# `Predicator.Conformance.JSON`
[🔗](https://github.com/riddler/predicator-ex/blob/v9.4.1/lib/predicator/conformance/json.ex#L1)

A deterministic JSON writer for the conformance corpus (`px-35i.4`).

The built-in `JSON.encode!/1` does not guarantee map key order, which makes
its output unusable for a byte-compared, checked-in corpus: two semantically
identical maps built by inserting keys in a different order could encode
differently, turning the staleness test red for no reason. `encode_canonical/1`
fixes that by walking the term itself, sorting object keys by codepoint, and
emitting no incidental whitespace - `JSON.encode!/1` is still used underneath
for scalar and string escaping, so the escaping rules stay the standard
library's.

`encode_lines/1` is the shipped corpus format: a list of objects, one per
line, so that changing a single case is a one-line diff instead of a
whole-file rewrite.

# `json_value`

```elixir
@type json_value() ::
  nil
  | boolean()
  | number()
  | binary()
  | [json_value()]
  | %{optional(binary() | atom()) => json_value()}
```

A term encodable as JSON: what `JSON.decode/1` can produce, plus atom map keys.

# `encode_canonical`

```elixir
@spec encode_canonical(json_value()) :: binary()
```

Encodes `term` as canonical JSON: object keys sorted by codepoint, no
incidental whitespace.

## Examples

    iex> Predicator.Conformance.JSON.encode_canonical(%{"b" => 1, "a" => 2})
    ~s({"a":2,"b":1})

    iex> Predicator.Conformance.JSON.encode_canonical([1, "two", true, nil])
    ~s([1,"two",true,null])

# `encode_lines`

```elixir
@spec encode_lines([json_value()]) :: binary()
```

Encodes a list of terms as the shipped corpus format: one canonical-JSON
object per line, each terminated by `\n`.

## Examples

    iex> Predicator.Conformance.JSON.encode_lines([%{"id" => 1}, %{"id" => 2}])
    ~s({"id":1}\n{"id":2}\n)

---

*Consult [api-reference.md](api-reference.md) for complete listing*