stdpipe 0.4.0__tar.gz → 0.4.1__tar.gz

{stdpipe-0.4.0/stdpipe.egg-info → stdpipe-0.4.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: stdpipe
-Version: 0.4.0
+Version: 0.4.1
 Summary: Simple Transient Detection Pipeline
 Author-email: Sergey Karpov <karpov.sv@gmail.com>
 Maintainer-email: Sergey Karpov <karpov.sv@gmail.com>

{stdpipe-0.4.0 → stdpipe-0.4.1}/stdpipe/__init__.py

@@ -6,4 +6,4 @@ detection related tasks, intended for quick and easy implementation of custom
 pipelines, as well as for interactive data analysis.
 """
 
-__version__ = "0.4.0"
+__version__ = "0.4.1"

{stdpipe-0.4.0 → stdpipe-0.4.1}/stdpipe/photometry.py

@@ -1082,7 +1082,7 @@ def get_objects_sep(
         # pixel outside each source's footprint, so the winpos aperture always
         # overlaps masked pixels by construction. Keep TRUNC / ALLMASKED /
         # NONPOSITIVE — those still convey real centroiding-quality info.
-        flag_win &= ~np.int_(sep.APER_HASMASKED)
+        flag_win &= ~int(sep.APER_HASMASKED)
     else:
         # Use initial positions from extraction
         xwin, ywin = obj0['x'], obj0['y']

{stdpipe-0.4.0 → stdpipe-0.4.1}/stdpipe/photometry_model.py

@@ -18,6 +18,12 @@ from statsmodels.tools.sm_exceptions import ConvergenceWarning
 
 from . import astrometry
 
+# Smallest fit scale still carrying usable scatter information. The fits are
+# performed on residuals normalized by the noise model, so the scale is
+# dimensionless and ~1 when the noise model describes the data; anything this
+# small means the fit is perfect down to floating-point roundoff.
+_MIN_FIT_SCALE = 1e-8
+
 
 def make_series(mul=1.0, x=1.0, y=1.0, order=1, sum=False, zero=True):
     x = np.atleast_1d(x)
@@ -592,7 +598,9 @@ def match(
         & np.isfinite(cmag)
         & np.isfinite(cmag_err)
         & (zero_err > 0)  # Zero errors would give infinite weights in the fit
-        & ((oflags & ~accept_flags) == 0)
+        # Cast to signed int: NumPy 2 (NEP 50) raises OverflowError when a
+        # negative Python int (~accept_flags) meets an unsigned flags column
+        & ((np.asarray(oflags, int) & ~int(accept_flags)) == 0)
     )  # initial mask
     if cat_color is not None and (fit_color_term or force_color_term is not None):
         idx0 &= np.isfinite(ccolor)
@@ -617,10 +625,8 @@ def match(
         else:
             C = sm.WLS(zero[idx], X[idx], weights=1 / total_err[idx] ** 2).fit()
 
-        # Perfect fit is fatal for RLM iterations, and whenever the scale
-        # is used for re-scaling the noise model
-        if not np.isfinite(C.scale) or (C.scale == 0.0 and (robust or scale_noise)):
-            log('Fit failed - fit scale collapsed to zero, model is degenerate')
+        if not np.isfinite(C.scale):
+            log('Fit failed - fit scale is not finite, model is degenerate')
             return None
 
         # RLM scale is a standard deviation while WLS one is a variance
@@ -650,8 +656,16 @@ def match(
 
         if scale_noise:
             # The fit scale measures residual scatter relative to the noise
-            # model used in this fit, so the applied scaling is cumulative
-            scale_err = scale_err * fit_scale
+            # model used in this fit, so the applied scaling is cumulative.
+            # Both fit branches normalize the residuals by the noise model, so
+            # fit_scale is dimensionless and ~1 for a well-described fit. A
+            # vanishing value means the model reproduces the data down to
+            # roundoff, leaving no scatter information to rescale with -- keep
+            # the noise model as is instead of collapsing it to zero.
+            if fit_scale > _MIN_FIT_SCALE:
+                scale_err = scale_err * fit_scale
+            else:
+                log('Fit scale is negligible - keeping the noise model unscaled')
 
         intrinsic_rms = (
             get_intrinsic_scatter((zero - zero_model)[idx], (zero_err * scale_err)[idx], max=max_intrinsic_rms)

{stdpipe-0.4.0 → stdpipe-0.4.1}/stdpipe/pipeline.py

@@ -601,10 +601,10 @@ def filter_transient_candidates(
                 xidx = vizier_checker_fn(xobj, xcat, catname)
                 xcat = xcat[xidx]
 
-            cand_idx &= ~np.in1d(obj[col_id], xcat[col_id])
+            cand_idx &= ~np.isin(obj[col_id], xcat[col_id])
 
             if remove == False:
-                obj_in['candidate_vizier_' + catname][np.in1d(obj[col_id], xcat[col_id])] = True
+                obj_in['candidate_vizier_' + catname][np.isin(obj[col_id], xcat[col_id])] = True
 
         log(
             np.sum(cand_idx),
@@ -626,10 +626,10 @@ def filter_transient_candidates(
                 obj[cand_idx], time=time, col_ra=obj_col_ra, col_dec=obj_col_dec, col_id=col_id
             )
             if xcat is not None and len(xcat):
-                cand_idx &= ~np.in1d(obj[col_id], xcat[col_id])
+                cand_idx &= ~np.isin(obj[col_id], xcat[col_id])
 
                 if remove == False:
-                    obj_in['candidate_skybot'][np.in1d(obj[col_id], xcat[col_id])] = True
+                    obj_in['candidate_skybot'][np.isin(obj[col_id], xcat[col_id])] = True
 
             log(np.sum(cand_idx), 'remains after matching with SkyBot')
 
@@ -647,10 +647,10 @@ def filter_transient_candidates(
             col_dec=obj_col_dec,
         )
         if xcat is not None and len(xcat):
-            cand_idx &= ~np.in1d(obj[col_id], xcat[col_id])
+            cand_idx &= ~np.isin(obj[col_id], xcat[col_id])
 
             if remove == False:
-                obj_in['candidate_ned'][np.in1d(obj[col_id], xcat[col_id])] = True
+                obj_in['candidate_ned'][np.isin(obj[col_id], xcat[col_id])] = True
 
         log(np.sum(cand_idx), 'remains after matching with NED')
 

{stdpipe-0.4.0 → stdpipe-0.4.1}/stdpipe/plots.py

@@ -1313,7 +1313,7 @@ def plot_mag_histogram(
     mag = obj[obj_col_mag]
     mag_sn = 1 / obj[obj_col_mag_err]
 
-    idx = (obj['flags'] & ~accept_flags) == 0
+    idx = (np.asarray(obj['flags'], int) & ~int(accept_flags)) == 0
 
     vmin = np.nanmin(mag)
     vmax = np.nanmax(mag)

{stdpipe-0.4.0 → stdpipe-0.4.1}/stdpipe/realbogus_features.py

@@ -1402,7 +1402,9 @@ def classify(
         if "flags" not in obj.colnames:
             obj.add_column(Column(np.zeros(len(obj), dtype=np.int32), name="flags"))
         # Reset existing bogus bit so output reflects this classification pass.
-        obj["flags"] &= ~FLAG_BOGUS
+        # Cast to signed int: NumPy 2 (NEP 50) raises OverflowError when a
+        # negative Python int (~FLAG_BOGUS) meets an unsigned flags column
+        obj["flags"] = np.asarray(obj["flags"], int) & ~int(FLAG_BOGUS)
         obj["flags"][predictions < 0] |= FLAG_BOGUS
 
     return obj

{stdpipe-0.4.0 → stdpipe-0.4.1/stdpipe.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: stdpipe
-Version: 0.4.0
+Version: 0.4.1
 Summary: Simple Transient Detection Pipeline
 Author-email: Sergey Karpov <karpov.sv@gmail.com>
 Maintainer-email: Sergey Karpov <karpov.sv@gmail.com>