Keep the current user prompts in the main, but also allow the user to give some command line arguments in order to spare them 4 user prompts every time they run this program.

See peracotta's main for inspiration.

For example in our case the WEEE-Open repo is useless since its commits are automatic.

Possible implementation:
... (-X|--exclude) WEEE-Open,other-repo,...

Since we're now using more than 1 library, add a requirements.txt with pip freeze > requirements.txt and replace the line in the README that says to install requests with pip install -r requirements.txt

Make the CLI UX better with the rich Python module

Automatically update the combined graph in the README every few days to reflect our progress on GitHub

Find an option for cloc to ignore the files contained in ignored_files.py. Also add bootstrap-dark.css to the ignored files.

The get_contributors_commits_stats HTTP request also gets data about lines added and removed (diff) by developers. Add these stats to the output and the graphs of #3

This function should return the number of the past year's commits, total and divided by repo.

Is it really our second biggest software project?
We should find out what file extensions to exclude from our SLOC count.

When using the options --no-commits --sloc --graphs the graph generation fails because it doesn't have data about commits.
Traceback:

Generating graphs...
Traceback (most recent call last):
  File "./main.py", line 624, in <module>
    main()
  File "./main.py", line 618, in main
    print_all_stats(commits_stats, lines_stats, contributors_stats, use_cloc, generate_graphs)
  File "./main.py", line 475, in print_all_stats
    generate_figure([repo_commits[graph], yearly_repo_commits[graph], sloc_by_repo[graph]], os.path.join(graph_dir, f'{graph}.svg'))
KeyError: 'ansible-pxe'

'ansible-pxe' is the key that fails because it is the first repo when sorting alphabetically.

Right now graph dimensions are determined with a constant value for pie charts and a value proportional to the number of bars for bar charts, with a completely arbitrary value added to them in order to account for graph title and axis name.

While this is good enough, it generates graphs for which the single column width is not of a fixed size, graphs in which the title space is tight or too large, and appearance variations which are undesirable. In some edge cases, it might even create an overlapped mess of colors and text.

Ideally, dimensions should be determined as follows:

• Accept an aspect ratio from the caller.
• Accept a DPI value from the caller.
• Accept title, legend and axis label font sizes.
• Accept a class width in pixels
• Accept a class spacing in pixels

Then use all the previous values to compute exact dimensions for:

• Figure
• Plot area
• Graph area
• Title area
• Axis label area
• Legend

The figure size should be the least value that can contain all its children.

The plot area (graph + title + legend + axis) should be the least value that can contain all its children, plus some whitespace on either side to have the aspect ratio match (unless the ratio is zero, in which case this last step is skipped).

The graph area should be calculated by using class and spacing values. For pie charts, the class is just one, for bar charts it's the number of bars. Spacing is in between classes (if there are many) and above the first class and below the last class.

This way the graph should be completely deterministic in aspect no matter the datasize or graph type.

The result should be a pie chart with e.g. 42% PHP 41.9% PHP, 10% Bash, 3% VHDL, 3.1% JavaScript

Since we have to make more than 60 GET requests, we get rate-limited by GitHub's servers.
To prevent this, implement a form of login (better without using a dedicated library, but it works as well)

See this StackOverflow thread to find an answer.

The resulting container has to automatically run ./main.py --ping every hour (0 * * * * in cron).

Instead of having fixed colors, the pie charts could use some variety.
In particular, we need to keep the same color order (since it allows for a readable chart), but we can randomize the first color of the list (e.g. normally "Pink", "Blue", "Red" can also become "Blue", "Red", "Pink" and "Red", "Pink", "Blue", but never change order and become "Blue", "Pink", "Red").

When running with only --ping option, after correctly getting stats about commits, the following error appears:

Traceback (most recent call last):
  File "/root/sardina/main.py", line 754, in <module>
    main()
  File "/root/sardina/main.py", line 745, in main
    language_total, language_repo = get_language_stats(repos, header) if (get_languages and not use_cloc) else (None, None)
UnboundLocalError: local variable 'get_languages' referenced before assignment

Set up Continuous Deployment with GitHub Actions. See WEEE-Open/peracotta#67 for more details

Using curl -H "Accept: application/vnd.github.v3+json" -H "Authorization: token <PAT>" https://api.github.com/rate_limit returns

{
  "message": "Bad credentials",
  "documentation_url": "https://docs.github.com/rest"
}

We need to fix the way we provide an authorization (currently using my PAT) to the APIs.

We can use the matplotlib library to plot some interesting graphs, such as:

• pie chart with percentage of commits (per repo, total this year, total all time)
• histogram with lines of code to compare repos
• other ideas? Feel free to implement them

The graphs should also be saved to a file.

Read dev_mode value from config file.

The development mode allows to save all the json requests contents to a json file per request and consequently reads it the at the next execution to save time.
This mode also skips the removal of the downloaded repos to save time for the lines stats section.