Bug: flush_buffer discards args - percent-style log interpolation broken for buffered records #8005
Description
Expected Behaviour
When flushing logs from buffer containing percent-style format arguments, logs should be formatted with said args, i.e
logger.debug("user=%s count=%d", "alice", 42) should be flushed as:
"user=alice count=42"
Current Behaviour
Logs flushed from buffer containing, percent-style format arguments (*args) are stored correctly in the buffer but discarded when the buffer is flushed.
i.e
logger.debug("user=%s count=%d", "alice", 42) Is flushed as:
"user=%s count=%d"
Code snippet
import logging
import os
os.environ["_X_AMZN_TRACE_ID"] = (
"Root=1-5759e988-bd862e3fe1be46a994272793;Parent=53995c3f42cd8ad8;Sampled=1"
)
from aws_lambda_powertools import Logger
from aws_lambda_powertools.logging.buffer import LoggerBufferConfig
class RecordCollector(logging.Handler):
"""Handler that captures emitted LogRecords for inspection."""
def __init__(self):
super().__init__()
self.records: list[logging.LogRecord] = []
def emit(self, record):
self.records.append(record)
def messages(self):
return [r.getMessage() for r in self.records]
collector = RecordCollector()
collector.setLevel(logging.DEBUG)
logger = Logger(
service="repro",
buffer_config=LoggerBufferConfig(
max_bytes=20480,
buffer_at_verbosity="DEBUG",
flush_on_error_log=True,
),
)
logger.addHandler(collector)
logger.debug("user=%s count=%d", "alice", 42)
assert collector.messages() == [], "Should be buffered, not emitted yet"
logger.flush_buffer()
print("Flushed messages:", collector.messages())
# BUG: message is "user=%s count=%d" instead of "user=alice count=42"
assert "user=alice count=42" in collector.messages(), (
f"Expected interpolated message, got: {collector.messages()}"
)Possible Solution
_create_and_flush_log_record hardcodes args=() instead of using the args saved in the buffer record.
The result is that flushed log messages contain the raw format string (e.g. "user=%s count=%d") instead of the interpolated message ("user=alice count=42").
When manually changing args=() to args=log_line.get("args", ()), at this line the flush works as expected
Steps to Reproduce
The above snippet proves the issue locally - the general issue reproduced in running lambdas as well
Powertools for AWS Lambda (Python) version
latest
AWS Lambda function runtime
3.12
Packaging format used
PyPi