Release/0.7.0 (#86)

* MAINT: Drop Python 3.6 (close #68) (#69)

  * Python 3.6 is no longer supported. Please update to >=3.7.

* MAINT: Add typing (#77)

* MAINT: Update README.md (close #81) (#82) (#83)

* MAINT: Rename mode 'full' to 'same' (close #55) (#84)

* CHORE: Update pytest-cov requirement from ^2.8.1 to ^3.0.0 (#78)

* CHORE: Update Makefile (#88)

Co-authored-by: GitHub Actions <action@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: GitHub Actions <action@github.com>

.github/workflows/ci.yml

@@ -26,7 +26,7 @@ jobs:
 
     strategy:
       matrix:
-        python-version: ['3.6', '3.7', '3.8', '3.9']
+        python-version: ['3.7', '3.8', '3.9']
 
     steps:
 
@@ -58,10 +58,12 @@ jobs:
     - uses: actions/checkout@v2
 
     - uses: actions/setup-python@v2
+      with:
+        python-version: '3.9'
 
     - run: pip install poetry && poetry install --extras torch
 
-    - run: make typecheck
+    - run: make type
 
   lint:
 

.github/workflows/doc.yml

@@ -26,6 +26,8 @@ jobs:
     - uses: actions/checkout@v2
 
     - uses: actions/setup-python@v2
+      with:
+        python-version: '3.9'
 
     - run: pip install poetry && poetry install --extras torch
 

Makefile

@@ -1,7 +1,8 @@
 PROJECT_NAME := fracdiff
+RUN := poetry run
 
 .PHONY: check
-check: test lint
+check: test lint type
 
 .PHONY: install
 install:
@@ -12,40 +13,41 @@ test: test-doctest test-pytest
 
 .PHONY: test-doctest
 test-doctest:
-	@poetry run pytest --doctest-modules $(PROJECT_NAME)
+	$(RUN) pytest --doctest-modules $(PROJECT_NAME)
 
 .PHONY: test-pytest
 test-pytest:
-	@poetry run pytest --doctest-modules --cov=$(PROJECT_NAME) tests
+	$(RUN) pytest --doctest-modules --cov=$(PROJECT_NAME) tests
 
 .PHONY: lint
 lint: lint-black lint-isort
 
 .PHONY: lint-black
 lint-black:
-	@poetry run python3 -m black --check .
+	$(RUN) python3 -m black --check --quiet .
 
 .PHONY: lint-isort
 lint-isort:
-	@poetry run python3 -m isort --check --force-single-line-imports .
+	$(RUN) run python3 -m isort --check --force-single-line-imports --quiet .
 
 .PHONY: format
 format: format-black format-isort
 
 .PHONY: format-black
 format-black:
-	@poetry run python3 -m black --quiet .
+	$(RUN) python3 -m black --quiet .
 
 .PHONY: format-isort
 format-isort:
-	@poetry run python3 -m isort --force-single-line-imports --quiet .
+	$(RUN) python3 -m isort --force-single-line-imports --quiet .
 
 .PHONY: doc
 doc:
 	@cd docs && make html
 
-.PHONY: typecheck
-	@poetry run mypy $(PROJECT_NAME)
+.PHONY: type
+type:
+	$(RUN) mypy $(PROJECT_NAME)
 
 .PHONY: publish
 publish:

README.md

@@ -23,18 +23,24 @@ See [M. L. Prado, "Advances in Financial Machine Learning"][prado].
 ## Installation
 
 ```sh
-$ pip install fracdiff
+pip install fracdiff
 ```
 
 ## Features
 
 ### Functionalities
 
-- [`fdiff`](https://fracdiff.github.io/fracdiff/generated/fracdiff.fdiff.html): A function that extends [`numpy.diff`](https://numpy.org/doc/stable/reference/generated/numpy.diff.html) to fractional differentiation.
-- [`Fracdiff`](https://fracdiff.github.io/fracdiff/generated/fracdiff.Fracdiff.html): A scikit-learn [transformer](https://scikit-learn.org/stable/modules/generated/sklearn.base.TransformerMixin.html) to compute fractional differentiation.
-- [`FracdiffStat`](https://fracdiff.github.io/fracdiff/generated/fracdiff.FracdiffStat.html): `Fracdiff` plus automatic choice of differentiation order that makes time-series stationary.
-- [`torch.fdiff`](https://fracdiff.github.io/fracdiff/generated/fracdiff.torch.fdiff.html): A functional that extends [`torch.diff`](https://pytorch.org/docs/stable/generated/torch.diff.html) to fractional differentiation.
-- [`torch.Fracdiff`](https://fracdiff.github.io/fracdiff/generated/fracdiff.torch.Fracdiff.html): A module that computes fractional differentiation.
+- [`fdiff`][doc-fdiff]: A function that extends [`numpy.diff`](https://numpy.org/doc/stable/reference/generated/numpy.diff.html) to fractional differentiation.
+- [`sklearn.Fracdiff`][doc-sklearn.Fracdiff]: A scikit-learn [transformer](https://scikit-learn.org/stable/modules/generated/sklearn.base.TransformerMixin.html) to compute fractional differentiation.
+- [`sklearn.FracdiffStat`][doc-sklearn.FracdiffStat]: `Fracdiff` plus automatic choice of differentiation order that makes time-series stationary.
+- [`torch.fdiff`][doc-torch.fdiff]: A functional that extends [`torch.diff`](https://pytorch.org/docs/stable/generated/torch.diff.html) to fractional differentiation.
+- [`torch.Fracdiff`][doc-torch.Fracdiff]: A module that computes fractional differentiation.
+
+[doc-fdiff]: https://fracdiff.github.io/fracdiff/generated/fracdiff.fdiff.html
+[doc-sklearn.Fracdiff]: https://fracdiff.github.io/fracdiff/generated/fracdiff.sklearn.Fracdiff.html
+[doc-sklearn.FracdiffStat]: https://fracdiff.github.io/fracdiff/generated/fracdiff.sklearn.FracdiffStat.html
+[doc-torch.fdiff]: https://fracdiff.github.io/fracdiff/generated/fracdiff.torch.fdiff.html
+[doc-torch.Fracdiff]: https://fracdiff.github.io/fracdiff/generated/fracdiff.torch.Fracdiff.html
 
 ### Speed
 
@@ -92,12 +98,14 @@ fdiff(a, 0.5, axis=-1)
 #        [0.    , 5.    , 3.5   , 4.375 ]])
 ```
 
-### Preprocessing by fractional differentiation
+### Scikit-learn API
+
+#### Preprocessing by fractional differentiation
 
 A transformer class [`Fracdiff`](https://fracdiff.github.io/fracdiff/#id1) performs fractional differentiation by its method `transform`.
 
 ```python
-from fracdiff import Fracdiff
+from fracdiff.sklearn import Fracdiff
 
 X = ...  # 2d time-series with shape (n_samples, n_features)
 
@@ -127,14 +135,14 @@ pipeline = Pipeline([
 pipeline.fit(X, y)
 ```
 
-### Fractional differentiation while preserving memory
+#### Fractional differentiation while preserving memory
 
 A transformer class [`FracdiffStat`](https://fracdiff.github.io/fracdiff/#fracdiffstat) finds the minumum order of fractional differentiation that makes time-series stationary.
 Differentiated time-series with this order is obtained by subsequently applying `transform` method.
 This series is interpreted as a stationary time-series keeping the maximum memory of the original time-series.
 
 ```python
-from fracdiff import FracdiffStat
+from fracdiff.sklearn import FracdiffStat
 
 X = ...  # 2d time-series with shape (n_samples, n_features)
 

fracdiff/fdiff.py

@@ -1,4 +1,5 @@
 from functools import partial
+from typing import Optional
 
 # found module but no type hints or library stubs
 import numpy as np  # type: ignore
@@ -7,7 +8,7 @@
 from scipy.special import binom  # type: ignore
 
 
-def fdiff_coef(d, window) -> np.ndarray:
+def fdiff_coef(d: float, window: int) -> np.ndarray:
     """Returns sequence of coefficients in fracdiff operator.
 
     Parameters
@@ -35,7 +36,13 @@ def fdiff_coef(d, window) -> np.ndarray:
 
 
 def fdiff(
-    a, n=1.0, axis=-1, prepend=np._NoValue, append=np._NoValue, window=10, mode="full"
+    a: np.ndarray,
+    n: float = 1.0,
+    axis: int = -1,
+    prepend: Optional[np.ndarray] = None,
+    append: Optional[np.ndarray] = None,
+    window: int = 10,
+    mode: str = "same",
 ) -> np.ndarray:
     """Calculate the `n`-th differentiation along the given axis.
 
@@ -59,13 +66,14 @@ def fdiff(
         Values to append.
     window : int, default=10
         Number of observations to compute each element in the output.
-    mode : {"full", "valid"}, default="full"
-        "full" (default) :
-            Return elements where at least one coefficient of fracdiff is used.
+    mode : {"same", "valid"}, default="same"
+        "same" (default) :
+            At the beginning of the time series,
+            return elements where at least one coefficient of fracdiff is used.
             Output size along ``axis`` is :math:`L_{\\mathrm{in}}`
             where :math:`L_{\\mathrm{in}}` is the length of ``a`` along ``axis``
             (plus the lengths of ``append`` and ``prepend``).
-            Boundary effects may be seen at the At the beginning of a time-series.
+            Boundary effects may be seen at the at the beginning of a time-series.
         "valid" :
             Return elements where all coefficients of fracdiff are used.
             Output size along ``axis`` is
@@ -114,18 +122,26 @@ def fdiff(
     array([[ 1. ,  3. ,  6. , 10. , 15. ],
            [-0.5,  3.5,  3. ,  3. ,  3.5]])
     """
+    if mode == "full":
+        mode = "same"
+        raise DeprecationWarning("mode 'full' was renamed to 'same'.")
+
     if isinstance(n, int) or n.is_integer():
+        prepend = np._NoValue if prepend is None else prepend  # type: ignore
+        append = np._NoValue if append is None else append  # type: ignore
         return np.diff(a, n=int(n), axis=axis, prepend=prepend, append=append)
 
     if a.ndim == 0:
         raise ValueError("diff requires input that is at least one dimensional")
 
     a = np.asanyarray(a)
-    axis = np.core.multiarray.normalize_axis_index(axis, a.ndim)
+    # Mypy complains:
+    # fracdiff/fdiff.py:135: error: Module has no attribute "normalize_axis_index"
+    axis = np.core.multiarray.normalize_axis_index(axis, a.ndim)  # type: ignore
     dtype = a.dtype if np.issubdtype(a.dtype, np.floating) else np.float64
 
     combined = []
-    if prepend is not np._NoValue:
+    if prepend is not None:
         prepend = np.asanyarray(prepend)
         if prepend.ndim == 0:
             shape = list(a.shape)
@@ -135,7 +151,7 @@ def fdiff(
 
     combined.append(a)
 
-    if append is not np._NoValue:
+    if append is not None:
         append = np.asanyarray(append)
         if append.ndim == 0:
             shape = list(a.shape)
@@ -149,7 +165,8 @@ def fdiff(
     if mode == "valid":
         D = partial(np.convolve, fdiff_coef(n, window).astype(dtype), mode="valid")
         a = np.apply_along_axis(D, axis, a)
-    elif mode == "full":
+    elif mode == "same":
+        # Convolve with the mode 'full' and cut last
         D = partial(np.convolve, fdiff_coef(n, window).astype(dtype), mode="full")
         s = tuple(
             slice(a.shape[axis]) if i == axis else slice(None) for i in range(a.ndim)

fracdiff/sklearn/fracdiff.py

@@ -16,7 +16,7 @@ class Fracdiff(TransformerMixin):
         The order of differentiation.
     window : int > 0 or None, default 10
         Number of observations to compute each element in the output.
-    mode : {"full", "valid"}, default "full"
+    mode : {"same", "valid"}, default "same"
         See :func:`fracdiff.fdiff` for details.
     window_policy : {"fixed"}, default "fixed"
         "fixed" (default) :
@@ -65,7 +65,7 @@ class Fracdiff(TransformerMixin):
            [-0.0625]])
     """
 
-    def __init__(self, d=1.0, window=10, mode="full", window_policy="fixed"):
+    def __init__(self, d=1.0, window=10, mode="same", window_policy="fixed"):
         self.d = d
         self.window = window
         self.mode = mode
@@ -76,7 +76,7 @@ def __repr__(self):
         Examples
         --------
         >>> Fracdiff(0.5)
-        Fracdiff(d=0.5, window=10, mode=full, window_policy=fixed)
+        Fracdiff(d=0.5, window=10, mode=same, window_policy=fixed)
         """
         name = self.__class__.__name__
         attrs = ["d", "window", "mode", "window_policy"]
@@ -104,7 +104,7 @@ def fit(self, X, y=None):
         self.coef_ = fdiff_coef(self.d, self.window)
         return self
 
-    def transform(self, X, y=None) -> numpy.array:
+    def transform(self, X, y=None) -> numpy.ndarray:
         """
         Return the fractional differentiation of `X`.
 

fracdiff/sklearn/fracdiffstat.py

@@ -20,7 +20,7 @@ class FracdiffStat(TransformerMixin, BaseEstimator):
     ----------
     window : int > 0 or None, default 10
         Number of observations to compute each element in the output.
-    mode : {"full", "valid"}, default "full"
+    mode : {"same", "valid"}, default "same"
         See :func:`fracdiff.fdiff` for details.
     window_policy : {"fixed"}, default "fixed"
         If "fixed" :
@@ -76,7 +76,7 @@ class FracdiffStat(TransformerMixin, BaseEstimator):
     def __init__(
         self,
         window=10,
-        mode="full",
+        mode="same",
         window_policy="fixed",
         stattest="ADF",
         pvalue=0.05,

fracdiff/sklearn/stat.py

@@ -1,3 +1,4 @@
+import numpy as np
 import statsmodels.tsa.stattools as stattools  # type: ignore
 
 
@@ -31,7 +32,7 @@ class StatTester:
     False
     """
 
-    def __init__(self, method="ADF"):
+    def __init__(self, method: str = "ADF"):
         self.method = method
 
     @property
@@ -41,7 +42,7 @@ def null_hypothesis(self) -> str:
         else:
             raise ValueError(f"Unknown method: {self.method}")
 
-    def pvalue(self, x) -> float:
+    def pvalue(self, x: np.ndarray) -> float:
         """
         Return p-value of the stationarity test.
 
@@ -61,7 +62,7 @@ def pvalue(self, x) -> float:
         else:
             raise ValueError(f"Unknown method: {self.method}")
 
-    def is_stat(self, x, pvalue=0.05) -> bool:
+    def is_stat(self, x: np.ndarray, pvalue: float = 0.05) -> bool:
         """
         Return whether stationarity test implies stationarity.
 

fracdiff/sklearn/tol.py

@@ -4,7 +4,7 @@
 from fracdiff.fdiff import fdiff_coef
 
 
-def window_from_tol_coef(n, tol_coef, max_window=2 ** 12) -> int:
+def window_from_tol_coef(n: float, tol_coef: float, max_window: int = 2 ** 12) -> int:
     """
     Return length of window determined from tolerance to memory loss.
 
@@ -40,10 +40,12 @@ def window_from_tol_coef(n, tol_coef, max_window=2 ** 12) -> int:
     -0.0625
     """
     coef = np.abs(fdiff_coef(n, max_window))
-    return np.searchsorted(-coef, -tol_coef) + 1  # index -> length
+    return int(np.searchsorted(-coef, -tol_coef) + 1)  # index -> length
 
 
-def window_from_tol_memory(n, tol_memory, max_window=2 ** 12) -> int:
+def window_from_tol_memory(
+    n: float, tol_memory: float, max_window: int = 2 ** 12
+) -> int:
     """
     Return length of window determined from tolerance to memory loss.
 
@@ -78,4 +80,4 @@ def window_from_tol_memory(n, tol_memory, max_window=2 ** 12) -> int:
     -0.20383...
     """
     lost_memory = np.abs(np.cumsum(fdiff_coef(n, max_window)))
-    return np.searchsorted(-lost_memory, -tol_memory) + 1  # index -> length
+    return int(np.searchsorted(-lost_memory, -tol_memory) + 1)  # index -> length

fracdiff/torch/functional.py

@@ -6,7 +6,7 @@
 from fracdiff.fdiff import fdiff_coef as fdiff_coef_numpy
 
 
-def fdiff_coef(d: float, window: int) -> torch.Tensor:
+def fdiff_coef(d: float, window: int) -> Tensor:
     """Returns sequence of coefficients in fracdiff operator.
 
     Args:
@@ -59,7 +59,6 @@ def fdiff(
           increases by the number of elements in each of these tensors.
 
     Examples:
-
         >>> from fracdiff.torch import fdiff
         >>> input = torch.tensor([1, 2, 4, 7, 0])
         >>> fdiff(input, 0.5, mode="same", window=3)