Skrub Add Estimated Memory Usage to TableReport

CHANGES.rst

@@ -41,6 +41,7 @@ New Features
   regular numeric format (``(432)`` becomes ``-432``). :pr:`1772` by :user:`Gabriela
   Gómez Jiménez <gabrielapgomezji>`.
 
+
 Changes
 -------
 - :meth:`choose_from` now transparently converts `outcomes` to a list when it is another type of sequence. :pr:`2100` by
@@ -54,6 +55,8 @@ Changes
   :pr:`2096` by :user:`Ayesha Siddiqua <siddiqua-tamk>`.
 - The :class:`TableReport` can now be exported in markdown format with ``.markdown``.
   :pr:`2048` by :user:`Riccardo Cappuzzo <rcap107>`.
+- The :class:`TableReport` can now be exported the estimated memory usage in TableReport when display data.
+  :pr:`2153` by :user:`Salam AlKaissi <salam-alkaissi> Sanae Janati Idrissi <sana-22>`.
 
 Bugfixes
 --------

skrub/_reporting/_data/templates/dataframe-sample.html

@@ -86,6 +86,13 @@
         <strong>{{ summary.n_rows | format_number }}</strong> rows ✕
         <strong data-manager="ColumnFilterMatchCount"
             data-test="n-columns-display">{{ summary.n_columns | format_number }}</strong> columns
+        {%- if summary.get("memory_usage_kb") is not none -%}
+            <span data-test="memory-usage-display"> (estimated memory usage: {{ "%.1f" | format(summary.get("memory_usage_kb")) }} KB)
+                {%- if summary.get("memory_estimate_unreliable") -%}
+                <em class="memory-warning">— estimate may be inaccurate for complex objects</em>
+                {%- endif -%}
+            </span>
+        {%- endif -%}
         {% if 'is_subsampled' in summary %}
          (subsampled from more rows)
         {% endif %}

skrub/_reporting/_data/templates/report.md

@@ -2,6 +2,12 @@
 
 The provided dataframe uses the {{ summary.dataframe_module }} library. It has
 **shape** {{ summary.n_rows }} rows × {{ summary.n_columns }} columns.
+{% if summary.get("memory_usage_kb") is not none %}
+**memory usage** {{ "%.1f" | format(summary.get("memory_usage_kb")) }} KB.
+{% if summary.get("memory_estimate_unreliable") %}
+_Note: memory estimate may be inaccurate for complex object columns._
+{% endif %}
+{% endif %}
 
 Columns are marked as "high cardinality" if they contain more than
 {{ summary.cardinality_threshold }} unique values.

skrub/_reporting/_summarize.py

@@ -18,6 +18,28 @@
 _N_TOP_ASSOCIATIONS = 1000
 
 
+def _memory_usage_kb(df):
+    if sbd.dataframe_module_name(df) == "pandas":
+        memory_usage_bytes = df.memory_usage(deep=False).sum()
+    else:
+        estimated_size = getattr(df, "estimated_size", None)
+        if estimated_size is None:
+            return None
+        memory_usage_bytes = estimated_size()
+    return memory_usage_bytes / 1024
+
+
+def _has_complex_objects(df):
+    """Return True when pandas has object-dtype columns.
+
+    The memory estimate is less reliable for object-dtype columns, so we warn
+    as soon as any are present.
+    """
+    if sbd.dataframe_module_name(df) != "pandas":
+        return False
+    return any(dtype == object for dtype in df.dtypes)
+
+
 def summarize_dataframe(
     df,
     *,
@@ -77,6 +99,7 @@ def summarize_dataframe(
         "dataframe_module": sbd.dataframe_module_name(df),
         "n_rows": n_rows,
         "n_columns": n_columns,
+        "memory_usage_kb": _memory_usage_kb(df),
         "columns": [],
         "dataframe_is_empty": not n_rows or not n_columns,
         "plots_skipped": not with_plots,
@@ -90,6 +113,11 @@ def summarize_dataframe(
     }
     if title is not None:
         summary["title"] = title
+    # detect complex objects that make memory estimates unreliable
+    try:
+        summary["memory_estimate_unreliable"] = _has_complex_objects(df)
+    except Exception:
+        summary["memory_estimate_unreliable"] = False
     if order_by is not None:
         df = sbd.sort(df, by=order_by)
         summary["order_by"] = order_by

skrub/_reporting/_table_report.py

@@ -240,7 +240,7 @@ class TableReport:
 
     >>> j = TableReport(df, plot_distributions=False).json()
     >>> print(j)
-    {"dataframe_module": "pandas", "n_rows": 2, "n_columns": 3, "columns": ...
+    {"dataframe_module": "pandas", "n_rows": 2, "n_columns": 3, "memory_usage_kb": 0.1 ...
 
 
     Advanced configuration: you can add custom column filters that will appear

skrub/_reporting/tests/test_markdown_template.py

@@ -37,6 +37,7 @@ def test_markdown_report_structure_and_titles(df_module):
     assert "# " in markdown_default  # Header should exist
     # Shape info should be present
     assert "**shape** 3 rows × 3 columns" in markdown
+    assert "**memory usage**" in markdown
     # Unique values should be present (default value)
     assert "40 unique values." in markdown
 

skrub/_reporting/tests/test_table_report.py

@@ -46,6 +46,7 @@ def test_report(air_quality):
     assert "With nulls" in html
     assert "First 10" in html
     assert "First 2" in html
+    assert "memory usage:" in html
     for col_name in sbd.column_names(air_quality):
         assert col_name in html
     report_id = get_report_id(html)