thomasgilgenast/aotd-graph repository overview

⁠aotd-graph

a graph visualization for "album of the day" selections

⁠Basic usage example

Construct a graph by running build_graph.py, then visualize it by running draw_graph.py.

⁠Premise

The goal is to create a meaningful graph of albums (identified by YouTube URL), so that algorithms "near" each other on the graph should be "similar."

Initially, every selected album is a vertex on the graph. How will we connect similar albums in meaningful ways?

The idea is to scrape the YouTube URL's to get related URL's for each album, and add a new vertex for each related album, with a corresponding new edge from the album it was related to. Thus, edges in the graph signify that two albums are similar enough that YouTube lists one as a related video from the other. These "explored" albums are the small vertices in the image above.

By recursing deeper into the tree of related videos, we hope to eventually flesh out the graph.

The final graph is rendered with a force-directed layout that should place albums which are more closely connected (shorter path of related videos) or more strongly connected (larger number of mutual related videos) closer together.

⁠Caveats

Ideally, we want to explore regions of YouTube video space which represent albums, not arbitary links such as cat videos. Therefore, we restrict our exploration to videos which contain the substring "album" in their title.

⁠Docker-specific notes

⁠Docker setup instructions (for PyCharm)

1. Install Docker

   • If Docker does not start, edit C:\Windows\System32\drivers\etc\hosts and add the machine name to a loopback IP (e.g., ::1 MY-COMPUTER-NAME)
   • In the General tab of the Docker Settings window, check the option for "Expose daemon on tcp://localhost:2375 without TLS"
   • In the Shared Drives tab of the Docker Settings window, check the option for whatever drive you cloned this project into

2. Pull the image from the Docker Cloud

   $ docker pull thomasgilgenast/aotd-graph
   

   Copy

3. Configure the Remote Interpreter

   • In PyCharm, under File > Settings, under Project Interpreter, click the gear and choose "Add Remote"
   • Select "Docker" from the radio buttons
   • Add a new Server if one doesn't exist yet (accept the defaults)
   • Select thomasgilgenast/aotd-graph:latest from the "Image name" dropdown
   • Change the "Python interpreter path" to python3

4. Test the installation

   • Right-click test.py and choose "Run 'test'"
   • A new file test.png should appear in the project directory

   

⁠Running scripts against Docker without using PyCharm

You can always run scripts directly within a Docker container by mapping the host directory where you cloned this repo to /aotd-graph on the container.

For example:

$ docker run -it -v <host directory>:/aotd-graph thomasgilgenast/aotd-graph python3 build_graph.py
$ docker run -it -v <host directory>:/aotd-graph thomasgilgenast/aotd-graph python3 draw_graph.py

⁠Manually updating the Docker image

This should happen automatically for pushes to master, version tags, and for PR's opened against this repo, but for posterity here is the "long way" to do it.

1. Edit the Dockerfile

2. Re-build the container

   $ docker build -t aotd-graph .
   

   Copy

3. Tag the container with a new version number

   $ docker tag aotd-graph <docker username>/aotd-graph:<version>
   

   Copy

4. Push the new version to Docker cloud

   $ docker push <docker username>/aotd-graph:<version>
   

   Copy