stack() overwrites state

[thorinaboenke]

Subject of the issue

When a detector calls stack() on another detector's serialized state, it overwrites its own learned data with the peer's instead of merging the two

from_binary returns None, which gets compared to a detector instance, so the error is:
IncompatibleFed Instances are incompatible

Your environment

0.5.3

Steps to reproduce

# stack() on another detectors to_binary() output overwrites the local detector's state and then raises. 
# using detectmatelibrary 0.5.3

import detectmatelibrary.schemas as schemas
from detectmatelibrary.detectors.new_value_detector import NewValueDetector

CONFIG = {"detectors": {"NewValueDetector": {
    "method_type": "new_value_detector",
    "auto_config": False,
    "params": {},
    "events": {1: {"i": {"params": {},
                         "variables": [{"pos": 0, "name": "user", "params": {}}]}}},
}}}


def train(*users):
    d = NewValueDetector(config=CONFIG)
    for u in users:
        d.train(schemas.ParserSchema({
            "parserType": "t", "EventID": 1, "template": "t", "variables": [u],
            "logID": "1", "parsedLogID": "1", "parserID": "p", "log": "l",
            "logFormatVariables": {},
        }))
    return d



a = train("A", "B")
b = train("C", "D")

print("A:", a.persistency.get_event_data(1)["user"].unique_set)
print("B:", b.persistency.get_event_data(1)["user"].unique_set)

try:
    a.stack(b.to_binary())   
except Exception as e:
    print("raised:", type(e).__name__, e)

print("A after:", a.persistency.get_event_data(1)["user"].unique_set, ", expected ['A', 'B', 'C', 'D']")

output:

A: {'B', 'A}
B: {'D', 'C'}
raised: IncompatibleFed Instances are incompatible
A after: {'D, 'C'} -- expected ['A', 'B', 'C', 'D']

Expected behaviour

state should not be overwritten but merged?

Actual behaviour

Tell us what happens instead