TdVisu

---

Visualization for dynamic programming on tree decompositions.

Create a graph object for each given graph that is of interest for the dynamic programming.

The visualization generates highlights and adds solution-tables for user defined time steps.

These snapshot of the graphs will be written in a graphviz-supported file-format to a folder of your choosing.

For the portable and light weight '.svg' format, all graphs for a timestep can be joined together to provide a thoroughly view on the process of dynamic programming.

With the '.svg' format the images are highly customizable, and even combining several timesteps together using svg animate would be an option in the future.

---

Using

Note: see also the steps prepared in the CI/CD .github/workflows/python-app.yml:

Graphviz (>=2.38). Be aware of changes in default layouts over different major versions of Graphviz. The project currently tests with graphviz-version: "12.2.1".

python-benedict[xml]

PostgreSQL adapter for Python: psycopg (3)

---

To register the graphviz plugins

https://gitlab.com/graphviz/graphviz/-/issues/1352

dot.exe -c

To install

In a command prompt with pip (to get pip see: https://pip.pypa.io/en/stable/) installed: Just run

pip install -h (for more information on install options)
pip install tdvisu

To download the latest version from the default branch:

git clone --depth 1 --single https://github.com/VaeterchenFrost/tdvisu

To isolate the dependencies

With virtualenv on the system installed you can isolate the environment, for example

virtualenv tdvisu_dir -p 3.13
cd tdvisu_dir/bin/
source activate
# Windows: ./tdvisu_dir/Scripts/activate

With Conda on the system installed the dependencies for this project can be automatically installed in a new environment:

Go to the projects base directory.

Open a conda-command-prompt with admin privileges and run the commands from the project folder

• to create a new environment with basic dependencies:

conda env create -f ./environment.yml

• to activate the environment:

conda activate tdvisu

Install from source

To clone the complete repository:

git clone https://github.com/VaeterchenFrost/tdvisu

To download only the latest version from the default branch:

git clone --depth 1 --single https://github.com/VaeterchenFrost/tdvisu

To install the project from the source folder:

pip install -h (for more information on install options)
pip install .

to confirm that the visualization finds all dependencies:

python ./tdvisu/visualization.py -h

to run all tests:

pip install .[test]
pytest ./test/

---

How to use

The visualization needs input in the form of the Json API. The creation of this file is exemplary implemented in construct_dpdb_visu.py or the fork GPUSAT and --visufile filename (optionally disabling preprocessing with -p).

Run the python file with the above dependencies installed: visualization.py

visualization.py takes two parameters, the json-infile to read from, and optionally one outputfolder. With both arguments a run might look like this:

python tdvisu/visualization.py visugpusat.json examplefolder

For #SAT it produces for example three different graphs suffixed with a running integer to represent timesteps:

• TDStep the tree decomposition with solved nodes
• PrimalGraphStep the primal graph with currently active variables highlighted
• IncidenceGraphStep the bipartite incidence graph with active clauses/variables highlighted

The graphs are images encoded in resolution independent .svg files (see https://www.lifewire.com/svg-file-4120603)

How to use construct_dpdb_visu.py

After installing the project dp_on_dbs with the there listed requirements, we need to

• edit the database.ini with our password to postgresql
• Solve a problem with python dpdb.py [GENERAL-OPTIONS] -f <INPUT-FILE> <PROBLEM> [PROBLEM-SPECIFIC-OPTIONS]
  • for the problem VertexCover
    • with flag --gr-file to store the htd Input (if the input was in a different format)
  • for the problem SharpSat
    • with flag --store-formula to store the formula in the database
• Run
  • Sat / SharpSat: python construct_dpdb_visu.py [PROBLEMNUMBER]
  • VertexCover: python construct_dpdb_visu.py [PROBLEMNUMBER] --twfile [TWFILE] with the file in DIMACS tw-format containing the edges of the graph.

Installation of PostgreSQL Dependencies

TdVisu requires PostgreSQL client libraries for database connectivity through psycopg.

Windows

Install PostgreSQL from postgresql.org or use the client-only installer:

# Option 1: Full PostgreSQL installation (includes server + client)
# Download from: https://www.postgresql.org/download/windows/

# Option 2: Client libraries only (recommended for development)
# Download "Command Line Tools" from PostgreSQL download page

macOS

# Using Homebrew
brew install postgresql@16
# OR just the client libraries
brew install libpq

Linux (Ubuntu/Debian)

# Client libraries and development headers
sudo apt-get install postgresql-client-16 libpq-dev

# OR just client libraries for runtime
sudo apt-get install libpq5

Linux (CentOS/RHEL/Fedora)

# Client libraries and development headers  
sudo dnf install postgresql16 postgresql16-devel
# OR
sudo yum install postgresql-client postgresql-devel

Conda Environment

# Cross-platform solution
conda install postgresql libpq

Verification

After installing PostgreSQL, verify the installation:

pg_config --version  # Should show PostgreSQL version
python -c "import psycopg; print('✅ psycopg works!')"

Note: Whatever version of libpq psycopg is compiled with, it will be possible to connect to PostgreSQL servers of any supported version. Install the most recent libpq version available for your platform.

For more details see: https://www.psycopg.org/psycopg3/docs/basic/install.html

---

New Release

Version

• Bump /version.py according to the changes made
• Change date to the release day, keep format

Requirements

In case dependencies have changed, or just to update some, check

• requirements.txt
• stable-requirements.txt (using pip freeze)
• setup.py

Write Changelog.md

• Add tag with link (see bottom for linking examples)
• Add changes, maybe some are already in Unreleased
• Update Unreleased with (No) unreleased changes

Review code

• Run tests (pytest)
• Check codestyle (pylint)

Push

• Push changes to main
• Wait for all automated checks! (All checks have passed...)

Create Release

• On the GitHub page go to: Release, Draft a new release
• Enter v'YOUR VERSION NUMBER' as the tag.
• Add a Release Title (could be just the version)
• Add some description (like in the CHANGELOG.md)
• Click on Publish release on the bottom

Should automatically release to PyPI

• For details see: Upload Python Package

Now you are set for the new release 🎉

---