Document workaround for slow HEAD

[Pyrolistical]

When this library loads a database file, it does a HEAD to get the total length of the database file. On some dev servers the etag is calculated, which in turns reads the entire database file. For large databases this is very slow, on the order of seconds.

I found this to be the case when using vite on windows. I wrote a small vite plugin which intercepts these HEAD calls and just returns Content-Length

import fs from "node:fs";

const SkipHeadEtag = () => ({
  name: "skip-head-etag",
  configureServer(server) {
    server.middlewares.use((req, res, next) => {
      if (req.method === "HEAD") {
        const { size } = fs.statSync(req.url.slice(1));
        res.setHeader("Accept-Ranges", "bytes");
        res.setHeader("Cache-Control", "no-cache");
        res.setHeader("Content-Length", String(size));
        res.end();
        return;
      }
      next();
    });
  },
});

That plugin is incomplete. It doesn't handle 404 properly, but shows the main idea of the workaround.