Routing of URL paths for nextstrain.org is a bit complicated at the moment. There are four primary sources of valid paths:
Server-side routes in the Express app (
src/app.js), which often respond with dynamic content.
Auspice client-side routes, which are activated when
auspice-client/dist/index.htmlis served. Auspice changes the URL path when the dataset or narrative changes.
Gatsby client-side routes, which are activated when any Gatsby HTML page is served from
static-site/public/. Gatsby changes the URL path when links between its pages are clicked and on page load to canonicalize the URL for a given page.
Server-side static HTML tree generated by Gatsby (
static-site/public/), primarily for assets and other behind-the-scenes resources Gatsby pages need.
Notably, these URL routers are all separate and largely do not know of each other. This means there can be some unexpected behaviour if routes clash, particularly between the server-side and client-side routes.
The server-side routes try to adhere to some basic organizational principles:
/charon/getDataset) go in
src/app.js. The longer we can keep all (server-side) routes in one place, the better.
Endpoints (“route handlers”, e.g.
charon.getDataset(req, res, next)) live in
src/endpoints/, grouped into arbitrary directories/files/modules as seems best for development.
The names in route paths and the endpoint source file paths do not necessarily need to match, although they may (as in the case of Charon). Routes are user-facing, endpoint source file paths are developer-facing.
There does not need to be a 1:1 correspondence between a single route and a single endpoint source file, although there may be (as in the case of Charon). For example, a single file
src/endpoints/groups.jsmay define several endpoints used for several routes under