Modifying a nested dictionary - python-3.x

Question

I'm using regex to look through a file to check for a username and whether their action caused an error or not (ERROR for error, INFO for successful action)
I am using a nested dictionary to keep track of actions, with the username as the primary key and the nested dictionary of how many INFO and ERROR lines they generate.
#!/usr/bin/env python3
import re
users = {}
with open('logfile.txt') as f:
for line in f:
regex_user = r"(INFO|ERROR): .* \((.+)\)$"
"""searches for users and if there message was info or error"""
user = re.serach(regex_user, line)
if user is None:
continue
name = user[2]
cat = user[1]
try:
# Method 1?
users[name][cat] = users[name].get(cat, 0) + 1
# Method 2?
users[name][cat] = users.get(name, {}).get(cat, 0) + 1
except KeyError:
print("Where are my keys?")
I am wondering which of the two methods (if any) are correctly modifying the dictionary to increase the count of the respective nested key.
Output should look like:
{'John': {'INFO': 22, 'ERROR': 3}}
if the log contained 22 lines of INFO and 3 lines of ERROR for user John.

Answer 1

You probably want to solve this with a nested collections.defaultdict or alternatively using setdefault() though the latter is not a clean in my opinion. collections.defaultdict() will allow you to reference a dictionary key that does not yet exist which is the crux of your issue.
Try something like:
import collections
import json
import random
users = collections.defaultdict(lambda: collections.defaultdict(int))
for i in range(100):
user = random.choice(["John", "Jane", "Sanjay"])
cat = random.choice(["INFO", "ERROR"])
users[user][cat] += 1
print(json.dumps(users, indent=2))
That will print out something like:
{
"Jane": {
"INFO": 16,
"ERROR": 19
},
"John": {
"ERROR": 21,
"INFO": 10
},
"Sanjay": {
"INFO": 18,
"ERROR": 16
}
}
though each run will be different.
As we noted by #Ryukashin, a defaultdict() does not print() as nicely as a dict(). So casting back is possilbe if it is really needed for some reason. The easiest (no not the most efficient) manner for that might be:
users = json.loads(json.dumps(users))