[WIP] Julia implementation of benchmarks

JuliaProject.toml

@@ -0,0 +1,12 @@
+[deps]
+DataFrames = "a93c6f00-e57d-5684-b7b6-d8193f3e46c0"
+Dates = "ade2ca70-3891-5945-98fb-dc099432e06a"
+GeoDataFrames = "62cb38b5-d8d2-4862-a48e-6a340996859f"
+GeoFormatTypes = "68eda718-8dee-11e9-39e7-89f7f65f511f"
+GeoInterface = "cf35fbd7-0cd7-5166-be24-54bfbe79505f"
+GeoParquet = "e99870d8-ce00-4fdd-aeee-e09192881159"
+GeometryOps = "3251bfac-6a57-4b6d-aa61-ac1fef2975ab"
+Parquet2 = "98572fba-bba0-415d-956f-fa77e587d26d"
+Statistics = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
+Tables = "bd369af6-aec1-5ad0-b16a-f7cc5008161c"
+WellKnownGeometry = "0f680547-7be7-4555-8820-bb198eeb646b"

geometryops.jl

@@ -0,0 +1,143 @@
+import GeometryOps as GO, GeoInterface as GI, GeoFormatTypes as GFT
+import GeoParquet, Parquet2, WellKnownGeometry
+
+import Tables, Dates, Statistics
+using DataFrames
+# function q1(data_paths::Dict{String, String})
+
+data_paths = Dict{String, String}()
+for file in ["trip", "building", "customer", "driver", "vehicle", "zone"]
+    data_paths[file] = joinpath(@__DIR__, "$file.parquet")
+end
+
+"""
+    q1(data_paths::Dict{String, String})::DataFrame
+
+Q1 (GeometryOps): Find trips starting within 50km of Sedona city center.
+"""
+q1(args...) = q1_naive(args...)
+
+# TODO: this needs a branch-and-bound is within distance algorithm.
+function q1_naive(data_paths::Dict{String, String})
+    trip_pq = Parquet2.Dataset(data_paths["trip"])#[!, ["t_tripkey", "t_pickuploc", "t_pickuptime"]]
+    trip_df = DataFrame(trip_pq; copycols = false)[!, ["t_tripkey", "t_pickuploc", "t_pickuptime"]]
+    trip_df.geometry = GO.tuples(GFT.WellKnownBinary.((GFT.Geom(),), trip_df.t_pickuploc); crs = GFT.EPSG(4326))
+    metadata!(trip_df, GI.GEOINTERFACE_CRS_KEY, GFT.EPSG(4326))
+    metadata!(trip_df, GI.GEOINTERFACE_GEOMETRYCOLUMNS_KEY, (:geometry,))
+
+    trip_df.pickup_lon = GI.x.(trip_df.geometry)
+    trip_df.pickup_lat = GI.y.(trip_df.geometry)
+    trip_df.distance_to_center = GO.distance.(trip_df.geometry, ((-111.7610, 34.8697),))
+
+    filtered = filter(:distance_to_center => x -> isfinite(x) && x <= 0.45, trip_df)
+    return sort!(filtered, [:distance_to_center, :t_tripkey])[!, ["t_tripkey", "pickup_lon", "pickup_lat", "t_pickuptime", "distance_to_center"]]
+end
+
+
+"""
+    q2(data_paths::Dict{String, String})::DataFrame
+
+Q2 (GeometryOps): Count trips starting within Coconino County zone.
+
+Finds the first zone row where z_name == 'Coconino County' and counts trips whose
+pickup point intersects that polygon. Returns single-row DataFrame with
+trip_count_in_coconino_county.
+
+TODO:
+- [ ] Implement tree accelerated point-in-polygon in GeometryOps, preferably with a prepared,
+      NaturallyIndexedRing implementation
+"""
+q2(args...) = q2_naive(args...)
+# TODO: this needs tree accelerated point-in-polygon - probably requires 
+# fleshing out the Prepared interface.
+function q2_naive(data_paths::Dict{String, String})
+    trip_pq = Parquet2.Dataset(data_paths["trip"])#[!, ["t_tripkey", "t_pickuploc", "t_pickuptime"]]
+    trip_df = DataFrame(trip_pq; copycols = false)[!, ["t_tripkey", "t_pickuploc", "t_pickuptime"]]
+    trip_df.geometry = GO.tuples(GFT.WellKnownBinary.((GFT.Geom(),), trip_df.t_pickuploc); crs = GFT.EPSG(4326))
+
+    zone_df = DataFrame(Parquet2.Dataset(data_paths["zone"]); copycols = false)
+    target_idx = findfirst(==("Coconino County"), zone_df.z_name)
+    if isnothing(target_idx)
+        return DataFrame(trip_count_in_coconino_county = [0])
+    end
+    target_poly = GFT.WellKnownBinary(GFT.Geom(), zone_df.z_boundary[target_idx]) |> GO.tuples
+
+    trip_loc_tree = GO.STRtree(trip_df.geometry)
+    target_extent = GI.extent(target_poly)
+
+    count = 0
+    GO.SpatialTreeInterface.depth_first_search(GO.intersects(target_extent), trip_loc_tree) do i
+        if GO.intersects(trip_df.geometry[i], target_poly)
+            count += 1
+        end
+    end
+
+    return DataFrame(trip_count_in_coconino_county = [count])
+end
+
+
+"""
+    q3(data_paths::Dict{String, String})::DataFrame
+
+Q3 (GeometryOps): Monthly trip stats within 15km (10km box + 5km buffer) of Sedona center.
+
+"""
+q3(args...) = q3_naive(args...)
+# TODO: this needs a branch-and-bound is within distance algorithm.
+function q3_naive(data_paths::Dict{String, String})
+    trip_pq = Parquet2.Dataset(data_paths["trip"])#[!, ["t_tripkey", "t_pickuploc", "t_pickuptime"]]
+    trip_df = DataFrame(trip_pq; copycols = false)
+    trip_df.geometry = GO.tuples(GFT.WellKnownBinary.((GFT.Geom(),), trip_df.t_pickuploc); crs = GFT.EPSG(4326))
+
+    base_poly = GI.Polygon(
+        [GI.LinearRing([
+            (-111.9060, 34.7347),
+            (-111.6160, 34.7347),
+            (-111.6160, 35.0047),
+            (-111.9060, 35.0047),
+            (-111.9060, 34.7347),
+        ])]
+    )
+
+    trip_df.distance = GO.distance.(trip_df.geometry, (base_poly,))
+
+    filtered = filter(:distance => <=(0.045), trip_df)
+
+    filtered.duration_seconds = (filtered.t_dropofftime - filtered.t_pickuptime)
+    filtered.pickup_month = (Dates.month.(filtered.t_pickuptime))
+
+    grouped = groupby(filtered, :pickup_month)
+    agg = combine(grouped, :t_tripkey => length => :total_trips, :distance => Statistics.mean => :avg_distance, :duration_seconds => Statistics.mean => :avg_duration, :t_fare => Statistics.mean => :avg_fare)
+    sort!(agg, :pickup_month)
+    return agg
+end
+
+"""
+    q4(data_paths::Dict{String, String})::DataFrame
+
+Q4 (GeometryOps): Zone distribution of top 1000 trips by tip amount.
+"""
+q4(args...) = q4_naive(args...)
+
+function q4_naive(data_paths::Dict{String, String})
+    # Algorithm to do coverage like things
+    # Do a dual query
+    # Store the result in a vector where the index is of the point and the value 
+    # is the index of the zone
+    # and have a sentinel value (0) for no zone yet
+    # if the point has already been assigned a zone, but intersection returns
+    # valid for another zone, then update the value to the new zone
+    # IF AND ONLY IF the new zone is less than the current zone
+    # OR: pseudorandom selection
+end
+
+"""
+    q5(data_paths::Dict{String, String})::DataFrame
+
+Q5 (GeometryOps): Monthly travel patterns for repeat customers (convex hull of dropoff locations)
+"""
+q5(args...) = q5_naive(args...)
+
+function q5_naive(data_paths::Dict{String, String})
+
+end