Pierre Sassoulas 1125296b53 Small performance/readability improvments when iterating dictionnary with `keys()` ...

Based on pylint's message ``consider-iterating-dictionary`` suggestion.
Surprisingly using a dict or set comprehension instead of a new temp var is
actually consistently slower here, which was not intuitive for me.

```python
from timeit import timeit

families = {1: {"testcase": [1, 2, 3, 5, 8]}}
attrs = {1: "a", 2: "b", 3: "c", 4: "d", 5: "e", 6: "f", 7: "g", 8: "h"}

class Old:
    def old(self):
        self.attrs = attrs
        temp_attrs = {}
        for key in self.attrs.keys():
            if key in families[1]["testcase"]:
                temp_attrs[key] = self.attrs[key]
        self.attrs = temp_attrs

class OldBis:
    def old(self):
        self.attrs = attrs
        temp_attrs = {}
        for key in self.attrs:
            if key in families[1]["testcase"]:
                temp_attrs[key] = self.attrs[key]
        self.attrs = temp_attrs

class New:
    def new(self):
        self.attrs = attrs
        self.attrs = { # Even worse with k: v for k in self.attrs.items()
            k: self.attrs[k] for k in self.attrs if k in families[1]["testcase"]
        }

if __name__ == "__main__":
    n = 1000000
    print(f"Old: {timeit(Old().old, number=n)}")
    print(f"Just removing the keys(): {timeit(OldBis().old, number=n)}")
    print(f"List comp, no temp var: {timeit(New().new, number=n)}")
```

Result:
Old: 0.9493889989680611
Just removing the keys(): 0.9042672360083088
List comp, no temp var: 0.9916125109884888

It's also true for the other example with similar benchmark, but the exact
code probably does not need to be in the commit message.

2024-03-31 14:43:07 +02:00

12 KiB

Raw Blame History

View Raw