message located last at the log line.

[ypolon7kiy]

Is your feature request related to a problem? Please describe.
I'm always frustrated when i try to read logs, but the message is located at the end of line, so thre is always need to open each line in CloudWatch.

In that way one cannot read the app flow logic easily.

Describe the solution you'd like
move message to the start of the line, or after timestamp.

Can be solved in:
logger/formatter.py
format(self, record)

PROBLEM EXAMPLE:
{"timestamp": "2020-08-27 08:55:33,265", "level": "INFO", "location": "lambda_handler:9", "service": "aaa", "sampling_rate": 0.0, "cold_start": true, "function_name": "TestPeriod", "function_memory_size": "128", "function_arn": "arn:aws:lambda:eu-central-1:508054367788:function:TestPeriod", "function_request_id": "d55d25b5-db25-446d-9c04-10cc79800981", "order_id": "aaa", "message": {"operation": "collect_payment"}}

[heitorlessa]

Hi Yuriy - Thanks for raising your first issue and feature request under the AWS Labs org :)

I see what you mean, and hopefully that's a quick fix:

• This is block of code that inserts the message key
• We'd need to move that block right after log_dict since Python 3.6+ preserves the order insertion

I've just tagged as a good first contribution issue in case you or anyone reading this later want to give this a try. Otherwise, I'll implement this later next week :)

[michaelbrewer]

@heitorlessa @ypolon7kiy I fixed this in #140. I think the order should be timestamp,level,location,message and the rest.

[michaelbrewer]

@heitorlessa @ypolon7kiy - otherwise if there is a strong preference i could make it timestamp,message,level,location.

But having the message after the location is handy to match things up to the implementation code.

[heitorlessa]

I think the challenge @ypolon7kiy is having is that he's using the standard CloudWatch Logs console, and the line preview shows only the first key (timestamp) hence the ask for bringing up as first.

That's why DAZN Lambda Powertools for JS does this: https://github.com/getndazn/dazn-lambda-powertools/blob/d870d55bbbe9851705b1e860e4a45d2968edaaf5/packages/lambda-powertools-logger/index.js#L82

I'd wait for @ypolon7kiy to chime back and confirm the challenge he's having.

Customers using ELK, CloudWatch Log Insights, and any advanced Log Analytics tool allows you to change the order as you see fit regardless of how the logs have been ingested, but I still suspect that's not the case with Yuriy

[ypolon7kiy]

Hi @heitorlessa and @michaelbrewer .
Thank you for the response.
I would put actually timestamp more closely to the end because in CloudWatch logs it looks like :

2020-08-27T12:21:21.894+03:00 | {"timestamp": "2020-08-27 09:21:21,894", "level": "INFO", "location": "lambda_handler:25", "service": "aaa", "sampling_rate": 0.0, "cold_start": false, "function_name": "TestPeriod", "function_memory_size
.....

So the timestamp appearing twice at the beginning.

But what you decide.

[michaelbrewer]

@ypolon7kiy @heitorlessa i updated it to be configurable as well and the default is "level", "location", "message", "timestamp" But if you want to change it just pass through format_key to the logger ie:

>>> logger = Logger(format_key=["message"])
>>> logger.info("Message")
{'message': 'Message', 'level': 'INFO', 'location': 'test_log_dict_key_custom_seq:139', 'timestamp': '2020-08-27 11:40:33,718', 'service': 'service_undefined', 'sampling_rate': 0.0}

[michaelbrewer]

@ypolon7kiy - you should check out #140 and see if this fits in with what you are expecting.

[heitorlessa]

@ypolon7kiy could you tell me whether the docs match your expectation?

https://github.com/awslabs/aws-lambda-powertools-python/pull/148/files

Once merged, this feature and docs will be available in the next release (1.5.0)