Add remaining aggregate funcs for hybrid mode

Author: mayagbarnes

README.md

@@ -484,10 +484,9 @@ st_pivot_table(
 
 **Auto thresholds:** In `"auto"` mode, server-side pre-aggregation activates when the dataset has at least 100K rows (high-cardinality layouts) or 250K rows (moderate layouts) and the estimated pivot shape exceeds the client-side comfort budget.
 
-**Supported aggregations:** `sum`, `count`, `min`, `max`, and `avg`. Configs using other aggregations (e.g. `median`, `count_distinct`) fall back to client-side computation automatically.
+**Supported aggregations:** All 10 aggregation types are supported in hybrid mode. `count` and `count_distinct` work on any column type; all other aggregations (`sum`, `avg`, `min`, `max`, `median`, `percentile_90`, `first`, `last`) coerce values to numeric and ignore non-numeric entries, consistent with client-only mode behavior. For non-decomposable aggregations (`avg`, `count_distinct`, `median`, `percentile_90`, `first`, `last`), the server computes correct totals and subtotals via a sidecar payload, ensuring accuracy that client-side re-aggregation alone cannot provide.
 
 **Limitations:**
-- Drill-down is disabled in hybrid mode because values are pre-aggregated on the server rather than built from raw rows.
 - Synthetic measures are not supported in hybrid mode (falls back to client-side).
 
 ### Locked Mode

e2e_playwright/e2e_utils.py

@@ -117,6 +117,8 @@
     "test_pivot_null_separate",
     "test_pivot_null_zero",
     "test_pivot_sparse_drilldown",
+    "test_pivot_hybrid_median",
+    "test_pivot_hybrid_count_distinct",
     "test_pivot_synthetic",
 ]
 

e2e_playwright/pivot_table_data_app.py

@@ -253,6 +253,35 @@ def render_app(data):
         on_config_change=noop,
     )
 
+    # --- Hybrid mode pivots with non-decomposable aggs ---
+    df_large = df.loc[df.index.repeat(50)].reset_index(drop=True)
+
+    st.subheader("Hybrid Median Pivot")
+    st_pivot_table(
+        df_large,
+        key="test_pivot_hybrid_median",
+        rows=["Region"],
+        columns=["Year"],
+        values=["Revenue"],
+        aggregation="median",
+        show_totals=True,
+        execution_mode="threshold_hybrid",
+        on_config_change=noop,
+    )
+
+    st.subheader("Hybrid Count Distinct Pivot")
+    st_pivot_table(
+        df_large,
+        key="test_pivot_hybrid_count_distinct",
+        rows=["Region"],
+        columns=["Year"],
+        values=["Category"],
+        aggregation="count_distinct",
+        show_totals=True,
+        execution_mode="threshold_hybrid",
+        on_config_change=noop,
+    )
+
     st.subheader("Synthetic Measures Pivot")
     st_pivot_table(
         df_synth,

e2e_playwright/pivot_table_data_test.py

@@ -534,3 +534,45 @@ def test_synthetic_menu_hides_show_values_as(page_at_app: Page):
         timeout=5000
     )
     expect(page.get_by_test_id("header-menu-display")).to_have_count(0)
+
+
+# ---------------------------------------------------------------------------
+# Hybrid mode tests (non-decomposable aggregations)
+# ---------------------------------------------------------------------------
+
+
+def test_hybrid_median_renders_cells(page_at_app: Page):
+    """Hybrid median pivot renders data cells and totals."""
+    page = page_at_app
+    container = get_pivot(page, "test_pivot_hybrid_median")
+    expect(container.get_by_test_id("pivot-table")).to_be_visible(timeout=15000)
+    cells = container.get_by_test_id("pivot-data-cell")
+    expect(cells.first).to_be_visible(timeout=5000)
+    cell_count = cells.count()
+    assert cell_count > 0, "Hybrid median pivot should render data cells"
+
+
+def test_hybrid_median_grand_total_not_dash(page_at_app: Page):
+    """Grand total for hybrid median is not the empty cell value."""
+    page = page_at_app
+    container = get_pivot(page, "test_pivot_hybrid_median")
+    expect(container.get_by_test_id("pivot-table")).to_be_visible(timeout=15000)
+    grand_total_cells = container.get_by_test_id("pivot-grand-total-cell")
+    if grand_total_cells.count() > 0:
+        text = grand_total_cells.first.inner_text().strip()
+        assert text != "-", f"Grand total should be a number, got '{text}'"
+
+
+def test_hybrid_count_distinct_renders_correct_values(page_at_app: Page):
+    """Hybrid count_distinct pivot shows actual distinct counts (not 1)."""
+    page = page_at_app
+    container = get_pivot(page, "test_pivot_hybrid_count_distinct")
+    expect(container.get_by_test_id("pivot-table")).to_be_visible(timeout=15000)
+    cells = container.get_by_test_id("pivot-data-cell")
+    expect(cells.first).to_be_visible(timeout=5000)
+    texts = [
+        c.inner_text().strip() for c in cells.all() if c.inner_text().strip() != "-"
+    ]
+    for t in texts:
+        val = float(t.replace(",", ""))
+        assert val >= 1, f"Count distinct value should be >= 1, got {val}"

streamlit_app.py

@@ -1183,6 +1183,71 @@
     "forced to threshold_hybrid mode for demonstration."
 )
 
+st.markdown(
+    """
+**Non-decomposable aggregations in hybrid mode:** All 10 aggregation types are
+supported in hybrid mode, including `median`, `count_distinct`, `percentile_90`,
+`first`, and `last`. The server computes correct totals via a sidecar payload.
+`count` and `count_distinct` work on any column type; all other aggregations
+coerce values to numeric and ignore non-numeric entries.
+"""
+)
+
+col_a, col_b = st.columns(2)
+with col_a:
+    st.markdown("**Hybrid Median Pivot**")
+    st_pivot_table(
+        df_hybrid,
+        key="hybrid_median_demo",
+        rows=["Region"],
+        columns=["Year"],
+        values=["Revenue"],
+        aggregation="median",
+        number_format={"Revenue": "$,.2f"},
+        show_totals=True,
+        execution_mode="threshold_hybrid",
+    )
+with col_b:
+    st.markdown("**Hybrid Count Distinct Pivot**")
+    st_pivot_table(
+        df_hybrid,
+        key="hybrid_count_distinct_demo",
+        rows=["Region"],
+        columns=["Year"],
+        values=["Category"],
+        aggregation="count_distinct",
+        show_totals=True,
+        execution_mode="threshold_hybrid",
+    )
+
+with st.expander("View Code — Non-decomposable Aggregations"):
+    st.code(
+        """
+# Median: server computes correct grand/row/col totals
+st_pivot_table(
+    df_hybrid,
+    rows=["Region"],
+    columns=["Year"],
+    values=["Revenue"],
+    aggregation="median",
+    show_totals=True,
+    execution_mode="threshold_hybrid",
+)
+
+# Count Distinct: works on any column type (e.g. Category strings)
+st_pivot_table(
+    df_hybrid,
+    rows=["Region"],
+    columns=["Year"],
+    values=["Category"],
+    aggregation="count_distinct",
+    show_totals=True,
+    execution_mode="threshold_hybrid",
+)
+""",
+        language="python",
+    )
+
 
 # ---------------------------------------------------------------------------
 # Section 15: Drag-and-Drop Field Configuration