anytrellis 0.1.0__py3-none-any.whl

anytrellis/__init__.py

@@ -0,0 +1,13 @@
+"""anytrellis: GPU-accelerated trellis, facet and small-multiple plots for notebooks."""
+
+from importlib.metadata import PackageNotFoundError
+from importlib.metadata import version as _version
+
+from anytrellis.linking import Link, link
+from anytrellis.plot import Trellis
+
+__all__ = ["Link", "Trellis", "link"]
+try:
+    __version__ = _version("anytrellis")
+except PackageNotFoundError:  # run from a source tree that is not installed
+    __version__ = "0.0.0"

anytrellis/codec.py

@@ -0,0 +1,71 @@
+"""Binary encodings for arrays sent to the browser.
+
+Arrays travel as raw little-endian typed buffers described by a small JSON descriptor, so
+the browser can view them without parsing or copying. The descriptor's ``id`` is a hash of
+the bytes and the dtype: identical data always has the same id and is sent once.
+"""
+
+from __future__ import annotations
+
+import hashlib
+from dataclasses import dataclass
+from typing import Any
+
+import numpy as np
+
+DTYPES = {
+    "float64": "<f8",
+    "float32": "<f4",
+    "int32": "<i4",
+    "uint32": "<u4",
+    "uint16": "<u2",
+    "uint8": "u1",
+}
+"""The dtypes the browser reads, and their NumPy spelling."""
+
+
+@dataclass(frozen=True)
+class Encoded:
+    """An encoded array: a JSON-able descriptor plus the raw bytes."""
+
+    descriptor: dict[str, Any]
+    data: bytes
+
+    @property
+    def id(self) -> str:
+        return self.descriptor["id"]
+
+
+def encode(values: Any, dtype: str, *, components: int = 1) -> Encoded:
+    """Encode an array of shape ``(n,)`` or ``(n, components)`` as ``dtype``."""
+    if dtype not in DTYPES:
+        raise ValueError(f"unknown dtype {dtype!r}; known: {list(DTYPES)}")
+    arr = np.asarray(values)
+    if arr.ndim == 2 and arr.shape[1] != components:
+        raise ValueError(f"expected {components} components, got shape {arr.shape}")
+    if arr.ndim == 1 and components > 1:
+        raise ValueError("pass a (n, components) array for interleaved data")
+    data = np.ascontiguousarray(arr.astype(DTYPES[dtype], copy=False)).tobytes()
+    count = int(arr.shape[0]) if arr.ndim else 1
+    digest = hashlib.blake2b(data, digest_size=12).hexdigest()
+    descriptor = {
+        "id": f"{digest}-{dtype}-{components}",
+        "dtype": dtype,
+        "count": count,
+        "components": components,
+        "byteLength": len(data),
+    }
+    return Encoded(descriptor, data)
+
+
+class Buffers:
+    """The buffers of a plot's spec, by id: what the browser may ask for."""
+
+    def __init__(self) -> None:
+        self.data: dict[str, bytes] = {}
+
+    def add(self, values: Any, dtype: str, *, components: int = 1) -> dict[str, Any]:
+        """Encode ``values``, keep the bytes, and return the descriptor for the spec."""
+        enc = encode(values, dtype, components=components)
+        self.data[enc.id] = enc.data
+        return enc.descriptor

anytrellis/data.py

@@ -0,0 +1,251 @@
+"""Reading columns from tables: Polars, PyArrow, pandas, DuckDB (via narwhals) or a dict of
+arrays. Every column is read as one of three kinds: numbers, times or categories."""
+
+from __future__ import annotations
+
+from collections.abc import Mapping
+from dataclasses import dataclass
+from typing import Any, Literal
+
+import numpy as np
+
+Kind = Literal["number", "time", "category"]
+
+MISSING_LABEL = "(missing)"
+"""The category of rows without a value."""
+
+
+class NotFoundError(KeyError):
+    """A missing column: a ``KeyError`` whose message reads as written."""
+
+    def __str__(self) -> str:
+        return str(self.args[0]) if self.args else ""
+
+
+@dataclass(frozen=True)
+class Column:
+    """A column read for plotting.
+
+    ``values`` are float64 for numbers (NaN where missing), float64 milliseconds since
+    1970 for times (wall-clock time in the column's zone; NaN where missing), and str for
+    categories (``MISSING_LABEL`` where missing). ``levels`` is the order of a category
+    column: declared (a Polars Enum, a pandas Categorical) or sorted.
+    """
+
+    name: str
+    kind: Kind
+    values: np.ndarray
+    levels: tuple[str, ...] = ()
+    codes: np.ndarray | None = None
+    """A category column's index of each row's value in ``levels`` (computed if not given)."""
+    log: bool = False
+    """Numbers in log10 units, for a log axis (the data's own values: ``10 ** values``)."""
+
+    def level_codes(self) -> np.ndarray:
+        """Each row's index in ``levels`` (−1: not one of them)."""
+        if self.codes is not None:
+            return self.codes
+        distinct, codes = _labels(self.values)
+        position = {lv: i for i, lv in enumerate(self.levels)}
+        level_of = np.array([position.get(t, -1) for t in distinct], dtype=np.int64)
+        out = level_of[codes] if len(codes) else codes
+        object.__setattr__(self, "codes", out)
+        return out
+
+    def __len__(self) -> int:
+        return len(self.values)
+
+
+class Table:
+    """A table whose columns are read once each."""
+
+    def __init__(self, data: Any) -> None:
+        if isinstance(data, Table):
+            data = data.native
+        self.native = data
+        self._frame = _frame(data)
+        self._cache: dict[str, Column] = {}
+        self.names = _column_names(self._frame)
+        self.n = _row_count(self._frame, self.names)
+
+    def __contains__(self, name: object) -> bool:
+        return name in self.names
+
+    def column(self, name: str) -> Column:
+        if name not in self._cache:
+            if name not in self.names:
+                raise NotFoundError(f"column {name!r} not found; available: {self.names}")
+            self._cache[name] = _read(self._frame, name)
+        return self._cache[name]
+
+    def log_scaled(self, names: set[str]) -> Table:
+        """This table with the columns ``names`` in log10 units, for a log axis: values of 0
+        or less are missing. Raw values (tooltips) and rows (``take``) are as they were."""
+        import copy
+
+        out = copy.copy(self)
+        out._cache = dict(self._cache)
+        for name in sorted(names):
+            col = self.column(name)
+            if col.kind != "number":
+                raise ValueError(f"a log scale needs numbers; {name!r} holds {col.kind}s")
+            with np.errstate(divide="ignore", invalid="ignore"):
+                values = np.where(col.values > 0, np.log10(col.values), np.nan)
+            out._cache[name] = Column(name, "number", values, log=True)
+        return out
+
+    def take(self, rows: np.ndarray) -> Any:
+        """These rows (indices), as the same kind of table as the data given."""
+        if isinstance(self._frame, Mapping):
+            return {k: np.asarray(v)[rows] for k, v in self._frame.items()}
+        import narwhals as nw
+
+        return nw.to_native(self._frame[rows])
+
+    def raw(self, name: str) -> np.ndarray:
+        """A column's values as they are (for tooltips)."""
+        if isinstance(self._frame, Mapping):
+            return np.asarray(self._frame[name]).ravel()
+        return self._frame[name].to_numpy()
+
+
+def _frame(data: Any) -> Any:
+    if isinstance(data, Mapping):
+        return data
+    import narwhals as nw
+
+    try:
+        out = nw.from_native(data)
+    except TypeError:
+        raise TypeError(
+            f"unsupported data of type {type(data).__name__}; pass a Polars, PyArrow, pandas "
+            "or DuckDB table, or a dict of arrays"
+        ) from None
+    if isinstance(out, nw.LazyFrame):
+        out = out.collect()
+    return out
+
+
+def _column_names(frame: Any) -> list[str]:
+    if isinstance(frame, Mapping):
+        return [str(k) for k in frame]
+    return list(frame.columns)
+
+
+def _row_count(frame: Any, names: list[str]) -> int:
+    if isinstance(frame, Mapping):
+        lengths = {len(np.asarray(frame[n]).ravel()) for n in names}
+        if len(lengths) > 1:
+            raise ValueError(f"the columns have different lengths: {sorted(lengths)}")
+        return lengths.pop() if lengths else 0
+    return len(frame)
+
+
+def _read(frame: Any, name: str) -> Column:
+    if isinstance(frame, Mapping):
+        return _from_array(name, frame[name])
+    import narwhals as nw
+
+    series = frame[name]
+    dtype = series.dtype
+    if isinstance(dtype, nw.Datetime):
+        if dtype.time_zone is not None:
+            series = series.dt.replace_time_zone(None)  # wall-clock time in its zone
+        ms = series.dt.timestamp("ms").cast(nw.Float64)
+        missing = series.is_null().to_numpy()
+        values = np.array(ms.to_numpy(), dtype=np.float64)  # a copy: it may be read-only
+        values[np.asarray(missing, dtype=bool)] = np.nan
+        return Column(name, "time", values)
+    if isinstance(dtype, nw.Date):
+        return _read(frame.with_columns(nw.col(name).cast(nw.Datetime("ms"))), name)
+    if dtype.is_numeric() or dtype == nw.Boolean:
+        values = series.cast(nw.Float64).to_numpy()
+        return Column(name, "number", np.asarray(values, dtype=np.float64))
+    levels: list[str] | None = None
+    if isinstance(dtype, nw.Enum):
+        levels = [str(c) for c in dtype.categories]
+    elif dtype == nw.Categorical:
+        native = nw.to_native(frame)
+        if hasattr(native, "iloc"):  # pandas keeps its declared order
+            levels = [str(c) for c in native[name].cat.categories]
+    return _categories(name, series.to_numpy(), levels)
+
+
+def _from_array(name: str, values: Any) -> Column:
+    cat = getattr(values, "cat", None)
+    declared = getattr(cat, "categories", None) if cat is not None else None
+    if declared is None and hasattr(values, "categories") and hasattr(values, "codes"):
+        declared = values.categories
+    arr = np.asarray(values).ravel()
+    if declared is not None:
+        return _categories(name, arr, [str(c) for c in declared])
+    if arr.dtype.kind == "M":
+        ms = arr.astype("datetime64[ms]")
+        out = ms.astype(np.int64).astype(np.float64)
+        out[np.isnat(ms)] = np.nan
+        return Column(name, "time", out)
+    if arr.dtype.kind in "biuf":
+        return Column(name, "number", arr.astype(np.float64))
+    if arr.dtype.kind == "O":
+        present = [v for v in arr if v is not None]
+        import datetime as dt
+
+        if present and all(isinstance(v, dt.datetime | dt.date) for v in present):
+            return _from_array(name, np.array(arr, dtype="datetime64[ms]"))
+        if present and all(isinstance(v, int | float | np.number) for v in present):
+            return Column(name, "number", np.array(arr, dtype=np.float64))
+    return _categories(name, arr, None)
+
+
+def _categories(name: str, values: np.ndarray, levels: list[str] | None) -> Column:
+    distinct, codes = _labels(values)
+    seen = set(distinct[np.unique(codes)]) if len(codes) else set()
+    if levels is None:
+        present = sorted((t for t in seen if t != MISSING_LABEL), key=_natural)
+    else:
+        present = [lv for lv in levels if lv in seen]
+    if MISSING_LABEL in seen:
+        present.append(MISSING_LABEL)
+    position = {lv: i for i, lv in enumerate(present)}
+    level_of = np.array([position.get(t, -1) for t in distinct], dtype=np.int64)
+    return Column(name, "category", distinct[codes], tuple(present), level_of[codes])
+
+
+def _labels(values: np.ndarray) -> tuple[np.ndarray, np.ndarray]:
+    """The distinct labels, and each value's index among them; each distinct value is
+    labelled once, so a million rows of a few categories cost a lookup each."""
+    arr = np.asarray(values).ravel()
+    try:
+        if arr.dtype.kind in "US":
+            uniq, inverse = np.unique(arr, return_inverse=True)
+            codes = inverse.ravel().astype(np.int64)
+        else:
+            index: dict[Any, int] = {}
+            codes = np.fromiter((index.setdefault(v, len(index)) for v in arr),
+                                dtype=np.int64, count=arr.size)  # fmt: skip
+            uniq = list(index)
+    except TypeError:  # unhashable values: one by one
+        uniq, codes = [_label(v) for v in arr], np.arange(arr.size, dtype=np.int64)
+    labels = [_label(u) for u in uniq]
+    # values with one label (1 and "1", or NaN objects): one index
+    first: dict[str, int] = {}
+    remap = np.array([first.setdefault(t, len(first)) for t in labels], dtype=np.int64)
+    distinct = np.array(list(first) or [""], dtype=object)
+    return distinct, remap[codes] if len(remap) else codes
+
+
+def _label(value: Any) -> str:
+    if value is None or (isinstance(value, float) and value != value):
+        return MISSING_LABEL
+    if isinstance(value, bytes):
+        return value.decode("utf-8", "replace")
+    if isinstance(value, np.generic):
+        value = value.item()
+    return str(value)
+
+
+def _natural(text: str) -> tuple[Any, ...]:
+    """Sort key: numbers in text by value ("site 2" before "site 10")."""
+    import re
+
+    return tuple(int(p) if p.isdigit() else p.lower() for p in re.split(r"(\d+)", text))

anytrellis/datasets/SOURCES.md

@@ -0,0 +1,17 @@
+# Sources of the bundled datasets
+
+The datasets in this folder (`anytrellis.datasets.load`) are not covered by anytrellis's MIT
+licence: each keeps the terms of its source.
+
+| dataset | source | licence |
+|---|---|---|
+| anscombe | F. J. Anscombe (1973), *Graphs in Statistical Analysis*, The American Statistician 27(1):17-21; as in [vega-datasets](https://github.com/vega/vega-datasets) 2.8.0 | published numbers, also in R's `datasets` package; no licence stated |
+| barley | Minnesota Agricultural Experiment Station; Immer, Hayes and Powers (1934); republished in Wiebe et al. (1940), *Yields of Barley Varieties in the United States and Canada, 1932-36*, USDA Technical Bulletin 735; as in vega-datasets 2.8.0 | a U.S. Government publication; also in R's `lattice` package |
+| cars | Quinlan, R. (1993). *Auto MPG* [Dataset]. UCI Machine Learning Repository. <https://doi.org/10.24432/C5859H> | [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/) |
+| gapminder | Gapminder Foundation: life expectancy, population and fertility, 1955-2005; as in vega-datasets 2.8.0 | [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/): based on free material from [GAPMINDER.ORG](https://www.gapminder.org/), CC-BY LICENSE |
+| penguins | Gorman, Williams and Fraser (2014), Palmer Station Antarctica LTER; [palmerpenguins](https://github.com/allisonhorst/palmerpenguins) by Horst, Hill and Gorman; as in vega-datasets 2.8.0 | [CC0 1.0](https://creativecommons.org/publicdomain/zero/1.0/) |
+| seattle_weather | NOAA National Climatic Data Center, daily records for Seattle, 2012-2015; as in vega-datasets 2.8.0 | a U.S. Government work: public domain |
+
+Changes made here: column names in snake_case; for `gapminder`, readable region names; for
+`cars`, the 406 cars of UCI's `auto-mpg.data-original`, with the model year as 1970-1982,
+the origin as USA, Europe or Japan, and missing values left empty.

anytrellis/datasets/__init__.py

@@ -0,0 +1,55 @@
+"""Small, well-known datasets for the examples: ``load("barley")``.
+
+=================  =====  ==============================================================
+name               rows   columns
+=================  =====  ==============================================================
+anscombe           44     dataset, x, y (Anscombe's quartet)
+barley             120    site, variety, year, yield (Immer et al. 1934; Cleveland 1993)
+cars               406    name, mpg, cylinders, displacement, horsepower, weight,
+                          acceleration, year, origin (the classic auto-mpg data)
+gapminder          693    country, region, year, pop, life_expect, fertility (1955-2005)
+penguins           344    species, island, bill_length_mm, bill_depth_mm,
+                          flipper_length_mm, body_mass_g, sex (Palmer penguins)
+seattle_weather    1461   date, precipitation, temp_max, temp_min, wind, weather
+=================  =====  ==============================================================
+
+``cars`` is the UCI Machine Learning Repository's Auto MPG (CC BY 4.0); the others are
+taken from vega-datasets 2.8.0, with snake_case column names and readable categories
+(``gapminder.region``). Each keeps its source's licence: see ``SOURCES.md`` here. The
+gapminder data is based on free material from GAPMINDER.ORG, CC-BY LICENSE.
+"""
+
+from __future__ import annotations
+
+from importlib import resources
+from typing import Any, Literal
+
+NAMES = ("anscombe", "barley", "cars", "gapminder", "penguins", "seattle_weather")
+_DATES = {"seattle_weather": "date"}
+
+Backend = Literal["polars", "pandas", "pyarrow"]
+
+
+def load(name: str, backend: Backend | None = None) -> Any:
+    """Dataset ``name`` as a table of ``backend`` (default: Polars, else pandas, else
+    PyArrow, whichever is installed)."""
+    import narwhals as nw
+
+    if name not in NAMES:
+        raise ValueError(f"no dataset {name!r}; there are {', '.join(NAMES)}")
+    path = resources.files(__name__) / f"{name}.csv"
+    with resources.as_file(path) as file:
+        frame = nw.read_csv(str(file), backend=backend or _backend())
+    column = _DATES.get(name)
+    if column is not None and frame.schema[column] == nw.String:  # PyArrow parses dates itself
+        frame = frame.with_columns(nw.col(column).str.to_datetime("%Y-%m-%d").cast(nw.Date))
+    return frame.to_native()
+
+
+def _backend() -> Backend:
+    import importlib.util
+
+    for backend in ("polars", "pandas", "pyarrow"):
+        if importlib.util.find_spec(backend) is not None:
+            return backend
+    raise ModuleNotFoundError("load() needs Polars, pandas or PyArrow: pip install polars")

anytrellis/datasets/anscombe.csv

@@ -0,0 +1,45 @@
+dataset,x,y
+I,10.0,8.04
+I,8.0,6.95
+I,13.0,7.58
+I,9.0,8.81
+I,11.0,8.33
+I,14.0,9.96
+I,6.0,7.24
+I,4.0,4.26
+I,12.0,10.84
+I,7.0,4.81
+I,5.0,5.68
+II,10.0,9.14
+II,8.0,8.14
+II,13.0,8.74
+II,9.0,8.77
+II,11.0,9.26
+II,14.0,8.1
+II,6.0,6.13
+II,4.0,3.1
+II,12.0,9.13
+II,7.0,7.26
+II,5.0,4.74
+III,10.0,7.46
+III,8.0,6.77
+III,13.0,12.74
+III,9.0,7.11
+III,11.0,7.81
+III,14.0,8.84
+III,6.0,6.08
+III,4.0,5.39
+III,12.0,8.15
+III,7.0,6.42
+III,5.0,5.73
+IV,8.0,6.58
+IV,8.0,5.76
+IV,8.0,7.71
+IV,8.0,8.84
+IV,8.0,8.47
+IV,8.0,7.04
+IV,8.0,5.25
+IV,19.0,12.5
+IV,8.0,5.56
+IV,8.0,7.91
+IV,8.0,6.89

anytrellis/datasets/barley.csv

@@ -0,0 +1,121 @@
+site,variety,year,yield
+University Farm,Manchuria,1931,27.0
+Waseca,Manchuria,1931,48.86667
+Morris,Manchuria,1931,27.43334
+Crookston,Manchuria,1931,39.93333
+Grand Rapids,Manchuria,1931,32.96667
+Duluth,Manchuria,1931,28.96667
+University Farm,Glabron,1931,43.06666
+Waseca,Glabron,1931,55.2
+Morris,Glabron,1931,28.76667
+Crookston,Glabron,1931,38.13333
+Grand Rapids,Glabron,1931,29.13333
+Duluth,Glabron,1931,29.66667
+University Farm,Svansota,1931,35.13333
+Waseca,Svansota,1931,47.33333
+Morris,Svansota,1931,25.76667
+Crookston,Svansota,1931,40.46667
+Grand Rapids,Svansota,1931,29.66667
+Duluth,Svansota,1931,25.7
+University Farm,Velvet,1931,39.9
+Waseca,Velvet,1931,50.23333
+Morris,Velvet,1931,26.13333
+Crookston,Velvet,1931,41.33333
+Grand Rapids,Velvet,1931,23.03333
+Duluth,Velvet,1931,26.3
+University Farm,Trebi,1931,36.56666
+Waseca,Trebi,1931,63.8333
+Morris,Trebi,1931,43.76667
+Crookston,Trebi,1931,46.93333
+Grand Rapids,Trebi,1931,29.76667
+Duluth,Trebi,1931,33.93333
+University Farm,No. 457,1931,43.26667
+Waseca,No. 457,1931,58.1
+Morris,No. 457,1931,28.7
+Crookston,No. 457,1931,45.66667
+Grand Rapids,No. 457,1931,32.16667
+Duluth,No. 457,1931,33.6
+University Farm,No. 462,1931,36.6
+Waseca,No. 462,1931,65.7667
+Morris,No. 462,1931,30.36667
+Crookston,No. 462,1931,48.56666
+Grand Rapids,No. 462,1931,24.93334
+Duluth,No. 462,1931,28.1
+University Farm,Peatland,1931,32.76667
+Waseca,Peatland,1931,48.56666
+Morris,Peatland,1931,29.86667
+Crookston,Peatland,1931,41.6
+Grand Rapids,Peatland,1931,34.7
+Duluth,Peatland,1931,32.0
+University Farm,No. 475,1931,24.66667
+Waseca,No. 475,1931,46.76667
+Morris,No. 475,1931,22.6
+Crookston,No. 475,1931,44.1
+Grand Rapids,No. 475,1931,19.7
+Duluth,No. 475,1931,33.06666
+University Farm,Wisconsin No. 38,1931,39.3
+Waseca,Wisconsin No. 38,1931,58.8
+Morris,Wisconsin No. 38,1931,29.46667
+Crookston,Wisconsin No. 38,1931,49.86667
+Grand Rapids,Wisconsin No. 38,1931,34.46667
+Duluth,Wisconsin No. 38,1931,31.6
+University Farm,Manchuria,1932,26.9
+Waseca,Manchuria,1932,33.46667
+Morris,Manchuria,1932,34.36666
+Crookston,Manchuria,1932,32.96667
+Grand Rapids,Manchuria,1932,22.13333
+Duluth,Manchuria,1932,22.56667
+University Farm,Glabron,1932,36.8
+Waseca,Glabron,1932,37.73333
+Morris,Glabron,1932,35.13333
+Crookston,Glabron,1932,26.16667
+Grand Rapids,Glabron,1932,14.43333
+Duluth,Glabron,1932,25.86667
+University Farm,Svansota,1932,27.43334
+Waseca,Svansota,1932,38.5
+Morris,Svansota,1932,35.03333
+Crookston,Svansota,1932,20.63333
+Grand Rapids,Svansota,1932,16.63333
+Duluth,Svansota,1932,22.23333
+University Farm,Velvet,1932,26.8
+Waseca,Velvet,1932,37.4
+Morris,Velvet,1932,38.83333
+Crookston,Velvet,1932,32.06666
+Grand Rapids,Velvet,1932,32.23333
+Duluth,Velvet,1932,22.46667
+University Farm,Trebi,1932,29.06667
+Waseca,Trebi,1932,49.2333
+Morris,Trebi,1932,46.63333
+Crookston,Trebi,1932,41.83333
+Grand Rapids,Trebi,1932,20.63333
+Duluth,Trebi,1932,30.6
+University Farm,No. 457,1932,26.43334
+Waseca,No. 457,1932,42.2
+Morris,No. 457,1932,43.53334
+Crookston,No. 457,1932,34.33333
+Grand Rapids,No. 457,1932,19.46667
+Duluth,No. 457,1932,22.7
+University Farm,No. 462,1932,25.56667
+Waseca,No. 462,1932,44.7
+Morris,No. 462,1932,47.0
+Crookston,No. 462,1932,30.53333
+Grand Rapids,No. 462,1932,19.9
+Duluth,No. 462,1932,22.5
+University Farm,Peatland,1932,28.06667
+Waseca,Peatland,1932,36.03333
+Morris,Peatland,1932,43.2
+Crookston,Peatland,1932,25.23333
+Grand Rapids,Peatland,1932,26.76667
+Duluth,Peatland,1932,31.36667
+University Farm,No. 475,1932,30.0
+Waseca,No. 475,1932,41.26667
+Morris,No. 475,1932,44.23333
+Crookston,No. 475,1932,32.13333
+Grand Rapids,No. 475,1932,15.23333
+Duluth,No. 475,1932,27.36667
+University Farm,Wisconsin No. 38,1932,38.0
+Waseca,Wisconsin No. 38,1932,58.16667
+Morris,Wisconsin No. 38,1932,47.16667
+Crookston,Wisconsin No. 38,1932,35.9
+Grand Rapids,Wisconsin No. 38,1932,20.66667
+Duluth,Wisconsin No. 38,1932,29.33333