added extra features

README.md

@@ -13,7 +13,7 @@ Roboquant is an open-source algorithmic trading platform. It is flexible, user-f
 So whether you are a beginning retail trader or an established trading firm, roboquant can help you to develop robust and fully automated trading strategies. You can find out more at [roboquant.org](https://roboquant.org).
 
 ## Usage
-The following code snippet shows the steps to run a full back-test on a number of stocks.
+The following code snippet shows all the steps to run a full back-test on a number of stocks.
 
 ```python
 import roboquant as rq
@@ -32,6 +32,9 @@ Make sure you have Python version 3.10 or higher installed.
 python3 -m pip install --upgrade roboquant
 ```
 
+You can also try roboquant in an online Jupyter Notebook [![Binder](https://mybinder.org/badge_logo.svg)](https://mybinder.org/v2/gh/neurallayer/roboquant-notebooks/HEAD?labpath=%2Fintro_roboquant.ipynb)
+
+
 The core of roboquant limits the number of dependencies. 
 But you can install roboquant including one or more of the optional dependencies if you require certain functionality:
 
@@ -76,8 +79,11 @@ python -m unittest discover -s tests/unit
 flake8 roboquant tests
 ```
 
-## Kotlin version
-Next to this Python version of `roboquant`, there is also a Kotlin version available.
-Both (will) share a similar API, just the used computer language is different.
+## License
+Roboquant is made available under the Apache 2.0 license. You can read more about the Apache 2.0 license on this page: https://www.apache.org/licenses/LICENSE-2.0.html
 
-Which one to use depends very much on personal preferences, skills and usage.
+## Disclaimer
+Absolutely no warranty is implied with this product. Use at your own risk. I provide no guarantee that it will be profitable, or that it won't lose all your money very quickly or doesn't contain bugs. All financial trading offers the possibility of loss. Leveraged trading, may result in you losing all your money, and still owing more. Backtested results are no guarantee of future performance. I can take no responsibility for any losses caused by live trading using roboquant. Use at your own risk. I am not registered or authorised by any financial regulator.
+
+## Kotlin version
+Next to this Python version of `roboquant`, there is also a Kotlin version available. Both (will) share a similar API, just the used computer language is different. Which one to use depends very much on personal preferences, skills and usage.

roboquant/strategies/features.py

@@ -25,6 +25,21 @@ def returns(self, period=1):
             return ReturnsFeature(self)
         return LongReturnsFeature(self, period)
 
+    def __getitem__(self, *args):
+        return SlicedFeature(self, args)
+
+
+class SlicedFeature(Feature):
+
+    def __init__(self, feature: Feature, args: tuple) -> None:
+        super().__init__()
+        self.args = args
+        self.feature = feature
+
+    def calc(self, evt):
+        values = self.feature.calc(evt)
+        return values[*self.args]
+
 
 class FixedValueFeature(Feature):
 
@@ -37,6 +52,7 @@ def calc(self, evt):
 
 
 class PriceFeature(Feature):
+    """Extract a single price for a symbol"""
 
     def __init__(self, symbol: str, price_type: str = "DEFAULT") -> None:
         self.symbol = symbol
@@ -50,6 +66,7 @@ def calc(self, evt):
 
 
 class CandleFeature(Feature):
+    """Extract the ohlcv values for a symbol"""
 
     def __init__(self, symbol: str) -> None:
 
@@ -64,7 +81,28 @@ def calc(self, evt):
         return np.full((5,), float("nan"))
 
 
+class FillFeature(Feature):
+    """If the feature returns nan, use the last complete values instead"""
+
+    def __init__(self, feature: Feature) -> None:
+        super().__init__()
+        self.history = None
+        self.feature: Feature = feature
+
+    def calc(self, evt):
+        values = self.feature.calc(evt)
+
+        if np.any(np.isnan(values)):
+            if self.history is not None:
+                return self.history
+            return values
+
+        self.history = values
+        return values
+
+
 class VolumeFeature(Feature):
+    """Extract the volume for a symbol"""
 
     def __init__(self, symbol, volume_type: str = "DEFAULT") -> None:
         super().__init__()
@@ -116,6 +154,50 @@ def calc(self, evt):
         return r
 
 
+class MaxReturnFeature(Feature):
+    """Calculate the maximum return over a certain period.
+    This will only work on features that return a single value.
+    """
+    def __init__(self, feature: Feature, period: int) -> None:
+        super().__init__()
+        self.history = deque(maxlen=period)
+        self.feature: Feature = feature
+
+    def calc(self, evt):
+        values = self.feature.calc(evt)
+        h = self.history
+
+        if len(h) < h.maxlen:  # type: ignore
+            h.append(values)
+            return np.full(values.shape, float("nan"))
+
+        r = max(h) / h[0] - 1.0
+        h.append(values)
+        return r
+
+
+class MinReturnFeature(Feature):
+    """Calculate the minimum return over a certain period.
+    This will only work on features that return a single value.
+    """
+    def __init__(self, feature: Feature, period: int) -> None:
+        super().__init__()
+        self.history = deque(maxlen=period)
+        self.feature: Feature = feature
+
+    def calc(self, evt):
+        values = self.feature.calc(evt)
+        h = self.history
+
+        if len(h) < h.maxlen:  # type: ignore
+            h.append(values)
+            return np.full(values.shape, float("nan"))
+
+        r = min(h) / h[0] - 1.0
+        h.append(values)
+        return r
+
+
 class SMAFeature(Feature):
 
     def __init__(self, feature, period) -> None:
@@ -141,6 +223,7 @@ def calc(self, evt):
 
 
 class DayOfWeekFeature(Feature):
+    """Calculate a one-hot-encoded day of week"""
 
     def __init__(self, tz=timezone.utc) -> None:
         self.tz = tz

tests/unit/test_features.py

@@ -1,6 +1,7 @@
 import unittest
 
 import numpy as np
+from roboquant.event import Event
 
 from roboquant.strategies.features import (
     PriceFeature,
@@ -36,6 +37,16 @@ def test_all_features(self):
                 result = feature.calc(evt)
                 self.assertTrue(len(result) > 0)
 
+    def test_core_feature(self):
+        f = FixedValueFeature(np.ones(10,))[2:5]
+        values = f.calc(Event.empty())
+        self.assertEqual(3, len(values))
+
+        f = FixedValueFeature(np.ones(10,)).returns()
+        values = f.calc(Event.empty())
+        values = f.calc(Event.empty())
+        self.assertEqual(0, values[0])
+
 
 if __name__ == "__main__":
     unittest.main()