rename attribute to batch_logl_bounds to avoid confusion with bounds

Author: segasai

CHANGELOG.md

@@ -16,6 +16,7 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 - Remove custom update_func
 - Remove hslice
 - Remove 'user-defined' proposal distribution
+- Change internal attribute from batch_bounds to batch_logl_bounds
 ### Fixed
 
 [2.1.5 - 2024-12-17]

py/dynesty/dynamicsampler.py

@@ -861,7 +861,7 @@ def results(self):
         for k in [
                 'nc', 'v', 'id', 'batch', 'it', 'u', 'n', 'logwt', 'logl',
                 'logvol', 'logz', 'logzvar', 'h', 'batch_nlive',
-                'batch_bounds', 'blob'
+                'batch_logl_bounds', 'blob'
         ]:
             d[k] = np.array(self.saved_run[k])
 
@@ -874,7 +874,7 @@ def results(self):
                 results.append(('samples_' + k, d[k]))
             for k in [
                     'logwt', 'logl', 'logvol', 'logz', 'batch_nlive',
-                    'batch_bounds', 'blob'
+                    'batch_logl_bounds', 'blob'
             ]:
                 results.append((k, d[k]))
             results.append(('logzerr', np.sqrt(d['logzvar'])))
@@ -1209,7 +1209,7 @@ def sample_initial(self,
                                            dtype=int)  # batch
 
         self.saved_run['batch_nlive'].append(self.nlive_init)  # initial nlive
-        self.saved_run['batch_bounds'].append(
+        self.saved_run['batch_logl_bounds'].append(
             (-np.inf, np.inf))  # initial bounds
 
         self.internal_state = DynamicSamplerStatesEnum.BASE_DONE
@@ -1473,7 +1473,7 @@ def combine_runs(self):
         nnew = len(new_d['n'])
         llmin, llmax = self.new_logl_min, self.new_logl_max
 
-        old_batch_bounds = self.saved_run['batch_bounds']
+        old_batch_logl_bounds = self.saved_run['batch_logl_bounds']
         old_batch_nlive = self.saved_run['batch_nlive']
         # Reset saved results.
         del self.saved_run
@@ -1587,7 +1587,8 @@ def combine_runs(self):
 
         # Saved batch quantities.
         self.saved_run['batch_nlive'] = old_batch_nlive + [(max(new_d['n']))]
-        self.saved_run['batch_bounds'] = old_batch_bounds + [((llmin, llmax))]
+        self.saved_run['batch_logl_bounds'] = old_batch_logl_bounds + [(
+            (llmin, llmax))]
 
     def run_nested(self,
                    nlive_init=None,

py/dynesty/dynesty.py

@@ -528,7 +528,6 @@ def _common_sampler_init(*,
             sample, InternalSampler):
         raise ValueError("Unknown sampling method: '{0}'".format(sample))
 
-    # TODO change this check to deal with new sampler interface
     if ncdim != ndim and (isinstance(sample, SliceSampler)
                           or isinstance(sample, RSliceSampler)):
         raise ValueError('ncdim unsupported for slice sampling')

py/dynesty/utils.py

@@ -283,7 +283,7 @@ def __init__(self, dynamic=False):
                 # these are special since their length
                 # is == the number of batches
                 'batch_nlive',  # number of live points added in batch
-                'batch_bounds'  # loglikelihood bounds used in batch
+                'batch_logl_bounds'  # loglikelihood bounds used in batch
             ])
         for k in keys:
             D[k] = []
@@ -606,7 +606,7 @@ def print_fn_fallback(results,
      'niter'),
     ('samples_batch', 'array[int]',
      "Tracks the batch during which the samples were proposed", 'niter'),
-    ('batch_bounds', 'array[tuple]',
+    ('batch_logl_bounds', 'array[tuple]',
      "The log-likelihood bounds used to run a batch.", 'nbatch'),
     ('batch_nlive', 'array[int]',
      "The number of live points used for  given batch", 'nbatch'),
@@ -1363,7 +1363,7 @@ def resample_run(res, rstate=None, return_idx=False):
         # Check if the number of live points explicitly changes.
         samples_n = res.samples_n
         samples_batch = res.samples_batch
-        batch_bounds = res.batch_bounds
+        batch_logl_bounds = res.batch_logl_bounds
         added_final_live = True
     else:
         # If the number of live points is constant, compute `samples_n` and
@@ -1380,8 +1380,8 @@ def resample_run(res, rstate=None, return_idx=False):
             raise ValueError("Final number of samples differs from number of "
                              "iterations and number of live points.")
         samples_batch = np.zeros(len(samples_n), dtype=int)
-        batch_bounds = np.array([(-np.inf, np.inf)])
-    batch_llmin = batch_bounds[:, 0]
+        batch_logl_bounds = np.array([(-np.inf, np.inf)])
+    batch_llmin = batch_logl_bounds[:, 0]
     # Identify unique particles that make up each strand.
     ids = np.unique(res.samples_id)
 
@@ -1629,7 +1629,7 @@ def unravel_run(res, print_progress=True):
         # Add on batch information (if available).
         try:
             rdict['samples_batch'] = res.samples_batch[strand]
-            rdict['batch_bounds'] = res.batch_bounds
+            rdict['batch_logl_bounds'] = res.batch_logl_bounds
         except AttributeError:
             pass
 
@@ -1858,15 +1858,15 @@ def _prepare_for_merge(res):
                              "iterations and number of live points in `res1`.")
 
     # Batch information (if available).
-    # note we also check for existance of batch_bounds
+    # note we also check for existance of batch_logl_bounds
     # because unravel_run makes 'static' runs of 1 livepoint
     # but some will have bounds
-    if res.isdynamic() or 'batch_bounds' in res.keys():
+    if res.isdynamic() or 'batch_logl_bounds' in res.keys():
         run_info['batch'] = res.samples_batch
-        run_info['bounds'] = res.batch_bounds
+        run_info['batch_logl_bounds'] = res.batch_logl_bounds
     else:
         run_info['batch'] = np.zeros(nrun, dtype=int)
-        run_info['bounds'] = np.array([(-np.inf, np.inf)])
+        run_info['batch_logl_bounds'] = np.array([(-np.inf, np.inf)])
     return run_nlive, run_info
 
 
@@ -1910,20 +1910,22 @@ def _merge_two(res1, res2, compute_aux=False):
 
     # These are merged batch bounds
     combined_bounds = np.unique(np.concatenate(
-        (base_info['bounds'], new_info['bounds'])),
+        (base_info['batch_logl_bounds'], new_info['batch_logl_bounds'])),
                                 axis=0)
     # Here we try to find where the new bounds are in the combined bounds
     new_bound_map = {}
     base_bound_map = {}
-    for i in range(len(new_info['bounds'])):
+    for i in range(len(new_info['batch_logl_bounds'])):
         new_bound_map[i] = np.where(
-            np.all(new_info['bounds'][i] == combined_bounds, axis=1))[0][0]
-    for i in range(len(base_info['bounds'])):
+            np.all(new_info['batch_logl_bounds'][i] == combined_bounds,
+                   axis=1))[0][0]
+    for i in range(len(base_info['batch_logl_bounds'])):
         base_bound_map[i] = np.where(
-            np.all(base_info['bounds'][i] == combined_bounds, axis=1))[0][0]
+            np.all(base_info['batch_logl_bounds'][i] == combined_bounds,
+                   axis=1))[0][0]
 
-    base_lowedge = np.min(base_info['bounds'][base_info['batch']])
-    new_lowedge = np.min(new_info['bounds'][new_info['batch']])
+    base_lowedge = np.min(base_info['batch_logl_bounds'][base_info['batch']])
+    new_lowedge = np.min(new_info['batch_logl_bounds'][new_info['batch']])
 
     # Iteratively walk through both set of samples to simulate
     # a combined run.
@@ -2021,7 +2023,7 @@ def _merge_two(res1, res2, compute_aux=False):
              samples=np.asarray(combined_info['v']),
              logl=np.asarray(combined_info['logl']),
              logvol=np.asarray(combined_info['logvol']),
-             batch_bounds=np.asarray(combined_bounds),
+             batch_logl_bounds=np.asarray(combined_bounds),
              blob=np.asarray(combined_info['blob']))
 
     for curk in ['id', 'it', 'n', 'u', 'batch']:

tests/test_misc.py

@@ -64,8 +64,8 @@ def test_maxcall():
     sampler.run_nested(dlogz_init=1, maxcall=1000, print_progress=printing)
 
 
-@pytest.mark.parametrize('dynamic,with_pool',
-                         itertools.product([True, False], [True, False]))
+@pytest.mark.parametrize('dynamic', [True, False])
+@pytest.mark.parametrize('with_pool', [True, False])
 def test_pickle(dynamic, with_pool):
     # test of pickling functionality
     ndim = 2
@@ -501,7 +501,7 @@ def test_maxiter_batch():
             # I am finding the the first iteration with the batch
             # [-inf, something]. Then I'm setting maxiter to be just above
             # that iteration
-            b1 = np.where(~np.isfinite(dres2.batch_bounds[:, 0]))[0][1]
+            b1 = np.where(~np.isfinite(dres2.batch_logl_bounds[:, 0]))[0][1]
             maxiter = np.min(
                 np.array(dsampler2.saved_run['it'])[
                     dsampler2.saved_run['batch'] == b1]) + nlive // 2
@@ -603,7 +603,8 @@ def test_verify_batch():
                                     1].min() > d0.results['samples_it'].max()
     # checke that the iterations are set correctly
     assert d1.ncall > d0.ncall
-    assert len(d1.results.batch_bounds) > len(d0.results.batch_bounds)
+    assert len(d1.results.batch_logl_bounds) > len(
+        d0.results.batch_logl_bounds)
 
 
 @pytest.mark.parametrize('dynamic', [False, True])