Add multithreading/async for all requests

[SlowMo24]

Requirement

The user wants to request one or more indicators for one ore more layers for multiple geometries.

Current Situation

The user has to do all optimisation at her end (e.g. asynchronous requests). Most users will use a very ineffective loop approach:

for indicator in indicators:
    for layer in layers:
        for geometry in geometries:
            request(oqt)

For an example see https://gitlab.gistools.geog.uni-heidelberg.de/giscience/ideal-vgi/osm-element-vectorisation/-/blob/6a65893fdaaa5cad97f4b1d7cf7fdc1701e89287/src/osm_element_vectorisation/indicators/surrounding/ohsome/oqt.py

Possible better (more performant) solution

Use group-by(geometry/indicator/layer) or for async(geometry/indicator/layer) approach to allow users to request multiple geometries at once. Multiple geometries are currently possible, what is the approach here?

This would require user friendly error handling, where the request only fails, if all geometries fail but returns partial responses, whenever possible.

Related: #151 and GIScience/ohsome-py#37

[matthiasschaub]

Thanks for creating this detailed issue! Some short notes from our discussion:

Currently possible:

• request multiple regions → FeatureCollection input.
  • not asynchronous yet (See Make calculation of indicators for a FeatureCollection or a Report asynchronous #305)
• request multiple layer/indicator combinations → Request a Report. But reports need to be defined by the OQT Team.
  • not asynchronous yet (See Make calculation of indicators for a FeatureCollection or a Report asynchronous #305)

Utilizing groupByBoundary queries to the ohsome API for a FeatureCollection input to OQT is not easily implemented. Also it is not clear if this will lead to big performance improvements.

For errors see this issue: #306

[matthiasschaub]

Asynchronous calculation of Indicators is now implemented. See previous comment.

[SlowMo24]

Great solution, thank you. One question: Are reports available via the API as JSON object and do they also handle FeatureCollections?

[matthiasschaub]

You are welcome!

In short the answer to your question is Yes and Yes.

Reports are also available via the API. The response is a GeoJSON representing the input GeoJSON. If the input is just a Feature the response will also be a Feature. If the input is a FeatureCollection the response will also be a FeatureCollection. The overall result and also the result of every Indicator will be part of the properties.
See also: https://oqt.ohsome.org/api/docs#/report/post_report_report_post