Feature/la data

[allentran]

addresses #114

Motivation and context

Speeds up pruning and parsing SM trees data (avoids two brute force N^2 computations) and some initial parsing for other LA datasets.

What I did

• LA city/county, city parsers for Matt Stiles data upto Beverly Hills (starting alphabetically)
• geohash street line segments and tree lat/lon points, match on same geohash while decrementing geohash digit precision (order(N) calculation)
• use GeoPandas sjoin() to do a spatial join based on the predicate contains (uses a RTree to avoid doing the N^2 calculation)

[emillipede]

question from @daniellezegelstein @Cdower @cajaks2

why is the variable df reassigned in every function? why isn't the variable renamed? @allentran

[allentran]

There is a parent class CityParser that does some useful things common to all cities.

Each cities actual parser is a subclass of CityParser, e.g BeverlyHillsParser. Each subclass reimplements/overloads (I forget the correct term) the method get_maximal_df() to do all the custom things required for that city. The name df refers to dataframe and each operation is just adding/changing stuff to that dataframe. I don't know if this is understood, but each get_maximal_df() is a different implementation specific to the city. (This is a common pattern in object oriented programming).

You could rename it but why would you when you're just operating on the same dataframe. I could rename every single instance of the dataframe, e.g df1, df2, ..., dfn but 1) since a lot of Pandas operations are not in-place, I'd be wasting memory and 2) even more confusing to someone reading the code as then I start to think why would the engineer create N different copies of the dataframe if they didn't need them. If I just see one variable/dataframe thats being operated on, the intent is clear (even without documentation) that the code just does stuff to a single dataframe and then returns it.

[Cdower]

Which geohash library is this? needs to be added to environment.yml.