Get bar oriented RSI working correctly

piker/fsp/_momo.py

@@ -1,10 +1,13 @@
 """
 Momentum bby.
 """
-from typing import AsyncIterator
+from typing import AsyncIterator, Optional
 
 import numpy as np
-from numba import jit, float64, optional
+from ringbuf import RingBuffer
+from numba import jit, float64, optional, int64
+
+from ..data._normalize import iterticks
 
 
 # TODO: things to figure the fuck out:
@@ -47,6 +50,9 @@ def ema(
                 s[0] = y[0]; t = 0
                 s[t] = a*y[t] + (1-a)*s[t-1], t > 0.
             }
+
+    More discussion here:
+    https://stackoverflow.com/questions/42869495/numpy-version-of-exponential-weighted-moving-average-equivalent-to-pandas-ewm
     """
     n = y.shape[0]
 
@@ -67,6 +73,7 @@ def ema(
         else:
             s[0] = ylast
 
+        print(s)
         for i in range(1, n):
             s[i] = y[i] * alpha + s[i-1] * (1 - alpha)
 
@@ -77,34 +84,40 @@ def ema(
 #     float64[:](
 #         float64[:],
 #         int64,
+#         float64,
+#         float64,
 #     ),
-#     # nopython=True,
+#     nopython=True,
 #     nogil=True
 # )
 def rsi(
     signal: 'np.ndarray[float64]',
-    period: int = 14,
+    period: int64 = 14,
     up_ema_last: float64 = None,
     down_ema_last: float64 = None,
 ) -> 'np.ndarray[float64]':
     alpha = 1/period
-    # print(signal)
 
     df = np.diff(signal)
-    up, down = np.where(df > 0, df, 0), np.where(df < 0, -df, 0)
+
+    up = np.where(df > 0, df, 0)
     up_ema = ema(up, alpha, up_ema_last)
+
+    down = np.where(df < 0, -df, 0)
     down_ema = ema(down, alpha, down_ema_last)
+
+    # avoid dbz errors
     rs = np.divide(
         up_ema,
         down_ema,
         out=np.zeros_like(up_ema),
-        where=down_ema!=0
+        where=down_ema != 0
     )
-    # print(f'up_ema: {up_ema}\ndown_ema: {down_ema}')
+
-    # print(f'rs: {rs}')
     # map rs through sigmoid (with range [0, 100])
     rsi = 100 - 100 / (1 + rs)
     # rsi = 100 * (up_ema / (up_ema + down_ema))
+
     # also return the last ema state for next iteration
     return rsi, up_ema[-1], down_ema[-1]
 
@@ -114,67 +127,96 @@ def rsi(
 # )
 async def _rsi(
     source: 'QuoteStream[Dict[str, Any]]',  # noqa
-    ohlcv: np.ndarray,
+    ohlcv: "ShmArray[T<'close'>]",
     period: int = 14,
 ) -> AsyncIterator[np.ndarray]:
     """Multi-timeframe streaming RSI.
 
     https://en.wikipedia.org/wiki/Relative_strength_index
     """
-    sig = ohlcv['close']
+    sig = ohlcv.array['close']
 
     # TODO: the emas here should be seeded with a period SMA as per
     # wilder's original formula..
-    rsi_h, up_ema_last, down_ema_last = rsi(sig, period, None, None)
+    rsi_h, last_up_ema_close, last_down_ema_close = rsi(sig, period, None, None)
 
     # deliver history
     yield rsi_h
 
-    _last = sig[-1]
+    index = ohlcv.index
 
     async for quote in source:
         # tick based updates
-        for tick in quote.get('ticks', ()):
+        for tick in iterticks(quote):
-            if tick.get('type') == 'trade':
+            # though incorrect below is interesting
-                curr = tick['price']
+            # sig = ohlcv.last(period)['close']
-                last = np.array([_last, curr])
+            sig = ohlcv.last(2)['close']
-                # await tractor.breakpoint()
+
+            # the ema needs to be computed from the "last bar"
+            if ohlcv.index > index:
+                last_up_ema_close = up_ema_last
+                last_down_ema_close = down_ema_last
+                index = ohlcv.index
+
             rsi_out, up_ema_last, down_ema_last = rsi(
-                    last,
+                sig,
                 period=period,
-                    up_ema_last=up_ema_last,
+                up_ema_last=last_up_ema_close,
-                    down_ema_last=down_ema_last,
+                down_ema_last=last_down_ema_close,
             )
-                _last = curr
+            print(f'rsi_out: {rsi_out}')
-                # print(f'last: {last}\n rsi: {rsi_out}')
+            yield rsi_out[-1:]
-                yield rsi_out[-1]
+
+def wma(
+    signal: np.ndarray,
+) -> np.ndarray:
+    if weights is None:
+        # default is a standard arithmetic mean
+        seq = np.full((length,), 1)
+        weights = seq / seq.sum()
+
+    assert length == len(weights)
 
 
 async def wma(
     source,  #: AsyncStream[np.ndarray],
+    length: int,
     ohlcv: np.ndarray,  # price time-frame "aware"
-    lookback: np.ndarray,  # price time-frame "aware"
+    weights: Optional[np.ndarray] = None,
-    weights: np.ndarray,
+) -> AsyncIterator[np.ndarray]:  # maybe something like like FspStream?
-) -> AsyncIterator[np.ndarray]:  # i like FinSigStream
+    """Streaming weighted moving average.
-    """Weighted moving average.
 
     ``weights`` is a sequence of already scaled values. As an example
     for the WMA often found in "techincal analysis":
     ``weights = np.arange(1, N) * N*(N-1)/2``.
     """
-    length = len(weights)
+    # deliver historical output as "first yield"
-    _lookback = np.zeros(length - 1)
+    yield np.convolve(ohlcv.array['close'], weights, 'valid')
 
-    ohlcv.from_tf('5m')
+    # begin real-time section
 
-    # async for frame_len, frame in source:
+    # fill length samples as lookback history
-    async for frame in source:
+    # ringbuf = RingBuffer(format='f', capacity=2*length)
-        wma = np.convolve(
+    # overflow = ringbuf.push(ohlcv['close'][-length + 1:])
-            ohlcv[-length:]['close'],
+    # assert overflow is None
-            # np.concatenate((_lookback, frame)),
+
-            weights,
+    # lookback = np.zeros((length,))
-            'valid'
+    # lookback[:-1] = ohlcv['close'][-length + 1:]
-        )
+
-        # todo: handle case where frame_len < length - 1
+    # async for frame in atleast(length, source):
-        _lookback = frame[-(length-1):]  # noqa
+    async for quote in source:
-        yield wma
+        for tick in iterticks(quote, type='trade'):
+            # writes no matter what
+            overflow = ringbuf.push(np.array([tick['price']]))
+            assert overflow is None
+
+            # history = np.concatenate(ringbuf.pop(length - 1), frame)
+
+            sig = ohlcv.last(length)
+            history = ringbuf.pop(ringbuf.read_available)
+            yield np.convolve(history, weights, 'valid')
+
+            # push back `length-1` datums as lookback in preparation
+            # for next minimum 1 datum arrival which will require
+            # another "window's worth" of history.
+            ringbuf.push(history[-length + 1:])