Merge pull request #161 from sot/allow-grid-model-extrapolation

Add tolerance args to clip_and_warn

chandra_aca/star_probs.py

@@ -389,11 +389,11 @@ def get_default_acq_prob_model_info(verbose=True):
     return info
 
 
-def clip_and_warn(name, val, val_lo, val_hi, model):
+def clip_and_warn(name, val, val_lo, val_hi, model, tol_lo=0.0, tol_hi=0.0):
     """
     Clip ``val`` to be in the range ``val_lo`` to ``val_hi`` and issue a
-    warning if clipping occurs.  The ``name`` and ``model`` are just used in
-    the warning.
+    warning if clipping occurs, subject to ``tol_lo`` and ``tol_hi`` expansions.
+    The ``name`` and ``model`` are just used in the warning.
 
     Parameters
     ----------
@@ -407,17 +407,26 @@ def clip_and_warn(name, val, val_lo, val_hi, model):
         Maximum
     model
         Model name
+    tol_lo
+        Tolerance below ``val_lo`` for issuing a warning (default=0.0)
+    tol_hi
+        Tolerance above ``val_hi`` for issuing a warning (default=0.0)
 
     Returns
     -------
     Clipped value
     """
     val = np.asarray(val)
-    if np.any((val > val_hi) | (val < val_lo)):
+
+    # Provide a tolerance for emitting a warning clipping
+    if np.any((val > val_hi + tol_hi) | (val < val_lo - tol_lo)):
         warnings.warn(
             f"\nModel {model} computed between {val_lo} <= {name} <= {val_hi}, "
             f"clipping input {name}(s) outside that range."
         )
+
+    # Now clip to the actual limits
+    if np.any((val > val_hi) | (val < val_lo)):
         val = np.clip(val, val_lo, val_hi)
 
     return val
@@ -621,7 +630,8 @@ def grid_model_acq_prob(
     model_filename = Path(gfm["info"]["data_file_path"]).name
 
     # Make sure inputs are within range of gridded model
-    mag = clip_and_warn("mag", mag, mag_lo, mag_hi, model_filename)
+    # TODO: run additional test cases on ASVT, make a new model, remove tol_hi for mag.
+    mag = clip_and_warn("mag", mag, mag_lo, mag_hi, model_filename, tol_hi=0.25)
     t_ccd = clip_and_warn("t_ccd", t_ccd, t_ccd_lo, t_ccd_hi, model_filename)
     halfwidth = clip_and_warn("halfw", halfwidth, halfw_lo, halfw_hi, model_filename)
 

chandra_aca/tests/test_star_probs.py

@@ -12,6 +12,7 @@
 from chandra_aca.star_probs import (
     acq_success_prob,
     binom_ppf,
+    clip_and_warn,
     conf,
     get_default_acq_prob_model_info,
     get_grid_func_model,
@@ -676,3 +677,29 @@ def test_grid_model_3_48(monkeypatch):
     assert info["version"] == "3.48"
     assert info["data_file_path"].endswith("grid-floor-2020-02.fits.gz")
     assert info["commit"] == "68a58099a9b51bef52ef14fbd0f1971f950e6ba3"
+
+
+def test_clip_and_warn():
+    """Test that we get a warning when clipping occurs"""
+    name = "mag"
+    model = "grid-floor"
+    val_lo = 5.0
+    val_hi = 11.75
+
+    # No warnings
+    clip_and_warn(name, 11.75, val_lo, val_hi, model)
+    clip_and_warn(name, 5.0, val_lo, val_hi, model)
+    clip_and_warn(name, 12.0, val_lo, val_hi, model, tol_hi=0.25)
+    clip_and_warn(name, 4.75, val_lo, val_hi, model, tol_lo=0.25)
+
+    # Expected warnings
+    match_re = rf"{model} computed between 5.0 <= {name} <= 11.75"
+    for val, tol_hi, tol_lo in [
+        (12.01, 0.25, 0.0),
+        (4.74, 0.0, 0.25),
+        ([4.74, 7.0, 12.01], 0.25, 0.25),
+    ]:
+        with pytest.warns(UserWarning, match=match_re):
+            clip_and_warn(
+                name, val, val_lo, val_hi, model, tol_hi=tol_hi, tol_lo=tol_lo
+            )