blob: eae4bef1a3c080cdf309d09d96277c8bad19cba3 [file] [log] [blame]
/*
* Copyright (C) 2022 The Android Open Source Project
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef SRC_TRACE_PROCESSOR_DB_COLUMN_STORAGE_H_
#define SRC_TRACE_PROCESSOR_DB_COLUMN_STORAGE_H_
#include <cstddef>
#include <cstdint>
#include <optional>
#include <vector>
#include "perfetto/base/compiler.h"
#include "perfetto/base/logging.h"
#include "perfetto/public/compiler.h"
#include "src/trace_processor/containers/bit_vector.h"
namespace perfetto::trace_processor {
// Base class for allowing type erasure when defining plug-in implementations
// of backing storage for columns.
class ColumnStorageBase {
public:
ColumnStorageBase() = default;
virtual ~ColumnStorageBase();
ColumnStorageBase(const ColumnStorageBase&) = delete;
ColumnStorageBase& operator=(const ColumnStorageBase&) = delete;
ColumnStorageBase(ColumnStorageBase&&) = default;
ColumnStorageBase& operator=(ColumnStorageBase&&) noexcept = default;
virtual const void* data() const = 0;
virtual const BitVector* bv() const = 0;
virtual uint32_t size() const = 0;
virtual uint32_t non_null_size() const = 0;
};
// Class used for implementing storage for non-null columns.
template <typename T>
class ColumnStorage final : public ColumnStorageBase {
public:
ColumnStorage() = default;
explicit ColumnStorage(const ColumnStorage&) = delete;
ColumnStorage& operator=(const ColumnStorage&) = delete;
ColumnStorage(ColumnStorage&&) = default;
ColumnStorage& operator=(ColumnStorage&&) noexcept = default;
T Get(uint32_t idx) const { return vector_[idx]; }
void Append(T val) { vector_.emplace_back(val); }
void Append(const std::vector<T>& vals) {
vector_.insert(vector_.end(), vals.begin(), vals.end());
}
void AppendMultiple(T val, uint32_t count) {
vector_.insert(vector_.end(), count, val);
}
void Set(uint32_t idx, T val) { vector_[idx] = val; }
PERFETTO_NO_INLINE void ShrinkToFit() { vector_.shrink_to_fit(); }
const std::vector<T>& vector() const { return vector_; }
const void* data() const final { return vector_.data(); }
const BitVector* bv() const final { return nullptr; }
uint32_t size() const final { return static_cast<uint32_t>(vector_.size()); }
uint32_t non_null_size() const final { return size(); }
template <bool IsDense>
static ColumnStorage<T> Create() {
static_assert(!IsDense, "Invalid for non-null storage to be dense.");
return ColumnStorage<T>();
}
// Create non-null storage from nullable storage without nulls.
static ColumnStorage<T> CreateFromAssertNonNull(
ColumnStorage<std::optional<T>> null_storage) {
PERFETTO_CHECK(null_storage.size() == null_storage.non_null_size());
ColumnStorage<T> x;
x.vector_ = std::move(null_storage).non_null_vector();
return x;
}
private:
std::vector<T> vector_;
};
// Class used for implementing storage for nullable columns.
template <typename T>
class ColumnStorage<std::optional<T>> final : public ColumnStorageBase {
public:
ColumnStorage() = default;
explicit ColumnStorage(const ColumnStorage&) = delete;
ColumnStorage& operator=(const ColumnStorage&) = delete;
ColumnStorage(ColumnStorage&&) = default;
ColumnStorage& operator=(ColumnStorage&&) noexcept = default;
std::optional<T> Get(uint32_t idx) const {
bool contains = valid_.IsSet(idx);
if (mode_ == Mode::kDense) {
return contains ? std::make_optional(data_[idx]) : std::nullopt;
}
return contains ? std::make_optional(data_[valid_.CountSetBits(idx)])
: std::nullopt;
}
void Append(T val) {
data_.emplace_back(val);
valid_.AppendTrue();
}
void Append(std::optional<T> val) {
if (val) {
Append(*val);
} else {
AppendNull();
}
}
void AppendMultipleNulls(uint32_t count) {
if (mode_ == Mode::kDense) {
data_.resize(data_.size() + static_cast<uint32_t>(count));
}
valid_.Resize(valid_.size() + static_cast<uint32_t>(count), false);
}
void AppendMultiple(T val, uint32_t count) {
data_.insert(data_.end(), count, val);
valid_.Resize(valid_.size() + static_cast<uint32_t>(count), true);
}
void Append(const std::vector<T>& vals) {
data_.insert(data_.end(), vals.begin(), vals.end());
valid_.Resize(valid_.size() + static_cast<uint32_t>(vals.size()), true);
}
void Set(uint32_t idx, T val) {
if (mode_ == Mode::kDense) {
valid_.Set(idx);
data_[idx] = val;
} else {
// Generally, we will be setting a null row to non-null so optimize for
// that path.
uint32_t row = valid_.CountSetBits(idx);
bool was_set = valid_.Set(idx);
if (PERFETTO_UNLIKELY(was_set)) {
data_[row] = val;
} else {
data_.insert(data_.begin() + static_cast<ptrdiff_t>(row), val);
}
}
}
bool IsDense() const { return mode_ == Mode::kDense; }
PERFETTO_NO_INLINE void ShrinkToFit() {
data_.shrink_to_fit();
valid_.ShrinkToFit();
}
// For dense columns the size of the vector is equal to size of the bit
// vector. For sparse it's equal to count set bits of the bit vector.
const std::vector<T>& non_null_vector() const& { return data_; }
const BitVector& non_null_bit_vector() const { return valid_; }
const void* data() const final { return non_null_vector().data(); }
const BitVector* bv() const final { return &non_null_bit_vector(); }
uint32_t size() const final { return valid_.size(); }
uint32_t non_null_size() const final {
return static_cast<uint32_t>(non_null_vector().size());
}
template <bool IsDense>
static ColumnStorage<std::optional<T>> Create() {
return IsDense ? ColumnStorage<std::optional<T>>(Mode::kDense)
: ColumnStorage<std::optional<T>>(Mode::kSparse);
}
std::vector<T> non_null_vector() && { return std::move(data_); }
private:
enum class Mode {
// Sparse mode is the default mode and ensures that nulls are stored using
// only
// a single bit (at the cost of making setting null entries to non-null
// O(n)).
kSparse,
// Dense mode forces the reservation of space for null entries which
// increases
// memory usage but allows for O(1) set operations.
kDense,
};
explicit ColumnStorage(Mode mode) : mode_(mode) {}
void AppendNull() {
if (mode_ == Mode::kDense) {
data_.emplace_back();
}
valid_.AppendFalse();
}
Mode mode_ = Mode::kSparse;
std::vector<T> data_;
BitVector valid_;
};
} // namespace perfetto::trace_processor
#endif // SRC_TRACE_PROCESSOR_DB_COLUMN_STORAGE_H_