> ## Documentation Index
> Fetch the complete documentation index at: https://docs.parmanasystems.com/llms.txt
> Use this file to discover all available pages before exploring further.

# parmana.crypto.canonical

Canonical Serializer (Python reimplementation).

Reproduces packages/crypto/src/CanonicalSerializer.ts byte-for-byte,
for already-JSON-decoded input, as the independent-language proof
requested by the PQC audit's Layer 5 ("deterministic serialization ...
Node.js, plus at least one other language").

This module deliberately does NOT reconstruct Date objects from ISO
strings before re-serializing: an Execution Trust Record read from a
JSON file (or an API response) already has every timestamp as a plain
string -- CanonicalSerializer.ts's own Date-handling branch only
matters on the TypeScript signing side, which starts from real Date
objects. Re-serializing already-JSON-decoded primitives sidesteps the
much harder problem of reproducing `Date.prototype.toISOString()`
exactly and is sufficient for verification, since the signer already
turned every Date into its canonical string form before ever hashing
or signing it.

Two concrete cross-language pitfalls this module corrects for --
found empirically while writing it, not assumed:

* JSON.stringify's default separators are compact (no space after
  "," or ":"); Python's json.dumps defaults to ", " and ": ". Left
  uncorrected, every object/array would serialize to different bytes
  in Python than in Node, and no signature would ever verify.
* JSON.stringify does not escape non-ASCII characters by default;
  Python's json.dumps defaults to ensure\_ascii=True, which \uXXXX-
  escapes them. Same effect: different bytes for any string containing
  non-ASCII content.

Known, narrow limitation: key sorting uses Python's default string
ordering (Unicode code point order), while CanonicalSerializer.ts
sorts by JavaScript's default string ordering (UTF-16 code unit
order). These are identical for every key actually used in this
codebase's real payloads (all ASCII identifiers) and for the entire
Basic Multilingual Plane; they diverge only for object keys containing
astral-plane Unicode characters (code points above U+FFFF), which
does not occur anywhere in this system's real payloads today.

#### canonical\_serialize

```python theme={null}
def canonical_serialize(value: Any) -> bytes
```

Serializes an already-JSON-decoded value into the same canonical
UTF-8 bytes CanonicalSerializer.ts would produce for the
equivalent JavaScript value.