Storing dictionary locations as a string and using eval - alternatives?

Victor Hooi victorhooi at gmail.com
Thu Aug 20 05:59:01 CEST 2015


Hi,

I have a textfile with a bunch of JSON objects, one per line.

I'm looking at parsing each of these, and extract some metrics from each line.

I have a dict called "metrics_to_extract", containing the metrics I'm looking at extracting. In this, I store a name used to identify the metric, along with the location in the parsed JSON object.

Below is my code:

>>>>>>>>>>>>>>>>>>>>>>>
metrics_to_extract = {
    'current_connections': "server_status_json['connections']['current']",
    'resident_memory': "server_status_json['mem']['resident']"
}


def add_point(name, value, timestamp, tags):
    return {
        "measurement": name,
        "tags": tags,
        # "time": timestamp.isoformat(),
        "time": timestamp,
        "fields": {
            "value": float(value)
        }
    }

with open(input_file, 'r') as f:
    json_points = []
    for line in f:
        if line.startswith("{"):
            server_status_json = json.loads(line)
            # pp.pprint(server_status_json)
            # import ipdb; ipdb.set_trace()
            timestamp = server_status_json['localTime']
            tags = {
                'project': project,
                'hostname': server_status_json['host'],
                'version': server_status_json['version'],
                'storage_engine': server_status_json['storageEngine']['name']
            }

            for key, value in metrics_to_extract.items():
                json_points.append(add_point(key, eval(value), timestamp, tags))
            # client.write_points(json_points)
        else:
            print("non matching line")
>>>>>>>>>>>>>>>>>>>>>>>

My question is - I'm using "eval" in the above, with the nested location (e.g. "server_status_json['mem']['resident']") stored as a string.

I get the feeling this isn't particularly idiomatic or a great way of doing it - and would be keen to hear alternative suggestions?

Thanks,
Victor


More information about the Python-list mailing list