Update app.py

app.py

@@ -1,3 +1,4 @@
+# app.py
 import os
 import re
 import zipfile
@@ -7,7 +8,7 @@ import osmnx as ox
 import folium
 import branca
 import pandas as pd
-import geopandas as gpd  # <- útil para alguns ambientes
+import geopandas as gpd
 
 # --- Initial settings ---
 geolocator = Nominatim(user_agent="gradio_osm_app")
@@ -16,78 +17,101 @@ ox.settings.log_console = False
 DOWNLOAD_DIR = "/tmp/download"
 os.makedirs(DOWNLOAD_DIR, exist_ok=True)
 
-def slugify(name):
+def slugify(name: str) -> str:
     return re.sub(r"[^0-9A-Za-z]+", "_", name).strip("_")
 
-# --------- Utilidades para salvar GPKG sem erros ----------
+# --------- Utils for safe GPKG writing ----------
+def _make_unique(names):
+    """Make column names unique by suffixing __1, __2 on duplicates."""
+    seen = {}
+    out = []
+    for n in names:
+        n0 = str(n)
+        cnt = seen.get(n0, 0)
+        out.append(n0 if cnt == 0 else f"{n0}__{cnt}")
+        seen[n0] = cnt + 1
+    return out
+
 def _is_complex_obj(v):
-    # Tipos que o OGR não aceita direto como campo (list, dict, set, tuple, bytes)
+    # Types OGR doesn't accept as field values
     return isinstance(v, (list, dict, tuple, set, bytes))
 
-def clean_for_gpkg(gdf):
-    """Sanitiza colunas e dtypes para gravação segura em GPKG."""
+def clean_for_gpkg(gdf: gpd.GeoDataFrame) -> gpd.GeoDataFrame:
+    """Sanitize columns and dtypes for safe GPKG write."""
     gdf = gdf.copy()
 
-    # CRS
+    # Ensure CRS
     if getattr(gdf, "crs", None) is None:
         try:
             gdf.set_crs(4326, inplace=True)
         except Exception:
             pass
 
-    # Nomes de colunas seguros e não muito longos
-    gdf.columns = (
-        pd.Index(gdf.columns)
-        .map(str)
-        .str.replace(r"[^0-9a-zA-Z_]", "_", regex=True)
-        .str.replace(r"_{2,}", "_", regex=True)
-        .str.strip("_")
-        .str.slice(0, 60)
-    )
-
-    # Converter dtypes problemáticos
+    # Normalize column names
+    cols = pd.Index(gdf.columns).map(str)
+    cols = cols.str.replace(r"[^0-9a-zA-Z_]", "_", regex=True)\
+               .str.replace(r"_{2,}", "_", regex=True)\
+               .str.strip("_")\
+               .str.slice(0, 60)
+
+    # Guarantee uniqueness (avoids gdf.dtypes[col] returning a Series)
+    if cols.duplicated().any() or isinstance(cols, pd.MultiIndex):
+        cols = pd.Index(_make_unique(cols))
+    gdf.columns = cols
+
+    geom_name = gdf.geometry.name if hasattr(gdf, "geometry") and gdf.geometry is not None else None
+
+    # Convert problematic dtypes
     for col in gdf.columns:
-        if col == gdf.geometry.name:
+        if col == geom_name:
             continue
 
-        # >>> correção: use a tabela de dtypes do DataFrame
         dt = gdf.dtypes[col]
+        # Defensive: if dt ever comes as Series, collapse to string
+        if isinstance(dt, pd.Series):
+            gdf[col] = gdf[col].astype("string").fillna("")
+            continue
+
+        # tz-aware datetimes -> naive
+        if pd.api.types.is_datetime64tz_dtype(dt):
+            gdf[col] = gdf[col].dt.tz_convert(None)
 
-        # Pandas tipo string nativo ajuda a evitar NaN/None
-        if dt == "object":
-            # Se houver qualquer valor complexo, vira string integral
+        # objects -> string (if complex) or pandas string dtype
+        if pd.api.types.is_object_dtype(dt):
             if gdf[col].apply(_is_complex_obj).any():
                 gdf[col] = gdf[col].astype(str)
             else:
-                # Objetos simples (str/num/None) -> string estável
                 gdf[col] = gdf[col].astype("string")
 
-        # Inteiros nulos (Int64) -> float64 (OGR lida melhor) ou string
+        # categorical -> string
+        if pd.api.types.is_categorical_dtype(dt):
+            gdf[col] = gdf[col].astype(str)
+
+        # nullable integer -> float64
         if pd.api.types.is_integer_dtype(dt) and str(dt).startswith("Int"):
             gdf[col] = gdf[col].astype("float64")
 
-        # Booleans com nulos -> string "true/false" para evitar falhas
+        # booleans with NaN -> string; pure bool stays
         if pd.api.types.is_bool_dtype(dt):
             if gdf[col].isna().any():
                 gdf[col] = gdf[col].astype("string").fillna("")
-            else:
-                pass  # bool puro costuma funcionar
 
-        # Preencher NaN em strings
+        # fill NaN in pandas string dtype
         if gdf[col].dtype == "string":
             gdf[col] = gdf[col].fillna("")
 
+    # Final friendly names
+    gdf.rename(columns=lambda c: str(c).strip("_")[:60], inplace=True)
+
     return gdf
 
-def try_to_file(gdf, path, driver="GPKG"):
-    """Tenta salvar; em caso de erro por campo, remove apenas o campo problemático e tenta novamente."""
+def try_to_file(gdf: gpd.GeoDataFrame, path: str, driver: str = "GPKG"):
+    """Try to save; on field error, drop only the offending field and retry once."""
     try:
         gdf.to_file(path, driver=driver)
         return
     except Exception as e:
         msg = str(e)
-        # Detecta qual campo quebrou (mensagem típica do pyogrio/OGR)
-        # Ex.: "Error adding field 'fixme' to layer"
         m = re.search(r"field '([^']+)'", msg, flags=re.IGNORECASE)
         if m:
             bad = m.group(1)
@@ -95,12 +119,16 @@ def try_to_file(gdf, path, driver="GPKG"):
                 gdf2 = gdf.drop(columns=[bad])
                 gdf2.to_file(path, driver=driver)
                 return
-        # Se não identificar, relança
         raise
 
-def ensure_saved(gdf, slug, layer):
+def ensure_saved(gdf: gpd.GeoDataFrame, slug: str, layer: str):
     if gdf is None or gdf.empty:
         return
+    # Extra guard: unique columns before cleaning/writing
+    if pd.Index(gdf.columns).duplicated().any():
+        gdf = gdf.copy()
+        gdf.columns = pd.Index(_make_unique(pd.Index(gdf.columns).map(str)))
+
     filename = f"{slug}_{layer}.gpkg"
     path = os.path.join(DOWNLOAD_DIR, filename)
     if os.path.exists(path):
@@ -108,24 +136,23 @@ def ensure_saved(gdf, slug, layer):
 
     gdf_clean = clean_for_gpkg(gdf)
 
-    # Fallback final: se ainda der erro, tenta remover colunas 'problemáticas' comuns em OSM
     try:
         try_to_file(gdf_clean, path, driver="GPKG")
     except Exception:
-        # Remove colunas muito propensas a erro
+        # Drop common problematic OSM fields as a second attempt
         drop_candidates = [c for c in gdf_clean.columns if c.lower() in {"fixme", "note", "source_ref"}]
         if drop_candidates:
             gdf_clean2 = gdf_clean.drop(columns=drop_candidates, errors="ignore")
             try_to_file(gdf_clean2, path, driver="GPKG")
         else:
-            # Sem pistas: tenta converter tudo não-geom para string como último recurso
+            # Last resort: stringify all non-geometry columns
             gdf_last = gdf_clean.copy()
             for c in gdf_last.columns:
                 if c != gdf_last.geometry.name:
                     gdf_last[c] = gdf_last[c].astype(str)
             try_to_file(gdf_last, path, driver="GPKG")
 
-# --------- UI / Lógica principal ----------
+# --------- UI / main logic ----------
 def make_legend(selected_layers):
     color_map = {
         "Highways": "yellow",
@@ -207,7 +234,7 @@ def map_with_layers(place_name, cb_highways, cb_buildings, cb_school,
         if flag:
             yield None, f"🔄 Downloading {name}..."
             gdf2 = ox.features_from_polygon(poly, tags)
-            # Apenas polígonos para essas camadas
+            # Only polygons for these layers
             gdf2 = gdf2[gdf2.geometry.type.isin(['Polygon', 'MultiPolygon'])]
             if not gdf2.empty:
                 layers[name] = gdf2
@@ -275,7 +302,7 @@ def download_data(place_name, cb_highways, cb_buildings, cb_school,
         print("Erro ao gerar o ZIP:", str(e))
         raise gr.Error(f"❌ Download failed: {e}")
 
-# --- Layout com abas ---
+# --- Layout with tabs ---
 with gr.Blocks(title="Geoeasy View") as demo:
     gr.HTML("""
     <style>
@@ -335,5 +362,3 @@ with gr.Blocks(title="Geoeasy View") as demo:
 
 if __name__ == '__main__':
     demo.launch()
-
-