The `logging.handlers.MemoryHandler` has a method `flush` which clears the buffer by assigning an empty list literal:
self.buffer = []
This forces the buffer to be a list instance.
My suggestion is to clear the buffer like this instead:
self.buffer.clear()
In this way it would be possible to implement a custom buffer or use the `collections.deque` when subclassing `MemoryHandler`. At the moment you must copy-past the `flush` method and modify it accordingly in the subclass:
```
def flush(self):
# (Docstring skipped)
self.acquire()
try:
if self.target:
for record in self.buffer:
self.target.handle(record)
self.buffer = [] # <-- change to `self.buffer.clear()`
finally:
self.release()
```
Example where this change is useful
===================================
This example creates a DequeMemoryHandler which uses the `collections.deque` as a buffer. Only the latest `capacity` number of records will stored in the buffer. The buffer is then flushed if and only if a record has a level greater or equal to `logging.ERROR`.
```
import collections
import logging
from logging.handlers import MemoryHandler
class DequeMemoryHandler(MemoryHandler):
def __init__(self, capacity, *args, **kwargs):
super().__init__(capacity, *args, **kwargs)
self.buffer = collections.deque(maxlen=capacity)
def shouldFlush(self, record):
return record.levelno >= self.flushLevel
handler = DequeMemoryHandler(capacity=5, target=logging.StreamHandler())
logging.basicConfig(level=logging.INFO, handlers=[handler])
for i in range(1, 21):
logging.info('Spam %d', i)
if i % 10 == 0:
logging.error(f'---> Eggs {i}')
```
Desired output (with the change):
---------------------------------
Spam 7
Spam 8
Spam 9
Spam 10
---> Eggs 10
Spam 17
Spam 18
Spam 19
Spam 20
---> Eggs 20
Actual output (without the change):
-----------------------------------
Spam 7
Spam 8
Spam 9
Spam 10
---> Eggs 10
Spam 11
Spam 12
Spam 13
Spam 14
Spam 15
Spam 16
Spam 17
Spam 18
Spam 19
Spam 20
---> Eggs 20
|
➜
