added more flexible binning

tjpcov/covariance_builder.py

@@ -724,6 +724,38 @@ def get_ell_eff(self):
 
         return ell
 
+    def get_binning_info(self):
+        """Return all ells used for the different bandpowers, their effective
+        ell and bandpower edges based on the SACC information.
+
+        It assume that all tracers have the same bandpower windows.
+
+        Args:
+            sacc_data (:obj:`sacc.sacc.Sacc`): Data Sacc instance
+
+        Returns:
+            ells: Array with the ells to pass to theory prediction.
+            ell_eff: Array with effective ells from sacc
+            ell_edges: Array with bandpower edges (assumed to be contiguous).
+        """
+        sacc_file = self.io.get_sacc_file()
+        tr1, tr2 = sacc_file.get_tracer_combinations()[0]
+        dt = sacc_file.get_data_types(tracers=(tr1, tr2))[0]
+        inds = sacc_file.indices(data_type=dt, tracers=(tr1, tr2))
+        bpw = sacc_file.get_bandpower_windows(inds)
+        if bpw is not None:
+            ells = np.repeat(bpw.values, bpw.weight.shape[1])
+            ells = ells.reshape(bpw.weight.shape)
+            ell_eff = np.average(ells, weights=bpw.weight, axis=0)
+            ell_edges = np.zeros(bpw.weight.shape[1] + 1)
+            for i in range(bpw.weight.shape[1]):
+                ell_edges[i] = bpw.values[bpw.weight[:, i] > 0][0]
+                if i == bpw.weight.shape[1] - 1:
+                    ell_edges[i + 1] = bpw.values[bpw.weight[:, i] > 0][-1]
+            return bpw.values, ell_eff, ell_edges
+        else:
+            return None
+
     def get_sacc_with_concise_dtypes(self):
         """Return a copy of the sacc file with concise data types.
 

tjpcov/covariance_gaussian_fsky.py

@@ -1,5 +1,6 @@
 import numpy as np
 import pyccl as ccl
+import warnings
 
 from .wigner_transform import bin_cov
 from .covariance_builder import CovarianceFourier, CovarianceProjectedReal
@@ -41,24 +42,37 @@ def get_binning_info(self, binning="linear"):
         """
         # TODO: This should be obtained from the sacc file or the input
         # configuration. Check how it is done in TXPipe:
-        # https://github.com/LSSTDESC/TXPipe/blob/a9dfdb7809ac7ed6c162fd3930c643a67afcd881/txpipe/covariance.py#L23
-        ell_eff = self.get_ell_eff()
-        nbpw = ell_eff.size
-
-        ellb_min, ellb_max = ell_eff.min(), ell_eff.max()
-        if binning == "linear":
-            del_ell = (ell_eff[1:] - ell_eff[:-1])[0]
-
-            ell_min = ellb_min - del_ell / 2
-            ell_max = ellb_max + del_ell / 2
-
-            ell_delta = (ell_max - ell_min) // nbpw
-            ell_edges = np.arange(ell_min, ell_max + 1, ell_delta)
-            ell = np.arange(ell_min, ell_max + ell_delta - 2)
+        # https://github.com/LSSTDESC/TXPipe/blob/a9dfdb7809ac7ed6c162fd3930c643a67a
+        out = self.get_binning_info()
+        if out is not None:
+            ell = out[0]
+            ell_eff = out[1]
+            ell_edges = out[2]
+            return ell, ell_eff, ell_edges
         else:
-            raise NotImplementedError(f"Binning {binning} not implemented yet")
+            warnings.warn(
+                "No bandpower windows found, \
+                           falling back to linear method"
+            )
+            ell_eff = self.get_ell_eff()
+            nbpw = ell_eff.size
+
+            ellb_min, ellb_max = ell_eff.min(), ell_eff.max()
+            if binning == "linear":
+                del_ell = (ell_eff[1:] - ell_eff[:-1])[0]
+
+                ell_min = ellb_min - del_ell / 2
+                ell_max = ellb_max + del_ell / 2
+
+                ell_delta = (ell_max - ell_min) // nbpw
+                ell_edges = np.arange(ell_min, ell_max + 1, ell_delta)
+                ell = np.arange(ell_min, ell_max + ell_delta - 2)
+            else:
+                raise NotImplementedError(
+                    f"Binning {binning} not implemented yet"
+                )
 
-        return ell, ell_eff, ell_edges
+            return ell, ell_eff, ell_edges
 
     def get_covariance_block(
         self,