LCR Indexing

This repository contains a framework for quickly evaluating and testing LCR indexes. The code has been used for my master thesis at the Technical University of Eindhoven, July 2021.

Building and running

The code has been written in C++-17. Numerous scripts have been included, which have all been written in python (3.7). The project has been tested on Windows and Linux. It should also run and compile on Mac.

Requirements

CMake version 3.10 or later
c++ compiler that supports C++-17
python 3.7 or later
boost library (must be present in folder /include/boost/)

Building the C++ code

linux and mac

cmake -DCMAKE_BUILD_TYPE=Release -DNO_GTEST=True -DNO_PRECOMPILED_HEADER=True -G"Unix Makefiles" ./

Remove -DNO_GTEST=True if the tests should be build. For CMake 3.16 or later, -DNO_PRECOMPILED_HEADER=True can be removed, which allows for faster re-build times. It is recommended to target Release builds. See CMake documentation for more options.

After running the cmake command, the following command should be used to build the executables:

make -j4

When rebuilding, only the above command is required. Unless files have been added/removed, then the cmake command must be re-executed.

Windows

for windows the following command can be used:

cmake -DCMAKE_BUILD_TYPE=Release -DNO_GTEST=True -DNO_PRECOMPILED_HEADER=True -G"Visual Studio 16 2019" -A x64 ./

Now open the visual studio projects to build and run the code.

Running

After building the executables the following command is available to run indexes:

MasterThesis[.exe] [reach|lcr] 
                  --graphFile [graphFile] 
                  --queryFile [queriesFile]                
                  --index [indexName] 
                  --indexParams [parameterList]
                  --timeLimit [timeLimitInSeconds] 
                  --memoryLimit [memoryLimitInMBs]
                  <[--control]>

In order to generate queries for a graph, the following command may be used:

LCRQueryGenerator[.exe] 
                  --graphFile [graphFileIn]
                  --randomQueries [count]
                  --connectedQueries [count]
                  <[--printStats]>
                  <[--allInOne]>
                  <[--splitRandomFromConnected]>

The last command can be used to import a graph for graph statistics. It is also possible to convert between graph representations. For example changing from .rdf file to .nt file, which can be achieved by specifying the --graphFileOut parameter.

GraphUtilities[.exe] [graphFileIn] 
                  [--printStats] 
                  [--graphFileOut] [graphFileOut]

Python scripts

Running the 'runner' scripts, requires the c++ executables to be build. The requirements for the python scripts can be found in runner/requirements.txt.

In graphProcessing, numerous scripts for downloading and processing real datasets can be found.
In runner, numerous scripts for running and creating benchmarks on real and synthetic datasets can be found.
In evaluation, numerous scripts for visualizing and evaluation benchmark results can be found.

lucdon/LCRIndexing