useMongocampSchema

Infers a table schema by sampling real documents — not by calling collectionApi.getJsonSchema(). This means the inferred shape always matches what's actually in the collection, including nested objects and arrays.

const { schemaFromSamples, schemaToTsInterface } = useMongocampSchema()

schemaFromSamples(records)

function schemaFromSamples(records: Array<Record<string, unknown>>): ColumnDefinition[]

interface ColumnDefinition {
  columnName: string
  columnKey: string
  columnType: string        // 'string' | 'number' | 'boolean' | 'date-time' | 'object' | 'array'
  required: boolean
  arrayItemType?: string    // only set when columnType === 'array'
  children?: ColumnDefinition[] // set for 'object', or 'array' with arrayItemType === 'object'
}

Usage:

const samples = await documentApi.listDocuments({ collectionName, rowsPerPage: 500, page: 1 })
const columns = schemaFromSamples(samples as Array<Record<string, unknown>>)

Type inference rules

• Every non-null value for a field across the sample is inspected; a single consistent type wins, mixed types fall back to 'string'.
• MongoDB extended JSON is recognized: { $date: ... } → 'date-time', { $oid: ... } → 'string', { $numberInt/$numberLong/$numberDouble/$numberDecimal: ... } → 'number'.
• Date-time strings are matched against a broad ISO 8601 pattern — date-only, no-seconds, milliseconds, Z, numeric timezone offsets, and space-separated date-times all count.
• Arrays are inspected element-by-element: a single consistent element type sets arrayItemType; mixed elements fall back to 'string'; an array that's always empty across the sample yields arrayItemType: 'unknown'.
• Objects (and arrays of objects) recurse: children is built by re-running inference over the sampled nested objects/items merged together.
• Field order: _id first, metaData last, everything else alphabetical.

required means "had a value", not "key existed"

A field only counts as present when its value is non-null/non-undefined in a given document — a key that exists but is always null is not marked required.

schemaToTsInterface(columns, name)

function schemaToTsInterface(columns: ColumnDefinition[], interfaceName: string): string

Generates a root interface {PascalCase(name)} { ... }, plus one nested interface per object / array-of-object field, and returns them all joined by blank lines — ready to paste into a .ts file.

Nested interface naming:

• object field → {Parent}{PascalCase(fieldKey)} — e.g. Cart + shippingAddress → CartShippingAddress
• array-of-object field → {Parent}{PascalCase(singularize(fieldKey))} — e.g. Cart + items → CartItem

Falls back to Record<string, unknown> / unknown[] when the field's shape was never observed (always empty in the sample). Optional fields (required: false) get a ? marker; non-identifier field names are automatically quoted.

This is what powers the "Copy as TS Interface" button in MongocampCollectionInfos.