Automate version bumping for Auspice

[tsibley]

The recipe for Auspice won't get automatic version bump PRs from Bioconda's "autobump" tooling because the tooling doesn't currently support NPM package sources.

We can add support by writing a class Npm(Hoster) for https://github.com/bioconda/bioconda-utils/blob/master/bioconda_utils/hosters.py and adding test cases to https://github.com/bioconda/bioconda-utils/blob/master/test/hoster_cases.yaml.

[victorlin]

Another option would be to automate this ourselves instead of relying on Bioconda's autobump tooling.

There could be a workflow in this fork repo or nextstrain/.github that can be called with parameters for feed (e.g. PyPI), feed package name (e.g. nextstrain-augur, and Bioconda recipe name (e.g. augur). This can be used by augur/release.yaml, etc.

Pros:

• Immediate auto-bump PRs instead of waiting on Bioconda's autobump schedule.
• Would not be impacted by Bioconda autobump delays/prolonged downtime (e.g. BiocondaBot auto-updater not creating PRs bioconda/bioconda-recipes#35414)

We'd still need to manually add dependency changes.

[tsibley]

I'm not necessarily against doing this ourselves—it might be the right option!—but considering this critically…

Is waiting for Bioconda's autobump schedule really a big deal? They seem to happen pretty quickly most of the time. Do we need same day turnaround on releases on all distribution channels? It seems to me that we don't really. If we implement this autobump ourselves outside of the Bioconda framework, then it's choosing to take on another operational thing to maintain. Instead of being impacted by Bioconda's autobump delays/downtime, we'll be impacted by our own. 🙃 One nice part of extending Bioconda's own framework for autoupdates is that we wouldn't be taking on all the operational work of it, just the initial implementation. (Plus, it benefits the rest of the Bioconda community by adding Node support.)

[corneliusroemer]

It is quite trivial to write a script to submit a PR, see Nextclade's way of doing things. It can be adapted quote easily I think.

Doesn't mean it wouldn't be good to contribute to the autobumper, but first we should check that they actually would consider it. They are quite busy with many potentially more important things like finally supporting Python 3.11 or keeping CI up - might not have bandwidth to extend autobumper.

[victorlin]

Is waiting for Bioconda's autobump schedule really a big deal? They seem to happen pretty quickly most of the time. Do we need same day turnaround on releases on all distribution channels?

It's a bigger deal now 2 years later with the addition of Nextstrain's Conda runtime, meaning the full release process for our tools includes waiting (1) for the Bioconda PR to be created and (2) for it to be manually merged. More discussion in nextstrain/conda-base#2