[Feat] Add master_field_gamma_method stub to Corr class

pyerrors/correlators.py

@@ -151,6 +151,67 @@ class Corr:
 
     gm = gamma_method
 
+    def master_field_gamma_method(self, **kwargs):
+        """Estimate the error of the Corr using the master-field approach.
+
+        This method accounts for correlations across the Corr index
+        (e.g., source positions x0) in addition to the standard Monte Carlo
+        autocorrelation, following the master-field idea by Lüscher,
+        EPJ Web Conf. 175 (2018) 01002.
+
+        Currently a stub that sets all error attributes to zero.
+        """
+        if self.N != 1:
+            raise ValueError("master_field_gamma_method is only implemented for single-valued correlators (N=1).")
+
+        obs_list = [item[0] for item in self.content if item is not None]
+
+        if not obs_list:
+            return
+
+        ref = obs_list[0]
+        for o in obs_list[1:]:
+            if o.names != ref.names:
+                raise ValueError("All Obs in the Corr must share the same ensemble names.")
+
+        for o in obs_list:
+            o.e_dvalue = {}
+            o.e_ddvalue = {}
+            o.e_tauint = {}
+            o.e_dtauint = {}
+            o.e_windowsize = {}
+            o.e_n_tauint = {}
+            o.e_n_dtauint = {}
+            o.e_rho = {}
+            o.e_drho = {}
+            o.S = {}
+            o.tau_exp = {}
+            o.N_sigma = {}
+
+            for e_name in o.mc_names:
+                o.e_dvalue[e_name] = 0.0
+                o.e_ddvalue[e_name] = 0.0
+                o.e_tauint[e_name] = 0.5
+                o.e_dtauint[e_name] = 0.0
+                o.e_windowsize[e_name] = 0
+                o.e_rho[e_name] = np.zeros(1)
+                o.e_drho[e_name] = np.zeros(1)
+                o.e_n_tauint[e_name] = np.zeros(1)
+                o.e_n_dtauint[e_name] = np.zeros(1)
+                o.S[e_name] = Obs.S_global
+                o.tau_exp[e_name] = Obs.tau_exp_global
+                o.N_sigma[e_name] = Obs.N_sigma_global
+
+            for e_name in o.cov_names:
+                o.e_dvalue[e_name] = np.sqrt(o.covobs[e_name].errsq())
+                o.e_ddvalue[e_name] = 0
+
+            o._dvalue = 0.0
+            for e_name in o.cov_names:
+                o._dvalue += o.e_dvalue[e_name] ** 2
+            o._dvalue = np.sqrt(o._dvalue)
+            o.ddvalue = 0.0
+
     def projected(self, vector_l=None, vector_r=None, normalize=False):
         """We need to project the Correlator with a Vector to get a single value at each timeslice.
 

tests/correlators_test.py

@@ -804,3 +804,32 @@ def test_prune_with_Nones():
 
     for t in range(T):
         assert np.all(pruned_then_padded.content[t] == padded_then_pruned.content[t])
+
+
+def test_master_field_gamma_method():
+    T = 8
+    N_cfg = 50
+    corr_content = []
+    for t in range(T):
+        data = np.random.normal(1.0, 0.1, N_cfg)
+        corr_content.append(pe.Obs([data], ['test_ens']))
+    corr = pe.Corr(corr_content)
+
+    corr.master_field_gamma_method()
+
+    for t in range(T):
+        assert corr[t].dvalue == 0.0
+        assert corr[t].ddvalue == 0.0
+        assert corr[t].e_tauint['test_ens'] == 0.5
+        assert corr[t].e_dvalue['test_ens'] == 0.0
+        assert corr[t].e_windowsize['test_ens'] == 0
+
+
+def test_master_field_gamma_method_matrix_raises():
+    N = 2
+    T = 4
+    corr_content = np.array([[[pe.pseudo_Obs(1.0, 0.1, 'test_ens') for _ in range(N)] for _ in range(N)] for _ in range(T)])
+    matrix_corr = pe.Corr(corr_content)
+
+    with pytest.raises(ValueError):
+        matrix_corr.master_field_gamma_method()