nextstrain/seasonal-flu

Create dated copies of 2y builds

huddlej opened this issue · 0 comments

Context

To enable SARS-CoV-2 narratives based on data at specific timepoints, we've maintained dated Auspice JSONs that are automatically generated as part of the standard Nextstrain workflow. These dated builds follow the naming format of auspice/ncov_{build_name}_{date}.json where the date is in YYYY-MM-DD format.

Description

To support similar narrative reports on the status of influenza, we should create dated copies of builds for the four lineages. SInce our reporting usually only needs recent context, we can start by creating these copies only for the 2y builds of HA and NA. With tree/meta, tip frequencies, and root sequence JSONs, this approach should produce 24 JSONs per day (4 lineages * 2 segments * 3 JSONs per build).

Possible solution

Following the ncov workflow example, the resulting dated JSONs would be named like:

flu_seasonal_h3n2_ha_2y_2021-04-20.json

And would produce a URL like https://nextstrain.org/flu/seasonal/h3n2/ha/2y/2021-04-20.

An alternate solution that would be more consistent with the hierarchical grouping standard of HTTP APIs would place the date earlier in the name such that all builds from that date could be displayed below. An example dated JSON would look like:

flu_seasonal_2021-04-20_h3n2_ha_2y.json

And the URL would look like https://nextstrain.org/flu/seasonal/2021-04-20/h3n2/ha/2y. When we eventually support browsing builds by exploring the URL's API, this hierarchical layout would enable a page at https://nextstrain.org/flu/seasonal/2021-04-20/ that could show all builds from that day.

Open question: Should we produce these dated builds for all trees on the private site, too, or only the public site?