core-lens 0.1.dev197__tar.gz → 0.1.dev198__tar.gz

{core_lens-0.1.dev197 → core_lens-0.1.dev198}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: core-lens
-Version: 0.1.dev197
+Version: 0.1.dev198
 Summary: Query, analyse, and visualise CoreStack's microwatershed and Earth science data through a clean, composable Python API.
 Project-URL: Homepage, https://github.com/ApoorvaKashyap/core-lens
 Project-URL: Issues, https://github.com/ApoorvaKashyap/core-lens/issues
@@ -21,7 +21,7 @@ Requires-Dist: core-lens[spatial]
 Provides-Extra: core
 Requires-Dist: awscrt>=0.23.0; extra == 'core'
 Requires-Dist: loguru>=0.7.3; extra == 'core'
-Requires-Dist: polars<2,>=1.39.0; extra == 'core'
+Requires-Dist: polars<3,>=2.0.0; extra == 'core'
 Requires-Dist: pyarrow<25,>=23.0.0; extra == 'core'
 Requires-Dist: pydantic<3,>=2.13.3; extra == 'core'
 Provides-Extra: full
@@ -33,7 +33,7 @@ Requires-Dist: geopandas<2,>=1.1.3; extra == 'full'
 Requires-Dist: loguru>=0.7.3; extra == 'full'
 Requires-Dist: lonboard<1,>=0.16.0; extra == 'full'
 Requires-Dist: matplotlib<4,>=3.10.9; extra == 'full'
-Requires-Dist: polars<2,>=1.39.0; extra == 'full'
+Requires-Dist: polars<3,>=2.0.0; extra == 'full'
 Requires-Dist: pyarrow<25,>=23.0.0; extra == 'full'
 Requires-Dist: pydantic<3,>=2.13.3; extra == 'full'
 Requires-Dist: pyogrio<1,>=0.10.0; extra == 'full'
@@ -41,39 +41,12 @@ Requires-Dist: pyproj<4,>=3.7.2; extra == 'full'
 Requires-Dist: scipy<2,>=1.17.1; extra == 'full'
 Requires-Dist: shapely<3,>=2.1.2; extra == 'full'
 Requires-Dist: statsmodels<1,>=0.14.6; extra == 'full'
-Provides-Extra: full-gpu
-Requires-Dist: awscrt>=0.23.0; extra == 'full-gpu'
-Requires-Dist: bokeh<4,>=3.6.0; extra == 'full-gpu'
-Requires-Dist: cudf-cu13==26.6.*; extra == 'full-gpu'
-Requires-Dist: cudf-polars-cu13==26.6.*; extra == 'full-gpu'
-Requires-Dist: duckdb<2,>=1.5.3; extra == 'full-gpu'
-Requires-Dist: geoarrow-rust-core>=0.3.0; extra == 'full-gpu'
-Requires-Dist: geopandas<2,>=1.1.3; extra == 'full-gpu'
-Requires-Dist: loguru>=0.7.3; extra == 'full-gpu'
-Requires-Dist: lonboard<1,>=0.16.0; extra == 'full-gpu'
-Requires-Dist: matplotlib<4,>=3.10.9; extra == 'full-gpu'
-Requires-Dist: polars<2,>=1.39.0; extra == 'full-gpu'
-Requires-Dist: pyarrow<25,>=23.0.0; extra == 'full-gpu'
-Requires-Dist: pydantic<3,>=2.13.3; extra == 'full-gpu'
-Requires-Dist: pyogrio<1,>=0.10.0; extra == 'full-gpu'
-Requires-Dist: pyproj<4,>=3.7.2; extra == 'full-gpu'
-Requires-Dist: scipy<2,>=1.17.1; extra == 'full-gpu'
-Requires-Dist: shapely<3,>=2.1.2; extra == 'full-gpu'
-Requires-Dist: statsmodels<1,>=0.14.6; extra == 'full-gpu'
-Provides-Extra: gpu
-Requires-Dist: awscrt>=0.23.0; extra == 'gpu'
-Requires-Dist: cudf-cu13==26.6.*; extra == 'gpu'
-Requires-Dist: cudf-polars-cu13==26.6.*; extra == 'gpu'
-Requires-Dist: loguru>=0.7.3; extra == 'gpu'
-Requires-Dist: polars<2,>=1.39.0; extra == 'gpu'
-Requires-Dist: pyarrow<25,>=23.0.0; extra == 'gpu'
-Requires-Dist: pydantic<3,>=2.13.3; extra == 'gpu'
 Provides-Extra: spatial
 Requires-Dist: awscrt>=0.23.0; extra == 'spatial'
 Requires-Dist: duckdb<2,>=1.5.3; extra == 'spatial'
 Requires-Dist: geopandas<2,>=1.1.3; extra == 'spatial'
 Requires-Dist: loguru>=0.7.3; extra == 'spatial'
-Requires-Dist: polars<2,>=1.39.0; extra == 'spatial'
+Requires-Dist: polars<3,>=2.0.0; extra == 'spatial'
 Requires-Dist: pyarrow<25,>=23.0.0; extra == 'spatial'
 Requires-Dist: pydantic<3,>=2.13.3; extra == 'spatial'
 Requires-Dist: pyogrio<1,>=0.10.0; extra == 'spatial'
@@ -82,7 +55,7 @@ Requires-Dist: shapely<3,>=2.1.2; extra == 'spatial'
 Provides-Extra: stats
 Requires-Dist: awscrt>=0.23.0; extra == 'stats'
 Requires-Dist: loguru>=0.7.3; extra == 'stats'
-Requires-Dist: polars<2,>=1.39.0; extra == 'stats'
+Requires-Dist: polars<3,>=2.0.0; extra == 'stats'
 Requires-Dist: pyarrow<25,>=23.0.0; extra == 'stats'
 Requires-Dist: pydantic<3,>=2.13.3; extra == 'stats'
 Requires-Dist: scipy<2,>=1.17.1; extra == 'stats'
@@ -94,7 +67,7 @@ Requires-Dist: geoarrow-rust-core>=0.3.0; extra == 'viz'
 Requires-Dist: loguru>=0.7.3; extra == 'viz'
 Requires-Dist: lonboard<1,>=0.16.0; extra == 'viz'
 Requires-Dist: matplotlib<4,>=3.10.9; extra == 'viz'
-Requires-Dist: polars<2,>=1.39.0; extra == 'viz'
+Requires-Dist: polars<3,>=2.0.0; extra == 'viz'
 Requires-Dist: pyarrow<25,>=23.0.0; extra == 'viz'
 Requires-Dist: pydantic<3,>=2.13.3; extra == 'viz'
 Description-Content-Type: text/markdown

{core_lens-0.1.dev197 → core_lens-0.1.dev198}/benchmarks/bench_aoi.py

@@ -116,7 +116,7 @@ view = aoi_bbox.mws
 
 t1 = time.perf_counter()
 print(f"First access     : {(t1 - t0) * 1000:.2f} ms")
-print(f"Keys shape       : {view.keys.collect().shape}")
+print(f"Keys shape       : {view.keys.collect(engine='in-memory').shape}")
 
 # Second access — should hit _scoped cache (near-zero cost).
 t0 = time.perf_counter()

{core_lens-0.1.dev197 → core_lens-0.1.dev198}/benchmarks/bench_entity.py

@@ -216,7 +216,7 @@ t0 = time.perf_counter()
 view_attr = entity.where(state="Bihar")
 t1 = time.perf_counter()
 print(f"where(state)         : {(t1 - t0) * 1000:.2f} ms")
-print(f"Bihar keys           : {view_attr.keys.collect().shape}")
+print(f"Bihar keys           : {view_attr.keys.collect(engine='in-memory').shape}")
 
 
 # ── 9. BaseEntity.where() — composite attribute filter ───────────────────────
@@ -225,7 +225,7 @@ t0 = time.perf_counter()
 view_composite = entity.where(state="Bihar", district="Banka", tehsil="Banka")
 t1 = time.perf_counter()
 print(f"where(3 cols)        : {(t1 - t0) * 1000:.2f} ms")
-print(f"Banka keys           : {view_composite.keys.collect().shape}")
+print(f"Banka keys           : {view_composite.keys.collect(engine='in-memory').shape}")
 
 
 # ── 10. BaseEntity.where() — list value (is_in) ──────────────────────────────
@@ -234,7 +234,7 @@ t0 = time.perf_counter()
 view_list = entity.where(state=["Bihar", "Jharkhand"])
 t1 = time.perf_counter()
 print(f"where(list)          : {(t1 - t0) * 1000:.2f} ms")
-print(f"Bihar+Jharkhand keys : {view_list.keys.collect().shape}")
+print(f"Bihar+Jharkhand keys : {view_list.keys.collect(engine='in-memory').shape}")
 
 
 # ── 11. BaseEntity.where() — unknown kwarg error ─────────────────────────────
@@ -253,7 +253,7 @@ t0 = time.perf_counter()
 view_sf = entity.spatial_filter(bbox=SMALL_BBOX)
 t1 = time.perf_counter()
 print(f"spatial_filter (bbox): {(t1 - t0) * 1000:.2f} ms")
-print(f"Keys                 : {view_sf.keys.collect().shape}")
+print(f"Keys                 : {view_sf.keys.collect(engine='in-memory').shape}")
 
 # geometry mode
 small_poly = sgeom.box(*SMALL_BBOX)
@@ -269,7 +269,7 @@ view_sf_area = entity.spatial_filter(
 )
 t1 = time.perf_counter()
 print(
-    f"spatial_filter (area): {(t1 - t0) * 1000:.2f} ms  keys={view_sf_area.keys.collect().shape}"
+    f"spatial_filter (area): {(t1 - t0) * 1000:.2f} ms  keys={view_sf_area.keys.collect(engine='in-memory').shape}"
 )
 
 

{core_lens-0.1.dev197 → core_lens-0.1.dev198}/benchmarks/bench_spatial.py

@@ -183,7 +183,7 @@ view = entity.spatial_filter(bbox=SMALL_BBOX)
 
 t1 = time.perf_counter()
 print(f"spatial_filter (bbox) : {(t1 - t0) * 1000:.2f} ms")
-print(f"View keys             : {view.keys.collect().shape}")
+print(f"View keys             : {view.keys.collect(engine='in-memory').shape}")
 
 t0 = time.perf_counter()
 view2 = entity.spatial_filter(geometry=small_poly)

{core_lens-0.1.dev197 → core_lens-0.1.dev198}/benchmarks/bench_view.py

@@ -54,7 +54,7 @@ t0 = time.perf_counter()
 view_all = aoi.mws
 t1 = time.perf_counter()
 print(f"aoi.mws          : {(t1 - t0) * 1000:.2f} ms")
-print(f"Keys             : {view_all.keys.collect().shape}")
+print(f"Keys             : {view_all.keys.collect(engine='in-memory').shape}")
 
 
 # ── 2. View.where() — single filter ──────────────────────────────────────────
@@ -63,7 +63,7 @@ t0 = time.perf_counter()
 view_bihar = view_all.where(state="Bihar")
 t1 = time.perf_counter()
 print(f"where(state)     : {(t1 - t0) * 1000:.2f} ms")
-print(f"Bihar keys       : {view_bihar.keys.collect().shape}")
+print(f"Bihar keys       : {view_bihar.keys.collect(engine='in-memory').shape}")
 
 
 # ── 3. View.where() — composite AND filter ────────────────────────────────────
@@ -72,7 +72,7 @@ t0 = time.perf_counter()
 view_banka = view_all.where(state="Bihar", district="Banka", tehsil="Banka")
 t1 = time.perf_counter()
 print(f"where(3 cols)    : {(t1 - t0) * 1000:.2f} ms")
-print(f"Banka keys       : {view_banka.keys.collect().shape}")
+print(f"Banka keys       : {view_banka.keys.collect(engine='in-memory').shape}")
 
 
 # ── 4. View.spatial_filter() — further narrow ─────────────────────────────────
@@ -82,7 +82,7 @@ t0 = time.perf_counter()
 view_spatial = view_all.spatial_filter(bbox=SMALL_BBOX)
 t1 = time.perf_counter()
 print(f"spatial_filter   : {(t1 - t0) * 1000:.2f} ms")
-print(f"Spatial keys     : {view_spatial.keys.collect().shape}")
+print(f"Spatial keys     : {view_spatial.keys.collect(engine='in-memory').shape}")
 
 
 # ── 5. View.between() — date range (no I/O) ───────────────────────────────────

{core_lens-0.1.dev197 → core_lens-0.1.dev198}/pyproject.toml

@@ -47,7 +47,7 @@ venv = ".venv"
 
 [project.optional-dependencies]
 core = [
-    "polars>=1.39.0,<2",
+    "polars>=2.0.0,<3",
     "pyarrow>=23.0.0,<25",
     "pydantic>=2.13.3,<3",
     "loguru>=0.7.3",
@@ -69,9 +69,10 @@ viz = [
     "bokeh>=3.6.0,<4",
 ]
 stats = ["core-lens[core]", "scipy>=1.17.1,<2", "statsmodels>=0.14.6,<1"]
-gpu = ["core-lens[core]", "cudf-cu13==26.6.*", "cudf-polars-cu13==26.6.*"]
+# GPU dependencies commented out until NVIDIA RAPIDS releases cudf-polars for Polars 2.0+
+# gpu = ["core-lens[core]", "cudf-cu13==26.6.*", "cudf-polars-cu13==26.6.*"]
 full = ["core-lens[spatial,viz,stats]"]
-full-gpu = ["core-lens[spatial,viz,stats,gpu]"]
+# full-gpu = ["core-lens[spatial,viz,stats,gpu]"]
 
 [dependency-groups]
 dev = [

{core_lens-0.1.dev197 → core_lens-0.1.dev198}/src/core_lens/_version.py

@@ -18,7 +18,7 @@ version_tuple: tuple[int | str, ...]
 commit_id: str | None
 __commit_id__: str | None
 
-__version__ = version = '0.1.dev197'
-__version_tuple__ = version_tuple = (0, 1, 'dev197')
+__version__ = version = '0.1.dev198'
+__version_tuple__ = version_tuple = (0, 1, 'dev198')
 
 __commit_id__ = commit_id = None

{core_lens-0.1.dev197 → core_lens-0.1.dev198}/src/core_lens/aoi.py

@@ -20,7 +20,11 @@ from loguru import logger
 
 from core_lens.base.entity import BaseEntity, EntityValidationError
 from core_lens.utils.paths import is_cloud_uri
-from core_lens.utils.polars_utils import cached_read_schema, parquet_scan_path
+from core_lens.utils.polars_utils import (
+    cached_read_schema,
+    cast_filter_values,
+    parquet_scan_path,
+)
 
 if TYPE_CHECKING:
     from core_lens.base.view import View
@@ -232,22 +236,28 @@ def _cached_resolve_boundary(
     schema_types = cached_read_schema(static_path, storage_options)
     for col, val in entity_kwargs.items():
         if col in schema.key_cols or col in schema.extra_static_cols:
-            is_list_col = isinstance(schema_types.get(col), pl.List)
+            col_dtype = schema_types.get(col)
+            is_list_col = isinstance(col_dtype, pl.List)
+            cast_val = cast_filter_values(val, col_dtype)
             if isinstance(val, list):
                 if is_list_col:
                     filter_expr = (
                         filter_expr
-                        & pl.col(col).list.eval(pl.element().is_in(val)).list.any()
+                        & pl.col(col).list.eval(pl.element().is_in(cast_val)).list.any()
                     )
                 else:
-                    filter_expr = filter_expr & pl.col(col).is_in(val)
+                    filter_expr = filter_expr & pl.col(col).is_in(cast_val)
             else:
                 if is_list_col:
-                    filter_expr = filter_expr & pl.col(col).list.contains(val)
+                    filter_expr = filter_expr & pl.col(col).list.contains(cast_val)
                 else:
-                    filter_expr = filter_expr & (pl.col(col) == val)
+                    filter_expr = filter_expr & (pl.col(col) == cast_val)
 
-    df = lf.filter(filter_expr).select(candidate.key_cols + [geom_col]).collect()
+    df = (
+        lf.filter(filter_expr)
+        .select(candidate.key_cols + [geom_col])
+        .collect(engine="in-memory")
+    )
 
     if df.is_empty():
         raise ValueError(

{core_lens-0.1.dev197 → core_lens-0.1.dev198}/src/core_lens/base/entity.py

@@ -29,7 +29,11 @@ from core_lens.utils.spatial import (
     build_bbox_index,
     exact_spatial_filter,
 )
-from core_lens.utils.polars_utils import cached_read_schema, parquet_scan_path
+from core_lens.utils.polars_utils import (
+    cached_read_schema,
+    cast_filter_values,
+    parquet_scan_path,
+)
 
 if TYPE_CHECKING:
     import shapely
@@ -487,23 +491,25 @@ class BaseEntity(ABC):
                 dtype = schema[col]
                 is_list_col = isinstance(dtype, pl.List) or dtype == pl.List
                 is_val_list = isinstance(val, (list, tuple, set))
+                cast_val = cast_filter_values(val, dtype)
 
                 if is_list_col:
                     if is_val_list:
                         filter_expr = filter_expr & (
-                            pl.col(col).list.set_intersection(list(val)).list.len() > 0
+                            pl.col(col).list.set_intersection(list(cast_val)).list.len()
+                            > 0
                         )
                     else:
-                        filter_expr = filter_expr & pl.col(col).list.contains(val)
+                        filter_expr = filter_expr & pl.col(col).list.contains(cast_val)
                 else:
                     if is_val_list:
-                        filter_expr = filter_expr & pl.col(col).is_in(list(val))
+                        filter_expr = filter_expr & pl.col(col).is_in(list(cast_val))
                     else:
-                        filter_expr = filter_expr & (pl.col(col) == val)
+                        filter_expr = filter_expr & (pl.col(col) == cast_val)
             lf = lf.filter(filter_expr)
 
         if not entity_kwargs:
-            keys = lf.select(self.key_cols).collect()
+            keys = lf.select(self.key_cols).collect(engine="in-memory")
         else:
             import shapely
             import shapely.wkb as swkb
@@ -538,7 +544,7 @@ class BaseEntity(ABC):
                 matched = (
                     other_lf.filter(match_expr)
                     .select([other_profile.geometry_col])
-                    .collect()
+                    .collect(engine="in-memory")
                 )
 
                 if matched.is_empty():
@@ -572,8 +578,9 @@ class BaseEntity(ABC):
                     candidates.select(self.key_cols).lazy(),
                     on=self.key_cols,
                     how="semi",
+                    maintain_order="left",
                 )
-                df = lf.collect()
+                df = lf.collect(engine="in-memory")
 
                 if df.is_empty():
                     keys = df.select(self.key_cols)

{core_lens-0.1.dev197 → core_lens-0.1.dev198}/src/core_lens/base/namespaces/plot.py

@@ -91,7 +91,7 @@ def _wkb_to_arrow_table(
     )
 
     # --- Fast path: pure Polygon / MultiPolygon rows (no Shapely) ---
-    poly_mask = geom_type_series.is_in([3, 6])
+    poly_mask = geom_type_series.is_in(pl.Series([3, 6], dtype=pl.UInt8).to_list())
     df_poly = df.filter(poly_mask)
 
     # --- Slow path: GeometryCollection rows — extract polygon parts via Shapely ---
@@ -545,7 +545,7 @@ class PlotNamespace:
         def _make_fig(pdf: Any, title: str) -> "BokehFigure":
             hover_cols = [c for c in pdf.columns if c not in ("geometry", "geom")]
             tooltips = [(c, f"@{{{c}}}") for c in hover_cols]
-            fig = figure(  # type: ignore[call-arg]
+            fig = figure(
                 width=900,
                 height=400,
                 x_axis_label=x,
@@ -658,7 +658,10 @@ class PlotNamespace:
         if len(unique_entities) > top_n:
             entities = unique_entities.limit(top_n).to_list()
             entity_df = entity_df.join(
-                pl.DataFrame({key_col: entities}), on=key_col, how="inner"
+                pl.DataFrame({key_col: entities}),
+                on=key_col,
+                how="inner",
+                maintain_order="left",
             )
         entity_pdf = entity_df.to_pandas()
 
@@ -716,12 +719,17 @@ class PlotNamespace:
         unique_entities = df[key_col].unique()
         if len(unique_entities) > top_n:
             entities = unique_entities.limit(top_n).to_list()
-            df = df.join(pl.DataFrame({key_col: entities}), on=key_col, how="inner")
+            df = df.join(
+                pl.DataFrame({key_col: entities}),
+                on=key_col,
+                how="inner",
+                maintain_order="left",
+            )
 
         def _make_scatter(y_col: str) -> "BokehFigure":
             from bokeh.models import ColumnDataSource
 
-            fig = figure(  # type: ignore[call-arg]
+            fig = figure(
                 width=800,
                 height=450,
                 x_axis_label=x,
@@ -815,12 +823,17 @@ class PlotNamespace:
         unique_entities = df[key_col].unique()
         if len(unique_entities) > top_n:
             entities = unique_entities.limit(top_n).to_list()
-            df = df.join(pl.DataFrame({key_col: entities}), on=key_col, how="inner")
+            df = df.join(
+                pl.DataFrame({key_col: entities}),
+                on=key_col,
+                how="inner",
+                maintain_order="left",
+            )
 
         df_pd = df.to_pandas()
 
         def _make_hist(x_col: str) -> "BokehFigure":
-            fig = figure(  # type: ignore[call-arg]
+            fig = figure(
                 width=800,
                 height=400,
                 x_axis_label=x_col,
@@ -902,7 +915,7 @@ class PlotNamespace:
 
         from bokeh.models import FactorRange
 
-        fig = figure(  # type: ignore[call-arg]
+        fig = figure(
             width=600,
             height=550,
             x_range=FactorRange(factors=cols),
@@ -993,7 +1006,7 @@ class PlotNamespace:
 
         mapper = LinearColorMapper(palette="Viridis256", low=low, high=high)
 
-        fig = figure(  # type: ignore[call-arg]
+        fig = figure(
             width=900,
             height=500,
             x_range=FactorRange(factors=x_vals),
@@ -1046,7 +1059,7 @@ class PlotNamespace:
         for row_col in num_cols:
             row: list[Any] = []
             for col_col in num_cols:
-                fig = figure(  # type: ignore[call-arg]
+                fig = figure(
                     width=200,
                     height=200,
                     x_axis_label=col_col if row_col == num_cols[-1] else "",

{core_lens-0.1.dev197 → core_lens-0.1.dev198}/src/core_lens/base/namespaces/stats.py

@@ -293,7 +293,7 @@ class StatsNamespace:
                     columns=["dw_precipitation", "ci_cropping_intensity"],
                     method=CorrelateMethod.PEARSON,
                 )
-                print(result.data.collect())
+                print(result.data.collect(engine="in-memory"))
 
             Per-MWS correlation, ranked by strength of dependence::
 
@@ -310,7 +310,7 @@ class StatsNamespace:
                     .sort("abs_corr", descending=True)
                     .select("mws_id", "correlation", "p_value", "n_obs")
                     .limit(5)
-                    .collect()
+                    .collect(engine="in-memory")
                 )
                 print(top5)
         """
@@ -378,18 +378,20 @@ class StatsNamespace:
                 )
 
                 # --- single collect boundary: scipy.t.sf needs eager numpy ---
-                result = lazy_result.collect()
+                result = lazy_result.collect(engine="in-memory")
                 t_vals = result["t_stat"].to_numpy()
                 df_vals = result["n_obs"].to_numpy() - 2
                 p_vals = 2.0 * sp.t.sf(np.abs(t_vals), df_vals)
-                result = result.with_columns(pl.Series("p_value", p_vals))
+                result = result.with_columns(
+                    pl.Series("p_value", np.atleast_1d(p_vals))
+                )
 
                 data = result.select([group_by, "correlation", "p_value", "n_obs"])
 
             else:
                 # Kendall: scipy has no batched/lazy equivalent, so this branch
                 # is inherently eager. Collect ONCE here — not scattered filters.
-                df = lf.collect()
+                df = lf.collect(engine="in-memory")
                 rows: list[dict[str, Any]] = []
                 for sub_full in df.partition_by(group_by):
                     key = sub_full[group_by][0]
@@ -422,7 +424,7 @@ class StatsNamespace:
 
         # Pooled mode — eager needed regardless (scipy.pearsonr/spearmanr/
         # kendalltau all take raw numpy arrays). Collect ONCE here.
-        df = lf.collect()
+        df = lf.collect(engine="in-memory")
         n_obs = len(df)
 
         rows = []
@@ -674,7 +676,9 @@ class StatsNamespace:
                 .select(join_keys + [column])
                 .rename({column: "value_to"})
             )
-            joined = from_df.join(to_df, on=join_keys, how="inner").with_columns(
+            joined = from_df.join(
+                to_df, on=join_keys, how="inner", maintain_order="left"
+            ).with_columns(
                 [
                     pl.lit(from_period).alias("year_from").cast(pl.Int32),
                     pl.lit(to_period).alias("year_to").cast(pl.Int32),
@@ -800,7 +804,7 @@ class StatsNamespace:
 
         df = self._r.df()
         if isinstance(df, pl.LazyFrame):
-            df = df.collect()
+            df = df.collect(engine="in-memory")
         key = self._r.key_cols[0]
 
         if mode == "cross_sectional":
@@ -945,20 +949,22 @@ class StatsNamespace:
                 n_dropped = counts.height - valid_keys.height
 
                 med_df = (
-                    base_df.join(valid_keys, on=key)
+                    base_df.join(valid_keys, on=key, maintain_order="left")
                     .group_by(key)
                     .agg(pl.col(column).median().alias("_med"))
                 )
                 mad_df = (
-                    base_df.join(valid_keys, on=key)
-                    .join(med_df, on=key)
+                    base_df.join(valid_keys, on=key, maintain_order="left")
+                    .join(med_df, on=key, maintain_order="left")
                     .with_columns(
                         (pl.col(column) - pl.col("_med")).abs().alias("_absdev")
                     )
                     .group_by(key)
                     .agg(pl.col("_absdev").median().alias("_mad"))
                 )
-                stats_df = med_df.join(mad_df, on=key).with_columns(
+                stats_df = med_df.join(
+                    mad_df, on=key, maintain_order="left"
+                ).with_columns(
                     (pl.col("_mad") * 1.4826).clip(lower_bound=1e-12).alias("_scale")
                 )
 
@@ -971,7 +977,7 @@ class StatsNamespace:
                 )
 
                 data = (
-                    eval_df.join(stats_df, on=key, how="inner")
+                    eval_df.join(stats_df, on=key, how="inner", maintain_order="left")
                     .with_columns(
                         ((pl.col(column) - pl.col("_med")) / pl.col("_scale")).alias(
                             "anomaly_score"
@@ -1203,7 +1209,12 @@ class StatsNamespace:
             if spec is None:
                 # Pull from result.data directly.
                 if col_name in df.columns:
-                    feat = feat.join(df.select([key, col_name]), on=key, how="left")
+                    feat = feat.join(
+                        df.select([key, col_name]),
+                        on=key,
+                        how="left",
+                        maintain_order="left",
+                    )
                 # else silently skip — not in data and no spec provided
                 continue
 
@@ -1286,7 +1297,12 @@ class StatsNamespace:
             else:
                 fetched = collect_lf(col_lf.group_by(key).agg(agg_expr))
 
-            feat = feat.join(fetched.select([key, col_name]), on=key, how="left")
+            feat = feat.join(
+                fetched.select([key, col_name]),
+                on=key,
+                how="left",
+                maintain_order="left",
+            )
 
         feature_cols = [c for c in feat.columns if c != key]
 

{core_lens-0.1.dev197 → core_lens-0.1.dev198}/src/core_lens/base/result.py

@@ -184,13 +184,24 @@ class Result:
         # 2. Push down the geometry filter into the Parquet reader using is_in
         # so we don't scan the entire static geometry file (which takes ~0.4s).
         geo_lf = self.entity.geometry_lazy
+        geo_schema = geo_lf.collect_schema()
+        cast_exprs = [
+            pl.col(k).cast(geo_schema[k])
+            for k in key_cols
+            if k in geo_schema and keys_df[k].dtype != geo_schema[k]
+        ]
+        if cast_exprs:
+            keys_df = keys_df.with_columns(cast_exprs)
+
         if len(key_cols) == 1 and keys_df.height < 50000:
             key = key_cols[0]
             geo_lf = geo_lf.filter(pl.col(key).is_in(keys_df[key].to_list()))
         else:
-            geo_lf = geo_lf.join(keys_df.lazy(), on=key_cols, how="semi")
+            geo_lf = geo_lf.join(
+                keys_df.lazy(), on=key_cols, how="semi", maintain_order="left"
+            )
 
-        joined = res.data.join(geo_lf, on=key_cols, how="left")
+        joined = res.data.join(geo_lf, on=key_cols, how="left", maintain_order="left")
 
         # 3. Cache the joined geometry so multiple plots are instant.
         return res._replace(data=joined, has_geometry=True).materialise()

{core_lens-0.1.dev197 → core_lens-0.1.dev198}/src/core_lens/base/view.py

@@ -14,6 +14,7 @@ from core_lens.utils.polars_utils import (
     scan_with_key_filter,
     collect_lf,
     cached_read_schema,
+    cast_filter_values,
 )
 
 if TYPE_CHECKING:
@@ -129,20 +130,22 @@ class View:
 
         filter_expr = pl.lit(True)
         for col, val in kwargs.items():
-            is_list_col = isinstance(schema_types.get(col), pl.List)
+            col_dtype = schema_types.get(col)
+            is_list_col = isinstance(col_dtype, pl.List)
+            cast_val = cast_filter_values(val, col_dtype)
             if isinstance(val, list):
                 if is_list_col:
                     filter_expr = (
                         filter_expr
-                        & pl.col(col).list.eval(pl.element().is_in(val)).list.any()
+                        & pl.col(col).list.eval(pl.element().is_in(cast_val)).list.any()
                     )
                 else:
-                    filter_expr = filter_expr & pl.col(col).is_in(val)
+                    filter_expr = filter_expr & pl.col(col).is_in(cast_val)
             else:
                 if is_list_col:
-                    filter_expr = filter_expr & pl.col(col).list.contains(val)
+                    filter_expr = filter_expr & pl.col(col).list.contains(cast_val)
                 else:
-                    filter_expr = filter_expr & (pl.col(col) == val)
+                    filter_expr = filter_expr & (pl.col(col) == cast_val)
 
         keys = lf.filter(filter_expr).select(self.entity.key_cols)
 
@@ -202,8 +205,10 @@ class View:
         candidates = bbox_intersects_geometry(self.entity._index, geometry)
         candidates = (
             candidates.lazy()
-            .join(self.keys, on=self.entity.key_cols, how="inner")
-            .collect()
+            .join(
+                self.keys, on=self.entity.key_cols, how="inner", maintain_order="left"
+            )
+            .collect(engine="in-memory")
         )
 
         keys_df = exact_spatial_filter(
@@ -521,8 +526,11 @@ class View:
                     data_df.select(self.entity.key_cols).lazy(),
                     on=self.entity.key_cols,
                     how="semi",
-                ).collect()
-                data_df = data_df.join(geom_df, on=self.entity.key_cols, how="left")
+                    maintain_order="left",
+                ).collect(engine="in-memory")
+                data_df = data_df.join(
+                    geom_df, on=self.entity.key_cols, how="left", maintain_order="left"
+                )
 
             from core_lens.utils.spatial import execute_spatial_join