Minor fixes to the dump PR

[shankari]

Minor fixes to:
MobilityNet/mobilitynet-analysis-scripts#54

• Add a comment explaining the reason for sys.maxsize
• rename run_full_pipeline
• in trajectory_evaluation.ipynb, the ground truth retrieval can use SpecDetails, doesn't need to use FileSpecDetails
• Ditto for trajectory_evaluation_spatio_temporal.ipynb
• And trajectory_evaluation_timestamp_metric_selection.ipynb

[singhish]

One thing to note regarding the last 3 points -- in order for the notebooks to run, it is necessary for SpecDetails.get_concat_trajectories, SpecDetails.get_shapes_for_leg, and SpecDetails.get_geojson_for_leg to be fixed to accommodate the updated start_loc/end_loc/route_coords/loc structures. They are currently handling these keys as dicts when they are now lists, which affects the logic that is currently in the notebooks. I was working on a fix today to address this but decided to scrap it as would probably make more sense to talk this over with you tomorrow @shankari.

[shankari]

@singhish yes you should change all the get_* methods to work with the updated structures.
By scrap, I assume you didn't actually throw out the code!

[singhish]

The trajectory_evaluation.ipynb notebook is now fully working. There are a few things I ran into with the other notebooks:

• trajectory_evaluation_spatio_temporal.ipynb: the spatial_algo_df is for some reason empty, and has nothing to do with get_ground_truth_for_leg. Blocked here, will touch base with you Monday @shankari.
• trajectory_evaluation_timestamp_metric_selection.ipynb: this notebook uses a spec that doesn't have reroutes (sd_la), need to make a design decision regarding get_ground_truth_for_leg to make start_ts and end_ts optional arguments in this case.

[shankari]

trajectory_evaluation_timestamp_metric_selection.ipynb: this notebook uses a spec that doesn't have reroutes (sd_la), need to make a design decision regarding get_ground_truth_for_leg to make start_ts and end_ts optional arguments in this case.

@singhish Use the start and end timestamps of the entire spec. Just like we did for spec creation.