XOR-based random number generator used to encrypt and decrypt a tree document

This is a two-part coding exercise. First, implement a pseudo-random number generator (PRNG) whose state update is built from bitwise XOR. Then use that generator to encrypt a tree-structured document, and to decrypt the result back into the original document.

Constraints and Clarifications

• Write the generator yourself from integer, shift and XOR operations; do not call a library random-number or cryptography function.
• The generator is seeded with an integer key. The same seed must always produce the same sequence, because decryption has to regenerate exactly the numbers that encryption used.
• The document is a tree: each node has a name, optional text content, and an ordered list of child nodes, similar to a simplified XML or JSON document. The exact node schema was not given; state the one you use.
• Decrypting the encrypted output with the same key must reproduce the original tree exactly: the same structure, child order, names and text.

Clarifying Questions

• Which output width should the generator produce (for example 32-bit or 64-bit unsigned integers), and do its statistical quality and period matter here?
• Should the encrypted result be an opaque blob, or must it keep the tree's shape, with each node's content encrypted in place?
• Is the scheme meant to resist a real attacker, or is the exercise about using a deterministic keystream correctly?
• Can node text contain arbitrary characters, including any delimiter you might want to use, and can the tree be very deep?

Part 1 — The generator

Implement a generator class that is constructed from an integer seed and returns a new pseudo-random unsigned integer of fixed width on each call, for example through a next() -> int method. Each output must come from updating an internal state with shifts and XOR.

What This Part Should Cover

• The state update, and how the state is kept at a fixed width
• Seed handling, including any seed that would break the generator
• Determinism: the same seed reproduces the same sequence
• What this kind of generator is and is not good enough for

Part 2 — Encrypt and decrypt a tree document

Using your generator, write encrypt(document, key) and decrypt(ciphertext, key) so that decrypt(encrypt(document, key), key) returns a tree equal to the original document.

What This Part Should Cover

• Turning the tree into something that can be encrypted, and back, without ambiguity, for example when text contains special characters
• How the generator's output is combined with the data, and why decryption inverts it
• Correctness on edge cases: empty text, non-ASCII text, a very deep or very wide tree
• Time and memory cost in terms of the document's size