add update method or implement save(update_fields) like django orm

[young118]

Add a functionality similar to Django ORM's save(update_fields=list) to the ojb.save() method. Currently, the save method writes all data in full to Redis. When two processes simultaneously save the same object, the later save operation will overwrite the attributes modified by the earlier save operation.

[abrookins]

Thanks for this request! You've actually identified two related but distinct problems:

1. Concurrent update safety (locking)

When two processes read the same model, modify different fields, and save - the last write wins and silently overwrites the first's changes. This is a concurrency control problem that could be addressed with:

• Optimistic locking - Use WATCH + version field to detect conflicts
• Pessimistic locking - Use Redis locks to prevent concurrent access

We're tracking locking options separately as it's a larger design decision.

2. Partial field updates

The immediate request - only save specific fields instead of the full document. This is what Django's save(update_fields=[]) provides.

---

Proposed: save(update_fields=[])

user.name = "New Name"
await user.save(update_fields=["name"])  # Only writes name field

Implementation:

• HashModel: HSET key name "New Name" (only specified fields)
• JsonModel: JSON.SET key $.name "New Name" (path-based update)

Pydantic validation runs on the specified fields before saving.

Constraints:

• update_fields with nx=True → Error (insert requires full document)
• update_fields containing pk → Error (can't change primary key)
• Unknown field names → ValueError

---

Related: FindQuery.update()

Bulk updates via FindQuery.update() already exist:

await User.find(User.status == "pending").update(status="active")

This updates matching records, but the current implementation is inefficient (loads full documents, writes full documents). See #777 for optimizing it to use partial field updates as well.