Build Aligned Arrays from Records

Task

Write build_aligned_arrays(records, fields). Each record is a dictionary with an id and the requested numeric fields. IDs must be unique and fields must contain at least one unique name.

Return (ids, values, status). On success, ids is a one-dimensional NumPy string array in record order. values is a two-dimensional float64 array with one row per record and one column per field, in field order. Return status "ok". An empty record list produces shapes (0,) and (0, len(fields)).

Check failures in this order: empty fields, duplicate field, duplicate record ID, missing field, then a value that cannot be converted to float. On failure return (None, None, reason), using "empty fields", "duplicate field: <name>", "duplicate id: <id>", "missing field: <id>.<field>", or "invalid value: <id>.<field>".

Example

For records [{"id":"a","x":1,"y":2}, {"id":"b","x":3,"y":4}] and fields ["y","x"], the IDs are array(["a","b"]) and the value rows are [[2.,1.],[4.,3.]].

Your implementation

You may import NumPy as np. Do not modify an input, print, or ask for input.