This repository has been archived on 2020-05-25. You can view files and clone it, but cannot push or open issues or pull requests.
tdlib-fork/tdutils/test/gzip.cpp
levlam 2f850bc834 Fix warnings.
GitOrigin-RevId: 25c5aaaf29fd1284427f354291a6e664847413ea
2020-03-20 17:32:58 +03:00

122 lines
3.9 KiB
C++

//
// Copyright Aliaksei Levin (levlam@telegram.org), Arseny Smirnov (arseny30@gmail.com) 2014-2020
//
// Distributed under the Boost Software License, Version 1.0. (See accompanying
// file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
//
#include "td/utils/buffer.h"
#include "td/utils/ByteFlow.h"
#include "td/utils/common.h"
#include "td/utils/Gzip.h"
#include "td/utils/GzipByteFlow.h"
#include "td/utils/logging.h"
#include "td/utils/Status.h"
#include "td/utils/tests.h"
#include "td/utils/Time.h"
static void encode_decode(td::string s) {
auto r = td::gzencode(s, 2);
ASSERT_TRUE(!r.empty());
ASSERT_EQ(s, td::gzdecode(r.as_slice()));
}
TEST(Gzip, gzencode_gzdecode) {
encode_decode(td::rand_string(0, 255, 1000));
encode_decode(td::rand_string('a', 'z', 1000000));
encode_decode(td::string(1000000, 'a'));
}
static void test_gzencode(td::string s) {
auto begin_time = td::Time::now();
auto r = td::gzencode(s, td::max(2, static_cast<int>(100 / s.size())));
ASSERT_TRUE(!r.empty());
LOG(INFO) << "Encoded string of size " << s.size() << " in " << (td::Time::now() - begin_time)
<< " with compression ratio " << static_cast<double>(r.size()) / static_cast<double>(s.size());
}
TEST(Gzip, gzencode) {
for (size_t len = 1; len <= 10000000; len *= 10) {
test_gzencode(td::rand_string('a', 'a', len));
test_gzencode(td::rand_string('a', 'z', len));
test_gzencode(td::rand_string(0, 255, len));
}
}
TEST(Gzip, flow) {
auto str = td::rand_string('a', 'z', 1000000);
auto parts = td::rand_split(str);
td::ChainBufferWriter input_writer;
auto input = input_writer.extract_reader();
td::ByteFlowSource source(&input);
td::GzipByteFlow gzip_flow(td::Gzip::Mode::Encode);
gzip_flow = td::GzipByteFlow(td::Gzip::Mode::Encode);
td::ByteFlowSink sink;
source >> gzip_flow >> sink;
ASSERT_TRUE(!sink.is_ready());
for (auto &part : parts) {
input_writer.append(part);
source.wakeup();
}
ASSERT_TRUE(!sink.is_ready());
source.close_input(td::Status::OK());
ASSERT_TRUE(sink.is_ready());
ASSERT_TRUE(sink.status().is_ok());
auto res = sink.result()->move_as_buffer_slice().as_slice().str();
ASSERT_TRUE(!res.empty());
ASSERT_EQ(td::gzencode(str, 2).as_slice().str(), res);
}
TEST(Gzip, flow_error) {
auto str = td::rand_string('a', 'z', 1000000);
auto zip = td::gzencode(str, 0.9).as_slice().str();
ASSERT_TRUE(!zip.empty());
zip.resize(zip.size() - 1);
auto parts = td::rand_split(zip);
auto input_writer = td::ChainBufferWriter();
auto input = input_writer.extract_reader();
td::ByteFlowSource source(&input);
td::GzipByteFlow gzip_flow(td::Gzip::Mode::Decode);
td::ByteFlowSink sink;
source >> gzip_flow >> sink;
ASSERT_TRUE(!sink.is_ready());
for (auto &part : parts) {
input_writer.append(part);
source.wakeup();
}
ASSERT_TRUE(!sink.is_ready());
source.close_input(td::Status::OK());
ASSERT_TRUE(sink.is_ready());
ASSERT_TRUE(!sink.status().is_ok());
}
TEST(Gzip, encode_decode_flow) {
auto str = td::rand_string('a', 'z', 1000000);
auto parts = td::rand_split(str);
td::ChainBufferWriter input_writer;
auto input = input_writer.extract_reader();
td::ByteFlowSource source(&input);
td::GzipByteFlow gzip_encode_flow(td::Gzip::Mode::Encode);
td::GzipByteFlow gzip_decode_flow(td::Gzip::Mode::Decode);
td::GzipByteFlow gzip_encode_flow2(td::Gzip::Mode::Encode);
td::GzipByteFlow gzip_decode_flow2(td::Gzip::Mode::Decode);
td::ByteFlowSink sink;
source >> gzip_encode_flow >> gzip_decode_flow >> gzip_encode_flow2 >> gzip_decode_flow2 >> sink;
ASSERT_TRUE(!sink.is_ready());
for (auto &part : parts) {
input_writer.append(part);
source.wakeup();
}
ASSERT_TRUE(!sink.is_ready());
source.close_input(td::Status::OK());
ASSERT_TRUE(sink.is_ready());
LOG_IF(ERROR, sink.status().is_error()) << sink.status();
ASSERT_TRUE(sink.status().is_ok());
ASSERT_EQ(str, sink.result()->move_as_buffer_slice().as_slice().str());
}