bug: Default zoom level for `px.scatter_map` is too high

[ndrezn]

using some slightly modified examples in https://plotly.com/python/tile-scatter-maps/.

If I try (FYI need to install geopandas<1.0, note removal of the zoom setting):

import plotly.express as px
import geopandas as gpd

geo_df = gpd.read_file(gpd.datasets.get_path("naturalearth_cities"))

fig = px.scatter_map(
 geo_df,
 lat=geo_df.geometry.y,
 lon=geo_df.geometry.x,
 hover_name="name",
)
fig.show()

Then I get:

Which is bad. We should set the Zoom to cover as many points as possible. Centroid on Africa seems reasonable for this example but not sure how it's determined.

Though I notice with some PX calls the centre point seems set automatically (note I removed the zoom level):

import plotly.express as px

df = px.data.carshare()
fig = px.scatter_map(
 df,
 lat="centroid_lat",
 lon="centroid_lon",
 color="peak_hour",
 size="car_hours",
 color_continuous_scale=px.colors.cyclical.IceFire,
)
fig.show()

Some investigation required to understand the current behaviour.

[emilykl]

@ndrezn I'm moving this to the plotly.py repo since this seems to be a Plotly Express issue, not plotly.js.

This code:

import plotly.express as px
import pandas as pd

df = pd.DataFrame({
 "city": ["San Marino", "Chicago", "Nairobi", "Istanbul", "Trondheim"],
 "lat": [43.9360958, 41.85003, -1.28333, 41.01384, 63.43049],
 "lon": [12.4417702, -87.65005, 36.81667, 28.94966, 10.39506],
})
fig = px.scatter_map(
 df,
 lat="lat",
 lon="lon",
 hover_name="city",
)
fig.show()

produces a map centered on the centroid but zoomed in way too far, such that none of the data is visible:

Center and zoom info of the final plot:

{
 center: {
 lat: 37.78942515999999,
 lon: 0.19062204000000377,
 },
 zoom: 8,
}

However, it seems the root cause is bad defaults being added by plotly.py, since the equivalent plot in plotly.js is fully zoomed out and shows all the data: Link to Codepen