Merge pull request #18 from espdev/refactoring-internals

Refactoring internals

Author: espdev

CHANGELOG.md

@@ -1,5 +1,13 @@
 # Changelog
 
+## v0.11.0
+
+* Internal re-design `SplinePPForm` and `NdGridSplinePPForm` classes [#17](https://github.com/espdev/csaps/issues/17):
+    - Remove `shape` and `axis` properties and reshaping data in these classes
+    - `NdGridSplinePPForm` coefficients array for 1D grid now is 1-d instead of 2-d
+* Refactoring the code and decrease memory consumption
+* Add `overload` type-hints for `csaps` function signatures
+
 ## v0.10.1
 
 * Fix call of `numpy.pad` function for numpy <1.17 [#15](https://github.com/espdev/csaps/issues/15)

README.md

@@ -14,7 +14,7 @@
 **csaps** is a Python package for univariate, multivariate and n-dimensional grid data approximation using cubic smoothing splines.
 The package can be useful in practical engineering tasks for data approximation and smoothing.
 
-## Installation
+## Installing
 
 Use pip for installing:
 
@@ -89,20 +89,22 @@ plt.show()
 
 ## Documentation
 
-More examples of usage and the full documentation can be found at ReadTheDocs.
-
-https://csaps.readthedocs.io
+More examples of usage and the full documentation can be found at https://csaps.readthedocs.io.
 
 ## Testing
 
 pytest, tox and Travis CI are used for testing. Please see [tests](tests).
 
-## Algorithms and implementations
+## Algorithm and Implementation
 
-**csaps** package is a Python modified port of MATLAB [CSAPS](https://www.mathworks.com/help/curvefit/csaps.html) function that is an implementation of 
+**csaps** Python package is inspired by MATLAB [CSAPS](https://www.mathworks.com/help/curvefit/csaps.html) function that is an implementation of 
 Fortran routine SMOOTH from [PGS](http://pages.cs.wisc.edu/~deboor/pgs/) (originally written by Carl de Boor).
 
-[csaps-cpp](https://github.com/espdev/csaps-cpp) C++11 Eigen based implementation of the algorithm.
+Also the algothithm implementation in other languages:
+
+* [csaps-rs](https://github.com/espdev/csaps-rs) Rust ndarray/sprs based implementation
+* [csaps-cpp](https://github.com/espdev/csaps-cpp) C++11 Eigen based implementation (incomplete)
+
 
 ## References
 

csaps/__init__.py

@@ -23,7 +23,6 @@
 )
 from csaps._types import (
     UnivariateDataType,
-    UnivariateVectorizedDataType,
     MultivariateDataType,
     NdGridDataType,
 )
@@ -46,7 +45,6 @@
 
     # Type-hints
     'UnivariateDataType',
-    'UnivariateVectorizedDataType',
     'MultivariateDataType',
     'NdGridDataType',
 ]

csaps/_shortcut.py

@@ -6,50 +6,93 @@
 """
 
 from collections import abc as c_abc
-from typing import Optional, Union, Sequence, NamedTuple
-
-import numpy as np
+from typing import Optional, Union, Sequence, NamedTuple, overload
 
 from ._base import ISmoothingSpline
 from ._sspumv import CubicSmoothingSpline
 from ._sspndg import ndgrid_prepare_data_sites, NdGridCubicSmoothingSpline
-from ._types import (
-    UnivariateDataType,
-    UnivariateVectorizedDataType,
-    NdGridDataType,
-)
-
-_XDataType = Union[UnivariateDataType, NdGridDataType]
-_YDataType = Union[UnivariateVectorizedDataType, np.ndarray]
-_XiDataType = Optional[Union[UnivariateDataType, NdGridDataType]]
-_WeightsDataType = Optional[Union[UnivariateDataType, NdGridDataType]]
-_SmoothDataType = Optional[Union[float, Sequence[Optional[float]]]]
+from ._types import UnivariateDataType, MultivariateDataType, NdGridDataType
 
 
 class AutoSmoothingResult(NamedTuple):
     """The result for auto smoothing for `csaps` function"""
 
-    values: _YDataType
+    values: MultivariateDataType
     """Smoothed data values"""
 
-    smooth: _SmoothDataType
+    smooth: Union[float, Sequence[Optional[float]]]
     """The calculated smoothing parameter"""
 
 
-_ReturnType = Union[
-    _YDataType,
-    AutoSmoothingResult,
-    ISmoothingSpline,
-]
+# **************************************
+# csaps signatures
+#
+@overload
+def csaps(xdata: UnivariateDataType,
+          ydata: MultivariateDataType,
+          *,
+          weights: Optional[UnivariateDataType] = None,
+          smooth: Optional[float] = None,
+          axis: Optional[int] = None) -> ISmoothingSpline: ...
+
+
+@overload
+def csaps(xdata: UnivariateDataType,
+          ydata: MultivariateDataType,
+          xidata: UnivariateDataType,
+          *,
+          weights: Optional[UnivariateDataType] = None,
+          axis: Optional[int] = None) -> AutoSmoothingResult: ...
+
+
+@overload
+def csaps(xdata: UnivariateDataType,
+          ydata: MultivariateDataType,
+          xidata: UnivariateDataType,
+          *,
+          smooth: float,
+          weights: Optional[UnivariateDataType] = None,
+          axis: Optional[int] = None) -> MultivariateDataType: ...
+
+
+@overload
+def csaps(xdata: NdGridDataType,
+          ydata: MultivariateDataType,
+          *,
+          weights: Optional[NdGridDataType] = None,
+          smooth: Optional[Sequence[float]] = None,
+          axis: Optional[int] = None) -> ISmoothingSpline: ...
+
+
+@overload
+def csaps(xdata: NdGridDataType,
+          ydata: MultivariateDataType,
+          xidata: NdGridDataType,
+          *,
+          weights: Optional[NdGridDataType] = None,
+          axis: Optional[int] = None) -> AutoSmoothingResult: ...
+
+
+@overload
+def csaps(xdata: NdGridDataType,
+          ydata: MultivariateDataType,
+          xidata: NdGridDataType,
+          *,
+          smooth: Sequence[float],
+          weights: Optional[NdGridDataType] = None,
+          axis: Optional[int] = None) -> MultivariateDataType: ...
+#
+# csaps signatures
+# **************************************
 
 
-def csaps(xdata: _XDataType,
-          ydata: _YDataType,
-          xidata: _XiDataType = None,
+def csaps(xdata: Union[UnivariateDataType, NdGridDataType],
+          ydata: MultivariateDataType,
+          xidata: Optional[Union[UnivariateDataType, NdGridDataType]] = None,
           *,
-          weights: _WeightsDataType = None,
-          smooth: _SmoothDataType = None,
-          axis: Optional[int] = None) -> _ReturnType:
+          weights: Optional[Union[UnivariateDataType, NdGridDataType]] = None,
+          smooth: Optional[Union[float, Sequence[float]]] = None,
+          axis: Optional[int] = None) -> Union[MultivariateDataType, ISmoothingSpline, AutoSmoothingResult]:
     """Smooths the univariate/multivariate/gridded data or computes the corresponding splines
 
     This function might be used as the main API for smoothing any data.

csaps/_sspndg.py

@@ -37,82 +37,64 @@ class NdGridSplinePPForm(SplinePPFormBase[ty.Sequence[np.ndarray], ty.Tuple[int,
 
     Parameters
     ----------
-    breaks : np.ndarray
-        Breaks values 1-D array
+    breaks : Sequence[np.ndarray]
+        The sequence of the breaks 1-D arrays for each dimension
+
     coeffs : np.ndarray
-        Spline coefficients N-D array
+        Tensor-product spline coefficients N-D array
+
     """
 
     def __init__(self, breaks: ty.Sequence[np.ndarray], coeffs: np.ndarray) -> None:
         self._breaks = breaks
         self._coeffs = coeffs
         self._pieces = tuple(x.size - 1 for x in breaks)
+        self._order = tuple(s // p for s, p in zip(coeffs.shape, self._pieces))
         self._ndim = len(breaks)
 
-        if self._ndim > 1:
-            self._order = tuple(s // p for s, p in zip(coeffs.shape, self._pieces))
-        else:
-            # the corner case for univariate spline that is represented as 1d-grid
-            self._order = (coeffs.shape[1] // self._pieces[0], )
-
     @property
     def breaks(self) -> ty.Sequence[np.ndarray]:
+        """Returns the sequence of the data breaks for each dimension"""
         return self._breaks
 
     @property
     def coeffs(self) -> np.ndarray:
+        """Returns n-d array of tensor-product n-d grid spline coefficients"""
         return self._coeffs
 
     @property
     def order(self) -> ty.Tuple[int, ...]:
+        """Returns the tuple of the spline orders for each dimension"""
         return self._order
 
     @property
     def pieces(self) -> ty.Tuple[int, ...]:
+        """Returns the tuple of the spline pieces for each dimension"""
         return self._pieces
 
     @property
     def ndim(self) -> int:
+        """Returns the dimensionality of n-d grid data"""
         return self._ndim
 
-    @property
-    def shape(self) -> ty.Tuple[int, ...]:
-        """Returns the original data shape
-        """
-        return tuple(x.size for x in self.breaks)
-
     def evaluate(self, xi: ty.Sequence[np.ndarray]) -> np.ndarray:
+        """Evaluates the spline for given data point(s) on the n-d grid"""
         shape = tuple(x.size for x in xi)
 
         coeffs = self.coeffs
-        coeffs_shape = list(coeffs.shape)
+        coeffs_shape = coeffs.shape
 
-        d = self.ndim - 1
-        permuted_axes = (d, *range(d))
+        ndim_m1 = self.ndim - 1
+        permuted_axes = (ndim_m1, *range(ndim_m1))
 
         for i in reversed(range(self.ndim)):
-            xii = xi[i]
-            ndim = int(np.prod(coeffs_shape[:d]))
-
-            if self.ndim > 2:
-                coeffs = coeffs.reshape((ndim, self.pieces[i] * self.order[i]))
-
-            spp = SplinePPForm(
-                breaks=self.breaks[i],
-                coeffs=coeffs,
-                pieces=self.pieces[i],
-                order=self.order[i],
-                shape=(ndim, xii.size)
-            )
+            umv_ndim = int(np.prod(coeffs_shape[:ndim_m1]))
+            coeffs = coeffs.reshape((umv_ndim, self.pieces[i] * self.order[i]))
 
-            coeffs = spp.evaluate(xii)
+            coeffs = SplinePPForm(breaks=self.breaks[i], coeffs=coeffs).evaluate(xi[i])
 
-            if self.ndim > 2:
-                coeffs = coeffs.reshape((*coeffs_shape[:d], shape[i]))
-
-            if self.ndim > 1:
-                coeffs = coeffs.transpose(permuted_axes)
-                coeffs_shape = list(coeffs.shape)
+            coeffs = coeffs.reshape((*coeffs_shape[:ndim_m1], shape[i])).transpose(permuted_axes)
+            coeffs_shape = coeffs.shape
 
         return coeffs.reshape(shape)
 
@@ -261,4 +243,6 @@ def _make_spline(self, smooth: ty.List[ty.Optional[float]]) -> ty.Tuple[NdGridSp
                 coeffs = coeffs.transpose(permute_axes)
                 coeffs_shape = list(coeffs.shape)
 
+        coeffs = coeffs.squeeze()
+
         return NdGridSplinePPForm(breaks=self._xdata, coeffs=coeffs), tuple(reversed(smooths))