# schemas/link-fields

## Purpose

Zod boundary schema asserting the structural shape of a `LinkFieldMap`. Used
to validate the optional `linkFields` argument passed to
`createBetterAuthSurrealAdapter` and to gate any callers that hand-roll their
own override map at runtime. Validation rejects non-string targets, empty
keys, and identifiers that would break the generated SurrealQL identifier
rule (`/^[a-zA-Z_][a-zA-Z0-9_]*$/`).

## Features

- Identifier-rule enforcement on model name, field name, and target table.
- Empty-string rejection (model/field/target must be `min(1)`).
- Composable: re-exported as `LinkFieldMapSchema` from `./schemas` and
  embedded inside `BetterAuthSurrealAdapterOptionsSchema`.
- Type-level `LinkFieldMapSchemaInput` / `LinkFieldMapSchemaOutput` exports
  for downstream consumers that derive their own option shapes.

## Usage: Basic

```ts
import { LinkFieldMapSchema } from "@suaveplan/better-auth-surrealdb/schemas/link-fields";

const parsed = LinkFieldMapSchema.safeParse({
    session: { userId: "user" },
});
if (!parsed.success) throw parsed.error;
```

## Usage: Advanced

```ts
import { LinkFieldMapSchema } from "@suaveplan/better-auth-surrealdb/schemas/link-fields";

// Reject non-string targets at the wire boundary
const bad = LinkFieldMapSchema.safeParse({
    session: { userId: 5 },
});
console.log(bad.success); // false

// Custom map merging
const merged = LinkFieldMapSchema.parse({
    ...{ session: { userId: "user" } },
    ...{ api_key: { userId: "user" } },
});
```

## API Reference

- `LinkFieldMapSchema: z.ZodType` — accepts `Record<string, Record<string, string>>`.
- `LinkFieldMapSchemaInput` — pre-parse type.
- `LinkFieldMapSchemaOutput` — post-parse type.

## Implementation Notes

The schema uses `z.record(IdentString, z.record(IdentString, IdentString))`.
Keys go through Zod's `z.record(keySchema, valueSchema)` 4-arg form so the
identifier check applies to the model/field names as well as the target
table. Empty-object inputs (`{}`) parse successfully — operators commonly
pass `linkFields: {}` to disable prefix manipulation entirely.

The `IdentString` schema delegates to `z.string().min(1).regex(...)` so
the inner regex never has to deal with empty strings (which would slip
through the regex via the implicit anchored-success path). The error
message attached to the regex check is human-readable and surfaces in
`ZodError.issues[i].message` for downstream error reporters, while the
issue path lets log aggregators pivot on offending fields without
parsing the message.

Type-level exports `LinkFieldMapSchemaInput` and `LinkFieldMapSchemaOutput`
are identical in this case (the schema is `transform`-free) but the
distinction is preserved so future extensions (e.g. coercing aliases to
canonical table names) can change one without breaking the other.