perf(route): 5% faster routing

[nigrosimone]

What's new

• Avoid unnecessary allocations and function calls
• Use a non-async next for most cases

Improvements

• ~20% less memory usage
• ~5% faster routing

Benchmark with 1000 middleware

benchmark\route.js

// run benchmark with: node benchmark\route.js --iterations=50000 --expose-gc

"use strict";

const { run, bench, group } = require("mitata");

const expressPr = require("../src/index");
const expressMaster = require("ultimate-express");

const appPr = expressPr();
const appMaster = expressMaster();

appPr.set("etag", false);
appPr.set("declarative responses", false);

appMaster.set("etag", false);
appMaster.set("declarative responses", false);

for (let index = 0; index < 1000; index++) {
  appPr.use(async(req, res, next) => next());
  appMaster.use(async(req, res, next) => next());
}

appPr.get("/", async (req, res) => res.send());
appMaster.get("/", async (req, res) => res.send());

appPr.listen(3001, async () => {
  appMaster.listen(3002, async () => {
    group("Express route benchmark", () => {
      bench("PR", () => fetch("http://localhost:3001/", { method: "HEAD", keepalive: true }));
      if (global.gc) global.gc();
      bench("MASTER", () => fetch("http://localhost:3002/", { method: "HEAD", keepalive: true }));
    });

    await run();

    process.exit(0);
  });
});

Pipeline benchmark

Side note: this benchmark has only 1 route, no benefit from my changes

Master

| Metric             | Ultimate Express   | Express         | Difference      |
|--------------------|--------------------|-----------------|-----------------|
| Requests/sec       | 73253.96           | 8004.12         | 9.15x faster    |
| Avg Latency        | 2.73            ms | 34.34        ms | 12.57x faster   |
| Transfer/sec       | 14.39           MB | 1.82         MB | 7.90x faster    |

PR

| Metric             | Ultimate Express   | Express         | Difference      |
|--------------------|--------------------|-----------------|-----------------|
| Requests/sec       | 73314.74           | 8019.90         | 9.14x faster    |
| Avg Latency        | 2.73            ms | 36.01        ms | 13.19x faster   |
| Transfer/sec       | 14.40           MB | 1.82         MB | 7.91x faster    |

[dimdenGD]

This does seems a little bit faster, but code quality becomes way worse, so I'm not sure if to merge or not... You should try improving the code somehow (more comments, less repetition, fix formatting)

[nigrosimone]

No problem, as soon as I have some free time I fix it

[nigrosimone]

@dimdenGD I fixed the formatting so that only the logic changes are more obvious now.

Do you think this is OK?

Updated bench

[dimdenGD]

Still weird format:

Also code repeating twice:

If possible without performance degradation, need to make a function

[dimdenGD]

What about declaring it with function instead

[dimdenGD]

You reformatted entire file so now its impossible to see changes

[nigrosimone]

@dimdenGD no way (without loose performance!) to make it a class function :(

I just added a comment to remind you to keep duplicated code parts the same 00b3a18

feel free to close the pr if you don't like the changes