Merge pull request #149 from ImperialCollegeLondon/tcai-profiling

Add small dataset for profiling

.github/workflows/pyrealm_ci.yaml

@@ -48,13 +48,6 @@ jobs:
       env:
         CODECOV_TOKEN: ${{ secrets.CODECOV_TOKEN }}
 
-    - name: Setup Git LFS
-      id: lfs
-      if: ${{ steps.codecov.outcome == 'success' && github.event_name != 'push' }}
-      run: |
-        git lfs pull
-        git lfs ls-files -n | xargs git lfs track  # track all LFS files
-
     - name: Run profiling
       id: profiling
       if: steps.lfs.outcome == 'success'

.gitignore

@@ -25,8 +25,12 @@ pyrealm.egg-info
 # Coverage
 .coverage
 
+# mypy
+.mypy_cache/
+
 # Data
-pyrealm_build_data/*.nc
+pyrealm_build_data/inputs_data_24.25.nc
+pyrealm_build_data/eda.py
 
 # Profiling
 prof/

CONTRIBUTING.md

@@ -204,16 +204,18 @@ depending on the os. Dedicated profiling tests have been created for PyRealm.
 Please see the relevant testing directory.
 
 ```bash
-poetry run pytest <test_path> --profile
+poetry run pytest --profile
 ```
 
 to generate a report. You can run
 
 ```bash
-poetry run pytest <test_path> --profile-svg
+poetry run pytest --profile-svg
 ```
 
-to generate a call graph. The graph is stored with the name `combined.svg`.
+to generate a call graph. The graph is saved at `prof/combined.svg`.
+
+To enable profiling of a test function or class, decorate it with `@pytest.mark.profiling`.
 
 #### Using 'pytest-coverage'
 

profiling/report.py

@@ -57,6 +57,7 @@
 ).str.extract(r"(.*?):(.*?)\((.*)\)", expand=True)
 dt = datetime.datetime.fromtimestamp(prof_path.stat().st_mtime)
 df.index = pd.Series([dt.strftime("%Y-%m-%d %H:%M:%S")] * len(df), name="timestamp")
+print(df)
 df["label"] = df["filename"].str.extract(r"(\w+).py").squeeze() + "." + df["function"]
 df["event"] = [args.event] * len(df)
 report_path = root / "profiling/prof-report.csv"

tests/test_profiling.py

@@ -21,7 +21,7 @@ class TestClass:
     def setup(self):
         """Setting up the pmodel and loading the test data set."""
         # Loading the dataset:
-        dpath = resources.files("pyrealm_build_data") / "inputs_data_24.25.nc"
+        dpath = resources.files("pyrealm_build_data") / "inputs_data_reduced_1y.nc"
 
         ds = xarray.load_dataset(dpath)
 
@@ -32,9 +32,10 @@ def setup(self):
         #        offset from the UTC times in the data. This step reduces the input data
         #        to complete daily blocks of data using local time
 
-        ds = ds.sel(time=slice("2000-01-01T01:59", "2001-12-31T01:59"))
-        self.local_offset = np.timedelta64(int((24.25 * (24 / 360)) * 60 * 60), "s")
-        self.local_time = ds["time"].to_numpy() - self.local_offset
+        ds = ds.sel(time=slice("2000-01-01T01:59", "2000-12-31T01:59"))
+        ds["local_time_offset"] = (ds["lon"] / 15 * 3600).astype("timedelta64[s]")
+        # 15° per hour, 3600 seconds per hour
+        ds["local_time"] = ds["time"] - ds["local_time_offset"]
 
         # Variable set up
         # Air temperature in Kelvin
@@ -49,7 +50,10 @@ def setup(self):
         # Gather PPFD µmole/m2/s1
         self.ppfd = ds["ppfd"].to_numpy()
         # Define atmospheric CO2 concentration (ppm)
-        self.co2 = np.ones_like(self.tc) * 400
+        self.co2 = ds["co2"].to_numpy()
+        # Define the local time at different longitudes
+        self.local_time = ds["local_time"].to_numpy().squeeze()
+        # TODO: make the code applicable to a dataset with multiple longitudes
 
         # Generate and check the PModelEnvironment
         self.pm_env = PModelEnvironment(