repository·master·Indexed 25 days ago
https://github.com/platformlab/nanologA high-performance, nanosecond-scale logging system for C++ that achieves massive throughput by extracting static log metadata at compile-time and deferring formatting to an offline decompression phase. It offers two versions: a C++17 library and a Preprocessor version for maximum performance. Features include a printf-like API with log levels (DEBUG, NOTICE, WARNING, ERROR), asynchronous logging, and a dedicated decompressor tool to convert binary logs into human-readable ASCII.
run_decompressionCosts.sh to evaluate decompression performance. The script generates a log file containing 1 of 6 specific log statements and measures the time required to decompress each variant.NanoLog produces binary log files to maintain high performance. To read them, you must use the decompressor application produced during the build process.
Run the following command:
./decompressor decompress <path_to_binary_log>
./decompressor decompress ./compressedLogNanoLog is designed for Linux-based systems and requires:
run_sortedDecompressionThreads.sh to study the impact of runtime logging concurrency on post-execution decompression. The script varies the number of runtime logging threads producing messages and measures the subsequent decompression time.The Preprocessor version is for advanced users seeking maximum performance. It requires integrating a Python script into your build chain and produces a non-portable library specific to each compilation.
NanoLogMakeFrag.USR_SRCS (source files) and USR_OBJS (object files).run-cxx macro instead of g++ to compile all .cc files into .o files.make will invoke a Python script to generate library code specific to your application.decompressor will be generated in your application directory.run_aggregation.sh to compare the processing performance of different tools. This script creates a NanoLog log file containing two different log statements in varying ratios and measures the time taken by Python, Awk, C++, and NanoLog to process them.benchmarks/ directory. These scripts handle machine profiling, execution, and result storage in the results/ directory (organized by timestamp).run_bench.sh script is the primary entry point for benchmarking. It profiles the host machine, runs the core benchmark application, and saves results to a timestamped subdirectory within results/. You can provide an optional suffix when invoking the script to identify your run.The C++17 version is the easiest to use and works like a traditional library.
runtime directory and run make. This produces libNanoLog.a and a decompressor executable.#include "NanoLogCpp17.h".-L ./runtime/ -lNanoLog -lrt -pthread.-Werror=format) to prevent silent log corruption.decompressor to turn the binary log into human-readable ASCII.NanoLog uses a printf-like API. You must include the appropriate header based on the version you chose (either NanoLogCpp17.h or NanoLog.h) and specify a log level.
Log Levels:
DEBUGNOTICEWARNINGERRORYou can control the active logging level using NanoLog::setLogLevel(...).
#include "NanoLogCpp17.h"
using namespace NanoLog::LogLevels;
int main()
{
NANO_LOG(NOTICE, "Hello World! This is an integer %d and a double %lf\r\n", 1, 2.0);
return 0;
}Use the NANO_LOG macro to record messages. It supports standard printf specifiers (except %n), including width and length specifiers.
To avoid prefixing every call with the NanoLog:: namespace, you can use using namespace NanoLog::LogLevels; to bring DEBUG, NOTICE, WARNING, and ERROR into the current scope.
You can control where NanoLog writes its compressed logs and the minimum severity level required for a message to be persisted.
NanoLog::setLogFile(path) to specify the output file location. The default is ./compressedLog.NanoLog::setLogLevel(level) to set the minimum LogLevel. Valid levels (from least to greatest) are DEBUG, NOTICE, WARNING, and ERROR.