memgraph

Memgraph is an ACID compliant high performance transactional distributed in-memory graph database featuring runtime native query compiling, lock free data structures, multi-version concurrency control and asynchronous IO.

dependencies

Memgraph can be compiled using any modern c++ compiler. It mostly relies on the standard template library, however, some things do require external libraries.

Some code contains linux-specific libraries and the build is only supported on a 64 bit linux kernel.

• linux
• clang 3.5 or Gcc 4.8 (good c++11 support, especially lock free atomics)
• boost 1.55 (or something, probably works with almost anything)
• lexertl (2015-07-14)
• lemon (parser generator)
• catch (for compiling tests)

Bolt

sudo apt-get install libssl-dev

build

cd build
cmake -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ -DRUNTIME_ASSERT=OFF -DTHROW_EXCEPTION_ON_ERROR=OFF -DNDEBUG=OFF ..

# Flags:
#   -DCMAKE_C_COMPILER=clang
#   -DCMAKE_CXX_COMPILER=clang++
#   -DRUNTIME_ASSERT=OFF
#   -DTHROW_EXCEPTION_ON_ERROR=OFF
#   -DNDEBUG=ON
#   -DCMAKE_BUILD_TYPE:STRING=debug
#   -DCMAKE_BUILD_TYPE:STRING=release

make
ctest
ctest -V
ctest -R test_name
ctest -R unit
ctest -R concurrent
ctest -R concurrent --parallel 4

Proof of concept

Dressipi

Run command:

cd build/poc
MEMGRAPH_CONFIG=/path/to/config/memgraph.yaml ./poc_astar -v "dressipi/graph_nodes_export.csv" -e "dressipi/graph_edges_export.csv"

Powerlinx

Run command:

cd build/poc
MEMGRAPH_CONFIG=/path/to/config/memgraph.yaml ./profile -d ";" -v "node_account.csv" -v "node_company.csv" -v "node_opportunity.csv"  -v "node_personnel.csv" -e "rel_created.csv" -e "rel_has_match.csv" -e "rel_interested_in.csv" -e "rel_is_employee.csv" -e "rel_partnered_with.csv" -e "rel_reached_to.csv" -e "rel_searched_and_clicked.csv" -e "rel_viewed.csv" -e "rel_works_in.csv"

Custom test build example

clang++ -std=c++1y -o concurrent_skiplist ../tests/concurrent/skiplist.cpp -pg -I../src/ -I../libs/fmt -Lfmt-prefix/src/fmt-build/fmt -lfmt -lpthread

TODO

• implement basic subset of queries:

  • create node
  • create edge
  • find node
  • find edge
  • update node
  • update edge
  • delete node
  • delete edge

• implement index

  • label index
  • type index
  • property index

• from header only approach to .hpp + .cpp

  • query engine has to be statically linked with the rest of the code

• unit test of queries that are manually implemented

  • src/query_engine/main_queries.cpp -> tests/unit/manual_queries

• unit test of query_engine

• console with history