diff options
author | Devtools Arcadia <arcadia-devtools@yandex-team.ru> | 2022-02-07 18:08:42 +0300 |
---|---|---|
committer | Devtools Arcadia <arcadia-devtools@mous.vla.yp-c.yandex.net> | 2022-02-07 18:08:42 +0300 |
commit | 1110808a9d39d4b808aef724c861a2e1a38d2a69 (patch) | |
tree | e26c9fed0de5d9873cce7e00bc214573dc2195b7 /library/cpp/codecs/static | |
download | ydb-1110808a9d39d4b808aef724c861a2e1a38d2a69.tar.gz |
intermediate changes
ref:cde9a383711a11544ce7e107a78147fb96cc4029
Diffstat (limited to 'library/cpp/codecs/static')
29 files changed, 863 insertions, 0 deletions
diff --git a/library/cpp/codecs/static/README b/library/cpp/codecs/static/README new file mode 100644 index 0000000000..1b07f02433 --- /dev/null +++ b/library/cpp/codecs/static/README @@ -0,0 +1 @@ +Support of static libraries in library/cpp/codecs. See library/cpp/codecs/static/example. diff --git a/library/cpp/codecs/static/builder.cpp b/library/cpp/codecs/static/builder.cpp new file mode 100644 index 0000000000..93e34a3edb --- /dev/null +++ b/library/cpp/codecs/static/builder.cpp @@ -0,0 +1,39 @@ +#include "builder.h" +#include "common.h" + +#include <library/cpp/codecs/static/static_codec_info.pb.h> + +#include <library/cpp/codecs/codecs.h> + +#include <util/generic/yexception.h> +#include <util/string/subst.h> + +namespace NCodecs { + TStaticCodecInfo BuildStaticCodec(const TVector<TString>& trainingData, const TCodecBuildInfo& info) { + TStaticCodecInfo result; + TCodecPtr codec = ICodec::GetInstance(info.CodecName); + Y_ENSURE_EX(codec, TCodecException() << "empty codec is not allowed"); + + codec->LearnX(trainingData.begin(), trainingData.end(), info.SampleSizeMultiplier); + { + TStringOutput sout{*result.MutableStoredCodec()}; + ICodec::Store(&sout, codec); + } + + auto& debugInfo = *result.MutableDebugInfo(); + debugInfo.SetStoredCodecHash(DataSignature(result.GetStoredCodec())); + debugInfo.SetCodecName(info.CodecName); + debugInfo.SetSampleSizeMultiplier(info.SampleSizeMultiplier); + debugInfo.SetTimestamp(info.Timestamp); + debugInfo.SetRevisionInfo(info.RevisionInfo); + debugInfo.SetTrainingSetComment(info.TrainingSetComment); + debugInfo.SetTrainingSetResId(info.TrainingSetResId); + return result; + } + + TString GetStandardFileName(const TStaticCodecInfo& info) { + TString cName = info.GetDebugInfo().GetCodecName(); + SubstGlobal(cName, ':', '.'); + return TStringBuilder() << cName << "." << info.GetDebugInfo().GetTimestamp() << ".codec_info"; + } +} diff --git a/library/cpp/codecs/static/builder.h b/library/cpp/codecs/static/builder.h new file mode 100644 index 0000000000..d7533be4d5 --- /dev/null +++ b/library/cpp/codecs/static/builder.h @@ -0,0 +1,29 @@ +#pragma once + +#include "static.h" + +#include <library/cpp/svnversion/svnversion.h> + +#include <util/datetime/base.h> +#include <util/generic/string.h> +#include <util/generic/vector.h> +#include <util/string/builder.h> + +namespace NCodecs { + struct TCodecBuildInfo { + // optimal values from SEARCH-1655 + TString CodecName = "solar-8k-a:zstd08d-1"; + float SampleSizeMultiplier = 1; + + // debug info: + time_t Timestamp = TInstant::Now().TimeT(); + TString RevisionInfo = (TStringBuilder() << "r" << ToString(GetProgramSvnRevision())); + TString TrainingSetComment; // a human comment on the training data + TString TrainingSetResId; // sandbox resid of the training set + }; + + TStaticCodecInfo BuildStaticCodec(const TVector<TString>& trainingData, const TCodecBuildInfo&); + + TString GetStandardFileName(const TStaticCodecInfo&); + +} diff --git a/library/cpp/codecs/static/common.h b/library/cpp/codecs/static/common.h new file mode 100644 index 0000000000..211de2a27d --- /dev/null +++ b/library/cpp/codecs/static/common.h @@ -0,0 +1,32 @@ +#pragma once + +#include <util/string/hex.h> +#include <util/digest/city.h> +#include <util/system/byteorder.h> + +namespace NCodecs { + template <class T> + ui64 DataSignature(const T& t) { + static_assert(!std::is_scalar<T>::value, "no scalars"); + return CityHash64(t.data(), t.size()); + } + + template <class T> + TString HexWriteScalar(T t) { + static_assert(std::is_scalar<T>::value, "scalars only"); + t = LittleToBig(t); + TString res = HexEncode(&t, sizeof(t)); + res.to_lower(); + return res; + } + + template <class T> + T HexReadScalar(TStringBuf s) { + static_assert(std::is_scalar<T>::value, "scalars only"); + T t = 0; + HexDecode(s.data(), Min(s.size(), sizeof(T)), &t); + t = BigToLittle(t); + return t; + } + +} diff --git a/library/cpp/codecs/static/example/example.cpp b/library/cpp/codecs/static/example/example.cpp new file mode 100644 index 0000000000..5b750b717e --- /dev/null +++ b/library/cpp/codecs/static/example/example.cpp @@ -0,0 +1,43 @@ +#include "example.h" + +#include <library/cpp/codecs/static/static.h> + +#include <util/generic/yexception.h> + +extern "C" { +extern const ui8 codec_info_huff_20160707[]; +extern const ui32 codec_info_huff_20160707Size; +extern const ui8 codec_info_sa_huff_20160707[]; +extern const ui32 codec_info_sa_huff_20160707Size; +}; + +namespace NStaticCodecExample { + static const NCodecs::TCodecConstPtr CODECS[] = { + nullptr, + NCodecs::RestoreCodecFromArchive(codec_info_huff_20160707, codec_info_huff_20160707Size), + NCodecs::RestoreCodecFromArchive(codec_info_sa_huff_20160707, codec_info_sa_huff_20160707Size), + }; + + static_assert(Y_ARRAY_SIZE(CODECS) == DV_COUNT, "bad array size"); + + void Encode(TBuffer& out, TStringBuf in, EDictVersion dv) { + Y_ENSURE(dv > DV_NULL && dv < DV_COUNT, "invalid dict version: " << (int)dv); + out.Clear(); + if (!in) { + return; + } + CODECS[dv]->Encode(in, out); + out.Append((char)dv); + } + + void Decode(TBuffer& out, TStringBuf in) { + out.Clear(); + if (!in) { + return; + } + EDictVersion dv = (EDictVersion)in.back(); + Y_ENSURE(dv > DV_NULL && dv < DV_COUNT, "invalid dict version: " << (int)dv); + in.Chop(1); + CODECS[dv]->Decode(in, out); + } +} diff --git a/library/cpp/codecs/static/example/example.h b/library/cpp/codecs/static/example/example.h new file mode 100644 index 0000000000..f9b3a7324b --- /dev/null +++ b/library/cpp/codecs/static/example/example.h @@ -0,0 +1,17 @@ +#pragma once + +#include <util/generic/strbuf.h> +#include <util/generic/buffer.h> + +namespace NStaticCodecExample { + enum EDictVersion : ui8 { + DV_NULL = 0, + DV_HUFF_20160707, + DV_SA_HUFF_20160707, + DV_COUNT + }; + + void Encode(TBuffer&, TStringBuf, EDictVersion dv = DV_SA_HUFF_20160707); + + void Decode(TBuffer&, TStringBuf); +} diff --git a/library/cpp/codecs/static/example/huffman.1467494385.codec_info b/library/cpp/codecs/static/example/huffman.1467494385.codec_info Binary files differnew file mode 100644 index 0000000000..5fc18270a6 --- /dev/null +++ b/library/cpp/codecs/static/example/huffman.1467494385.codec_info diff --git a/library/cpp/codecs/static/example/solar-8k-a.huffman.1467494385.codec_info b/library/cpp/codecs/static/example/solar-8k-a.huffman.1467494385.codec_info Binary files differnew file mode 100644 index 0000000000..d36d8e24ec --- /dev/null +++ b/library/cpp/codecs/static/example/solar-8k-a.huffman.1467494385.codec_info diff --git a/library/cpp/codecs/static/example/ya.make b/library/cpp/codecs/static/example/ya.make new file mode 100644 index 0000000000..ca6c5fd900 --- /dev/null +++ b/library/cpp/codecs/static/example/ya.make @@ -0,0 +1,24 @@ +LIBRARY() + +OWNER(velavokr) + +SRCS( + GLOBAL example.cpp +) + +PEERDIR( + library/cpp/codecs + library/cpp/codecs/static +) + +ARCHIVE_ASM( + "solar-8k-a.huffman.1467494385.codec_info" + NAME codec_info_sa_huff_20160707 +) + +ARCHIVE_ASM( + "huffman.1467494385.codec_info" + NAME codec_info_huff_20160707 +) + +END() diff --git a/library/cpp/codecs/static/static.cpp b/library/cpp/codecs/static/static.cpp new file mode 100644 index 0000000000..44a07dd73a --- /dev/null +++ b/library/cpp/codecs/static/static.cpp @@ -0,0 +1,98 @@ +#include "static.h" +#include "common.h" + +#include <library/cpp/codecs/static/static_codec_info.pb.h> +#include <library/cpp/archive/yarchive.h> + +#include <util/draft/datetime.h> + +#include <util/string/builder.h> +#include <util/stream/buffer.h> +#include <util/stream/mem.h> +#include <util/string/hex.h> +#include <util/ysaveload.h> + +namespace NCodecs { + static constexpr TStringBuf STATIC_CODEC_INFO_MAGIC = "CodecInf"; + + static TStringBuf GetStaticCodecInfoMagic() { + return STATIC_CODEC_INFO_MAGIC; + } + + void SaveCodecInfoToStream(IOutputStream& out, const TStaticCodecInfo& info) { + TBufferOutput bout; + info.SerializeToArcadiaStream(&bout); + ui64 hash = DataSignature(bout.Buffer()); + out.Write(GetStaticCodecInfoMagic()); + ::Save(&out, hash); + ::Save(&out, bout.Buffer()); + } + + TStaticCodecInfo LoadCodecInfoFromStream(IInputStream& in) { + { + TBuffer magic; + magic.Resize(GetStaticCodecInfoMagic().size()); + Y_ENSURE_EX(in.Read(magic.Data(), GetStaticCodecInfoMagic().size()) == GetStaticCodecInfoMagic().size(), + TCodecException() << "bad codec info"); + Y_ENSURE_EX(TStringBuf(magic.data(), magic.size()) == GetStaticCodecInfoMagic(), + TCodecException() << "bad codec info"); + } + + ui64 hash; + ::Load(&in, hash); + TBuffer info; + ::Load(&in, info); + Y_ENSURE_EX(hash == DataSignature(info), TCodecException() << "bad codec info"); + + TStaticCodecInfo result; + Y_ENSURE_EX(result.ParseFromArray(info.data(), info.size()), TCodecException() << "bad codec info"); + + return result; + } + + TString SaveCodecInfoToString(const TStaticCodecInfo& info) { + TStringStream s; + SaveCodecInfoToStream(s, info); + return s.Str(); + } + + TStaticCodecInfo LoadCodecInfoFromString(TStringBuf data) { + TMemoryInput m{data.data(), data.size()}; + return LoadCodecInfoFromStream(m); + } + + TString FormatCodecInfo(const TStaticCodecInfo& ci) { + TStringBuilder s; + s << "codec name: " << ci.GetDebugInfo().GetCodecName() << Endl; + s << "codec hash: " << HexWriteScalar(ci.GetDebugInfo().GetStoredCodecHash()) << Endl; + s << "dict size: " << ci.GetStoredCodec().Size() << Endl; + s << "sample mult: " << ci.GetDebugInfo().GetSampleSizeMultiplier() << Endl; + s << "orig.compress: " << ci.GetDebugInfo().GetCompression() * 100 << " %" << Endl; + s << "timestamp: " << ci.GetDebugInfo().GetTimestamp() << " (" + << NDatetime::TSimpleTM::NewLocal(ci.GetDebugInfo().GetTimestamp()).ToString() + << ")" << Endl; + s << "revision: " << ci.GetDebugInfo().GetRevisionInfo() << Endl; + s << "training set comment: " << ci.GetDebugInfo().GetTrainingSetComment() << Endl; + s << "training set resId: " << ci.GetDebugInfo().GetTrainingSetResId() << Endl; + return s; + } + + TString LoadStringFromArchive(const ui8* begin, size_t size) { + TArchiveReader ar(TBlob::NoCopy(begin, size)); + Y_VERIFY(ar.Count() == 1, "invalid number of entries"); + auto blob = ar.ObjectBlobByKey(ar.KeyByIndex(0)); + return TString{blob.AsCharPtr(), blob.Size()}; + } + + TCodecConstPtr RestoreCodecFromCodecInfo(const TStaticCodecInfo& info) { + return NCodecs::ICodec::RestoreFromString(info.GetStoredCodec()); + } + + TCodecConstPtr RestoreCodecFromArchive(const ui8* begin, size_t size) { + const auto& data = LoadStringFromArchive(begin, size); + const auto& info = LoadCodecInfoFromString(data); + const auto& codec = RestoreCodecFromCodecInfo(info); + Y_ENSURE_EX(codec, TCodecException() << "null codec"); + return codec; + } +} diff --git a/library/cpp/codecs/static/static.h b/library/cpp/codecs/static/static.h new file mode 100644 index 0000000000..c1eaed2a74 --- /dev/null +++ b/library/cpp/codecs/static/static.h @@ -0,0 +1,34 @@ +#pragma once + +#include <library/cpp/codecs/codecs.h> + +#include <util/generic/strbuf.h> +#include <util/generic/string.h> +#include <util/stream/output.h> + +namespace NCodecs { + class TStaticCodecInfo; + + // load + + TCodecConstPtr RestoreCodecFromCodecInfo(const TStaticCodecInfo&); + + TStaticCodecInfo LoadCodecInfoFromString(TStringBuf data); + + TString LoadStringFromArchive(const ui8* begin, size_t size); + + TCodecConstPtr RestoreCodecFromArchive(const ui8* begin, size_t size); + + // save + + TString SaveCodecInfoToString(const TStaticCodecInfo&); + + void SaveCodecInfoToStream(IOutputStream& out, const TStaticCodecInfo&); + + // misc + + TStaticCodecInfo LoadCodecInfoFromStream(IInputStream& in); + + TString FormatCodecInfo(const TStaticCodecInfo&); + +} diff --git a/library/cpp/codecs/static/static_codec_info.proto b/library/cpp/codecs/static/static_codec_info.proto new file mode 100644 index 0000000000..362abb4dad --- /dev/null +++ b/library/cpp/codecs/static/static_codec_info.proto @@ -0,0 +1,17 @@ +package NCodecs; + +message TStaticCodecInfo { + message TDebugInfo { + optional string CodecName = 1; // the exact codec variant name + optional uint64 Timestamp = 2; // when the codec was built + optional string RevisionInfo = 3; // the arcadia revision info + optional float SampleSizeMultiplier = 4; // how the default sample size was modified to improve compression + optional float Compression = 5; // the compression on the training set ((raw_size - coded_size) / raw_size) + optional string TrainingSetComment = 6; // a human readable description of the training set + optional string TrainingSetResId = 7; // the training set sandbox resource id + optional uint64 StoredCodecHash = 8; // cityhash64(data) + } + + optional bytes StoredCodec = 1; // the data of the codec + optional TDebugInfo DebugInfo = 2; // misc debug info which could be useful in finding whereabouts later +} diff --git a/library/cpp/codecs/static/tools/common/ct_common.cpp b/library/cpp/codecs/static/tools/common/ct_common.cpp new file mode 100644 index 0000000000..fe77691280 --- /dev/null +++ b/library/cpp/codecs/static/tools/common/ct_common.cpp @@ -0,0 +1,74 @@ +#include "ct_common.h" + +#include <library/cpp/codecs/codecs.h> +#include <library/cpp/codecs/static/static_codec_info.pb.h> +#include <library/cpp/string_utils/base64/base64.h> + +#include <util/stream/output.h> +#include <util/string/builder.h> +#include <util/system/hp_timer.h> + +namespace NCodecs { + TString TComprStats::Format(const TStaticCodecInfo& info, bool checkMode) const { + TStringBuilder s; + s << "raw size/item: " << RawSizePerRecord() << Endl; + s << "enc.size/item: " << EncSizePerRecord() << Endl; + if (checkMode) { + s << "orig.enc.size/item: " << OldEncSizePerRecord(info.GetDebugInfo().GetCompression()) << Endl; + } + s << "enc time us/item: " << EncTimePerRecordUS() << Endl; + s << "dec time us/item: " << DecTimePerRecordUS() << Endl; + s << "dict size: " << info.GetStoredCodec().Size() << Endl; + s << "compression: " << AsPercent(Compression()) << " %" << Endl; + if (checkMode) { + s << "orig.compression: " << AsPercent(info.GetDebugInfo().GetCompression()) << " %" << Endl; + } + return s; + } + + TComprStats TestCodec(const ICodec& c, const TVector<TString>& input) { + TComprStats stats; + + TBuffer encodeBuffer; + TBuffer decodeBuffer; + for (const auto& data : input) { + encodeBuffer.Clear(); + decodeBuffer.Clear(); + + stats.Records += 1; + stats.RawSize += data.size(); + + THPTimer timer; + c.Encode(data, encodeBuffer); + stats.EncSize += encodeBuffer.size(); + stats.EncSeconds += timer.PassedReset(); + + c.Decode(TStringBuf{encodeBuffer.data(), encodeBuffer.size()}, decodeBuffer); + stats.DecSeconds += timer.PassedReset(); + Y_ENSURE(data == TStringBuf(decodeBuffer.data(), decodeBuffer.size()), "invalid encoding at record " << stats.Records); + } + + return stats; + } + + void ParseBlob(TVector<TString>& result, EDataStreamFormat fmt, const TBlob& blob) { + TStringBuf bin(blob.AsCharPtr(), blob.Size()); + TStringBuf line; + TString buffer; + while (bin.ReadLine(line)) { + if (DSF_BASE64_LF == fmt) { + Base64Decode(line, buffer); + line = buffer; + } + if (!line) { + continue; + } + result.emplace_back(line.data(), line.size()); + } + } + + TBlob GetInputBlob(const TString& dataFile) { + return dataFile && dataFile != "-" ? TBlob::FromFile(dataFile) : TBlob::FromStream(Cin); + } + +} diff --git a/library/cpp/codecs/static/tools/common/ct_common.h b/library/cpp/codecs/static/tools/common/ct_common.h new file mode 100644 index 0000000000..9d3dcbda93 --- /dev/null +++ b/library/cpp/codecs/static/tools/common/ct_common.h @@ -0,0 +1,75 @@ +#pragma once + +#include <util/generic/string.h> +#include <util/generic/vector.h> +#include <util/memory/blob.h> +#include <cmath> + +namespace NCodecs { + class TStaticCodecInfo; + class ICodec; + + struct TComprStats { + double EncSeconds = 0; + double DecSeconds = 0; + size_t Records = 0; + size_t RawSize = 0; + size_t EncSize = 0; + + static double Round(double n, size_t decPlaces = 2) { + double p = pow(10, decPlaces); + return round(n * p) / p; + } + + static double AsPercent(double n) { + return Round(n * 100); + } + + static double AsMicroSecond(double s) { + return s * 1000000; + } + + double PerRecord(double n) const { + return Round((double)(Records ? n / Records : 0)); + } + + double Compression() const { + return ((double)RawSize - (double)EncSize) / RawSize; + } + + double EncTimePerRecordUS() const { + return PerRecord(AsMicroSecond(EncSeconds)); + } + + double DecTimePerRecordUS() const { + return PerRecord(AsMicroSecond(DecSeconds)); + } + + double RawSizePerRecord() const { + return PerRecord(RawSize); + } + + double EncSizePerRecord() const { + return PerRecord(EncSize); + } + + double OldEncSizePerRecord(double compr) const { + return PerRecord((1 - compr) * RawSize); + } + + TString Format(const TStaticCodecInfo&, bool checkMode) const; + }; + + TComprStats TestCodec(const ICodec&, const TVector<TString>& data); + + enum EDataStreamFormat { + DSF_NONE, + DSF_PLAIN_LF /* "plain" */, + DSF_BASE64_LF /* "base64" */, + }; + + void ParseBlob(TVector<TString>&, EDataStreamFormat, const TBlob&); + + TBlob GetInputBlob(const TString& dataFile); + +} diff --git a/library/cpp/codecs/static/tools/common/ya.make b/library/cpp/codecs/static/tools/common/ya.make new file mode 100644 index 0000000000..d624222dad --- /dev/null +++ b/library/cpp/codecs/static/tools/common/ya.make @@ -0,0 +1,19 @@ +LIBRARY() + +OWNER(velavokr) + +SRCS( + ct_common.cpp +) + +PEERDIR( + library/cpp/codecs + library/cpp/codecs/static + library/cpp/getopt/small + library/cpp/string_utils/base64 + util/draft +) + +GENERATE_ENUM_SERIALIZATION(ct_common.h) + +END() diff --git a/library/cpp/codecs/static/tools/static_codec_checker/README b/library/cpp/codecs/static/tools/static_codec_checker/README new file mode 100644 index 0000000000..723a68300b --- /dev/null +++ b/library/cpp/codecs/static/tools/static_codec_checker/README @@ -0,0 +1,4 @@ +This is a viewer for generated codec and utility for verification of the compression quality on a new data. + +Usage: +static_codec_checker -t -c 029b29ff64a74927.codec_info -f plain samples.txt diff --git a/library/cpp/codecs/static/tools/static_codec_checker/static_codec_checker.cpp b/library/cpp/codecs/static/tools/static_codec_checker/static_codec_checker.cpp new file mode 100644 index 0000000000..9c8d568d82 --- /dev/null +++ b/library/cpp/codecs/static/tools/static_codec_checker/static_codec_checker.cpp @@ -0,0 +1,73 @@ +#include <library/cpp/codecs/static/tools/common/ct_common.h> +#include <library/cpp/codecs/static/static.h> +#include <library/cpp/codecs/static/static_codec_info.pb.h> +#include <library/cpp/codecs/codecs.h> +#include <library/cpp/getopt/small/last_getopt.h> + +#include <util/digest/city.h> +#include <util/generic/yexception.h> +#include <util/stream/file.h> +#include <util/stream/buffer.h> +#include <util/stream/format.h> +#include <util/string/builder.h> + +int main(int argc, char** argv) { + NCodecs::TCodecPtr codecPtr; + NCodecs::EDataStreamFormat fmt = NCodecs::DSF_NONE; + TString codecFile; + bool testCompression = false; + + auto opts = NLastGetopt::TOpts::Default(); + opts.SetTitle("Prints a .codec_info file and optionally checks its performance on new data. See also static_codec_generator."); + opts.SetCmdLineDescr("-c 9089f3e9b7a0f0d4.codec_info -t -f base64 qtrees.sample.txt"); + NCodecs::TStaticCodecInfo codec; + + opts.AddLongOption('c', "codec-info").RequiredArgument("codec_info").Handler1T<TString>([&codecFile, &codec, &codecPtr](TString name) { + codecFile = name; + codec.CopyFrom(NCodecs::LoadCodecInfoFromString(TUnbufferedFileInput(name).ReadAll())); + codecPtr = NCodecs::ICodec::RestoreFromString(codec.GetStoredCodec()); + }) + .Required() + .Help(".codec_info file with serialized static data for codec"); + + opts.AddLongOption('t', "test").NoArgument().StoreValue(&testCompression, true).Optional().Help("test current performance"); + + opts.AddLongOption('f', "format").RequiredArgument(TStringBuilder() << "(" << NCodecs::DSF_PLAIN_LF << "|" << NCodecs::DSF_BASE64_LF << ")").StoreResult(&fmt).Optional().Help("test set input file format"); + + opts.SetFreeArgsMin(0); + opts.SetFreeArgTitle(0, "testing_set_input_file", "testing set input files"); + + NLastGetopt::TOptsParseResult res(&opts, argc, argv); + + Cout << codecFile << Endl; + Cout << NCodecs::FormatCodecInfo(codec) << Endl; + + if (testCompression) { + if (NCodecs::DSF_NONE == fmt) { + Cerr << "Specify format (-f|--format) for testing set input" << Endl; + exit(1); + } + + Cout << "Reading testing set data ... " << Flush; + + TVector<TString> allData; + for (const auto& freeArg : res.GetFreeArgs()) { + NCodecs::ParseBlob(allData, fmt, NCodecs::GetInputBlob(freeArg)); + } + + if (!res.GetFreeArgs()) { + NCodecs::ParseBlob(allData, fmt, NCodecs::GetInputBlob("-")); + } + + Cout << "Done" << Endl << Endl; + + Cout << "records: " << allData.size() << Endl; + Cout << "raw size: " << NCodecs::GetInputSize(allData.begin(), allData.end()) << " bytes" << Endl << Endl; + + Cout << "Testing compression ... " << Flush; + auto stats = NCodecs::TestCodec(*codecPtr, allData); + Cout << "Done" << Endl << Endl; + + Cout << stats.Format(codec, true) << Endl; + } +} diff --git a/library/cpp/codecs/static/tools/static_codec_checker/ya.make b/library/cpp/codecs/static/tools/static_codec_checker/ya.make new file mode 100644 index 0000000000..90e06ca448 --- /dev/null +++ b/library/cpp/codecs/static/tools/static_codec_checker/ya.make @@ -0,0 +1,16 @@ +PROGRAM() + +OWNER(velavokr) + +SRCS( + static_codec_checker.cpp +) + +PEERDIR( + library/cpp/codecs + library/cpp/codecs/static + library/cpp/codecs/static/tools/common + library/cpp/getopt/small +) + +END() diff --git a/library/cpp/codecs/static/tools/static_codec_generator/README b/library/cpp/codecs/static/tools/static_codec_generator/README new file mode 100644 index 0000000000..e6bb52b959 --- /dev/null +++ b/library/cpp/codecs/static/tools/static_codec_generator/README @@ -0,0 +1,4 @@ +This is a utility for reproducible teaching of a codec. And also for saving it into a file with a unique name for a static compilation as a resource. + +Usage: +static_codec_generator -t -m 'the training data description' -f plain samples.txt diff --git a/library/cpp/codecs/static/tools/static_codec_generator/static_codec_generator.cpp b/library/cpp/codecs/static/tools/static_codec_generator/static_codec_generator.cpp new file mode 100644 index 0000000000..45fdb5c5fe --- /dev/null +++ b/library/cpp/codecs/static/tools/static_codec_generator/static_codec_generator.cpp @@ -0,0 +1,82 @@ +#include <library/cpp/codecs/static/tools/common/ct_common.h> +#include <library/cpp/codecs/static/static_codec_info.pb.h> +#include <library/cpp/codecs/static/builder.h> +#include <library/cpp/codecs/codecs.h> + +#include <library/cpp/getopt/small/last_getopt.h> + +#include <util/generic/yexception.h> +#include <util/stream/file.h> +#include <util/string/builder.h> + +int main(int argc, char** argv) { + NCodecs::TCodecBuildInfo info; + NCodecs::EDataStreamFormat fmt = NCodecs::DSF_NONE; + + auto opts = NLastGetopt::TOpts::Default(); + opts.SetCmdLineDescr("-m 'Training set: 100000 qtrees taken from web mmeta logs' -f base64 qtrees.sample.txt"); + opts.SetTitle("Teaches the codec and serializes it as a file named CODECNAME.hash(CODECDATA).bin"); + + opts.AddLongOption('m', "message").RequiredArgument("training_set_comment").StoreResult(&info.TrainingSetComment).Required().Help("a human description for the training set"); + + opts.AddLongOption('r', "resource").RequiredArgument("training_set_res_id").StoreResult(&info.TrainingSetResId).Optional().Help("sandbox resource id for the training set"); + + opts.AddLongOption('c', "codec").RequiredArgument("codec_name").StoreResult(&info.CodecName).Optional().DefaultValue(info.CodecName); + + opts.AddLongOption('s', "sample-multiplier").RequiredArgument("multiplier").StoreResult(&info.SampleSizeMultiplier).Optional().DefaultValue(ToString(info.SampleSizeMultiplier)).Help("multiplier for default sample size"); + + opts.AddLongOption('f', "format").RequiredArgument(TStringBuilder() << "(" << NCodecs::DSF_PLAIN_LF << "|" << NCodecs::DSF_BASE64_LF << ")").StoreResult(&fmt).Required().Help("training set input file format"); + + opts.AddLongOption("list-codecs").NoArgument().Handler0([]() { + Cout << JoinStrings(NCodecs::ICodec::GetCodecsList(), "\n") << Endl; + exit(0); + }) + .Optional() + .Help("list available codecs"); + + opts.AddLongOption("fake-revision").RequiredArgument("revision").StoreResult(&info.RevisionInfo).Optional().Hidden(); // replace static_codec_generator revision in debug info + + opts.AddLongOption("fake-timestamp").RequiredArgument("timestamp").StoreResult(&info.Timestamp).Optional().Hidden(); // replace generating timestamp in debug info + + opts.SetFreeArgsMin(0); + opts.SetFreeArgTitle(0, "training_set_input_file", "training set input files"); + + NLastGetopt::TOptsParseResult res(&opts, argc, argv); + + Cout << "Reading training set data ... " << Flush; + TVector<TString> allData; + for (const auto& freeArg : res.GetFreeArgs()) { + NCodecs::ParseBlob(allData, fmt, NCodecs::GetInputBlob(freeArg)); + } + + if (!res.GetFreeArgs()) { + NCodecs::ParseBlob(allData, fmt, NCodecs::GetInputBlob("-")); + } + Cout << "Done" << Endl << Endl; + + Cout << "records: " << allData.size() << Endl; + Cout << "raw size: " << NCodecs::GetInputSize(allData.begin(), allData.end()) << " bytes" << Endl << Endl; + + Cout << "Training " << info.CodecName << " , sample size multiplier is " << info.SampleSizeMultiplier << " ... " << Flush; + auto codec = NCodecs::BuildStaticCodec(allData, info); + Cout << "Done" << Endl; + + TString codecName = NCodecs::GetStandardFileName(codec); + NCodecs::TCodecPtr codecPtr = NCodecs::ICodec::RestoreFromString(codec.GetStoredCodec()); + + Cout << "Testing compression ... " << Flush; + auto stats = NCodecs::TestCodec(*codecPtr, allData); + Cout << "Done" << Endl << Endl; + + codec.MutableDebugInfo()->SetCompression(stats.Compression()); + + Cout << stats.Format(codec, false) << Endl; + + Cout << "Saving as " << codecName << " ... " << Flush; + { + TUnbufferedFileOutput fout{codecName}; + NCodecs::SaveCodecInfoToStream(fout, codec); + fout.Finish(); + } + Cout << "Done" << Endl << Endl; +} diff --git a/library/cpp/codecs/static/tools/static_codec_generator/ya.make b/library/cpp/codecs/static/tools/static_codec_generator/ya.make new file mode 100644 index 0000000000..efbc440dd1 --- /dev/null +++ b/library/cpp/codecs/static/tools/static_codec_generator/ya.make @@ -0,0 +1,17 @@ +PROGRAM() + +OWNER(velavokr) + +SRCS( + static_codec_generator.cpp +) + +PEERDIR( + library/cpp/codecs + library/cpp/codecs/static + library/cpp/codecs/static/tools/common + library/cpp/digest/md5 + library/cpp/getopt/small +) + +END() diff --git a/library/cpp/codecs/static/tools/tests/canondata/result.json b/library/cpp/codecs/static/tools/tests/canondata/result.json new file mode 100644 index 0000000000..7a637c6763 --- /dev/null +++ b/library/cpp/codecs/static/tools/tests/canondata/result.json @@ -0,0 +1,6 @@ +{ + "static_codec_tools.test_static_codec_tools": { + "checksum": "960e3c8c57fb846ab53ccbd07e287233", + "uri": "sbr://144512644/static_codec_tools.test_static_codec_tools/solar-8k-a.huffman.1467494385.codec_info" + } +}
\ No newline at end of file diff --git a/library/cpp/codecs/static/tools/tests/static_codec_tools.py b/library/cpp/codecs/static/tools/tests/static_codec_tools.py new file mode 100644 index 0000000000..db4140e370 --- /dev/null +++ b/library/cpp/codecs/static/tools/tests/static_codec_tools.py @@ -0,0 +1,18 @@ +#!/usr/bin/env python + +import yatest.common as tt +import os.path as op + +def test_static_codec_tools(): + tt.execute([tt.binary_path("library/cpp/codecs/static/tools/static_codec_generator/static_codec_generator")] + + ["-m", "test codec", "-r", "sbr://143310406", "-f", "plain", "-c", "solar-8k-a:huffman", "-s", "1", + "--fake-revision", "r2385905", "--fake-timestamp", "1467494385", "sample.txt"], + timeout=60) + assert(op.exists("solar-8k-a.huffman.1467494385.codec_info")) + tt.canonical_execute(tt.binary_path("library/cpp/codecs/static/tools/static_codec_checker/static_codec_checker"), + args=["-c", "solar-8k-a.huffman.1467494385.codec_info"], + timeout=60) + tt.execute([tt.binary_path("library/cpp/codecs/static/tools/static_codec_checker/static_codec_checker")] + + ["-c", "solar-8k-a.huffman.1467494385.codec_info", "-f", "plain", "-t", "sample.txt"], + timeout=60) + return tt.canonical_file("solar-8k-a.huffman.1467494385.codec_info") diff --git a/library/cpp/codecs/static/tools/tests/ya.make b/library/cpp/codecs/static/tools/tests/ya.make new file mode 100644 index 0000000000..c5324eaf53 --- /dev/null +++ b/library/cpp/codecs/static/tools/tests/ya.make @@ -0,0 +1,20 @@ +PY2TEST() + +OWNER(velavokr) + +TEST_SRCS(static_codec_tools.py) + +DATA(sbr://143310406) + +TIMEOUT(4200) + +TAG(ya:not_autocheck) + +DEPENDS( + library/cpp/codecs/static/tools/static_codec_checker + library/cpp/codecs/static/tools/static_codec_generator +) + + + +END() diff --git a/library/cpp/codecs/static/tools/ya.make b/library/cpp/codecs/static/tools/ya.make new file mode 100644 index 0000000000..dd3e8437aa --- /dev/null +++ b/library/cpp/codecs/static/tools/ya.make @@ -0,0 +1,5 @@ +RECURSE( + common + static_codec_generator + static_codec_checker +) diff --git a/library/cpp/codecs/static/ut/builder_ut.cpp b/library/cpp/codecs/static/ut/builder_ut.cpp new file mode 100644 index 0000000000..b47c279ed1 --- /dev/null +++ b/library/cpp/codecs/static/ut/builder_ut.cpp @@ -0,0 +1,57 @@ +#include <library/cpp/testing/unittest/registar.h> +#include <library/cpp/codecs/static/builder.h> +#include <library/cpp/codecs/static/static_codec_info.pb.h> +#include <util/string/vector.h> + +class TStaticCodecInfoBuilderTest: public NUnitTest::TTestBase { + UNIT_TEST_SUITE(TStaticCodecInfoBuilderTest) + UNIT_TEST(TestBuild) + UNIT_TEST_SUITE_END(); + +private: + TVector<TString> PrepareData() { + TVector<TString> data; + for (ui32 i = 'a'; i <= 'z'; ++i) { + data.push_back(TString(1, (char)i)); + } + return data; + } + + void TestBuild() { + TVector<TString> data; + NCodecs::TCodecBuildInfo info; + info.CodecName = "huffman"; + info.SampleSizeMultiplier = 2; + info.Timestamp = 1467494385; + info.RevisionInfo = "r2385905"; + info.TrainingSetComment = "some dummy data"; + info.TrainingSetResId = "sbr://1234"; + auto res = NCodecs::BuildStaticCodec(PrepareData(), info); + UNIT_ASSERT_VALUES_EQUAL(res.ShortUtf8DebugString(), + "StoredCodec: \"\\007\\000huffman@S\\000a" + "\\006b\\005c\\005d\\005e\\005f\\005g\\005h\\005i\\005j\\005k\\005l\\005m\\005n\\005o" + "\\005p\\005q\\005r\\005s\\005t\\005u\\004v\\004w\\004x\\004y\\004z\\004\xC7?\xC8>" + "\xC9=\xCA<\xCB;\xCC:\3159\3168\3177\3206\3215\3224\3233\3242\3251\3260\xD7/\xD8." + "\xD9-\xDA,\xDB+\xDC*\xDD)\xDE(\xDF\\'\xE0&\xE1%\xE2$\xE3#\xE4\\\"\xE5!\xE6 \xE7" + "\\037\xE8\\036\xE9\\035\xEA\\034\xEB\\033\xEC\\032\xED\\031\xEE\\030\xEF\\027\xF0" + "\\026\xF1\\025\xF2\\024\xF3\\023\xF4\\022\xF5\\021\xF6\\020\xF7\\017\xF8\\016\xF9" + "\\r\xFA\\014\xFB\\013\xFC\\n\xFD\\t\xFE\\010\xFF\\007\" " + "DebugInfo { " + "CodecName: \"huffman\" " + "Timestamp: 1467494385 " + "RevisionInfo: \"r2385905\" " + "SampleSizeMultiplier: 2 " + "TrainingSetComment: \"some dummy data\" " + "TrainingSetResId: \"sbr://1234\" " + "StoredCodecHash: 2509195835471488613 " + "}"); + + UNIT_ASSERT_VALUES_EQUAL(NCodecs::GetStandardFileName(res), "huffman.1467494385.codec_info"); + UNIT_ASSERT_VALUES_EQUAL(res.GetDebugInfo().GetStoredCodecHash(), 2509195835471488613ULL); + + auto res1 = NCodecs::LoadCodecInfoFromString(NCodecs::SaveCodecInfoToString(res)); + UNIT_ASSERT_VALUES_EQUAL(res1.ShortUtf8DebugString(), res.ShortUtf8DebugString()); + } +}; + +UNIT_TEST_SUITE_REGISTRATION(TStaticCodecInfoBuilderTest); diff --git a/library/cpp/codecs/static/ut/static_ut.cpp b/library/cpp/codecs/static/ut/static_ut.cpp new file mode 100644 index 0000000000..57e1e62887 --- /dev/null +++ b/library/cpp/codecs/static/ut/static_ut.cpp @@ -0,0 +1,27 @@ +#include <library/cpp/testing/unittest/registar.h> +#include <library/cpp/codecs/static/example/example.h> + +class TStaticCodecUsageTest: public NUnitTest::TTestBase { + UNIT_TEST_SUITE(TStaticCodecUsageTest) + UNIT_TEST(TestUsage) + UNIT_TEST_SUITE_END(); + +private: + void DoTestUsage(NStaticCodecExample::EDictVersion dv, size_t expectedSize) { + const TStringBuf letov = "Всё идёт по плану"; + + TBuffer outEnc, outDec; + NStaticCodecExample::Encode(outEnc, letov, dv); + NStaticCodecExample::Decode(outDec, TStringBuf{outEnc.data(), outEnc.size()}); + + UNIT_ASSERT_VALUES_EQUAL(outEnc.Size(), expectedSize); + UNIT_ASSERT_EQUAL(TStringBuf(outDec.data(), outDec.size()), letov); + } + + void TestUsage() { + DoTestUsage(NStaticCodecExample::DV_HUFF_20160707, 18u); + DoTestUsage(NStaticCodecExample::DV_SA_HUFF_20160707, 22u); + } +}; + +UNIT_TEST_SUITE_REGISTRATION(TStaticCodecUsageTest) diff --git a/library/cpp/codecs/static/ut/ya.make b/library/cpp/codecs/static/ut/ya.make new file mode 100644 index 0000000000..b9116097d8 --- /dev/null +++ b/library/cpp/codecs/static/ut/ya.make @@ -0,0 +1,14 @@ +UNITTEST_FOR(library/cpp/codecs/static) + +OWNER(velavokr) + +SRCS( + builder_ut.cpp + static_ut.cpp +) + +PEERDIR( + library/cpp/codecs/static/example +) + +END() diff --git a/library/cpp/codecs/static/ya.make b/library/cpp/codecs/static/ya.make new file mode 100644 index 0000000000..00e00fd8d4 --- /dev/null +++ b/library/cpp/codecs/static/ya.make @@ -0,0 +1,18 @@ +LIBRARY() + +OWNER(velavokr) + +SRCS( + builder.cpp + static_codec_info.proto + static.cpp +) + +PEERDIR( + library/cpp/codecs + library/cpp/archive + library/cpp/svnversion + util/draft +) + +END() |