first commit

This commit is contained in:
2026-05-29 08:20:56 +01:00
commit 949299f58d
6368 changed files with 527176 additions and 0 deletions

View File

@@ -0,0 +1,78 @@
# Run-length encoding (`compressRle` / `decompressRle`)
<!-- begin:auto-docs -->
_(no submodules)_
<!-- end:auto-docs --> - rle.schema — Zod schema for the per-call `Compressi…
- [rle.schema](./rle.schema.md) — Zod schema for the per-call `CompressionOptions` accepted by `compressRle` / `decompressRle`. The RLE codec is byte-onl…
- rle.schema — Zod schema for the per-call `Compressi…
- [rle.schema](./rle.schema.md) — Zod schema for the per-call `CompressionOptions` accepted by `compressRle` / `decompressRle`. The RLE codec is byte-onl…
- rle.schema — Zod schema for the per-call `Compressi…
- [rle.schema](./rle.schema.md) — Zod schema for the per-call `CompressionOptions` accepted by `compressRle` / `decompressRle`. The RLE codec is byte-onl…
- rle.schema — Zod schema for the per-call `Compressi…
- [rle.schema](./rle.schema.md) — Zod schema for the per-call `CompressionOptions` accepted by `compressRle` / `decompressRle`. The RLE codec is byte-onl…
- [rle.schema](./rle.schema.md) — Zod schema for the per-call `CompressionOptions` accepted by `compressRle` / `decompressRle`. The RLE codec is byte-onl…
## Purpose
`compressRle` / `decompressRle` are byte-level run-length encoding helpers shipped as a separate module within `@suaveplan/bitpacker-compressed` so embedded / no-streams environments can import them without pulling in `CompressionStream`-using gzip/deflate adapters or the `CompressedBitPacker` class. They are not part of the `CompressionRegistry` framing — bytes produced by `compressRle` are NOT the same shape as bytes produced by the registry's `"none"`/`"gzip"`/`"deflate"` adapters, which include the `CompressedBitPacker` header. If you need framed RLE, write a thin `RleCompressionAdapter` that internally calls these helpers.
## Features
- Bytes-only fast path; no streams required.
- Single-byte runs longer than 3 are emitted as `[byte][0][count]`; everything else passes through verbatim.
- Symmetric round-trip with `decompressRle`.
- Zero allocation overhead beyond the output `Uint8Array`.
## Usage: Basic Example
```ts
import { compressRle, decompressRle } from "@suaveplan/bitpacker-compressed";
const data = new Uint8Array([1, 1, 1, 1, 1, 2, 3]);
const compressed = compressRle(data); // [1, 0, 5, 2, 3]
const restored = decompressRle(compressed); // back to original
```
## Usage: Advanced Example
```ts
import {
compressRle,
decompressRle,
type CompressionAdapter,
} from "@suaveplan/bitpacker-compressed";
// Wrap RLE in a CompressionAdapter shape so it plugs into the registry.
class RleAdapter implements CompressionAdapter {
readonly name = "rle";
async compress(data: Uint8Array) { return compressRle(data); }
async decompress(data: Uint8Array) { return decompressRle(data); }
}
```
## API Reference
### `compressRle(data: Uint8Array): Uint8Array`
- Encodes runs of >3 identical bytes as `[byte][0][count]`.
- Returns a fresh `Uint8Array`; input is not mutated.
### `decompressRle(data: Uint8Array): Uint8Array`
- Reverses `compressRle`. Looks ahead for `[0][count]` markers after each byte.
- Returns a fresh `Uint8Array`.
### `CompressionOptions`
Type alias for `{ readonly level?: number }` — accepted but ignored by the bytes-only RLE path; included so adapter authors can re-export the canonical compression-options shape.
## Implementation Notes
- The marker byte is `0`, which would collide with a literal `0` in input data; the encoder handles this by always emitting the run-length form when `byte === 0` runs.
- This is not the most aggressive RLE possible — it does not pack runs of distinct bytes into literal blocks. The intent is "smaller than verbatim with a single linear pass", not "competitive with gzip on text".
- Use this when you have known byte-pattern data (e.g. graphics with large flat regions, sparse network frames). For general-purpose data, use the gzip / deflate adapters.