Files
nushell/crates/nu-command/src/formats/to/msgpackz.rs
Devyn Cairns 28e33587d9 msgpackz: increase default compression level (#13035)
# Description
Increase default compression level for brotli on msgpackz to 3. This has
the best compression time generally. Level 0 and 1 give weird results
and sometimes cause extremely inflated outputs rather than being
compressed. So far this hasn't really been a problem for the plugin
registry file, but has been for other data.

The `$example` is the web-app example from https://json.org/example.html

Benchmarked with:

```nushell
seq 0 11 | each { |level|
  let compressed = ($example | to msgpackz --quality $level)
  let time = (timeit { $example | to msgpackz --quality $level })
  {
    level: $level
    time: $time
    length: ($compressed | bytes length)
    ratio: (($uncompressed_length | into float) / ($compressed | bytes length))
  }
}
```

```
╭────┬───────┬─────────────────┬────────┬───────╮
│  # │ level │      time       │ length │ ratio │
├────┼───────┼─────────────────┼────────┼───────┤
│  0 │     0 │ 4ms 611µs 875ns │   3333 │  0.72 │
│  1 │     1 │ 1ms 334µs 500ns │   3333 │  0.72 │
│  2 │     2 │     190µs 333ns │   1185 │  2.02 │
│  3 │     3 │      184µs 42ns │   1128 │  2.12 │
│  4 │     4 │      245µs 83ns │   1098 │  2.18 │
│  5 │     5 │     265µs 584ns │   1040 │  2.30 │
│  6 │     6 │     270µs 792ns │   1040 │  2.30 │
│  7 │     7 │     444µs 708ns │   1040 │  2.30 │
│  8 │     8 │       1ms 801µs │   1040 │  2.30 │
│  9 │     9 │     843µs 875ns │   1037 │  2.31 │
│ 10 │    10 │ 4ms 128µs 375ns │    984 │  2.43 │
│ 11 │    11 │ 6ms 352µs 834ns │    986 │  2.43 │
╰────┴───────┴─────────────────┴────────┴───────╯
```

cc @maxim-uvarov
2024-06-04 17:19:10 -07:00

89 lines
2.5 KiB
Rust

use std::io::Write;
use nu_engine::command_prelude::*;
use super::msgpack::write_value;
const BUFFER_SIZE: usize = 65536;
const DEFAULT_QUALITY: u32 = 3; // 1 can be very bad
const DEFAULT_WINDOW_SIZE: u32 = 20;
#[derive(Clone)]
pub struct ToMsgpackz;
impl Command for ToMsgpackz {
fn name(&self) -> &str {
"to msgpackz"
}
fn signature(&self) -> Signature {
Signature::build(self.name())
.input_output_type(Type::Any, Type::Binary)
.named(
"quality",
SyntaxShape::Int,
"Quality of brotli compression (default 3)",
Some('q'),
)
.named(
"window-size",
SyntaxShape::Int,
"Window size for brotli compression (default 20)",
Some('w'),
)
.category(Category::Formats)
}
fn usage(&self) -> &str {
"Convert Nu values into brotli-compressed MessagePack."
}
fn extra_usage(&self) -> &str {
"This is the format used by the plugin registry file ($nu.plugin-path)."
}
fn run(
&self,
engine_state: &EngineState,
stack: &mut Stack,
call: &Call,
input: PipelineData,
) -> Result<PipelineData, ShellError> {
fn to_u32(n: Spanned<i64>) -> Result<Spanned<u32>, ShellError> {
u32::try_from(n.item)
.map_err(|err| ShellError::CantConvert {
to_type: "u32".into(),
from_type: "int".into(),
span: n.span,
help: Some(err.to_string()),
})
.map(|o| o.into_spanned(n.span))
}
let quality = call
.get_flag(engine_state, stack, "quality")?
.map(to_u32)
.transpose()?;
let window_size = call
.get_flag(engine_state, stack, "window-size")?
.map(to_u32)
.transpose()?;
let value_span = input.span().unwrap_or(call.head);
let value = input.into_value(value_span)?;
let mut out_buf = vec![];
let mut out = brotli::CompressorWriter::new(
&mut out_buf,
BUFFER_SIZE,
quality.map(|q| q.item).unwrap_or(DEFAULT_QUALITY),
window_size.map(|w| w.item).unwrap_or(DEFAULT_WINDOW_SIZE),
);
write_value(&mut out, &value, 0)?;
out.flush().err_span(call.head)?;
drop(out);
Ok(Value::binary(out_buf, call.head).into_pipeline_data())
}
}