perf: Optimize resolving X-Forwarded-For addresses

[dmkng]

Hello, I have a suggestion. When looking up how trustProxy option in Fastify works in detail, I found out that the proxy-addr dependency used by Fastify, specifically the .all() function, checks if the trust argument is provided, and if it's not, then it just returns all adresses, skipping the trust checking logic alltogether, which can speed up things greatly, especially if there's a lot of X-Forwarded-For addresses. So, I made this PR, replacing the return true function with undefined, which has the same effect in the .ips property getter of Request, but because the proxy-addr default function doesn't support skipping trust argument, I replaced it with proxy-addr .all() function in the .ip property getter and did the same thing there as the proxy-addr default function does, so everything behaves exactly the same as before. If something needs further explanation please let me know.

Checklist

• run npm run test and npm run benchmark
• tests and/or benchmarks are included
• documentation is changed or added
• commit message and code follows the Developer's Certification of Origin
  and the Code of conduct

[mcollina]

can you adjust the pr title according to the rules?

[mcollina]

how much perf does this give us?

[dmkng]

can you adjust the pr title according to the rules?

I added perf: at the beggining, is there something else to adjust?

how much perf does this give us?

I made a simple test that sends 11 addresses in X-Forwarded-For to the Fastify server with trustProxy: true and a single endpoint printing .ip and .ips properties of Request, I can add it to the PR or to the separate branch of my fork.

Without the optimization:

[1] Running 30s test @ http://localhost:3000/
[1] 100 connections with 10 pipelining factor
[1]
[1]
[1] ┌─────────┬───────┬───────┬───────┬───────┬──────────┬──────────┬────────┐
[1] │ Stat    │ 2.5%  │ 50%   │ 97.5% │ 99%   │ Avg      │ Stdev    │ Max    │
[1] ├─────────┼───────┼───────┼───────┼───────┼──────────┼──────────┼────────┤
[1] │ Latency │ 23 ms │ 26 ms │ 34 ms │ 83 ms │ 27.37 ms │ 14.94 ms │ 402 ms │
[1] └─────────┴───────┴───────┴───────┴───────┴──────────┴──────────┴────────┘
[1] ┌───────────┬─────────┬─────────┬─────────┬────────┬──────────┬─────────┬─────────┐
[1] │ Stat      │ 1%      │ 2.5%    │ 50%     │ 97.5%  │ Avg      │ Stdev   │ Min     │
[1] ├───────────┼─────────┼─────────┼─────────┼────────┼──────────┼─────────┼─────────┤
[1] │ Req/Sec   │ 26 415  │ 26 415  │ 37 055  │ 37 503 │ 35 877,6 │ 2460,9  │ 26 404  │
[1] ├───────────┼─────────┼─────────┼─────────┼────────┼──────────┼─────────┼─────────┤
[1] │ Bytes/Sec │ 12.7 MB │ 12.7 MB │ 17.8 MB │ 18 MB  │ 17.2 MB  │ 1.18 MB │ 12.7 MB │
[1] └───────────┴─────────┴─────────┴─────────┴────────┴──────────┴─────────┴─────────┘
[1]
[1] Req/Bytes counts sampled once per second.
[1] # of samples: 30
[1]
[1] 1077k requests in 30.05s, 516 MB read

With the optimization:

[1] Running 30s test @ http://localhost:3000/
[1] 100 connections with 10 pipelining factor
[1]
[1]
[1] ┌─────────┬───────┬───────┬───────┬───────┬──────────┬──────────┬────────┐
[1] │ Stat    │ 2.5%  │ 50%   │ 97.5% │ 99%   │ Avg      │ Stdev    │ Max    │
[1] ├─────────┼───────┼───────┼───────┼───────┼──────────┼──────────┼────────┤
[1] │ Latency │ 22 ms │ 25 ms │ 29 ms │ 81 ms │ 26.49 ms │ 14.54 ms │ 261 ms │
[1] └─────────┴───────┴───────┴───────┴───────┴──────────┴──────────┴────────┘
[1] ┌───────────┬─────────┬─────────┬─────────┬────────┬───────────┬────────┬─────────┐
[1] │ Stat      │ 1%      │ 2.5%    │ 50%     │ 97.5%  │ Avg       │ Stdev  │ Min     │
[1] ├───────────┼─────────┼─────────┼─────────┼────────┼───────────┼────────┼─────────┤
[1] │ Req/Sec   │ 29 055  │ 29 055  │ 37 407  │ 37 631 │ 37 050,94 │ 1512,6 │ 29 044  │
[1] ├───────────┼─────────┼─────────┼─────────┼────────┼───────────┼────────┼─────────┤
[1] │ Bytes/Sec │ 13.9 MB │ 13.9 MB │ 17.9 MB │ 18 MB  │ 17.7 MB   │ 723 kB │ 13.9 MB │
[1] └───────────┴─────────┴─────────┴─────────┴────────┴───────────┴────────┴─────────┘
[1]
[1] Req/Bytes counts sampled once per second.
[1] # of samples: 30
[1]
[1] 1112k requests in 30.05s, 531 MB read

[gurgunday]

lgtm

[gurgunday]

Just leaving a note here that this won't slow down the performance if only one IP exists, as proxyAddr itself also calls proxyAddr.all, and returns the last element of the returned array, so it does the same thing

[mcollina]

lgtm