jarvisplot 1.3.8__tar.gz → 1.3.10__tar.gz

{jarvisplot-1.3.8 → jarvisplot-1.3.10}/JarvisPLOT.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: JarvisPLOT
-Version: 1.3.8
+Version: 1.3.10
 Summary: Jarvis-PLOT: YAML-driven plotting engine
 Requires-Python: >=3.10
 Description-Content-Type: text/markdown
@@ -117,6 +117,7 @@ and evidence summary behavior.
 
 - Figures are saved automatically to the output paths defined in the YAML configuration.
 - Common output formats include PNG and PDF (backend-dependent).
+- Saved figures include file metadata such as `Creator: Jarvis-PLOT, powered by Jarvis-HEP` and `Jarvis-PLOT version: X.Y.Z`; the PNG `Description` also includes both fields for macOS Finder compatibility.
 - Jarvis-PLOT works in headless environments (SSH, batch jobs) without any GUI backend.
 
 ---

{jarvisplot-1.3.8 → jarvisplot-1.3.10}/JarvisPLOT.egg-info/SOURCES.txt

@@ -26,14 +26,17 @@ jarvisplot/Figure/adapters_ternary.py
 jarvisplot/Figure/colorbar_runtime.py
 jarvisplot/Figure/config_runtime.py
 jarvisplot/Figure/data_pipelines.py
+jarvisplot/Figure/density_cell_runtime.py
 jarvisplot/Figure/dynesty_runtime.py
 jarvisplot/Figure/figure.py
 jarvisplot/Figure/helper.py
+jarvisplot/Figure/interp_2d_runtime.py
 jarvisplot/Figure/interp_natural_neighbor.py
 jarvisplot/Figure/interp_natural_neighbor_exact.py
 jarvisplot/Figure/layer_runtime.py
 jarvisplot/Figure/layout_runtime.py
 jarvisplot/Figure/method_registry.py
+jarvisplot/Figure/posterior_hpd.py
 jarvisplot/Figure/preprocessor.py
 jarvisplot/Figure/preprocessor_runtime.py
 jarvisplot/Figure/profile_runtime.py
@@ -68,11 +71,14 @@ jarvisplot/utils/expression.py
 jarvisplot/utils/interpolator.py
 jarvisplot/utils/pathing.py
 tests/test_data_loader_contracts.py
+tests/test_density_cell_transform.py
 tests/test_docs_status_labels.py
 tests/test_figure_contracts.py
+tests/test_interp_2d_transform.py
 tests/test_interp_natural_neighbor.py
 tests/test_interpolator_contracts.py
 tests/test_jpcontour_methods.py
 tests/test_parse_data_metadata.py
 tests/test_pathing_contracts.py
+tests/test_posterior_hpd.py
 tests/test_template_json.py

{jarvisplot-1.3.8 → jarvisplot-1.3.10}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: JarvisPLOT
-Version: 1.3.8
+Version: 1.3.10
 Summary: Jarvis-PLOT: YAML-driven plotting engine
 Requires-Python: >=3.10
 Description-Content-Type: text/markdown
@@ -117,6 +117,7 @@ and evidence summary behavior.
 
 - Figures are saved automatically to the output paths defined in the YAML configuration.
 - Common output formats include PNG and PDF (backend-dependent).
+- Saved figures include file metadata such as `Creator: Jarvis-PLOT, powered by Jarvis-HEP` and `Jarvis-PLOT version: X.Y.Z`; the PNG `Description` also includes both fields for macOS Finder compatibility.
 - Jarvis-PLOT works in headless environments (SSH, batch jobs) without any GUI backend.
 
 ---

{jarvisplot-1.3.8 → jarvisplot-1.3.10}/README.md

@@ -97,6 +97,7 @@ and evidence summary behavior.
 
 - Figures are saved automatically to the output paths defined in the YAML configuration.
 - Common output formats include PNG and PDF (backend-dependent).
+- Saved figures include file metadata such as `Creator: Jarvis-PLOT, powered by Jarvis-HEP` and `Jarvis-PLOT version: X.Y.Z`; the PNG `Description` also includes both fields for macOS Finder compatibility.
 - Jarvis-PLOT works in headless environments (SSH, batch jobs) without any GUI backend.
 
 ---

{jarvisplot-1.3.8 → jarvisplot-1.3.10}/jarvisplot/Figure/adapters_rect.py

@@ -11,8 +11,9 @@ from matplotlib.axes import Axes
 from matplotlib.collections import PolyCollection
 #
 from .helper import _auto_clip, _mask_by_extend, voronoi_finite_polygons_2d, _clip_poly_to_rect
-from .profile_runtime import grid_profile_mesh
+from .profile_runtime import regular_grid_mesh
 from .interp_natural_neighbor import resolve_backend
+from .posterior_hpd import prepare_hpd_contour_style
 
 DEFAULT_JP_SAMPLE_GRID = 500
 
@@ -371,19 +372,20 @@ class StdAxesAdapter:
         artists = self.ax.scatter(x, y, **kw)
         return _auto_clip(artists, self.ax, self._clip_path)
 
-    def grid_profile(self, **kwargs):
-        """Grid-cell partition rendering for profiled scalar fields.
+    def pcolormesh(self, *args, **kwargs):
+        """Render regular-grid fields, including compact tables with __grid_* metadata."""
+        if args:
+            kw = self._merge("pcolormesh", kwargs)
+            artist = self.ax.pcolormesh(*args, **kw)
+            return _auto_clip(artist, self.ax, self._clip_path)
 
-        Compared with tripcolor, this draws axis-aligned cells (like a regular
-        partition of the plane), closer to Voronoi-style region coloring.
-        """
         x = np.asarray(kwargs.pop("x"), dtype=float)
         y = np.asarray(kwargs.pop("y"), dtype=float)
         z = np.asarray(kwargs.pop("z", np.zeros_like(x)), dtype=float)
         df = kwargs.pop("__df__", None)
-        kw = self._merge("grid_profile", kwargs)
+        kw = self._merge("pcolormesh", kwargs)
 
-        # Strip kwargs that are not used by grid_profile rendering.
+        # Strip kwargs that are not used by pcolormesh rendering.
         for k in ("shading", "levels", "extend", "space", "marker", "s", "c"):
             kw.pop(k, None)
 
@@ -411,7 +413,7 @@ class StdAxesAdapter:
             xlim = [float(self.ax.get_xlim()[0]), float(self.ax.get_xlim()[1])]
         if ylim is None:
             ylim = [float(self.ax.get_ylim()[0]), float(self.ax.get_ylim()[1])]
-        mesh = grid_profile_mesh(
+        mesh = regular_grid_mesh(
             x,
             y,
             z,
@@ -441,7 +443,6 @@ class StdAxesAdapter:
         if zorder is not None:
             kw.setdefault("zorder", zorder)
 
-        # histo2d-like rendering via pcolormesh
         if norm is not None:
             artist = self.ax.pcolormesh(
                 x_edges,
@@ -482,7 +483,13 @@ class StdAxesAdapter:
 
     def contour(self, *args, **kwargs):
         kw = self._merge("contour", kwargs)
+        label_map = kw.pop("_hpd_label_map", None)
         artists = self.ax.contour(*args, **kw)
+        if label_map:
+            try:
+                self.ax.clabel(artists, fmt=label_map)
+            except Exception:
+                pass
         return _auto_clip(artists, self.ax, self._clip_path)
 
     def contourf(self, *args, **kwargs):
@@ -510,6 +517,7 @@ class StdAxesAdapter:
         """Scattered-data contour plot rendered in axes-fraction coordinates."""
         kw = self._merge("contour", kwargs)
         kw.pop("transform", None)
+        logger = kw.pop("_logger", None)
         X, Y, Z = self._interpolate_jpcontour_grid(
             x,
             y,
@@ -524,6 +532,8 @@ class StdAxesAdapter:
             diagnostics=diagnostics,
             backend_options=backend_options,
         )
+        kw = prepare_hpd_contour_style(Z, X, Y, kw, logger=logger)
+        label_map = kw.pop("_hpd_label_map", None)
         artists = self.ax.contour(
             X,
             Y,
@@ -532,6 +542,11 @@ class StdAxesAdapter:
             transform=self.ax.transAxes,
             **kw,
         )
+        if label_map:
+            try:
+                self.ax.clabel(artists, fmt=label_map)
+            except Exception:
+                pass
         return _auto_clip(artists, self.ax, self._clip_path)
 
     def jpcontourf(

{jarvisplot-1.3.8 → jarvisplot-1.3.10}/jarvisplot/Figure/adapters_ternary.py

@@ -54,14 +54,6 @@ class TernaryAxesAdapter(StdAxesAdapter):
             kwargs['y'] = y
         return super().scatter(**kwargs)
 
-    def grid_profile(self, **kwargs):
-        if {"left", "right", "bottom"}.issubset(kwargs.keys()):
-            x, y = self._lbr_to_xy(kwargs.pop('left'), kwargs.pop('right'), kwargs.pop('bottom'))
-            kwargs['x'] = x
-            kwargs['y'] = y
-        return super().grid_profile(**kwargs)
-
-
     def plot(self, **kwargs):
         if {"left", "right", "bottom"}.issubset(kwargs.keys()):
             x, y = self._lbr_to_xy(kwargs.pop('left'), kwargs.pop('right'), kwargs.pop('bottom'))

{jarvisplot-1.3.8 → jarvisplot-1.3.10}/jarvisplot/Figure/colorbar_runtime.py

@@ -144,7 +144,6 @@ _COLORED_Z_METHODS = frozenset({
     "tripcolor", "tripcolor_axes",
     "pcolor", "pcolormesh", "imshow",
     "voronoi", "voronoif",
-    "grid_profile",
 })
 
 _CONTOUR_LIKE_METHODS = frozenset({

jarvisplot-1.3.10/jarvisplot/Figure/density_cell_runtime.py

@@ -0,0 +1,427 @@
+from __future__ import annotations
+
+import re
+from typing import Any, Mapping, Sequence, Tuple
+
+import numpy as np
+import pandas as pd
+
+from ..memtrace import memtrace_checkpoint
+from ..utils.expression import eval_dataframe_expression
+
+
+_DENSITY_CELL_TYPES = {"make_density_core"}
+
+
+class _ScaledBwMethod:
+    def __init__(self, factor: float, base: str):
+        self.factor = float(factor)
+        self.base = str(base).strip().lower()
+
+    def __call__(self, kde) -> float:
+        if self.base == "silverman":
+            return self.factor * float(kde.silverman_factor())
+        return self.factor * float(kde.scotts_factor())
+
+    def __str__(self) -> str:
+        return f"{self.factor:g}*{self.base}"
+
+
+def is_density_cell_transform(step: Any) -> bool:
+    if not isinstance(step, Mapping):
+        return False
+    if any(name in step for name in _DENSITY_CELL_TYPES):
+        return True
+    return str(step.get("type", "")).strip().lower() in _DENSITY_CELL_TYPES
+
+
+def density_cell_config(step: Mapping[str, Any]) -> Mapping[str, Any]:
+    for name in _DENSITY_CELL_TYPES:
+        if name in step:
+            cfg = step.get(name)
+            return cfg if isinstance(cfg, Mapping) else {}
+    if str(step.get("type", "")).strip().lower() in _DENSITY_CELL_TYPES:
+        return step
+    return {}
+
+
+def _logger_emit(logger, level: str, message: str) -> None:
+    if logger is None:
+        return
+    try:
+        getattr(logger, level)(message)
+    except Exception:
+        pass
+
+
+def _coord_cfg(cfg: Mapping[str, Any], key: str) -> Mapping[str, Any]:
+    coors = cfg.get("coordinates", {})
+    if not isinstance(coors, Mapping):
+        coors = {}
+    spec = coors.get(key, cfg.get(key, None))
+    if isinstance(spec, Mapping):
+        return spec
+    if spec is None:
+        return {}
+    return {"name": str(spec)}
+
+
+def _output_name(cfg: Mapping[str, Any], key: str, default: str) -> str:
+    name = _coord_cfg(cfg, key).get("name", default)
+    text = str(name).strip()
+    return text or default
+
+
+def _resolve_array(df: pd.DataFrame, cfg: Mapping[str, Any], key: str, default: str, logger, *, required: bool) -> np.ndarray:
+    spec = _coord_cfg(cfg, key)
+    expr = spec.get("expr")
+    if expr is not None:
+        return np.asarray(eval_dataframe_expression(df, expr, logger=logger, allow_column=True), dtype=float)
+    name = str(spec.get("name", default)).strip()
+    if not name:
+        if required:
+            raise ValueError(f"make_density_core requires a non-empty coordinate for '{key}'.")
+        return np.ones(int(df.shape[0]), dtype=float)
+    if name not in df.columns:
+        if required:
+            raise KeyError(f"make_density_core missing required column '{name}'.")
+        return np.ones(int(df.shape[0]), dtype=float)
+    return np.asarray(df[name], dtype=float)
+
+
+def _axis_lim(values: np.ndarray, cfg: Mapping[str, Any], key: str) -> Tuple[float, float]:
+    spec = _coord_cfg(cfg, key)
+    lim = spec.get("lim", spec.get("limits", None))
+    domain = cfg.get("domain", {})
+    if not isinstance(domain, Mapping):
+        domain = {}
+    if lim is None:
+        lim = domain.get(f"{key}lim", cfg.get(f"{key}lim", None))
+    if lim is None:
+        finite = np.asarray(values, dtype=float)
+        finite = finite[np.isfinite(finite)]
+        if finite.size == 0:
+            raise ValueError(f"make_density_core cannot infer {key} limits from non-finite data.")
+        lo = float(np.min(finite))
+        hi = float(np.max(finite))
+        if lo == hi:
+            hi = lo + 1.0
+        return lo, hi
+    if not isinstance(lim, Sequence) or len(lim) != 2:
+        raise ValueError(f"make_density_core coordinates.{key}.lim must have two values.")
+    lo, hi = float(lim[0]), float(lim[1])
+    if not (np.isfinite(lo) and np.isfinite(hi) and hi > lo):
+        raise ValueError(f"make_density_core coordinates.{key}.lim requires finite lo < hi.")
+    return lo, hi
+
+
+def _axis_scale(cfg: Mapping[str, Any], key: str) -> str:
+    spec = _coord_cfg(cfg, key)
+    domain = cfg.get("domain", {})
+    if not isinstance(domain, Mapping):
+        domain = {}
+    return str(spec.get("scale", domain.get(f"{key}scale", cfg.get(f"{key}scale", "linear")))).strip().lower()
+
+
+def _project_axis(values: np.ndarray, lim: Tuple[float, float], scale: str, key: str) -> np.ndarray:
+    lo, hi = lim
+    values = np.asarray(values, dtype=float)
+    if scale == "log":
+        if lo <= 0 or hi <= 0:
+            raise ValueError(f"make_density_core {key} log scale requires positive limits.")
+        out = np.full(values.shape, np.nan, dtype=float)
+        mask = values > 0
+        out[mask] = (np.log(values[mask]) - np.log(lo)) / (np.log(hi) - np.log(lo))
+        return out
+    return (values - lo) / (hi - lo)
+
+
+def _unproject_axis(values: np.ndarray, lim: Tuple[float, float], scale: str) -> np.ndarray:
+    lo, hi = lim
+    values = np.asarray(values, dtype=float)
+    if scale == "log":
+        return np.exp(np.log(lo) + values * (np.log(hi) - np.log(lo)))
+    return lo + values * (hi - lo)
+
+
+def _grid_shape(cfg: Mapping[str, Any], default: int = 128) -> Tuple[int, int]:
+    grid = cfg.get("grid", {})
+    if not isinstance(grid, Mapping):
+        grid = {}
+    bins = grid.get("bins", grid.get("bin", cfg.get("bins", cfg.get("bin", None))))
+    if bins is not None:
+        nx = ny = int(bins)
+    else:
+        nx = int(grid.get("nx", cfg.get("nx", default)))
+        ny = int(grid.get("ny", cfg.get("ny", default)))
+    if nx < 2 or ny < 2:
+        raise ValueError("make_density_core grid requires nx >= 2 and ny >= 2.")
+    return nx, ny
+
+
+def _bw_method(cfg: Mapping[str, Any]):
+    bw = cfg.get("bw_method", cfg.get("bandwidth", None))
+    if bw is None:
+        return None
+    if isinstance(bw, str):
+        text = bw.strip().lower()
+        if text in {"", "default", "none"}:
+            return None
+        if text in {"scott", "silverman"}:
+            return text
+        match = re.fullmatch(r"([0-9]*\.?[0-9]+)\s*\*\s*(scott|silverman)", text)
+        if match:
+            factor = float(match.group(1))
+            if not np.isfinite(factor) or factor <= 0:
+                raise ValueError("make_density_core scaled bw_method factor must be positive and finite.")
+            return _ScaledBwMethod(factor, match.group(2))
+        value = float(text)
+        if not np.isfinite(value) or value <= 0:
+            raise ValueError("make_density_core float bw_method must be positive and finite.")
+        return value
+    value = float(bw)
+    if not np.isfinite(value) or value <= 0:
+        raise ValueError("make_density_core float bw_method must be positive and finite.")
+    return value
+
+
+def _prepare_samples(df: pd.DataFrame, cfg: Mapping[str, Any], logger):
+    x = _resolve_array(df, cfg, "x", "x", logger, required=True)
+    y = _resolve_array(df, cfg, "y", "y", logger, required=True)
+    weight_required = bool(_coord_cfg(cfg, "weight"))
+    w = _resolve_array(df, cfg, "weight", "weight", logger, required=weight_required)
+    n = min(len(x), len(y), len(w), int(df.shape[0]))
+    x = np.asarray(x[:n], dtype=float)
+    y = np.asarray(y[:n], dtype=float)
+    w = np.asarray(w[:n], dtype=float)
+    valid = np.isfinite(x) & np.isfinite(y) & np.isfinite(w) & (w > 0)
+    if not np.any(valid):
+        raise ValueError("make_density_core got no finite samples with positive finite weights.")
+    if not np.all(valid):
+        _logger_emit(logger, "warning", f"make_density_core dropped {int((~valid).sum())} invalid or non-positive samples.")
+    return x[valid], y[valid], w[valid], n
+
+
+def _normalize(values: np.ndarray, *, enabled: bool) -> np.ndarray:
+    values = np.asarray(values, dtype=float)
+    if not enabled:
+        return values
+    total = float(np.sum(values))
+    if not np.isfinite(total) or total <= 0:
+        raise ValueError("make_density_core cannot normalize non-positive or non-finite total weight.")
+    return values / total
+
+
+def _grid_density_cell(x, y, w, cfg, logger):
+    xlim = _axis_lim(x, cfg, "x")
+    ylim = _axis_lim(y, cfg, "y")
+    xscale = _axis_scale(cfg, "x")
+    yscale = _axis_scale(cfg, "y")
+    nx, ny = _grid_shape(cfg)
+    xp = _project_axis(x, xlim, xscale, "x")
+    yp = _project_axis(y, ylim, yscale, "y")
+    inside = np.isfinite(xp) & np.isfinite(yp) & (xp >= 0.0) & (xp <= 1.0) & (yp >= 0.0) & (yp <= 1.0)
+    if not np.any(inside):
+        raise ValueError("make_density_core grid found no positive-mass samples inside the domain.")
+    if not np.all(inside):
+        _logger_emit(logger, "warning", f"make_density_core grid ignored {int((~inside).sum())} samples outside the domain.")
+    ix = np.clip(xp[inside], 0.0, 1.0 - 1e-12)
+    iy = np.clip(yp[inside], 0.0, 1.0 - 1e-12)
+    ix = (ix * nx).astype(np.int32)
+    iy = (iy * ny).astype(np.int32)
+    mass = np.zeros((ny, nx), dtype=float)
+    np.add.at(mass, (iy, ix), w[inside])
+    gx = (np.arange(nx, dtype=float) + 0.5) / float(nx)
+    gy = (np.arange(ny, dtype=float) + 0.5) / float(ny)
+    xg, yg = np.meshgrid(_unproject_axis(gx, xlim, xscale), _unproject_axis(gy, ylim, yscale))
+    return xg.ravel(), yg.ravel(), mass.ravel()
+
+
+def _bridson_points(radius: float, *, seed: int | None = None, k: int = 30) -> np.ndarray:
+    rng = np.random.default_rng(seed)
+    cell = radius / np.sqrt(2.0)
+    nx = max(1, int(np.ceil(1.0 / cell)))
+    ny = max(1, int(np.ceil(1.0 / cell)))
+    grid = -np.ones((ny, nx), dtype=np.int64)
+    points: list[tuple[float, float]] = []
+    active: list[int] = []
+
+    def _grid_pos(pt):
+        gx = min(nx - 1, max(0, int(pt[0] / cell)))
+        gy = min(ny - 1, max(0, int(pt[1] / cell)))
+        return gx, gy
+
+    def _far_enough(pt) -> bool:
+        gx, gy = _grid_pos(pt)
+        r2 = radius * radius
+        for yy in range(max(0, gy - 2), min(ny, gy + 3)):
+            for xx in range(max(0, gx - 2), min(nx, gx + 3)):
+                idx = int(grid[yy, xx])
+                if idx < 0:
+                    continue
+                px, py = points[idx]
+                if (pt[0] - px) ** 2 + (pt[1] - py) ** 2 < r2:
+                    return False
+        return True
+
+    first = (float(rng.random()), float(rng.random()))
+    points.append(first)
+    gx, gy = _grid_pos(first)
+    grid[gy, gx] = 0
+    active.append(0)
+    while active:
+        slot = int(rng.integers(0, len(active)))
+        idx = active[slot]
+        base = points[idx]
+        accepted = False
+        for _ in range(int(k)):
+            rr = radius * (1.0 + rng.random())
+            theta = 2.0 * np.pi * rng.random()
+            cand = (base[0] + rr * np.cos(theta), base[1] + rr * np.sin(theta))
+            if cand[0] < 0.0 or cand[0] > 1.0 or cand[1] < 0.0 or cand[1] > 1.0:
+                continue
+            if not _far_enough(cand):
+                continue
+            points.append((float(cand[0]), float(cand[1])))
+            new_idx = len(points) - 1
+            gx, gy = _grid_pos(cand)
+            grid[gy, gx] = new_idx
+            active.append(new_idx)
+            accepted = True
+            break
+        if not accepted:
+            active.pop(slot)
+    return np.asarray(points, dtype=float)
+
+
+def _bridson_density_cell(x, y, w, cfg, logger):
+    xlim = _axis_lim(x, cfg, "x")
+    ylim = _axis_lim(y, cfg, "y")
+    xscale = _axis_scale(cfg, "x")
+    yscale = _axis_scale(cfg, "y")
+    xp = _project_axis(x, xlim, xscale, "x")
+    yp = _project_axis(y, ylim, yscale, "y")
+    inside = np.isfinite(xp) & np.isfinite(yp) & (xp >= 0.0) & (xp <= 1.0) & (yp >= 0.0) & (yp <= 1.0)
+    if not np.any(inside):
+        raise ValueError("make_density_core bridson found no positive-mass samples inside the domain.")
+    if not np.all(inside):
+        _logger_emit(logger, "warning", f"make_density_core bridson ignored {int((~inside).sum())} samples outside the domain.")
+    bridson = cfg.get("bridson", {})
+    if not isinstance(bridson, Mapping):
+        bridson = {}
+    bin_value = int(bridson.get("bin", bridson.get("bins", cfg.get("bin", 32))))
+    if bin_value <= 0:
+        raise ValueError("make_density_core bridson bin must be positive.")
+    radius = 1.0 / float(bin_value)
+    seed = bridson.get("seed", cfg.get("seed", None))
+    seed = None if seed is None else int(seed)
+    k = int(bridson.get("k", bridson.get("candidates", 30)))
+    support = _bridson_points(radius, seed=seed, k=k)
+    samples = np.column_stack([xp[inside], yp[inside]])
+    try:
+        from scipy.spatial import cKDTree
+
+        _, idx = cKDTree(support).query(samples, k=1)
+    except Exception:
+        diff = samples[:, None, :] - support[None, :, :]
+        idx = np.argmin(np.sum(diff * diff, axis=2), axis=1)
+    mass = np.zeros(int(support.shape[0]), dtype=float)
+    np.add.at(mass, np.asarray(idx, dtype=np.int64), w[inside])
+    xs = _unproject_axis(support[:, 0], xlim, xscale)
+    ys = _unproject_axis(support[:, 1], ylim, yscale)
+    return xs, ys, mass
+
+
+def _kde_core(x, y, w, cfg, logger):
+    xlim = _axis_lim(x, cfg, "x")
+    ylim = _axis_lim(y, cfg, "y")
+    xscale = _axis_scale(cfg, "x")
+    yscale = _axis_scale(cfg, "y")
+    if xscale != "linear" or yscale != "linear":
+        raise ValueError("make_density_core method=kde currently requires linear x/y scales.")
+    nx, ny = _grid_shape(cfg)
+    if x.shape[0] < 3:
+        raise ValueError("make_density_core method=kde requires at least three finite 2D samples.")
+    pts = np.vstack([x, y])
+    if np.linalg.matrix_rank(pts - pts.mean(axis=1, keepdims=True)) < 2:
+        raise ValueError("make_density_core method=kde requires non-collinear 2D samples.")
+    weights = _normalize(w, enabled=True)
+    try:
+        from scipy.stats import gaussian_kde
+    except Exception as e:
+        raise ImportError("make_density_core method=kde requires scipy.stats.gaussian_kde.") from e
+    kde = gaussian_kde(dataset=pts, bw_method=_bw_method(cfg), weights=weights)
+    xmin, xmax = xlim
+    ymin, ymax = ylim
+    dx = (xmax - xmin) / float(nx)
+    dy = (ymax - ymin) / float(ny)
+    x_centers = np.linspace(xmin + 0.5 * dx, xmax - 0.5 * dx, nx)
+    y_centers = np.linspace(ymin + 0.5 * dy, ymax - 0.5 * dy, ny)
+    xg, yg = np.meshgrid(x_centers, y_centers)
+    density = np.asarray(kde(np.vstack([xg.ravel(), yg.ravel()])), dtype=float).reshape(ny, nx)
+    density = np.maximum(np.where(np.isfinite(density), density, 0.0), 0.0)
+    mass = density.ravel() * dx * dy
+    return xg.ravel(), yg.ravel(), mass
+
+
+def _diagnostics(*, method: str, input_count: int, output_count: int, normalize: bool, names: Tuple[str, str, str], total: float) -> str:
+    return (
+        "make_density_core diagnostics:\n"
+        f"\t method \t-> {method}\n"
+        f"\t input_samples \t-> {input_count}\n"
+        f"\t output_support \t-> {output_count}\n"
+        f"\t normalize \t-> {normalize}\n"
+        f"\t output_columns \t-> {names[0]}, {names[1]}, {names[2]}\n"
+        f"\t output_weight_sum \t-> {total:.17g}"
+    )
+
+
+def density_cell(df: pd.DataFrame, cfg: Mapping[str, Any], logger=None) -> pd.DataFrame:
+    """
+    Construct a minimal posterior density-support table from raw samples.
+
+    The output contains exactly three columns: support x, support y, and
+    aggregated posterior mass/weight. Geometry, density division, smoothing, and
+    plotting preparation are intentionally left to later transforms/rendering.
+    """
+    if not isinstance(df, pd.DataFrame):
+        raise TypeError("make_density_core expects a pandas DataFrame.")
+    cfg = cfg if isinstance(cfg, Mapping) else {}
+    method = str(cfg.get("method", "grid")).strip().lower()
+    if method not in {"grid", "bridson", "kde"}:
+        raise ValueError("make_density_core method must be one of: grid, bridson, kde.")
+    memtrace_checkpoint(logger, "make_density_core.before", df)
+    x, y, w, input_count = _prepare_samples(df, cfg, logger)
+    if method == "grid":
+        xs, ys, mass = _grid_density_cell(x, y, w, cfg, logger)
+    elif method == "bridson":
+        xs, ys, mass = _bridson_density_cell(x, y, w, cfg, logger)
+    else:
+        xs, ys, mass = _kde_core(x, y, w, cfg, logger)
+    normalize = bool(cfg.get("normalize", True))
+    mass = _normalize(mass, enabled=normalize)
+    names = (
+        _output_name(cfg, "x", "x"),
+        _output_name(cfg, "y", "y"),
+        _output_name(cfg, "weight", "weight"),
+    )
+    out = pd.DataFrame({names[0]: np.asarray(xs, dtype=float), names[1]: np.asarray(ys, dtype=float), names[2]: mass})
+    if list(out.columns) != list(names):
+        out = out.loc[:, list(names)]
+    msg = _diagnostics(
+        method=method,
+        input_count=int(input_count),
+        output_count=int(out.shape[0]),
+        normalize=normalize,
+        names=names,
+        total=float(np.sum(out[names[2]].to_numpy(dtype=float))),
+    )
+    if bool(cfg.get("diagnostics", True)):
+        _logger_emit(logger, "warning", msg)
+    memtrace_checkpoint(
+        logger,
+        "make_density_core.after",
+        out,
+        extra={"method": method, "rows": int(out.shape[0]), "weight_sum": float(np.sum(out[names[2]]))},
+    )
+    return out