We need to sort the sed.info before saving the data. The reason is that the...

We need to sort the sed.info before saving the data. The reason is that the dictionary orders the keys with a hash so it appears random. In addition the order also changes from one process to another (I suspect the hash also depends on a seed). Thanks to Yannick for pointing out this issue.

pcigale/analysis_modules/pdf_analysis/utils.py

@@ -170,6 +170,8 @@ def save_table_best(filename, obsid, chi2, chi2_red, variables, fluxes,
     best_model_table.add_column(Column(np_chi2, name="chi_square"))
     best_model_table.add_column(Column(np_chi2_red, name="reduced_chi_square"))
 
+    info_keys = list(info_keys)
+    info_keys.sort()
     for index, name in enumerate(info_keys):
         column = Column(np_variables[:, index], name=name)
         best_model_table.add_column(column)

pcigale/analysis_modules/pdf_analysis/workers.py

@@ -124,7 +124,7 @@ def init_analysis(params, filters, analysed, redshifts, fluxes, variables,
     global gbl_redshifts, gbl_w_redshifts, gbl_analysed_averages
     global gbl_analysed_std, gbl_best_fluxes, gbl_best_parameters
     global gbl_best_chi2, gbl_best_chi2_red, gbl_save, gbl_n_obs
-    global gbl_lim_flag, gbl_phase
+    global gbl_lim_flag, gbl_phase, gbl_keys
 
     gbl_analysed_averages = np.ctypeslib.as_array(analysed_averages[0])
     gbl_analysed_averages = gbl_analysed_averages.reshape(analysed_averages[1])
@@ -151,6 +151,7 @@ def init_analysis(params, filters, analysed, redshifts, fluxes, variables,
 
     gbl_n_obs = n_obs
     gbl_phase = phase
+    gbl_keys = None
 
 
 def sed(idx):
@@ -210,6 +211,7 @@ def analysis(idx, obs):
 
     """
     # Tolerance threshold under which any flux or error is considered as 0.
+    global gbl_keys
     tolerance = 1e-12
 
     obs_fluxes = np.array([obs[name] for name in gbl_filters])
@@ -382,7 +384,11 @@ def analysis(idx, obs):
 
         gbl_best_fluxes[idx, :] = gbl_model_fluxes[wz[0][wvalid[0][best_index]], :] \
                       *norm_facts[best_index]
-        gbl_best_parameters[idx, :] = list(sed.info.values())
+
+        if gbl_keys is None:
+            gbl_keys = list(sed.info.keys())
+            gbl_keys.sort()
+        gbl_best_parameters[idx, :] = np.array([sed.info[k] for k in gbl_keys])
         gbl_best_chi2[idx] = chi2_[best_index]
         gbl_best_chi2_red[idx] = chi2_[best_index] / obs_fluxes.size
 

pcigale/analysis_modules/savefluxes/__init__.py

@@ -118,7 +118,7 @@ class SaveFluxes(AnalysisModule):
         # Retrieve an arbitrary SED to obtain the list of output parameters
         warehouse = SedWarehouse()
         sed = warehouse.get_sed(creation_modules, params.from_index(0))
-        info = sed.info
+        info = list(sed.info.keys())
         n_info = len(sed.info)
         del warehouse, sed
 

pcigale/analysis_modules/savefluxes/workers.py

@@ -37,7 +37,7 @@ def init_fluxes(params, filters, save_sed, fluxes, info, t_begin, n_computed):
     """
     global gbl_model_fluxes, gbl_model_info, gbl_n_computed, gbl_t_begin
     global gbl_params, gbl_previous_idx, gbl_filters, gbl_save_sed
-    global gbl_warehouse
+    global gbl_warehouse, gbl_keys
 
     gbl_model_fluxes = np.ctypeslib.as_array(fluxes[0])
     gbl_model_fluxes = gbl_model_fluxes.reshape(fluxes[1])
@@ -58,6 +58,7 @@ def init_fluxes(params, filters, save_sed, fluxes, info, t_begin, n_computed):
 
     gbl_warehouse = SedWarehouse()
 
+    gbl_keys = None
 
 def fluxes(idx):
     """Worker process to retrieve a SED and affect the relevant data to shared
@@ -70,7 +71,7 @@ def fluxes(idx):
         handler.
 
     """
-    global gbl_previous_idx
+    global gbl_previous_idx, gbl_keys
     if gbl_previous_idx > -1:
         gbl_warehouse.partial_clear_cache(
             gbl_params.index_module_changed(gbl_previous_idx, idx))
@@ -89,7 +90,10 @@ def fluxes(idx):
                                  gbl_filters])
 
     gbl_model_fluxes[idx, :] = model_fluxes
-    gbl_model_info[idx, :] = list(sed.info.values())
+    if gbl_keys is None:
+        gbl_keys = list(sed.info.keys())
+        gbl_keys.sort()
+    gbl_model_info[idx, :] = np.array([sed.info[k] for k in gbl_keys])
 
     with gbl_n_computed.get_lock():
         gbl_n_computed.value += 1