Free library idea: drawing maps and geodataframes in py5

[villares]

I know it can sound a bit ridiculous, but, maybe if someone with the right skills comes by...

I noticed that geocoded data in a geodataframe can contain shapely geometry, and previously I have played with Unfolding Maps in Processing.py so, some geo/map viz would be great to have.

[hx2A]

Yes, definitely, I'm interested in exploring map data and py5. I've wanted to do this for a long time but simply haven't had the time. Making progress with py5+shapely would be a step in the right direction, and make it easier for someone else to take the next step do something with mapping data.

[villares]

If I had the time I would study @marceloprates's prettymaps code.

[hx2A]

Also, here's a link to the open issue:

processing/processing4#643

I pasted in the wrong link before.

I'll make a PR if my fix holds up.

[hx2A]

I made the PR - fingers crossed it gets merged

[marceloprates]

Hey @hx2A ! Sure, I'd happy to! I don't have much free time at the moment, but could collaborate on that as soon as I finish some other activities.

[hx2A]

@marceloprates , I'm looking forward to working on this! I'm currently adding a feature to py5 that we'll want to use to interface your code to py5. I'll let you know when that is ready.

[marceloprates]

Awesome, please let me know when it's ready!

[villares]

My crude explorations, some progress :)

# Draws ways and buildings from data © OpenStreetMap contributors
import py5

from shapely import Polygon, MultiPolygon
from shapely import LineString, MultiLineString, LinearRing
from shapely import Point, MultiPoint
from shapely import GeometryCollection
from shapely.ops import transform

import osmnx as ox
ox.settings.log_console=True
ox.settings.use_cache=True

from geopandas import GeoDataFrame

bgcolor = '#343434'
edge_color = '#CCCCCC'
bldg_color = '#008888'
point = ox.geocode('Avenida Paulista, São Paulo, Brasil')
dist = 2000

bbox = ox.utils_geo.bbox_from_point(point, dist=dist)
fp = ox.features_from_point(point, tags={'building':True}, dist=dist)
G = ox.graph_from_point(point, network_type='walk', dist=dist, truncate_by_edge=True, retain_all=True)
nodes, edges = ox.graph_to_gdfs(G)
m_edges = edges.to_crs(epsg=3857)
m_fp = fp.to_crs(epsg=3857)
#m_edges = edges.set_crs('EPSG:4326').to_crs('EPSG:3857')
paulista = m_edges[m_edges['name'] == 'Avenida Paulista']
#selected_edges_gdf = gdf_meters[gdf_meters['edge_id'].isin(selected_edges)]
paulista_length = paulista.geometry.length.sum()
print(paulista_length)  # value seems wrong!!!

# for debugging:
import matplotlib.pyplot as plt
fig, ax = ox.plot_graph(G, bgcolor=bgcolor, node_size=0, edge_color=edge_color, show=False)
fig, ax = ox.plot_footprints(fp, ax=ax, bbox=bbox, color=bldg_color, show=False) #, save=True)
paulista.plot(ax=ax, color='red')
# an then call plt.show() at the REPL...

def setup():
    py5.size(800, 800)
    py5.background(bgcolor)
    #py5.translate(400, 400)
    minx, miny, maxx, maxy = m_edges.total_bounds
    scale_x = py5.width / (maxx - minx)
    scale_y = py5.height / (maxy - miny)
    translate_x = -minx * scale_x
    translate_y = -miny * scale_y
    matrix = [scale_x, 0, 0, scale_y, translate_x, translate_y]
    screen_edges = m_edges['geometry'].affine_transform(matrix).apply(flip_y)
    screen_buildings = m_fp['geometry'].affine_transform(matrix).apply(flip_y)
    screen_paulista = paulista.affine_transform(matrix).apply(flip_y)
    py5.no_stroke()
    py5.fill(bldg_color)
    draw_shapely(b for b in screen_buildings if b.area)    
    py5.stroke(edge_color)
    draw_shapely(screen_edges)
    py5.stroke(255, 0, 0)
    draw_shapely(screen_paulista)
    py5.save(f'{__file__[:-3]}.png')

def flip_y(geom):
    return transform(lambda x, y: (x, -y + py5.height), geom)
 
def draw_shapely(shps, sketch: py5.Sketch=None):
    """
    Draw most shapely objects with py5.
    This will use the "current" py5 sketch as default.
    """
    s = sketch or py5.get_current_sketch()
    if isinstance(shps, (MultiPolygon, MultiLineString, GeometryCollection)):
        for shp in shps.geoms:
            draw_shapely(shp)
    elif isinstance(shps, Polygon):
        with s.begin_closed_shape():
            s.vertices(shps.exterior.coords)
            for hole in shps.interiors:
                with s.begin_contour():
                    s.vertices(hole.coords)
    elif isinstance(shps, (LineString, LinearRing)):
        # no need to uses begin_closed_shape() because LinearRing repeats the start/end coordinates
        with s.push_style():
            s.no_fill()
            with s.begin_shape():
                s.vertices(shps.coords)
    elif isinstance(shps, Point):
        s.point(shps.x, shps.y) 
    elif isinstance(shps, MultiPoint):
        s.points((p.x, p.y) for p in shps.geoms)
    elif isinstance(shps, GeoDataFrame):
        for shp in shps.geometry:
                draw_shapely(shp)
    else:
        try:
            for shp in shps:
                draw_shapely(shp)
        except TypeError as e:
            print(f"Unable to draw: {shps}")
        
py5.run_sketch(block=False)

[villares]

Flipping stuff upfront seems costly, so the example linked below would bring it into the (rudimentary) drawing function, but, as @hx2A anticipated, thinking in 2 coordinates systems at once is hard...

https://github.com/villares/sketch-a-day/blob/main/2023/sketch_2023_08_28/sketch_2023_08_28.py

[hx2A]

Flipping stuff upfront seems costly, so the example linked below would bring it into the (rudimentary) drawing function, but, as @hx2A anticipated, thinking in 2 coordinates systems at once is hard...

My current strategy is to alter the Shapely shape to convert it's coordinate system. Something like

from shapely import affinity

obj = affinity.scale(obj, yfact=-1, origin="center")

I'm not sure if that will most correctly address the issue though, need more time with this.

In any case, I have a complete prototype of the convert functions for all Shapely geometry types. I need Processing to accept my PR though.

[hx2A]

In any case, I have a complete prototype of the convert functions for all Shapely geometry types. I need Processing to accept my PR though.

Things are moving along here. Processing accepted my PR and I have merged the py5 branch with the shapely and trimesh code. This week I'll take a look at the prettymaps code to get ideas of how py5 and prettymaps can work together and I'll report back.

[Maltiec]

Hmm, I have some GPS data of my bicycle riding in a city park - may be it's time to build a map of bicycle routes.)

[hx2A]

That'd be a great idea!

[hx2A]

Here's a map made in py5 with prettymaps:

Do you recognize the location?

This is a prototype but it is in pretty good shape. While building this I found a bug in Processing and made a PR and also found a bug in py5's new shapely integration code. Working on this has been a useful test of the new py5 feature.

@marceloprates , the plotter / vsketch mode seems to be commented out and is not working? I need to support the "background" functionality and maybe some other things I am missing. The py5 code is still a bit messy and I will make sure there are no bugs on my end before asking you to do anything.

py5 does not have hatching so we'll want to make some presets that are well suited for py5. I'll enlist @villares and others for help with that when the prototype is ready for it.

[villares]

wow... Francesca Franco would have enjoyed this I guess ;-)

[hx2A]

you got it!