Update aggregate_spatial process implementation

.gitignore

@@ -1,3 +1,5 @@
+temporary_trashcan/
+
 # Byte-compiled / optimized / DLL files
 __pycache__/
 *.py[cod]

benchmarks/REPORT.MD

@@ -0,0 +1,287 @@
+# Aggregate Spatial Analysis
+
+## Disclaimer
+
+**A lot has changed in the testing setup so the documentation needs to be updated.**
+
+The latest scripts for `xvec` and `exactextract` can be found in:
+
+- `benchmarks/xvec/`
+- `benchmarks/exactextract/`
+
+Data still remains in:
+
+- `benchmarks/data/`
+
+Although not maintained anymore `memory_profiler` is still working and was used.
+
+Profiling decorator has been added to the `aggregate_spatial` function to enable line by line profiling.
+
+`openeo-processes-dask/openeo_processes_dask/process_implementations/cubes/aggregate.py`
+
+Example to run the scripts with `memory_profiler`:
+
+```bash
+mprof run xvec_small_sample.py
+```
+
+Dont forget to install all extras when installing with poetry to get the `benchmark` extras and all other requirements.
+
+```bash
+poetry install --all-extras
+```
+
+## Table of Contents
+
+1. [Introduction](#10-introduction)
+2. [Methodology](#20-methodology)
+3. [Datasets](#30-datasets)
+4. [Initial Results (OUTDATED)](#40-initial-results-outdated)
+    1. [Small Dataset (2 polygons)](#41-small-dataset-2-polygons)
+        1. [OpenEO xvec](#411-openeo-xvec)
+        2. [OpenEO exactextract](#412-openeo-exactextract)
+    2. [Large Dataset (116 polygons)](#42-large-dataset-116-polygons)
+        1. [OpenEO xvec](#421-openeo-xvec)
+        2. [OpenEO exactextract](#422-openeo-exactextract)
+5. [Conclusion and Comments](#50-conclusion-and-comments)
+
+## 1.0 Introduction
+
+The purpose of this benchmark is to quantify the performance of the OpenEO aggregate_spatial process. Current implementation uses the `xvec` library. Issues arise if a large vector dataset is used. The kernel tends to crash, suggesting the current implementation is not efficient enough. The benchmark will test the performance of the current implementation and compare it to a new implementation using the `exactextract` library.
+
+## 2.0 Methodology
+
+TODO
+
+## 3.0 Datasets
+
+For the purpose of testing the performance we have obtained two datasets. The first is a very small dataset containing two polygons in the Bolzano area. The second dataset is a larger dataset containing the Alto Adige region.
+
+The first dataset: `sample_polygons.geojson` has been provided by @clausmichele
+
+- 2 polygons
+
+The second dataset: `alto_adige.geojson` comes from the Openpolis github repository: `https://github.com/openpolis/geojson-italy`
+
+- 116 polygons
+
+## 4.0 Initial Results (Outdated)
+
+The sections below provide some results for the `small` and `large` dataset. We are comparing the current OpenEO `xvec` implementation first, followed by utilizing `exactextract` to perform the same operation.
+
+Quick summary of the results:
+
+| Metric | Small Dataset (2 polygons) | Large Dataset (116 polygons) |
+|  ---  |  ---  |  ---  |
+|OpenEO xvec| 20-27 seconds | 1057 seconds / 17 minutes |
+|OpenEO exactextract| 15-25 seconds | 1085 seconds / 18 minutes |
+
+### 4.1 Small Dataset (2 polygons)
+
+#### 4.1.1 OpenEO xvec
+
+```python
+from profiler import Profiler
+
+import json
+import fiona
+import geopandas as gpd
+from openeo.local import LocalConnection
+
+local_conn = LocalConnection("./")
+
+url = "https://earth-search.aws.element84.com/v1/collections/sentinel-2-l2a"
+spatial_extent = {"east": 11.40, "north": 46.52, "south": 46.46, "west": 11.25}
+temporal_extent = ["2022-06-01", "2022-06-30"]
+bands = ["red"]
+properties = {"eo:cloud_cover": dict(lt=80)}
+
+s2_datacube = local_conn.load_stac(
+    url=url,
+    spatial_extent=spatial_extent,
+    temporal_extent=temporal_extent,
+    bands=bands,
+    properties=properties,
+)
+
+s2_datacube = s2_datacube.resample_spatial(
+    projection="EPSG:4326", resolution=0.0001
+).drop_dimension("band")
+
+polys_path = "./data/sample_polygons.geojson"
+
+polys = gpd.read_file(polys_path)
+
+# Don't know why this is needed but it is.
+# https://geopandas.org/en/stable/docs/reference/api/geopandas.GeoDataFrame.__geo_interface__.html
+polys = polys.__geo_interface__ 
+
+aggregate = s2_datacube.aggregate_spatial(geometries=polys, reducer="mean")
+
+@Profiler(reruns=1, sample_interval=1, log_file="sm_xvec.csv")
+def run_aggregate():
+    aggregate.execute()
+
+run_aggregate()
+```
+
+| Metric | Value |
+|  ---  |  ---  |
+|Run_ID| X |
+|Timestamp| X |
+|Sample_Timestamp| X |
+|CPU_Usage_%| 24% |
+|Time_Taken| 20-27 seconds |
+|Median_Memory_Usage_MB| 45 - 65 MB |
+
+#### 4.1.2 OpenEO exactextract
+
+```python
+from profiler import Profiler
+
+import json
+import geopandas as gpd
+from openeo.local import LocalConnection
+from exactextract import exact_extract
+
+local_conn = LocalConnection("./")
+
+url = "https://earth-search.aws.element84.com/v1/collections/sentinel-2-l2a"
+spatial_extent = {"east": 11.40, "north": 46.52, "south": 46.46, "west": 11.25}
+temporal_extent = ["2022-06-01", "2022-06-30"]
+bands = ["red"]
+properties = {"eo:cloud_cover": dict(lt=80)}
+
+s2_datacube = local_conn.load_stac(
+    url=url,
+    spatial_extent=spatial_extent,
+    temporal_extent=temporal_extent,
+    bands=bands,
+    properties=properties,
+)
+
+s2_datacube = s2_datacube.resample_spatial(projection="EPSG:4326",resolution=0.0001).drop_dimension("band")
+data = s2_datacube.execute()
+
+polys = gpd.read_file("./data/sample_polygons.geojson")
+
+@Profiler(reruns=1, sample_interval=1)
+def run_extract():
+    exact_extract(data, polys, 'mean')
+
+run_extract()
+```
+
+| Metric | Value |
+|  ---  |  ---  |
+|Run_ID| X |
+|Timestamp| X |
+|Sample_Timestamp| X |
+|CPU_Usage_%| 28% |
+|Time_Taken| 15-25 seconds |
+|Median_Memory_Usage_MB| 40 - 60 MB |
+
+### 4.2 Large Dataset (116 polygons)
+
+#### 4.2.1 OpenEO xvec
+
+```python
+from profiler import Profiler
+
+import json
+import geopandas as gpd
+from openeo.local import LocalConnection
+
+local_conn = LocalConnection("./")
+
+URL = "https://earth-search.aws.element84.com/v1/collections/sentinel-2-l2a"
+SPATIAL_EXTENT = {"east": 11.8638, "north": 46.7135, "south": 46.3867, "west": 10.7817}
+TEMPORAL_EXTENT = ["2022-06-01", "2022-06-30"]
+BANDS = ["red"]
+PROPERTIES = {"eo:cloud_cover": dict(lt=80)}
+
+s2_datacube = local_conn.load_stac(
+    url=URL,
+    spatial_extent=SPATIAL_EXTENT,
+    temporal_extent=TEMPORAL_EXTENT,
+    bands=BANDS,
+    properties=PROPERTIES,
+)
+
+s2_datacube = s2_datacube.resample_spatial(
+    projection="EPSG:4326", resolution=0.0001).drop_dimension("band")
+
+polys = gpd.read_file("./data/alto_adige.geojson")
+polys = polys.__geo_interface__
+
+aggregate = s2_datacube.aggregate_spatial(geometries=polys, reducer="mean")
+
+
+@Profiler(reruns=1, sample_interval=1)
+def run_aggregate():
+    aggregate.execute()
+
+run_aggregate()
+```
+
+| Metric | Value |
+|  ---  |  ---  |
+|Run_ID| X |
+|Timestamp| X |
+|Sample_Timestamp| X |
+|CPU_Usage_%| 59% |
+|Time_Taken| 1057 seconds / 17 minutes |
+|Median_Memory_Usage_MB| 360 MB |
+
+#### 4.2.2 OpenEO exactextract
+
+```python
+from profiler import Profiler
+
+import json
+import geopandas as gpd
+from openeo.local import LocalConnection
+from exactextract import exact_extract
+
+local_conn = LocalConnection("./")
+
+URL = "https://earth-search.aws.element84.com/v1/collections/sentinel-2-l2a"
+SPATIAL_EXTENT = {"east": 11.8638, "north": 46.7135, "south": 46.3867, "west": 10.7817}
+TEMPORAL_EXTENT = ["2022-06-01", "2022-06-30"]
+BANDS = ["red"]
+PROPERTIES = {"eo:cloud_cover": dict(lt=80)}
+
+s2_datacube = local_conn.load_stac(
+    url=URL,
+    spatial_extent=SPATIAL_EXTENT,
+    temporal_extent=TEMPORAL_EXTENT,
+    bands=BANDS,
+    properties=PROPERTIES,
+)
+
+s2_datacube = s2_datacube.resample_spatial(projection="EPSG:4326",resolution=0.0001).drop_dimension("band")
+data = s2_datacube.execute()
+
+polys = gpd.read_file("./data/alto_adige.geojson")
+
+@Profiler(reruns=1, sample_interval=1)
+def run_extract():
+    exact_extract(data, polys, 'mean')
+
+run_extract()
+```
+
+| Metric | Value |
+|  ---  |  ---  |
+|Run_ID| X |
+|Timestamp| X |
+|Sample_Timestamp| X |
+|CPU_Usage_%| 63% |
+|Time_Taken| 1085 seconds / 18 minutes |
+|Median_Memory_Usage_MB| 360 MB |
+
+### 5.0 Conclusion and Comments
+
+So based on running the two comparisons, the efficiency is actually not that much
+different. I am not exactly sure why yet, perhaps the OpenEO input cube causes it
+to be slow even on exactextract.