Python dependencies are hard. Especially when the package you're publishing is sort of a weird hybrid between "library" (where one should not pin things!) and an end-product (where reproducibility is very important, and therefore one should definitely pin things).
This change doesn't actually change any of the resolved dependencies (as of today). But it attempts to organize things into logical groups:
Dependencies that define compatible operating environments (e.g. mkdocs and markdown)
Dependencies that we reflect back out as features of our own package (e.g. mdx sane lists)
Dependencies that are like category 1, except they're only in place temporarily to work around an underlying python dependency issue. (We no longer have these, but we were carrying around a couple for a long time see #91 & #78)
So that we can more easily reason about dependency updates from renovate.
What / Why
Python dependencies are hard. Especially when the package you're publishing is sort of a weird hybrid between "library" (where one should not pin things!) and an end-product (where reproducibility is very important, and therefore one should definitely pin things).
This change doesn't actually change any of the resolved dependencies (as of today). But it attempts to organize things into logical groups:
So that we can more easily reason about dependency updates from
renovate
.