Migrating from ruamel.yaml to PyYAML before 1.0 release

[hanxiao]

Describe the feature

For histororical reason, ruamel.yaml was used, it was one of the best YAML python parser by the time 2017. Unfortunately, I recently have growing concerns over its long-term support and may affect Jina:

• The author of ruamel.yaml moved the project codebase from Github to Gitlab and now to SourceForge. Making it challenge to track the issues and discuss the problems.
• So far ruamel.yaml does not seem to support Python 3.9, but Jina has 3.9 support in the roadmap (add Python 3.9 support #1253). PyYAML already add 3.9 wheels and working on Python 3.10 nightly!
• Lower usage compare to PyYAML : daily download at 111,561 vs. PyYAML daily download 1,278,847

In general, it is not about popularity, performance or technology. Yaml parsing is one of the five core depencies to Jina. Extremely extremely important, we need a project that is sustainable enough in the long run, with a robust and open govenenance mode. PyYAML seems to be better on that direction.

Your proposal

Replace ruamel.yaml to PyYAML before 1.0. It must be done before 1.0. I will lead on this.

---

Environment

Screenshots

[hanxiao]

Compatability issues:

• ruamel.yaml support list as key but PyYAML does not.
  PyYaml can't parse mappings with lists as keys yaml/pyyaml#339
• ruamel.yaml interprests on as string, while PyYAML interprets it as bool: True
on is loaded as True instead of a string yaml/pyyaml#376