Open sanmai-NL opened 1 year ago
Regardless of the practices by its maintainer, ruamel-yaml
is still functionally better than PyYAML.
Would it be possible to adopt https://github.com/eyeseast/python-frontmatter? This is more feature complete than the self-developed source code.
You'd still transitively depend on ruamel-yaml
, I surmise, but at least the choice of YAML library will no longer be your concern.
Sorry for not getting back to you.. I'll have a look shortly
My original issue with Pyyaml were some issues around ordering and duplicate keys. It's been 2-3 years since I had that issue (with a different project).
I think based on the requirements for #22 I'll need to do some testing.
The fundamental challenge I have here is that we lose:
Effectively: the nature of the opinionation changes). I'd look for feedback as to whether that is something of interest.
@butler54 I assume you're referring to PyYAML
, rather than python-frontmatter
? From an outside point of view, adopting the latter seems the easiest migration path. And the matter of switching from ruamel-yaml
to PyYAML
can be considered within the scope of python-frontmatter
.
Unfortunately python-frontmatter
's api today would not deliver consistent results due to behaviour differences in the yaml parses (e.g. pyyaml will enforce alphabetical ordering of keys when writing out which is inconsistent with current behaviour).
That said - the benefits of being able to support yaml/toml/json outweigh that change. Will most likely build it as a major release shorlty.
This dependency is highly idiosyncratic. It still uses
setup.py
and seems to generate one, indeed. It also generates source code within a Dockerfile. It uses custom C source code. The sole maintainer hosts its source code on SourceForge, an outdated platform. He categorizes these packages on PyPI as beta quality.Can you not use the more popular PyYAML?