ddn.encoding.varint

ddn.encoding.varint — checked unsigned base-128 varints (LEB128).

Canonical little-endian base-128 encoding over caller-provided buffers: each byte carries 7 value bits, the high bit marks continuation. 32-bit values take 1–5 bytes (5th byte ≤ 0x0F), 64-bit values 1–10 (10th byte ≤ 0x01).

Contract highlights (design record doc/DDN-ENCODING.md):

• Encode checks capacity first and is non-committing:

  on INSUFFICIENT_CAPACITY the output buffer is untouched.

• Decode is strict-canonical: overlong encodings (trailing

  zero final byte where a shorter form exists) and payloads beyond the width are rejected, not silently accepted.

• Errors are VarintError values in a ddn.result.Result:

  bounded context (the byte count examined when the failure was detected), no allocation, no throwing — @nogc nothrow throughout. TRUNCATED and INSUFFICIENT_CAPACITY are retryable; OVERLONG and OVERFLOW mean the data is wrong.

Signed integers: compose ZigZag with these functions (ddn.encoding.zigzag), the sint32/sint64 family — see the N009 notes in the design record.

Example:

import ddn.encoding.varint;

ubyte[10] buf;
auto r = varint.encodeUint64(300, buf[]);
assert(r.ok && r.value == 2 && buf[0 .. 2] == [0xAC, 0x02]);
auto d = varint.decodeUint64(buf[0 .. 2]);
assert(d.ok && d.value.value == 300 && d.value.length == 2);