feat(tableau): export Excel via DuckDB (AST->SQL) (#41)

Rewrite download_data to read filterModel/columnState from the AG Grid
component and add export_dataframe (compiles filterModel to SQL via
filtermodel_to_ast/ast_to_sql, builds ORDER BY, excludes hidden columns,
queries DuckDB via query_marches). Replaces the old Polars filter_query
pipeline tied to the removed DataTable.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

src/pages/tableau.py

@@ -5,7 +5,6 @@ import uuid
 from datetime import datetime
 
 import dash_bootstrap_components as dbc
-import polars as pl
 from dash import (
     ClientsideFunction,
     Input,
@@ -20,7 +19,7 @@ from dash import (
 )
 from flask_login import current_user
 
-from src.db import query_marches, schema
+from src.db import schema
 from src.figures import ag_grid, make_column_picker
 from src.pages._compte_shell import current_user_has_subscription
 from src.saved_views import db as saved_views_db
@@ -31,10 +30,8 @@ from src.utils.seo import META_CONTENT
 from src.utils.table import (
     COLUMNS,
     build_view_query,
-    filter_table_data,
     get_default_hidden_columns,
     invert_columns,
-    sort_table_data,
     write_styled_excel,
 )
 from src.utils.tracking import track_search
@@ -467,27 +464,25 @@ def update_meta(total):
 @callback(
     Output("download-data", "data"),
     Input("btn-download-data", "n_clicks"),
-    State("tableau_datatable", "filter_query"),
-    State("tableau_datatable", "sort_by"),
-    State("tableau_datatable", "hidden_columns"),
+    State("tableau_grid", "filterModel"),
+    State("tableau_grid", "columnState"),
     prevent_initial_call=True,
 )
-def download_data(n_clicks, filter_query, sort_by, hidden_columns: list | None = None):
-    lff: pl.LazyFrame = query_marches().lazy()
+def download_data(n_clicks, filter_model, column_state):
+    from src.utils.grid import export_dataframe
 
-    # Les colonnes masquées sont supprimées
-    if hidden_columns:
-        lff = lff.drop(hidden_columns)
-
-    if filter_query:
-        track_search(filter_query, "tab download")
-        lff = filter_table_data(lff, filter_query)
-
-    if sort_by and len(sort_by) > 0:
-        lff = sort_table_data(lff, sort_by)
+    sort_model = [
+        {"colId": c["colId"], "sort": c["sort"]}
+        for c in (column_state or [])
+        if c.get("sort")
+    ]
+    hidden_columns = [c["colId"] for c in (column_state or []) if c.get("hide")]
+    if filter_model:
+        track_search(json.dumps(filter_model), "tab download")
+    df = export_dataframe(filter_model, sort_model, hidden_columns)
 
     def to_bytes(buffer):
-        write_styled_excel(lff.collect(engine="streaming"), buffer)
+        write_styled_excel(df, buffer)
 
     date = datetime.now().strftime("%Y-%m-%d_%H:%M:%S")
     return dcc.send_bytes(to_bytes, filename=f"decp_{date}.xlsx")

src/utils/grid.py

@@ -37,6 +37,23 @@ def fetch_grid_page(
     return page.to_dicts(), total
 
 
+def export_dataframe(filter_model, sort_model, hidden_columns) -> pl.DataFrame:
+    """Renvoie les lignes filtrées/triées pour l'export Excel.
+
+    Colonnes masquées exclues, valeurs brutes (non post-traitées HTML).
+    """
+    ast = filtermodel_to_ast(filter_model, schema)
+    filter_sql, params = ast_to_sql(ast, schema)
+    order_by = sort_model_to_sql(sort_model, schema) or None
+    visible = [c for c in schema.names() if c not in set(hidden_columns or [])]
+    return query_marches(
+        where_sql=filter_sql,
+        params=params,
+        columns=visible,
+        order_by=order_by,
+    )
+
+
 _LINK_COLUMNS = {
     "marche",
     "uid",

tests/test_grid.py

@@ -1,4 +1,4 @@
-from src.utils.grid import fetch_grid_page, grid_column_defs
+from src.utils.grid import export_dataframe, fetch_grid_page, grid_column_defs
 
 
 def test_column_defs_have_field_and_filter():
@@ -40,3 +40,14 @@ def test_fetch_grid_page_filter_reduces_count():
 def test_fetch_grid_page_offset_slicing():
     rows, _ = fetch_grid_page(None, None, 0, 5)
     assert len(rows) <= 5
+
+
+def test_export_dataframe_excludes_hidden_columns():
+    df = export_dataframe(None, None, hidden_columns=["objet"])
+    assert "objet" not in df.columns
+
+
+def test_export_dataframe_applies_filter():
+    fm = {"objet": {"filterType": "text", "type": "contains", "filter": "zzzzzznope"}}
+    df = export_dataframe(fm, None, hidden_columns=[])
+    assert df.height == 0

tests/test_main.py

@@ -99,7 +99,7 @@ def test_003_tableau_download(dash_duo: DashComposite):
     print(app.server.name)
 
     outputs = [
-        download_data(1, "", [], None),
+        download_data(1, None, None),
         download_acheteur_data(1, "/acheteurs/123", "2025", "ACHETEUR 1"),
         download_titulaire_data(1, "/titulaires/345", "2025", "TITULAIRE 1"),
     ]