Download all data from a Facebook Group: posts, comments (and subcomments), likes, reactions, users, shares, user mentions and shared links using the official facebook-sdk module for Python 2.7. WARNING! Under development, not all features are implemented or to be considered stable.

Set a virtual environment: virtualenv facebookenv && source facebookenv/bin/activate. I suggest to use the virtualenvwrapper utility: mkvirtualenv facebookenv && workon.

Install dependencies: pip install -r requirements.txt.

Copy example.cnf to your_group_name.cnf.

Edit your_group_name.cnf adding your access_token (or app_id and app_secret) and your group_id. You must create a Facebook App to obtain the first three parameters: https://developers.facebook.com/. If you don't know your group id, you can find it in the source of the page.

You can choose between two modes: the archive one downloads all elements in the time window selected, the update one takes an existing gexf file and download only new elements since the last post found and until the until_datetime or now.

Other optional options are since_datetime and until_datetime to narrow the time window and filter harvested posts. You can use your favorite format to write datetime, describing it in datetime_format option using the strftime syntax: https://strftime.org/. Please escape % char with another %, so use double %% (sic).

You can also enable or disable two features: download of all members (and not only the active ones in the time window) with all_members, and layout computing (Fruchterman-Reingold force-directed layout) with calc_layout.

All harvested data will be saved in two files: file_name.gexf and file_name.json. You can customize file_name value, but the default value is {YYYYmmddHHMM}_{group name}.

Run: python fb_group.py your_group_name.cnf. Standard output is quite verbose.

Harvested data are organized in a graph structure:

• nodes: users, posts, comments, domains;
• edges:
  • user -|is author of|-> post,
  • user -|is author of|-> comment,
  • user -|reacts to|-> post,
  • user -|reacts to|-> comment,
  • post -|mentions|-> user,
  • post -|mentions|-> domain,
  • comment -|in reply to|-> post,
  • comment -|in reply to|-> comment,
  • comment -|mentions|-> user,
  • comment -|mentions|-> domain.

There are four types of nodes and four types of edges. Nodes have additional attributes (ie. name of user, message for posts and comments). The -|reacts to|-> edge has an attribute for both posts and comments, the type of reaction (LIKE, LOVE, WOW, HAHA, SAD, ANGRY, THANKFUL[, SHARE]), and the -|mentions|-> one directed to a domain has many attributs, including original link.

This graph is saved in two formats: GEXF and JSON.

It's a XML based format suitable to be imported and managed by Gephi.

It's a simple JSON representation of the network: { "nodes": [...], "edges": [...] }. Single nodes are objects with an id and some additional attributes. Singles edges are objects with a source and a target containing nodes' ids. You can use it directly in javascript visualizations, ie. powered by the d3js library.

Marco Goldin.