fix error when report input contains lists skrub-data#1066 (skrub-dat…

…a#1087)

CHANGES.rst

@@ -58,6 +58,10 @@ Minor changes
   the "sample" tab panel by specifying ``n_rows``.
   :pr:`1083` by :user:`Jérôme Dockès <jeromedockes>`.
 
+* the `TableReport` used to raise an exception when the dataframe contained
+  unhashable types such as python lists. This has been fixed in :pr:`1087` by
+  :user:`Jérôme Dockès <jeromedockes>`.
+
 Release 0.3.0
 =============
 

skrub/_dataframe/_common.py

@@ -719,10 +719,10 @@ def _to_string_pandas(col):
 
 @to_string.specialize("polars", argument_type="Column")
 def _to_string_polars(col):
-    if col.dtype != pl.Object:
+    # polars raises an error when trying to cast those types to string directly
+    # so we have to use map_elements
+    if col.dtype not in (pl.Object, pl.List, pl.Array):
         return _cast_polars(col, pl.String)
-    # Objects are mere passengers in polars dataframes and we can't do
-    # anything with them; cast raises an exception.
     # polars emits a performance warning when using map_elements
     with warnings.catch_warnings():
         warnings.filterwarnings(

skrub/_reporting/_data/templates/column-summary.html

@@ -86,7 +86,7 @@ <h3 class="margin-r-t">
                 <summary>Most frequent values</summary>
                 <div class="shrink">
                     <div class="copybutton-grid">
-                        {% for value in column.value_counts %}
+                        {% for (value, count) in column.value_counts %}
                         {% set val_id = "{}-freq-value-{}".format(col_id, loop.index0) %}
                         <div class="box" data-test="frequent-value-{{ loop.index0 }}">
                             <pre id="{{ val_id }}"
@@ -99,7 +99,7 @@ <h3 class="margin-r-t">
                         {% set val_id = "{}-freq-value-list".format(col_id) %}
                         <div class="box">
                             <pre id="{{ val_id }}"
-                                data-copy-text="{{ column.value_counts | list }}">{{ column.value_counts.keys() | list }}</pre>
+                                data-copy-text="{{ column.most_frequent_values }}">{{ column.most_frequent_values }}</pre>
                             {{ buttons.copybutton(val_id) }}
                         </div>
                     </div>

skrub/_reporting/_html.py

@@ -33,6 +33,13 @@
 }
 
 
+def _is_null(value):
+    isna = pd.isna(value)
+    if isinstance(isna, bool):
+        return isna
+    return False
+
+
 def _get_jinja_env():
     env = jinja2.Environment(
         loader=jinja2.FileSystemLoader(
@@ -47,7 +54,7 @@ def _get_jinja_env():
         "svg_to_img_src",
     ]:
         env.filters[function_name] = getattr(_utils, function_name)
-    env.filters["is_null"] = pd.isna
+    env.filters["is_null"] = _is_null
     env.globals["random_string"] = random_string
     return env
 

skrub/_reporting/_plotting.py

@@ -109,9 +109,8 @@ def value_counts(value_counts, n_unique, n_rows, color=COLOR_0):
 
     Parameters
     ----------
-    value_counts : dict
-        The keys are values, and values are counts. Must be sorted from most to
-        least frequent.
+    value_counts : list
+        Pairs of (value, count). Must be sorted from most to least frequent.
 
     n_unique : int
         Cardinality of the plotted column, used to determine if all unique
@@ -129,8 +128,8 @@ def value_counts(value_counts, n_unique, n_rows, color=COLOR_0):
     str
         The plot as a XML string.
     """
-    values = [_utils.ellide_string_short(s) for s in value_counts.keys()][::-1]
-    counts = list(value_counts.values())[::-1]
+    values = [_utils.ellide_string_short(v) for v, _ in value_counts][::-1]
+    counts = [c for _, c in value_counts][::-1]
     if n_unique > len(value_counts):
         title = f"{len(value_counts)} most frequent"
     else:

skrub/_reporting/_summarize.py

@@ -154,10 +154,11 @@ def _add_value_counts(summary, column, *, dataframe_summary, with_plots):
     summary["unique_proportion"] = n_unique / max(1, dataframe_summary["n_rows"])
     summary["high_cardinality"] = n_unique >= _HIGH_CARDINALITY_THRESHOLD
     summary["value_counts"] = value_counts
+    summary["most_frequent_values"] = [v for v, _ in value_counts]
 
     if n_unique == 1:
         summary["value_is_constant"] = True
-        summary["constant_value"] = next(iter(value_counts.keys()))
+        summary["constant_value"] = value_counts[0][0]
     else:
         summary["value_is_constant"] = False
         if with_plots:

skrub/_reporting/_utils.py

@@ -37,7 +37,7 @@ def top_k_value_counts(column, k):
     n_unique = sbd.shape(counts)[0]
     counts = sbd.sort(counts, by="count", descending=True)
     counts = sbd.slice(counts, k)
-    return n_unique, dict(zip(*to_dict(counts).values()))
+    return n_unique, list(zip(*to_dict(counts).values()))
 
 
 def quantiles(column):

skrub/_reporting/tests/test_summarize.py

@@ -74,7 +74,8 @@ def test_summarize(monkeypatch, df_module, air_quality, order_by, with_plots):
         "plot_names": [],
         "position": 0,
         "unique_proportion": 0.118,
-        "value_counts": {"London": 8, "Paris": 9},
+        "value_counts": [("Paris", 9), ("London", 8)],
+        "most_frequent_values": ["Paris", "London"],
         "value_is_constant": False,
     }
 

skrub/_reporting/tests/test_table_report.py

@@ -74,3 +74,10 @@ def test_empty_dataframe(df_module):
 def test_open(pd_module, browser_mock):
     TableReport(pd_module.example_dataframe, title="the title").open()
     assert b"the title" in browser_mock.content
+
+
+def test_non_hashable_values(df_module):
+    # non-regression test for #1066
+    df = df_module.make_dataframe(dict(a=[[1, 2, 3], None, [4]]))
+    html = TableReport(df).html()
+    assert "[1, 2, 3]" in html