_t PART_SIZE_BITS = 8;
using Transform = RadixSortFloatTransform<KeyBits>;
using Allocator = RadixSortMallocAllocator;
static Key & extractKey(Element & elem) { return elem.mean; }
};
private:
UInt32 max_bins;
// quantity of stored weighted-values
UInt32 size;
// calculated lower and upper bounds of seen points
Mean lower_bound;
Mean upper_bound;
static constexpr Float32 epsilon = 1e-8;
static constexpr Mean epsilon = 1e-8;
// Weighted values representation of histogram.
// We allow up to max_bins * 2 values stored in intermediate states
// Is allocated in arena, so there is no explicit management.
WeightedValue* points;
private:
void sort()
{
RadixSort<RadixSortTraits>::execute(points, size);
std::sort(points, points + size,
[](const WeightedValue& first, const WeightedValue& second)
{
return first.mean < second.mean;
});
}
/**
* Repeatedly fuse until max_bins bins left
* Repeatedly fuse most close values until max_bins bins left
*/
void compress()
{
sort();
while (size > max_bins)
{
size_t min_index = 0;
auto quality = [&](size_t i)
{
return points[i + 1].mean - points[i].mean;
};
for (size_t i = 0; i + 1 != size; ++i)
{
if (quality(min_index) > quality(i))
{
min_index = i;
}
}
auto newsz = size;
if (size <= max_bins) return;
points[min_index] = points[min_index] + points[min_index + 1];
for (size_t i = min_index + 1; i + 1 < size; ++i)
{
points[i] = points[i + 1];
}
size--;
// Maintain doubly-linked list of "active" points
// and store neighbour pairs in priority queue by distance
AutoArray<UInt32> previous(size + 1);
AutoArray<UInt32> next(size + 1);
AutoArray<bool> active(size + 1, true);
active[size] = false;
auto delete_node = [&](UInt32 i)
{
previous[next[i]] = previous[i];
next[previous[i]] = next[i];
active[i] = false;
};
for (size_t i = 0; i <= size; i++)
{
previous[i] = i - 1;
next[i] = i + 1;
}
next[size] = 0;
previous[0] = size;
using QueueItem = std::pair<Mean, UInt32>;
std::vector<QueueItem> storage;
storage.reserve(2 * size - max_bins);
std::priority_queue<QueueItem, std::vector<QueueItem>, std::greater<QueueItem>> queue;
auto quality = [&](UInt32 i) { return points[next[i]].mean - points[i].mean; };
for (size_t i = 0; i + 1 < size; i++)
queue.push({quality(i), i});
while (newsz > max_bins && !queue.empty())
{
auto min_item = queue.top();
queue.pop();
auto l = min_item.second;
auto r = next[l];
if (!active[l] || !active[r] || quality(l) > min_item.first)
continue;
points[l] = points[l] + points[r];
delete_node(r);
if (active[next[l]])
queue.push({quality(l), l});
if (active[previous[l]])
queue.push({quality(previous[l]), previous[l]});
newsz--;
}
size_t l = 0;
for (size_t r = 0; r < size; r++)
if (active[r])
points[l++] = points[r];
size = newsz;
}
/***
* Delete too close points from histogram.
* Assume that points are sorted.
* Assumes that points are sorted.
*/
void unique()
{
size_t l = 0;
for (auto r = l + 1; r != size; r++)
for (auto r = l + 1; r < size; r++)
{
if (abs(points[l].mean - points[r].mean) < epsilon)
if (points[l].mean + epsilon >= points[r].mean)
{
points[l] = points[l] + points[r];
}
@ -139,11 +172,15 @@ public:
AggregateFunctionHistogramData(UInt32 max_bins)
: max_bins(max_bins)
, size(0)
, lower_bound(std::numeric_limits<Mean>::max())
, upper_bound(std::numeric_limits<Mean>::lowest())
, points(nullptr)
{
}
void insertResultInto(ColumnVector<Mean>& to_lower, ColumnVector<Mean>& to_upper, ColumnVector<Weight>& to_weights) {
if (!points) return;
compress();
unique();
@ -167,7 +204,7 @@ public:
lower_bound = std::min(lower_bound, value);
upper_bound = std::max(upper_bound, value);
if (size > max_bins * 2)
if (size >= max_bins * 2)
{
compress();
}
@ -199,7 +236,7 @@ public:
readVarUInt(size, buf);
if (size > max_bins)
if (size > max_bins * 2)
throw Exception("Too many bins", ErrorCodes::TOO_LARGE_ARRAY_SIZE);
if (!points)