Strategy & Quality Levels

There is no CLevel enum — quality levels are plain u32.

Levels

Quality levels are u32 in range 0 to 11, with default 11:

pub const MIN_QUALITY: u32 = 0;
pub const MAX_QUALITY: u32 = 11;
pub const DEFAULT_QUALITY: u32 = 11;

pub fn versionString() []const u8 // "0.0.2"
pub fn versionNumber() u32        // 2

Use with brotli.compressWithOptions or StreamingCompressor:

// Numeric quality levels
const c1 = try brotli.compressWithOptions(allocator, data, .{ .quality = 1 });  // fastest
const c5 = try brotli.compressWithOptions(allocator, data, .{ .quality = 5 });  // balanced
const c9 = try brotli.compressWithOptions(allocator, data, .{ .quality = 9 });  // good ratio
const c11 = try brotli.compressWithOptions(allocator, data, .{});               // best (default)

// Via streaming
var sc = brotli.StreamingCompressor.init(allocator, .{ .quality = 9 });
defer sc.deinit();

Quality trade-off (measured on pseudo-text, 512 KiB):

Quality	Ratio	Character
0–1	~21%	Fastest; greedy matching
5	~18.5%	Balanced; lazy matching
9	~17.5%	Deep hash-chain search
11	~18%	Best analysis (context modeling heuristics)

Window Size

Backward-reference distances are bounded by lgwin (10..24):

const opts = brotli.CompressionOptions{
    .quality = 11,
    .lgwin = 22, // default; max distance = (1 << 22) - 16
};

Larger windows find more distant matches at a small memory cost. Use large_window = true to accept values up to 30 for decoding interop.

Mode

Mode selects text-analysis hints (mirrors BROTLI_MODE_*):

pub const Mode = enum(u3) {
    generic = 0,
    text = 1,
    font = 2,
};

Set via options:

const opts = brotli.CompressionOptions{
    .quality = 11,
    .mode = .text,
    .size_hint = input.len,
};