Accept real pandas frames in the Mandela, OTIS and SIU estimators; doctor points at download-bootstrap

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GVZR4QjrTXCv55j4qdo6E2

Author: rootcoder007

README.md

@@ -21,7 +21,7 @@ morie is a Python (and R) package — once Python is present it is `pip install
 
 - **Windows** — install Python from [python.org](https://www.python.org/downloads/) (on the first screen tick **Add python.exe to PATH**), then `pip install morie`. Full walkthrough: [Windows](#recommended--windows) below. Windows has no `curl`/`bash`, so the one-liner does not apply there.
 - **macOS / Linux** — the one-liner below sets up everything. It needs `curl` and `bash`, which macOS has built in and most Linux ships.
-- **Already have Python ≥3.10** — just `pip install morie`.
+- **Already have Python ≥3.10** — just `pip install morie`. The estimators take a pandas DataFrame, a CSV path, or a dict of columns; pandas is optional (morie ships its own frame core).
 
 ### For terminal users — one-liner (Linux / macOS / WSL)
 

src/morie/doctor.py

@@ -111,7 +111,7 @@ def _check_datasets() -> tuple[bool, str]:
 
         db_path = morie_db()
         if not db_path.exists():
-            return False, "morie.db not found -- reinstall morie"
+            return False, "morie.db not found -- it is fetched, not shipped: run `morie download-bootstrap`"
         size_mb = db_path.stat().st_size // (1024 * 1024)
         ds = list_datasets()
         cached = [d for d in ds if d["cached"]]
@@ -353,7 +353,7 @@ def _heal(results: dict[str, Any]) -> bool:
         elif label == "morie version":
             print("  [hint] run `morie update` to upgrade morie itself.")
         elif label == "Built-in datasets":
-            print("  [hint] reinstall to restore the built-in DB: pip install --force-reinstall morie")
+            print("  [hint] the built-in DB is downloaded, not shipped: run `morie download-bootstrap`.")
         elif label == "R (Rscript)":
             print("  [hint] install R from https://www.r-project.org/ (optional -- only the R bridge needs it).")
         else:

src/morie/fn/_frame_core.py

@@ -17,6 +17,7 @@
 import csv as _csv
 import datetime as _dt
 import math as _math
+import os
 
 from . import _array_core as _ac
 
@@ -280,7 +281,7 @@ def _binop(self, other, fn):
         if isinstance(other, Series):
             other = other._data
         # A native array (marr/oarr) is a sequence, not a scalar: without this
-        # it was broadcast whole against each element, so 
+        # it was broadcast whole against each element, so
         # produced a column OF ARRAYS instead of an elementwise sum.
         if not isinstance(other, (str, bytes)) and hasattr(other, "__len__")                 and hasattr(other, "__iter__") and not isinstance(other, dict):
             other = list(other)
@@ -1551,9 +1552,7 @@ def sql_type(values):
             for v in values:
                 if v is None or _isnan(v):
                     continue
-                if isinstance(v, bool):
-                    seen.add("INTEGER")
-                elif isinstance(v, int):
+                if isinstance(v, bool) or isinstance(v, int):
                     seen.add("INTEGER")
                 elif isinstance(v, float):
                     seen.add("REAL")
@@ -2917,7 +2916,6 @@ def close(self):
 
 class ExcelFile:
     def __init__(self, path):
-        import xml.etree.ElementTree as ET
         import zipfile
         self._path = path
         zf = zipfile.ZipFile(path)
@@ -3017,11 +3015,46 @@ def _assert_series_equal(left, right, **kw):
                         DataFrame({"v": list(right)}), **kw)
 
 
-class _TestingNamespace(object):
+class _TestingNamespace:
     """Mirrors `pandas.testing`."""
 
     assert_frame_equal = staticmethod(_assert_frame_equal)
     assert_series_equal = staticmethod(_assert_series_equal)
 
 
 testing = _TestingNamespace()
+
+
+def coerce_frame(data, what: str = "data") -> DataFrame:
+    """Return a native DataFrame for whatever a caller hands an estimator.
+
+    Accepts a native DataFrame (returned as is), a pandas DataFrame or any
+    object exposing ``columns`` and column access, a path to a CSV file,
+    or a sequence of row mappings. The estimators index with native masks;
+    a pandas frame reaching them directly treats those masks as labels and
+    fails with a KeyError (found on the flagship's own bundled sample,
+    2026-09-18), so every public entry point coerces here first.
+    """
+    if isinstance(data, DataFrame):
+        return data
+    if isinstance(data, (str, os.PathLike)):
+        path = os.fspath(data)
+        if not os.path.exists(path):
+            raise FileNotFoundError(f"{what}: no such file {path!r}")
+        return read_csv(path)
+    if hasattr(data, "columns") and hasattr(data, "__getitem__"):
+        cols = list(data.columns)
+        out = {}
+        for c in cols:
+            v = data[c]
+            out[c] = list(v.tolist()) if hasattr(v, "tolist") else list(v)
+        return DataFrame(out)
+    if isinstance(data, dict):
+        return DataFrame(data)
+    if isinstance(data, (list, tuple)):
+        rows = list(data)
+        if rows and all(isinstance(r, dict) for r in rows):
+            cols = list(rows[0].keys())
+            return DataFrame({c: [r.get(c) for r in rows] for c in cols})
+    raise TypeError(f"{what} must be a DataFrame (native or pandas), a CSV path, "
+                    f"a dict of columns or a list of row dicts, got {type(data).__name__}")

src/morie/mrm_mandela_spectrum.py

@@ -104,7 +104,7 @@ def mrm_otis_mandela_spectrum(
         ...                     columns="contact_proxy",
         ...                     values="pct")
     """
-    df = data.copy()
+    df = pd.coerce_frame(data).copy()
     dur = pd.to_numeric(df[duration_col], errors="coerce")
     df["_dur"] = dur
     df["_long"] = (dur > threshold_days).fillna(False)
@@ -158,7 +158,8 @@ def _eligible_mask(proxy: str) -> pd.Series:
                     ids_m = df.loc[elig].groupby(id_col)["_dur"].sum()
                     cum_long = cum > threshold_days
                     # restrict cum_long to ids that had alert-proxy-active placements
-                    cum_long_proxy = cum_long.index.isin(ids_m.index) & cum_long.values
+                    in_proxy = cum_long.index.isin(ids_m.index)
+                    cum_long_proxy = [bool(a) and bool(b) for a, b in zip(in_proxy, list(cum_long.values))]
                     n_d = int(cum.size)
                     if proxy == "none":
                         n_m = int(cum_long.sum())

src/morie/mrm_otis.py

@@ -93,7 +93,7 @@ def mrm_otis_placement_concentration(
         n_individuals, n_placements, mean_per_individual, gini,
         hill_alpha, top_pct_share.
     """
-    df = data.copy()
+    df = pd.coerce_frame(data).copy()
     if gender_col and gender_keep is not None:
         df = df[df[gender_col].isin(list(gender_keep))]
     df["_midpt"] = df[band_col].map(_band_to_midpoint)

src/morie/mrm_siu.py

@@ -71,7 +71,7 @@ def mrm_siu_case_to_decision_km(
     min_n: int = 5,
 ) -> SIUCaseDecisionResult:
     """KM-style time-from-incident-to-Director's-decision summary."""
-    df = data.copy()
+    df = pd.coerce_frame(data).copy()
     inc = pd.to_datetime(df[incident_col], errors="coerce")
     dec = pd.to_datetime(df[decision_col], errors="coerce")
     svc = df[service_col].astype(str)
@@ -116,7 +116,7 @@ def mrm_siu_per_service_rate(
     stratify_col: str | None = None,
 ) -> pd.DataFrame:
     """Per-police-service case counts by year (and optional stratum)."""
-    df = data.copy()
+    df = pd.coerce_frame(data).copy()
     df["_year"] = pd.to_datetime(df[incident_col], errors="coerce").dt.year
     svc = df[service_col].astype(str)
     df = df[df["_year"].notna() & (svc.str.len() > 0) & (svc != "nan")]
@@ -139,7 +139,7 @@ def mrm_siu_outcome_classifier(
     service_col: str = "police_service",
 ) -> pd.DataFrame:
     """Tabulate SIU Director's-decision outcomes by service."""
-    df = data.copy()
+    df = pd.coerce_frame(data).copy()
     if outcome_col not in df.columns:
         for alt in [
             "director_decision",

tests/test_mrm_mandela_spectrum.py

@@ -0,0 +1,32 @@
+import pytest
+
+
+def test_flagship_accepts_pandas_frames_csv_paths_and_rows(tmp_path):
+    """The Mandela spectrum crashed on a real pandas frame (KeyError on a
+    native boolean mask) and on a CSV path; every entry point now coerces."""
+    import os
+
+    import morie
+    from morie.fn import _frame_core as fc
+    from morie.mrm_mandela_spectrum import mrm_otis_mandela_spectrum as f
+    p = os.path.join(os.path.dirname(morie.__file__), "data/samples/otis_b01_sample.csv")
+    native = f(p)
+    assert len(native) > 0
+    rows = fc.read_csv(p)
+    class FakePandas:  # duck-typed like pandas: .columns and column access with .tolist()
+        def __init__(self, frame):
+            self.columns = list(frame.columns); self._f = frame
+        def __getitem__(self, c):
+            class Col:
+                def __init__(self, v): self.v = list(v)
+                def tolist(self): return self.v
+            return Col(self._f[c])
+    fake = f(FakePandas(rows))
+    assert fake.to_dict("records") == native.to_dict("records")
+    pd = pytest.importorskip("pandas")
+    real = f(pd.read_csv(p))
+    assert real.to_dict("records") == native.to_dict("records")
+    with pytest.raises(FileNotFoundError):
+        f(str(tmp_path / "missing.csv"))
+    with pytest.raises(TypeError):
+        f(12345)