Improve storing of the package configs

[mfocko]

Related to packit/packit#2306

This topic came up during an arch discussion about logging of the package configs. Currently we store the configs as JSON in the database with each of the project events, this “may” be a bit wasteful, given that the configs should not change that often.

To optimize this, we should be able to hash a canonical JSON as a string. This would allow us to store the JSON only once (in a separate table) and reference it from the “project event model”.

As for the canonical from, only nested dictionaries do not guarantee an order, but json.dump provides an option to resolve this, therefore it should be possible to obtaine a canonical JSON by following:

json.dumps(
  package_config,
  separators=(',', ':'),  # minimize whitespace
  sort_keys=True,         # sort the keys in dictionaries
)

Permalink to the field as of the time of creating this issue:

packit-service/packit_service/models.py

Line 1595 in 99ff2c5

packages_config = Column(JSON)