split_proxydb: clear the tags of any previous split

cfr/reconjob.py

@@ -289,7 +289,7 @@ def clear_proxydb_tags(self, verbose=False):
             verbose (bool, optional): print verbose information. Defaults to False.
         '''
         for pid, pobj in self.proxydb.records.items():
-            pobj.tags = []
+            pobj.tags = set()
 
         if verbose:
             p_success(f'>>> job.proxydb updated with tags cleared')
@@ -304,6 +304,12 @@ def split_proxydb(self, tag='calibrated', assim_frac=None, seed=0, verbose=False
             verbose (bool, optional): print verbose information. Defaults to False.
         '''
         assim_frac = self.io_cfg('proxy_assim_frac', assim_frac, default=0.75, verbose=verbose)
+
+        # remove the tags of any previous split, so that repeated calls (e.g., one per seed
+        # in run_da_mc) give independent splits instead of accumulating 'assim' records
+        for pid, pobj in self.proxydb.records.items():
+            pobj.tags = set(pobj.tags) - {'assim', 'eval'}
+
         target_pdb = self.proxydb.filter(by='tag', keys=[tag])
 
         nrec_assim = int(target_pdb.nrec * assim_frac)

tests/test_reconjob.py

@@ -252,6 +252,18 @@ def test_same_seed_gives_same_split(self):
         tags2 = {pid: frozenset(r.tags) for pid, r in job.proxydb.records.items()}
         assert tags1 == tags2
 
+    def test_resplit_replaces_previous_split(self):
+        # A second split (e.g., the next seed in run_da_mc) must not keep the
+        # 'assim'/'eval' tags of the first one
+        job = self._make_calibrated_job(n=4)
+        for seed in range(10):
+            job.split_proxydb(seed=seed, assim_frac=0.5)
+            assim = {pid for pid, r in job.proxydb.records.items() if 'assim' in r.tags}
+            evals = {pid for pid, r in job.proxydb.records.items() if 'eval' in r.tags}
+            assert len(assim) == 2
+            assert len(evals) == 2
+            assert assim.isdisjoint(evals)
+
     def test_all_tagged_records_assigned(self):
         # Every calibrated record gets either assim or eval — none left unassigned
         job = self._make_calibrated_job(n=4)