gleamunison/sap

Schema-Aligned Parsing (gap analysis concept B).

Liberal-in-what-you-accept repair of near-miss model output: fences, yapping, comments, trailing commas, single quotes, unquoted keys, Python literals — then strict decode, then validation against a JSON-native Schema (bridged from core ast.Type). Pure Gleam, no FFI.

Types

SapError

</>

pub type SapError {
  NoJsonFound
  StrictError(reason: String)
  ValidationFailed(paths: List(String))
}

Constructors

```
NoJsonFound
```
```
StrictError(reason: String)
```
```
ValidationFailed(paths: List(String))
```

Schema

</>

pub type Schema {
  SAny
  SInt
  SFloat
  SText
  SBool
  SList(of: Schema)
  SObj(fields: List(#(String, Schema)), required: List(String))
  SEnum(values: List(String))
}

Constructors

```
SAny
```
```
SInt
```
```
SFloat
```
```
SText
```
```
SBool
```
```
SList(of: Schema)
```

SObj(fields: List(#(String, Schema)), required: List(String))

```
SEnum(values: List(String))
```

Values

decode_as

</>

pub fn decode_as(
  text: String,
  schema: Schema,
) -> Result(dynamic.Dynamic, SapError)

Repair, decode, and validate against a schema.

parse

</>

pub fn parse(text: String) -> Result(String, SapError)

Repair near-miss model output into strict JSON text.

schema_from_type

</>

pub fn schema_from_type(t: ast.Type) -> Schema

Bridge from core types; unresolved constructors degrade to SAny.