Optimize URI.to_string/1

[preciz]

Assisted by: Antigravity CLI & Gemini 3.5 Flash (hand rolled in the end, don't blame the LLM 😄 )

What this does:

• Avoids the map update.
• Destructures the URI once.
• Defers a condition to extract_authority.

It is a small performance win in most cases and it uses less memory in almost all cases.

Bench:

Mix.install([
  {:benchee, "~> 1.0"}
])

defmodule OldURI do
  def to_string(%{scheme: scheme, port: port, path: path, query: query, fragment: fragment} = uri) do
    uri =
      case scheme && URI.default_port(scheme) do
        ^port -> %{uri | port: nil}
        _ -> uri
      end

    authority = extract_authority(uri)

    IO.iodata_to_binary([
      if(scheme, do: [scheme, ?:], else: []),
      if(authority, do: ["//" | authority], else: []),
      if(path, do: path, else: []),
      if(query, do: ["?" | query], else: []),
      if(fragment, do: ["#" | fragment], else: [])
    ])
  end

  defp extract_authority(%{host: nil, authority: authority}) do
    authority
  end

  defp extract_authority(%{host: host, userinfo: userinfo, port: port}) do
    [
      if(userinfo, do: [userinfo | "@"], else: []),
      if(String.contains?(host, ":"), do: ["[", host | "]"], else: host),
      if(port, do: [":" | Integer.to_string(port)], else: [])
    ]
  end
end

defmodule NewURI do
  def to_string(%{
        host: host,
        authority: authority,
        userinfo: userinfo,
        scheme: scheme,
        port: port,
        path: path,
        query: query,
        fragment: fragment
      }) do
    # Based on https://tools.ietf.org/html/rfc3986#section-5.3
    authority = extract_authority(host, authority, scheme, userinfo, port)

    IO.iodata_to_binary([
      if(scheme, do: [scheme, ?:], else: []),
      if(authority, do: ["//" | authority], else: []),
      if(path, do: path, else: []),
      if(query, do: ["?" | query], else: []),
      if(fragment, do: ["#" | fragment], else: [])
    ])
  end

  defp extract_authority(_host = nil, authority, _scheme, _userinfo, _port) do
    authority
  end

  defp extract_authority(host, _authority, scheme, userinfo, port) do
    [
      if(userinfo, do: [userinfo | "@"], else: []),
      if(String.contains?(host, ":"), do: ["[", host | "]"], else: host),
      if(port && (is_nil(scheme) || URI.default_port(scheme) != port),
        do: [":" | Integer.to_string(port)],
        else: []
      )
    ]
  end
end

inputs = %{
  "Simple URI" => URI.parse("https://elixir-lang.org/getting-started"),
  "URI with default port" => URI.parse("https://elixir-lang.org:443/getting-started"),
  "URI with non-default port" => URI.parse("http://elixir-lang.org:8080/getting-started"),
  "URI with userinfo" => URI.parse("postgres://user:pass@db.example.com:5432/mydb"),
  "URI with IPv6 host" => URI.parse("http://[2001:db8::1]:80/index.html"),
  "URI with query & fragment" => URI.parse("https://google.com/search?q=elixir#result")
}

Benchee.run(
  %{
    "Main (Old) to_string/1" => fn uri -> OldURI.to_string(uri) end,
    "Latest (Optimized) to_string/1" => fn uri -> NewURI.to_string(uri) end
  },
  inputs: inputs,
  time: 2,
  warmup: 1,
  memory_time: 2
)

Benchmark results on my noisy heat throttling mini PC:

Operating System: Linux
CPU Information: AMD Ryzen 7 8845HS w
Number of Available Cores: 16
Available memory: 54.72 GB
Elixir 1.20.0
Erlang 29.0.1
JIT enabled: true

Benchmark suite executing with the following configuration:
warmup: 1 s
time: 2 s
memory time: 2 s
reduction time: 0 ns
parallel: 1
Estimated total run time: 1 min
Excluding outliers: false

##### With input Simple URI #####
Name                                     ips        average  deviation         median         99th %
Latest (Optimized) to_string/1        2.79 M      357.93 ns  ±1897.20%         321 ns         551 ns
Main (Old) to_string/1                2.58 M      387.04 ns  ±1376.56%         351 ns         581 ns

Comparison:
Latest (Optimized) to_string/1        2.79 M
Main (Old) to_string/1                2.58 M - 1.08x slower +29.11 ns

Memory usage statistics:

Name                              Memory usage
Latest (Optimized) to_string/1           312 B
Main (Old) to_string/1                   408 B - 1.31x memory usage +96 B

**All measurements for memory usage were the same**

##### With input URI with IPv6 host #####
Name                                     ips        average  deviation         median         99th %
Latest (Optimized) to_string/1        2.66 M      376.57 ns   ±980.21%         360 ns         591 ns
Main (Old) to_string/1                2.46 M      406.81 ns   ±951.47%         381 ns         611 ns

Comparison:
Latest (Optimized) to_string/1        2.66 M
Main (Old) to_string/1                2.46 M - 1.08x slower +30.24 ns

Memory usage statistics:

Name                              Memory usage
Latest (Optimized) to_string/1           360 B
Main (Old) to_string/1                   456 B - 1.27x memory usage +96 B

**All measurements for memory usage were the same**

##### With input URI with default port #####
Name                                     ips        average  deviation         median         99th %
Latest (Optimized) to_string/1        2.76 M      361.70 ns  ±1870.83%         331 ns         531 ns
Main (Old) to_string/1                2.56 M      390.72 ns  ±1392.85%         360 ns         591 ns

Comparison:
Latest (Optimized) to_string/1        2.76 M
Main (Old) to_string/1                2.56 M - 1.08x slower +29.02 ns

Memory usage statistics:

Name                              Memory usage
Latest (Optimized) to_string/1           312 B
Main (Old) to_string/1                   408 B - 1.31x memory usage +96 B

**All measurements for memory usage were the same**

##### With input URI with non-default port #####
Name                                     ips        average  deviation         median         99th %
Latest (Optimized) to_string/1        2.46 M      406.52 ns  ±1425.70%         371 ns         601 ns
Main (Old) to_string/1                2.36 M      423.44 ns  ±1725.83%         381 ns         631 ns

Comparison:
Latest (Optimized) to_string/1        2.46 M
Main (Old) to_string/1                2.36 M - 1.04x slower +16.92 ns

Memory usage statistics:

Name                              Memory usage
Latest (Optimized) to_string/1           376 B
Main (Old) to_string/1                   376 B - 1.00x memory usage +0 B

**All measurements for memory usage were the same**

##### With input URI with query & fragment #####
Name                                     ips        average  deviation         median         99th %
Latest (Optimized) to_string/1        2.50 M      399.44 ns  ±1470.98%         371 ns         602 ns
Main (Old) to_string/1                2.46 M      406.96 ns  ±1131.66%         371 ns         621 ns

Comparison:
Latest (Optimized) to_string/1        2.50 M
Main (Old) to_string/1                2.46 M - 1.02x slower +7.52 ns

Memory usage statistics:

Name                              Memory usage
Latest (Optimized) to_string/1           352 B
Main (Old) to_string/1                   448 B - 1.27x memory usage +96 B

**All measurements for memory usage were the same**

##### With input URI with userinfo #####
Name                                     ips        average  deviation         median         99th %
Main (Old) to_string/1                2.82 M      354.24 ns   ±805.43%         341 ns         491 ns
Latest (Optimized) to_string/1        2.80 M      357.48 ns   ±743.11%         341 ns         511 ns

Comparison:
Main (Old) to_string/1                2.82 M
Latest (Optimized) to_string/1        2.80 M - 1.01x slower +3.24 ns

Memory usage statistics:

Name                              Memory usage
Main (Old) to_string/1                   304 B
Latest (Optimized) to_string/1           304 B - 1.00x memory usage +0 B

**All measurements for memory usage were the same**

[josevalim]

💚 💙 💜 💛 ❤️