Analyzer: Optimize `area_to_pandas` method

EwoutH commented 3 months ago

The area_to_pandas function has been optimized to significantly improve its execution speed.

Set lookups: Areas are converted to sets for O(1) membership checks, making the isin operations more efficient.
Vectorization: Replaced loop-based calculations with vectorized operations for total_travel_time and total_free_time, leveraging the power of numpy and pandas to perform these operations in bulk.
Single DataFrame access: Minimized repeated access to DataFrame columns, reducing the overhead of these operations.
Efficient grouping: Used unique() and nunique() for faster unique element counting.

In my testing of a large simulation, this sped up the function from ~75 seconds to ~2.5 seconds (about 30x).

EwoutH commented 2 months ago

@toruseo I think this PR is relatively straightforward, it just makes this function faster without modifying it's behavior.

EwoutH commented 2 months ago

Thanks for merging! This PR was fully tested and validated to have identical output.

toruseo / UXsim