Merge pull request #110 from FAST-HEP/BK_issue_109-fix_explode_strings

Fix explode method for string arrays

CHANGELOG.md

@@ -5,6 +5,10 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
 ## [Unreleased]
+## [0.16.1] - 2020-02-16
+### Fixed
+- Unit test that was broken by Pandas >1.0.0
+- Bug in explode function when an dimension contains strings, issue #109, PR #110 [@BenKrikler](https://github.com/benkrikler)
 
 ## [0.16.0] - 2019-11-1
 ### Added

fast_carpenter/summary/binned_dataframe.py

@@ -262,6 +262,9 @@ def _bin_values(data, dimensions, binnings, weights, out_dimensions=None, out_we
     return histogram
 
 
+_explodable_types = (tuple, list, np.ndarray)
+
+
 def explode(df):
     """
     Based on this answer:
@@ -270,6 +273,8 @@ def explode(df):
     """
     # get the list columns
     lst_cols = [col for col, dtype in df.dtypes.items() if is_object_dtype(dtype)]
+    # Be more specific about which objects are ok
+    lst_cols = [col for col in lst_cols if isinstance(df[col][0], _explodable_types)]
     if not lst_cols:
         return df
 

fast_carpenter/version.py

@@ -12,5 +12,5 @@ def split_version(version):
     return tuple(result)
 
 
-__version__ = '0.16.0'
+__version__ = '0.16.1'
 version_info = split_version(__version__) # noqa

setup.cfg

@@ -1,5 +1,5 @@
 [bumpversion]
-current_version = 0.16.0
+current_version = 0.16.1
 commit = True
 tag = False
 
@@ -18,4 +18,3 @@ test = pytest
 
 [tool:pytest]
 collect_ignore = ['setup.py']
-

tests/summary/test_binned_dataframe.py

@@ -130,11 +130,15 @@ def test_binneddataframe_run_twice_data_mc(run_twice_data_mc, dataset_col, pad_m
     results = collector._prepare_output(dataset_readers_list)
 
     assert results.index.nlevels == 2 + int(dataset_col)
-    if pad_missing or not dataset_col:
+    if tuple(map(int, pd.__version__.split("."))) >= (1, 0, 0):
         length = (4 * 12) * (1 + int(dataset_col))
     else:
-        length = 95  # When dataset_col True and pad_missing False one bin is missing
-        assert len(results) == length
+        # Pre Pandas 1.0.0 the following lengths were needed.
+        if pad_missing or not dataset_col:
+            length = (4 * 12) * (1 + int(dataset_col))
+        else:
+            length = 95  # When dataset_col True and pad_missing False one bin is missing
+    assert len(results) == length
 
     totals = results.sum()
     # Based on: events->Draw("Jet_Py", "", "goff")
@@ -233,3 +237,12 @@ def test_explode():
                         })
     exploded = bdf.explode(df2)
     assert len(exploded) == 8
+
+    df = pd.DataFrame({'number': [1, 8, 3], 'string': ['one', 'eight', 'three']})
+    exploded = bdf.explode(df)
+    assert len(exploded) == 3
+
+    df["list"] = [list(range(i)) for i in df.number]
+    exploded = bdf.explode(df)
+    assert len(exploded) == 1 + 8 + 3
+    assert np.array_equal(exploded.list, [0, 0, 1, 2, 3, 4, 5, 6, 7, 0, 1, 2])