Switch to JSON-based default ETags instead of repr-based in the API.

Review Request #9734 — Created March 4, 2018 and submitted — Latest diff uploaded

Information

Djblets
release-2.0.x
b26be03...

Reviewers

When generating a default ETag for a resource, we were making use of
repr() and generating a hash based on that. This was fine when just
using Python 2.x, but given a number of differences in 3.x (ordering,
string prefixes, and other changes), we ended up with very different
ETags, and these differences weren't even consistent across the various
Python 3.x releases.

Instead of repr(), we're now just simply serializing using
json.dumps(), sorting keys to ensure order. This creates a consistent
representation.

It's worth noting that we're ultimately serializing to JSON twice, and
at the moment there's no getting around that. The response serialization
lives in a different part of the API codebase, and works quite
differently. Down the road, we may want to give more control over ETags
bo the response code that otherwise performs the serialization.

Unit tests pass on Python 2.7, 3.4, 3.5, and 3.6.