From 035d57febdaacf31c21d8c48dac57c108bfbf36f Mon Sep 17 00:00:00 2001
From: Felix Claessen <felix@seita.nl>
Date: Fri, 14 Aug 2020 18:00:58 +0200
Subject: [PATCH 01/15] Call finalize in all constructors to inherit metadata.
 Prepare tests for upcoming pandas==1.1.1 functionality: inherit metadata when
 resampling and grouping.

---
 requirements.txt                       |  0
 timely_beliefs/beliefs/__init__.py     |  5 ++-
 timely_beliefs/beliefs/classes.py      | 21 +++++++-----
 timely_beliefs/tests/test_belief_io.py | 44 ++++++++++++++++++++++++++
 4 files changed, 61 insertions(+), 9 deletions(-)
 create mode 100644 requirements.txt

diff --git a/requirements.txt b/requirements.txt
new file mode 100644
index 00000000..e69de29b
diff --git a/timely_beliefs/beliefs/__init__.py b/timely_beliefs/beliefs/__init__.py
index 54310b32..469a6f08 100644
--- a/timely_beliefs/beliefs/__init__.py
+++ b/timely_beliefs/beliefs/__init__.py
@@ -8,6 +8,7 @@
 from datetime import datetime, timedelta
 from typing import List
 
+import pandas as pd
 from pandas.api.extensions import register_dataframe_accessor
 
 
@@ -120,7 +121,9 @@ def number_of_sources(self):
     @property
     def number_of_probabilistic_beliefs(self) -> int:
         """Return the number of beliefs in the BeliefsDataFrame that are probabilistic (more than 1 unique value)."""
-        df = self._obj.for_each_belief(df=self._obj).nunique(dropna=True)
+        df = self._obj.for_each_belief(
+            df=self._obj, fnc=pd.DataFrame.nunique, dropna=True
+        )
         return len(df[df > 1].max(axis=1).dropna())
 
     @property
diff --git a/timely_beliefs/beliefs/classes.py b/timely_beliefs/beliefs/classes.py
index bc706217..94a9e5dc 100644
--- a/timely_beliefs/beliefs/classes.py
+++ b/timely_beliefs/beliefs/classes.py
@@ -375,7 +375,11 @@ class BeliefsSeries(pd.Series):
 
     @property
     def _constructor(self):
-        return BeliefsSeries
+        def f(*args, **kwargs):
+            """ Call __finalize__() after construction to inherit metadata. """
+            return BeliefsSeries(*args, **kwargs).__finalize__(self, method="inherit")
+
+        return f
 
     @property
     def _constructor_expanddim(self):
@@ -436,7 +440,13 @@ class BeliefsDataFrame(pd.DataFrame):
 
     @property
     def _constructor(self):
-        return BeliefsDataFrame
+        def f(*args, **kwargs):
+            """ Call __finalize__() after construction to inherit metadata. """
+            return BeliefsDataFrame(*args, **kwargs).__finalize__(
+                self, method="inherit"
+            )
+
+        return f
 
     @property
     def _constructor_sliced(self):
@@ -1017,12 +1027,7 @@ def resample_events(
                         "cumulative_probability": "prod",  # assume independent variables
                     }
                 )
-                # make a new BeliefsDataFrame, because agg() doesn't behave nicely for subclassed DataFrames
-                df = BeliefsDataFrame(
-                    df.reset_index(),
-                    sensor=self.sensor,
-                    event_resolution=event_resolution,
-                )
+                df.event_resolution = event_resolution
             else:
                 # upsample
                 new_index = pd.date_range(
diff --git a/timely_beliefs/tests/test_belief_io.py b/timely_beliefs/tests/test_belief_io.py
index a275aff2..1925af71 100644
--- a/timely_beliefs/tests/test_belief_io.py
+++ b/timely_beliefs/tests/test_belief_io.py
@@ -385,6 +385,50 @@ def test_slicing_retains_metadata(drop_level):
     metadata = {md: getattr(example_df, md) for md in METADATA}
     df = df.xs("2000-01-03 10:00:00+00:00", level="event_start", drop_level=drop_level)
     print(df)
+    assert isinstance(df, tb.BeliefsDataFrame)
+    for md in metadata:
+        assert getattr(df, md) == metadata[md]
+
+
+@pytest.mark.parametrize("resolution", [timedelta(minutes=30), timedelta(hours=2)])
+def test_agg_resampling_retains_metadata(resolution):
+    """
+    Test whether aggregate resampling retains the metadata.
+
+    Fails with pandas==1.0.0
+    Succeeds with pandas==1.1.0
+    """
+    df = example_df
+    metadata = {md: getattr(example_df, md) for md in METADATA}
+    df = df.resample(resolution, level="event_start").mean()
+    print(df)
+    assert isinstance(df, tb.BeliefsDataFrame)
+    for md in metadata:
+        # if md == "event_resolution":
+        #     assert df.event_resolution == resolution
+        # else:  # todo: the event_resolution metadata is only updated when resampling using df.resample_events(). A reason to override the original resample method, or otherwise something to document.
+        assert getattr(df, md) == metadata[md]
+
+
+def test_groupby_retains_metadata():
+    """ Test whether grouping by index level retains the metadata.
+
+    Succeeds with pandas==1.0.0
+    Fails with pandas==1.1.0
+    Should be fixed with https://github.com/pandas-dev/pandas/pull/35688
+    """
+    df = example_df
+    metadata = {md: getattr(example_df, md) for md in METADATA}
+
+    def assert_function(x):
+        print(x)
+        assert isinstance(x, tb.BeliefsDataFrame)
+        for md in metadata:
+            assert getattr(x, md) == metadata[md]
+        return x
+
+    df = df.groupby(level="event_start").apply(lambda x: assert_function(x))
+    assert isinstance(df, tb.BeliefsDataFrame)
     for md in metadata:
         assert getattr(df, md) == metadata[md]
 

From c56d7cccb45155f3bcce1b22fe28cf2d846e89bd Mon Sep 17 00:00:00 2001
From: "F.N. Claessen" <felix@seita.nl>
Date: Thu, 26 Nov 2020 16:22:56 +0100
Subject: [PATCH 02/15] Update reference in test. Add pandas test.

---
 timely_beliefs/tests/test_belief_io.py | 14 ++++++++++++--
 1 file changed, 12 insertions(+), 2 deletions(-)

diff --git a/timely_beliefs/tests/test_belief_io.py b/timely_beliefs/tests/test_belief_io.py
index 1925af71..6a8f611b 100644
--- a/timely_beliefs/tests/test_belief_io.py
+++ b/timely_beliefs/tests/test_belief_io.py
@@ -411,11 +411,11 @@ def test_agg_resampling_retains_metadata(resolution):
 
 
 def test_groupby_retains_metadata():
-    """ Test whether grouping by index level retains the metadata.
+    """Test whether grouping by index level retains the metadata.
 
     Succeeds with pandas==1.0.0
     Fails with pandas==1.1.0
-    Should be fixed with https://github.com/pandas-dev/pandas/pull/35688
+    Should be fixed with https://github.com/pandas-dev/pandas/pull/37461
     """
     df = example_df
     metadata = {md: getattr(example_df, md) for md in METADATA}
@@ -477,3 +477,13 @@ def test_init_from_beliefs_series():
         bdf, df_copy
     )  # new bdf retains altered column of original bdf
     pd.testing.assert_series_equal(s, s_copy)  # input BeliefsSeries was not altered
+
+
+def test_groupby_retains_attribute():
+    df = pd.DataFrame([[1, 2], [3, 4]], columns=["x", "y"])
+    df.a = "b"
+    assert df.a == "b"
+    df = df.groupby("x").apply(lambda x: x)
+    assert df.a == "b"
+    df = df.groupby("x").sum()
+    assert df.a == "b"

From 5b24bec904c0a59188b2f6e01e51314ffd715314 Mon Sep 17 00:00:00 2001
From: "F.N. Claessen" <felix@seita.nl>
Date: Mon, 7 Dec 2020 17:21:37 +0100
Subject: [PATCH 03/15] Finish rebase.

---
 requirements.txt | 0
 1 file changed, 0 insertions(+), 0 deletions(-)
 delete mode 100644 requirements.txt

diff --git a/requirements.txt b/requirements.txt
deleted file mode 100644
index e69de29b..00000000

From 96163ee1f90b0652a4f7aa835399028ae15f824c Mon Sep 17 00:00:00 2001
From: "F.N. Claessen" <felix@seita.nl>
Date: Mon, 7 Dec 2020 20:35:13 +0100
Subject: [PATCH 04/15] Complete the ordering of BeliefSources by name.

---
 timely_beliefs/sources/classes.py | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/timely_beliefs/sources/classes.py b/timely_beliefs/sources/classes.py
index fc7d609b..feae8d89 100644
--- a/timely_beliefs/sources/classes.py
+++ b/timely_beliefs/sources/classes.py
@@ -1,3 +1,4 @@
+from functools import total_ordering
 from typing import Union
 
 from sqlalchemy import Column, Integer, String
@@ -5,10 +6,11 @@
 from timely_beliefs.db_base import Base
 
 
+@total_ordering
 class BeliefSource(object):
 
     """
-    A belief source is any data-creating entitiy such as a user, a ML model or a script.
+    A belief source is any data-creating entity such as a user, a ML model or a script.
     """
 
     name: str

From 939e023173e34e19b73b9fa2fa345e47b3ca753c Mon Sep 17 00:00:00 2001
From: "F.N. Claessen" <felix@seita.nl>
Date: Wed, 9 Dec 2020 17:16:41 +0100
Subject: [PATCH 05/15] Fix computation order.

---
 timely_beliefs/beliefs/classes.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/timely_beliefs/beliefs/classes.py b/timely_beliefs/beliefs/classes.py
index 94a9e5dc..952954f4 100644
--- a/timely_beliefs/beliefs/classes.py
+++ b/timely_beliefs/beliefs/classes.py
@@ -1032,7 +1032,7 @@ def resample_events(
                 # upsample
                 new_index = pd.date_range(
                     start=df.index[0],
-                    periods=len(df) * self.event_resolution // event_resolution,
+                    periods=len(df) * (self.event_resolution // event_resolution),
                     freq=event_resolution,
                     name="event_start",
                 )

From 1007af85c189465f6842b23685f3f71e312a0222 Mon Sep 17 00:00:00 2001
From: "F.N. Claessen" <felix@seita.nl>
Date: Wed, 9 Dec 2020 17:18:29 +0100
Subject: [PATCH 06/15] Workaround for aggregation function when resampling.

---
 timely_beliefs/beliefs/classes.py | 54 ++++++++++++++++++++++---------
 1 file changed, 39 insertions(+), 15 deletions(-)

diff --git a/timely_beliefs/beliefs/classes.py b/timely_beliefs/beliefs/classes.py
index 952954f4..872a554f 100644
--- a/timely_beliefs/beliefs/classes.py
+++ b/timely_beliefs/beliefs/classes.py
@@ -1,9 +1,8 @@
 import math
 from datetime import datetime, timedelta
-from typing import Any, Callable, List, Optional, Tuple, Union
+from typing import Any, Callable, Dict, List, Optional, Tuple, Union
 
 import altair as alt
-import numpy as np
 import pandas as pd
 from pandas.core.groupby import DataFrameGroupBy
 from sqlalchemy import Column, DateTime, Float, ForeignKey, Integer, Interval
@@ -1012,24 +1011,23 @@ def resample_events(
             and keep_only_most_recent_belief
             and df.lineage.number_of_sources == 1
         ):
-            df = df.reset_index(
-                level=[belief_timing_col, "source", "cumulative_probability"]
-            )
             if event_resolution > self.event_resolution:
                 # downsample
-                df = df.resample(event_resolution).agg(
-                    {
-                        "event_value": np.nanmean,
-                        "source": "first",  # keep the only source
-                        belief_timing_col: "max"
-                        if belief_timing_col == "belief_time"
-                        else "min",  # keep only most recent belief
-                        "cumulative_probability": "prod",  # assume independent variables
-                    }
-                )
+                column_functions = {
+                    "event_value": "mean",
+                    "source": "first",  # keep the only source
+                    belief_timing_col: "max"
+                    if belief_timing_col == "belief_time"
+                    else "min",  # keep only most recent belief
+                    "cumulative_probability": "prod",  # assume independent variables
+                }
+                df = resample_beliefs_data_frame(df, event_resolution, column_functions)
                 df.event_resolution = event_resolution
             else:
                 # upsample
+                df = df.reset_index(
+                    level=[belief_timing_col, "source", "cumulative_probability"]
+                )
                 new_index = pd.date_range(
                     start=df.index[0],
                     periods=len(df) * (self.event_resolution // event_resolution),
@@ -1472,3 +1470,29 @@ def assign_sensor_and_event_resolution(df, sensor, event_resolution):
         if sensor
         else None
     )
+
+
+def resample_beliefs_data_frame(
+    df: BeliefsDataFrame, event_resolution: timedelta, col_att_dict: Dict[str, str]
+) -> BeliefsDataFrame:
+    """Because df.resample().agg() doesn't behave nicely for subclassed DataFrames,
+    we aggregate each index level and column separately against the resampled event_start level,
+    and then recombine them afterwards.
+    """
+    belief_timing_col = (
+        "belief_time" if "belief_time" in df.index.names else "belief_horizon"
+    )
+    event_timing_col = "event_start" if "event_start" in df.index.names else "event_end"
+    return pd.concat(
+        [
+            getattr(
+                df.reset_index()
+                .set_index(event_timing_col)[col]
+                .to_frame()
+                .resample(event_resolution),
+                att,
+            )()
+            for col, att in col_att_dict.items()
+        ],
+        axis=1,
+    ).set_index([belief_timing_col, "source", "cumulative_probability"], append=True)

From e72ef14bd115c619b4aea8cf201f0bf9a7631ac7 Mon Sep 17 00:00:00 2001
From: "F.N. Claessen" <felix@seita.nl>
Date: Wed, 9 Dec 2020 17:21:17 +0100
Subject: [PATCH 07/15] Fix tests: - Separate tests for mean resampling and
 aggregate resampling. - Separate tests for temporary attributes and subclass
 attributes. - Test metadata propagation for groupby of subclassed DataFrames.

---
 timely_beliefs/tests/test_belief_io.py | 112 +++++++++++++++++++++++--
 1 file changed, 106 insertions(+), 6 deletions(-)

diff --git a/timely_beliefs/tests/test_belief_io.py b/timely_beliefs/tests/test_belief_io.py
index 6a8f611b..236f4312 100644
--- a/timely_beliefs/tests/test_belief_io.py
+++ b/timely_beliefs/tests/test_belief_io.py
@@ -1,6 +1,7 @@
 import os
 from datetime import datetime, timedelta
 
+import numpy as np
 import pandas as pd
 import pytest
 import pytz
@@ -391,9 +392,9 @@ def test_slicing_retains_metadata(drop_level):
 
 
 @pytest.mark.parametrize("resolution", [timedelta(minutes=30), timedelta(hours=2)])
-def test_agg_resampling_retains_metadata(resolution):
+def test_mean_resampling_retains_metadata(resolution):
     """
-    Test whether aggregate resampling retains the metadata.
+    Test whether mean resampling retains the metadata.
 
     Fails with pandas==1.0.0
     Succeeds with pandas==1.1.0
@@ -410,6 +411,34 @@ def test_agg_resampling_retains_metadata(resolution):
         assert getattr(df, md) == metadata[md]
 
 
+@pytest.mark.parametrize("resolution", [timedelta(minutes=30), timedelta(hours=2)])
+def _test_agg_resampling_retains_metadata(resolution):
+    """
+    Test whether aggregate resampling retains the metadata.
+
+    Fails with pandas==1.1.5
+    """
+    df = example_df
+    metadata = {md: getattr(example_df, md) for md in METADATA}
+    df = df.reset_index(level=["belief_time", "source", "cumulative_probability"])
+    df = df.resample(resolution).agg(
+        {
+            "event_value": np.nanmean,
+            "source": "first",  # keep the only source
+            "belief_time": "max",  # keep the latest belief
+            "cumulative_probability": "prod",  # assume independent variables
+        }
+    )
+    df = df.set_index(["belief_time", "source", "cumulative_probability"], append=True)
+    print(df)
+    assert isinstance(df, tb.BeliefsDataFrame)
+    for md in metadata:
+        # if md == "event_resolution":
+        #     assert df.event_resolution == resolution
+        # else:  # todo: the event_resolution metadata is only updated when resampling using df.resample_events(). A reason to override the original resample method, or otherwise something to document.
+        assert getattr(df, md) == metadata[md]
+
+
 def test_groupby_retains_metadata():
     """Test whether grouping by index level retains the metadata.
 
@@ -479,11 +508,82 @@ def test_init_from_beliefs_series():
     pd.testing.assert_series_equal(s, s_copy)  # input BeliefsSeries was not altered
 
 
-def test_groupby_retains_attribute():
+def test_groupby_does_not_retain_temporary_attribute():
     df = pd.DataFrame([[1, 2], [3, 4]], columns=["x", "y"])
     df.a = "b"
     assert df.a == "b"
-    df = df.groupby("x").apply(lambda x: x)
-    assert df.a == "b"
-    df = df.groupby("x").sum()
+    df2 = df.groupby("x").apply(lambda x: x)
+    assert not hasattr(df2, "a")
+    df3 = df.groupby("x").sum()
+    assert not hasattr(df3, "a")
+
+
+@pytest.mark.parametrize(
+    "att, args",
+    [
+        # ("all", []),
+        # ("any", []),
+        # ("count", []),
+        ("first", []),
+        ("last", []),
+        ("max", []),
+        # ("mean", []),
+        # ("median", []),
+        ("min", []),
+        ("prod", []),
+        # ("sem", []),
+        # ("size", []),
+        # ("std", []),
+        ("sum", []),
+        # ("var", []),
+        # ("apply", [lambda x: x]),
+        # ("apply", [np.max])
+        # ("apply", [np.min])
+        # ("apply", [np.nanmean]),
+        ("agg", ["first"]),
+        ("agg", ["max"]),
+        # ("agg", ["mean"]),
+        ("agg", ["min"]),
+        ("agg", ["sum"]),
+        # ("agg", [{"y": "min"}]),
+        # ("agg", [{"x": "min", "y": "max"}]),
+    ],
+)
+def test_groupby_retains_subclass_attribute(att, args):
+    """Checks on metadata propagation for subclassed DataFrames under groupby operations.
+
+    Commented-out parameter combinations fail with pandas==1.1.5
+    """
+
+    METADATA = ["a"]
+
+    class SubclassedSeries(pd.Series):
+
+        _metadata = METADATA
+
+        @property
+        def _constructor(self):
+            return SubclassedSeries
+
+        @property
+        def _constructor_expanddim(self):
+            return SubclassedDataFrame
+
+    class SubclassedDataFrame(pd.DataFrame):
+
+        _metadata = METADATA
+
+        @property
+        def _constructor(self):
+            return SubclassedDataFrame
+
+        @property
+        def _constructor_sliced(self):
+            return SubclassedSeries
+
+    df = SubclassedDataFrame([[1, 2], [3, 4]], columns=["x", "y"])
+    df.a = "b"
     assert df.a == "b"
+    df2 = getattr(df.groupby("x"), att)(*args)
+    print(df2)
+    assert df2.a == "b"

From 74379967e38ff01ee6c9f331773da9e8cfa0b39b Mon Sep 17 00:00:00 2001
From: "F.N. Claessen" <felix@seita.nl>
Date: Thu, 10 Dec 2020 11:44:19 +0100
Subject: [PATCH 08/15] Update pandas dependency. Drop support for
 pandas<1.1.5. Bump timely-beliefs version with major release.

---
 README.md | 4 +++-
 setup.py  | 4 ++--
 2 files changed, 5 insertions(+), 3 deletions(-)

diff --git a/README.md b/README.md
index bfe6eb71..def0322c 100644
--- a/README.md
+++ b/README.md
@@ -148,6 +148,8 @@ This will create an interactive Vega-Lite chart like the one in the screenshot a
 
 ## Development
 
-We welcome other contributions to timely_beliefs.
+The `timely_beliefs` package runs on `pandas>=1.1.5`.
+Contact us if you need support for older versions.
+We welcome other contributions to `timely_beliefs`.
 
 [See our developer docs for details.](dev/dev.md)
diff --git a/setup.py b/setup.py
index a22d0af5..c830522e 100644
--- a/setup.py
+++ b/setup.py
@@ -14,10 +14,10 @@
         "uncertainty",
         "lineage",
     ],
-    version="0.1.3",
+    version="1.0.0",
     install_requires=[
         "pytz",
-        "pandas>=0.24,<1.1",  # test_groupby_preserves_metadata fails on 1.1
+        "pandas>=1.1.5",
         "numpy",
         "pyerf",
         "SQLAlchemy",

From 5dce22dde55e148c1e6a44aacfcf13e01728b4fd Mon Sep 17 00:00:00 2001
From: "F.N. Claessen" <felix@seita.nl>
Date: Thu, 10 Dec 2020 16:23:20 +0100
Subject: [PATCH 09/15] Prepare dtype test for empty frames.

---
 timely_beliefs/beliefs/classes.py         |  2 +-
 timely_beliefs/tests/test_belief_query.py | 15 +++++++++++++++
 2 files changed, 16 insertions(+), 1 deletion(-)

diff --git a/timely_beliefs/beliefs/classes.py b/timely_beliefs/beliefs/classes.py
index 872a554f..43748dcb 100644
--- a/timely_beliefs/beliefs/classes.py
+++ b/timely_beliefs/beliefs/classes.py
@@ -1457,7 +1457,7 @@ def set_columns_and_indices_for_empty_frame(df, columns, indices, default_types)
         elif default_types[col] in (int, float):
             df[col] = pd.to_numeric(df[col])
 
-    df.set_index(indices, inplace=True)
+    df.set_index(indices, inplace=True)  # todo: pandas GH30517
 
 
 def assign_sensor_and_event_resolution(df, sensor, event_resolution):
diff --git a/timely_beliefs/tests/test_belief_query.py b/timely_beliefs/tests/test_belief_query.py
index 5242d631..22c7a94e 100644
--- a/timely_beliefs/tests/test_belief_query.py
+++ b/timely_beliefs/tests/test_belief_query.py
@@ -263,3 +263,18 @@ def test_upsample(time_slot_sensor, rolling_day_ahead_beliefs_about_time_slot_ev
     belief_df = belief_df.resample_events(new_resolution)
     assert belief_df.sensor.event_resolution == timedelta(minutes=15)
     assert belief_df.event_resolution == new_resolution
+
+
+def _test_empty_frame(time_slot_sensor):
+    """ pandas GH30517 """
+    bdf = DBTimedBelief.query(
+        session=session,
+        sensor=time_slot_sensor,
+        belief_before=datetime(1900, 1, 1, 13, tzinfo=utc),
+    )
+    assert len(bdf) == 0  # no data expected
+    assert pd.api.types.is_datetime64_dtype(bdf.index.get_level_values("belief_time"))
+    bdf = bdf.convert_index_from_belief_time_to_horizon()
+    assert pd.api.types.is_timedelta64_dtype(
+        bdf.index.get_level_values("belief_horizon")
+    )  # dtype of belief_horizon is timedelta64[ns], so the minimum horizon on an empty BeliefsDataFrame is NaT instead of NaN

From 4987ff55c67b7300a14f047db47a1e534c390dd3 Mon Sep 17 00:00:00 2001
From: "F.N. Claessen" <felix@seita.nl>
Date: Thu, 17 Dec 2020 18:04:01 +0100
Subject: [PATCH 10/15] Add comment about which pandas version fixed
 test_groupby_retains_metadata.

---
 timely_beliefs/tests/test_belief_io.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/timely_beliefs/tests/test_belief_io.py b/timely_beliefs/tests/test_belief_io.py
index 236f4312..29459e4b 100644
--- a/timely_beliefs/tests/test_belief_io.py
+++ b/timely_beliefs/tests/test_belief_io.py
@@ -444,7 +444,7 @@ def test_groupby_retains_metadata():
 
     Succeeds with pandas==1.0.0
     Fails with pandas==1.1.0
-    Should be fixed with https://github.com/pandas-dev/pandas/pull/37461
+    Fixed with pandas==1.1.5
     """
     df = example_df
     metadata = {md: getattr(example_df, md) for md in METADATA}

From 6b7907c1b96fc45a071ee00bbd091e268f6f591f Mon Sep 17 00:00:00 2001
From: "F.N. Claessen" <felix@seita.nl>
Date: Thu, 17 Dec 2020 18:06:11 +0100
Subject: [PATCH 11/15] Rename downsampling function.

---
 timely_beliefs/beliefs/classes.py | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/timely_beliefs/beliefs/classes.py b/timely_beliefs/beliefs/classes.py
index 43748dcb..d4aa066d 100644
--- a/timely_beliefs/beliefs/classes.py
+++ b/timely_beliefs/beliefs/classes.py
@@ -1021,7 +1021,9 @@ def resample_events(
                     else "min",  # keep only most recent belief
                     "cumulative_probability": "prod",  # assume independent variables
                 }
-                df = resample_beliefs_data_frame(df, event_resolution, column_functions)
+                df = downsample_beliefs_data_frame(
+                    df, event_resolution, column_functions
+                )
                 df.event_resolution = event_resolution
             else:
                 # upsample
@@ -1472,7 +1474,7 @@ def assign_sensor_and_event_resolution(df, sensor, event_resolution):
     )
 
 
-def resample_beliefs_data_frame(
+def downsample_beliefs_data_frame(
     df: BeliefsDataFrame, event_resolution: timedelta, col_att_dict: Dict[str, str]
 ) -> BeliefsDataFrame:
     """Because df.resample().agg() doesn't behave nicely for subclassed DataFrames,

From 5f7abf14c8e8f2efbe56adc4b094fe77a1c4339c Mon Sep 17 00:00:00 2001
From: "F.N. Claessen" <felix@seita.nl>
Date: Thu, 17 Dec 2020 20:00:06 +0100
Subject: [PATCH 12/15] Simplify calls to for_each_belief.

---
 timely_beliefs/beliefs/__init__.py    | 6 ++----
 timely_beliefs/visualization/utils.py | 8 +++-----
 2 files changed, 5 insertions(+), 9 deletions(-)

diff --git a/timely_beliefs/beliefs/__init__.py b/timely_beliefs/beliefs/__init__.py
index 469a6f08..ee31aca2 100644
--- a/timely_beliefs/beliefs/__init__.py
+++ b/timely_beliefs/beliefs/__init__.py
@@ -105,7 +105,7 @@ def number_of_belief_times(self):
     def number_of_beliefs(self) -> int:
         """Return the total number of beliefs in the BeliefsDataFrame, including both deterministic beliefs (which
         require a single row) and probabilistic beliefs (which require multiple rows)."""
-        return len(self._obj.for_each_belief(df=self._obj))
+        return len(self._obj.for_each_belief())
 
     @property
     def sources(self) -> List[int]:
@@ -121,9 +121,7 @@ def number_of_sources(self):
     @property
     def number_of_probabilistic_beliefs(self) -> int:
         """Return the number of beliefs in the BeliefsDataFrame that are probabilistic (more than 1 unique value)."""
-        df = self._obj.for_each_belief(
-            df=self._obj, fnc=pd.DataFrame.nunique, dropna=True
-        )
+        df = self._obj.for_each_belief(fnc=pd.DataFrame.nunique, dropna=True)
         return len(df[df > 1].max(axis=1).dropna())
 
     @property
diff --git a/timely_beliefs/visualization/utils.py b/timely_beliefs/visualization/utils.py
index 5f5898c6..5e6ca3a1 100644
--- a/timely_beliefs/visualization/utils.py
+++ b/timely_beliefs/visualization/utils.py
@@ -250,20 +250,18 @@ def prepare_df_for_plotting(
         df["lower_value"] = df["event_value"]
     else:
         df_ci0 = (
-            df.for_each_belief(get_nth_percentile_belief, n=(1 - ci) * 100 / 2, df=df)
+            df.for_each_belief(get_nth_percentile_belief, n=(1 - ci) * 100 / 2)
             .rename(columns={"event_value": "lower_value"})
             .droplevel("cumulative_probability")
             .drop("belief_horizon", axis=1)
         )
         df_exp = (
-            df.for_each_belief(get_nth_percentile_belief, n=50, df=df)
+            df.for_each_belief(get_nth_percentile_belief, n=50)
             .rename(columns={"event_value": "expected_value"})
             .droplevel("cumulative_probability")
         )
         df_ci1 = (
-            df.for_each_belief(
-                get_nth_percentile_belief, n=100 - (1 - ci) * 100 / 2, df=df
-            )
+            df.for_each_belief(get_nth_percentile_belief, n=100 - (1 - ci) * 100 / 2)
             .rename(columns={"event_value": "upper_value"})
             .droplevel("cumulative_probability")
             .drop("belief_horizon", axis=1)

From a1acd4877cd1a528480d1bf2f6e3b32d0519ee87 Mon Sep 17 00:00:00 2001
From: "F.N. Claessen" <felix@seita.nl>
Date: Thu, 17 Dec 2020 20:01:14 +0100
Subject: [PATCH 13/15] Fix plotting integer values (int64 is not JSON
 serializable).

---
 timely_beliefs/visualization/utils.py | 1 +
 1 file changed, 1 insertion(+)

diff --git a/timely_beliefs/visualization/utils.py b/timely_beliefs/visualization/utils.py
index 5e6ca3a1..b275e1d4 100644
--- a/timely_beliefs/visualization/utils.py
+++ b/timely_beliefs/visualization/utils.py
@@ -40,6 +40,7 @@ def plot(
 
     # Set up data source
     bdf = bdf.copy()
+    bdf["event_value"] = bdf["event_value"].astype(float)
     sensor_name = bdf.sensor.name
     sensor_unit = bdf.sensor.unit if bdf.sensor.unit != "" else "a.u."  # arbitrary unit
     plottable_df, belief_horizon_unit = prepare_df_for_plotting(

From 0db3886f03f57a18c431d28d3f632c0cfd43a184 Mon Sep 17 00:00:00 2001
From: "F.N. Claessen" <felix@seita.nl>
Date: Mon, 21 Dec 2020 11:46:57 +0100
Subject: [PATCH 14/15] Add reference in test docs to relevant pandas issue.

---
 timely_beliefs/tests/test_belief_io.py | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/timely_beliefs/tests/test_belief_io.py b/timely_beliefs/tests/test_belief_io.py
index 29459e4b..b968ab45 100644
--- a/timely_beliefs/tests/test_belief_io.py
+++ b/timely_beliefs/tests/test_belief_io.py
@@ -553,6 +553,8 @@ def test_groupby_retains_subclass_attribute(att, args):
     """Checks on metadata propagation for subclassed DataFrames under groupby operations.
 
     Commented-out parameter combinations fail with pandas==1.1.5
+    The relevant issue has to do with calling finalize after operations:
+    see https://github.com/pandas-dev/pandas/issues/28283
     """
 
     METADATA = ["a"]

From 41a06abd3c0bfdc2692dd012cad90fb6800a4b87 Mon Sep 17 00:00:00 2001
From: Felix Claessen <30658763+Flix6x@users.noreply.github.com>
Date: Mon, 21 Dec 2020 11:51:13 +0100
Subject: [PATCH 15/15] Issue 35 metadata lost on multiplication (#43)

This PR adds no additional logic, only a test to check whether Issue #35 is successfully resolved (plus some refactoring of test util functions). Note that this is a merge into resample-while-keeping-metadata, and that branch (with PR #23) actually contains the logic that resolves this issue (as a side effect, because I was actually addressing deeper issues there: #22 and #26).

* Refactor metadata propagation checks to util function.

* Add test for metadata propagation upon multiplication (GH 35).

Co-authored-by: F.N. Claessen <felix@seita.nl>
---
 timely_beliefs/tests/test_belief_io.py | 67 +++++++++++++-------------
 timely_beliefs/tests/utils.py          | 32 +++++++++++-
 2 files changed, 64 insertions(+), 35 deletions(-)

diff --git a/timely_beliefs/tests/test_belief_io.py b/timely_beliefs/tests/test_belief_io.py
index b968ab45..298d1e05 100644
--- a/timely_beliefs/tests/test_belief_io.py
+++ b/timely_beliefs/tests/test_belief_io.py
@@ -7,8 +7,8 @@
 import pytz
 
 import timely_beliefs as tb
-from timely_beliefs.beliefs.classes import METADATA
 from timely_beliefs.examples import example_df
+from timely_beliefs.tests.utils import assert_metadata_is_retained
 
 
 @pytest.fixture(scope="module")
@@ -360,21 +360,16 @@ def test_converting_between_data_frame_and_series_retains_metadata():
     Test whether expanding dimensions of a BeliefsSeries into a BeliefsDataFrame retains the metadata.
     """
     df = example_df
-    metadata = {md: getattr(example_df, md) for md in METADATA}
     series = df["event_value"]
-    for md in metadata:
-        assert getattr(series, md) == metadata[md]
+    assert_metadata_is_retained(series, original_df=example_df, is_series=True)
     df = series.to_frame()
-    for md in metadata:
-        assert getattr(df, md) == metadata[md]
+    assert_metadata_is_retained(df, original_df=example_df)
 
 
 def test_dropping_index_levels_retains_metadata():
     df = example_df.copy()
-    metadata = {md: getattr(example_df, md) for md in METADATA}
     df.index = df.index.get_level_values("event_start")  # drop all other index levels
-    for md in metadata:
-        assert getattr(df, md) == metadata[md]
+    assert_metadata_is_retained(df, original_df=example_df)
 
 
 @pytest.mark.parametrize("drop_level", [True, False])
@@ -383,12 +378,9 @@ def test_slicing_retains_metadata(drop_level):
     Test whether slicing the index of a BeliefsDataFrame retains the metadata.
     """
     df = example_df
-    metadata = {md: getattr(example_df, md) for md in METADATA}
     df = df.xs("2000-01-03 10:00:00+00:00", level="event_start", drop_level=drop_level)
     print(df)
-    assert isinstance(df, tb.BeliefsDataFrame)
-    for md in metadata:
-        assert getattr(df, md) == metadata[md]
+    assert_metadata_is_retained(df, original_df=example_df)
 
 
 @pytest.mark.parametrize("resolution", [timedelta(minutes=30), timedelta(hours=2)])
@@ -400,15 +392,13 @@ def test_mean_resampling_retains_metadata(resolution):
     Succeeds with pandas==1.1.0
     """
     df = example_df
-    metadata = {md: getattr(example_df, md) for md in METADATA}
     df = df.resample(resolution, level="event_start").mean()
     print(df)
-    assert isinstance(df, tb.BeliefsDataFrame)
-    for md in metadata:
-        # if md == "event_resolution":
-        #     assert df.event_resolution == resolution
-        # else:  # todo: the event_resolution metadata is only updated when resampling using df.resample_events(). A reason to override the original resample method, or otherwise something to document.
-        assert getattr(df, md) == metadata[md]
+    assert_metadata_is_retained(
+        df,
+        original_df=example_df,
+        event_resolution=example_df.event_resolution,
+    )  # todo: the event_resolution metadata is only updated when resampling using df.resample_events(). A reason to override the original resample method, or otherwise something to document.
 
 
 @pytest.mark.parametrize("resolution", [timedelta(minutes=30), timedelta(hours=2)])
@@ -419,7 +409,6 @@ def _test_agg_resampling_retains_metadata(resolution):
     Fails with pandas==1.1.5
     """
     df = example_df
-    metadata = {md: getattr(example_df, md) for md in METADATA}
     df = df.reset_index(level=["belief_time", "source", "cumulative_probability"])
     df = df.resample(resolution).agg(
         {
@@ -431,12 +420,11 @@ def _test_agg_resampling_retains_metadata(resolution):
     )
     df = df.set_index(["belief_time", "source", "cumulative_probability"], append=True)
     print(df)
-    assert isinstance(df, tb.BeliefsDataFrame)
-    for md in metadata:
-        # if md == "event_resolution":
-        #     assert df.event_resolution == resolution
-        # else:  # todo: the event_resolution metadata is only updated when resampling using df.resample_events(). A reason to override the original resample method, or otherwise something to document.
-        assert getattr(df, md) == metadata[md]
+    assert_metadata_is_retained(
+        df,
+        original_df=example_df,
+        event_resolution=example_df.event_resolution,
+    )  # todo: the event_resolution metadata is only updated when resampling using df.resample_events(). A reason to override the original resample method, or otherwise something to document.
 
 
 def test_groupby_retains_metadata():
@@ -447,19 +435,14 @@ def test_groupby_retains_metadata():
     Fixed with pandas==1.1.5
     """
     df = example_df
-    metadata = {md: getattr(example_df, md) for md in METADATA}
 
     def assert_function(x):
         print(x)
-        assert isinstance(x, tb.BeliefsDataFrame)
-        for md in metadata:
-            assert getattr(x, md) == metadata[md]
+        assert_metadata_is_retained(x, original_df=example_df)
         return x
 
     df = df.groupby(level="event_start").apply(lambda x: assert_function(x))
-    assert isinstance(df, tb.BeliefsDataFrame)
-    for md in metadata:
-        assert getattr(df, md) == metadata[md]
+    assert_metadata_is_retained(df, original_df=example_df)
 
 
 def test_copy_series_retains_name_and_metadata():
@@ -589,3 +572,19 @@ def _constructor_sliced(self):
     df2 = getattr(df.groupby("x"), att)(*args)
     print(df2)
     assert df2.a == "b"
+
+
+@pytest.mark.parametrize("constant", [1, -1, 3.14, timedelta(hours=1), ["TiledString"]])
+def test_multiplication_with_constant_retains_metadata(constant):
+    """ Check whether the metadata is still there after multiplication. """
+    # GH 35
+    df = example_df * constant
+    assert_metadata_is_retained(df, original_df=example_df)
+
+    # Also check suggested workarounds from GH 35
+    if constant == -1:
+        df = -example_df
+        assert_metadata_is_retained(df, original_df=example_df)
+
+        df = example_df.abs()
+        assert_metadata_is_retained(df, original_df=example_df)
diff --git a/timely_beliefs/tests/utils.py b/timely_beliefs/tests/utils.py
index 05fc7d9c..3e368c8c 100644
--- a/timely_beliefs/tests/utils.py
+++ b/timely_beliefs/tests/utils.py
@@ -1,7 +1,37 @@
-from typing import Union
+from datetime import timedelta
+from typing import Optional, Union
 
 import numpy as np
 
+import timely_beliefs as tb
+from timely_beliefs.beliefs.classes import METADATA
+
 
 def equal_lists(list_a: Union[list, np.ndarray], list_b: Union[list, np.ndarray]):
     return all(np.isclose(a, b) for a, b in zip(list_a, list_b))
+
+
+def assert_metadata_is_retained(
+    result_df: Union[tb.BeliefsDataFrame, tb.BeliefsSeries],
+    original_df: tb.BeliefsDataFrame,
+    is_series: bool = False,
+    event_resolution: Optional[timedelta] = None,
+):
+    """Fail if result_df is not a BeliefsDataFrame with the same metadata as the original BeliefsDataFrame.
+
+    Can also be used to check for a BeliefsSeries (using is_series=True).
+
+    :param result_df: BeliefsDataFrame or BeliefsSeries to be checked for metadata propagation
+    :param original_df: BeliefsDataFrame containing the original metadata
+    :param is_series: if True, we check that the result is a BeliefsSeries rather than a BeliefsDataFrame
+    :param event_resolution: optional timedelta in case we expect a different event_resolution than the original
+    """
+    metadata = {md: getattr(original_df, md) for md in METADATA}
+    assert isinstance(
+        result_df, tb.BeliefsDataFrame if not is_series else tb.BeliefsSeries
+    )
+    for md in metadata:
+        if md == "event_resolution" and event_resolution is not None:
+            assert result_df.event_resolution == event_resolution
+        else:
+            assert getattr(result_df, md) == metadata[md]