Add multi-symbol-buffer increment support

bar_select
Tyler Goodlet 2020-09-25 15:16:58 -04:00
parent 41e85ccaa9
commit e3e219aa4b
3 changed files with 83 additions and 55 deletions

View File

@ -23,10 +23,13 @@ from ._sharedmem import (
maybe_open_shm_array,
attach_shm_array,
open_shm_array,
SharedArray,
ShmArray,
get_shm_token,
)
from ._buffer import incr_buffer
from ._buffer import (
increment_ohlc_buffer,
subscribe_ohlc_for_increment
)
__all__ = [
@ -35,7 +38,7 @@ __all__ = [
'attach_shm_array',
'open_shm_array',
'get_shm_token',
'incr_buffer',
'subscribe_ohlc_for_increment',
]
@ -115,7 +118,7 @@ class Feed:
"""
name: str
stream: AsyncIterator[Dict[str, Any]]
shm: SharedArray
shm: ShmArray
_broker_portal: tractor._portal.Portal
_index_stream: Optional[AsyncIterator[Dict[str, Any]]] = None
@ -129,7 +132,7 @@ class Feed:
# created for all practical purposes
self._index_stream = await self._broker_portal.run(
'piker.data',
'incr_buffer',
'increment_ohlc_buffer',
shm_token=self.shm.token,
topics=['index'],
)

View File

@ -1,54 +1,73 @@
"""
Data buffers for fast shared humpy.
"""
from typing import Tuple, Callable
import time
from typing import Tuple, Callable, Dict
# import time
import tractor
import trio
from ._sharedmem import attach_shm_array
from ._sharedmem import ShmArray
_shms: Dict[int, ShmArray] = {}
@tractor.msg.pub
async def incr_buffer(
async def increment_ohlc_buffer(
shm_token: dict,
get_topics: Callable[..., Tuple[str]],
# delay_s: Optional[float] = None,
):
"""Task which inserts new bars into the provide shared memory array
every ``delay_s`` seconds.
This task fulfills 2 purposes:
- it takes the subscribed set of shm arrays and increments them
on a common time period
- broadcast of this increment "signal" message to other actor
subscribers
Note that if **no** actor has initiated this task then **none** of
the underlying buffers will actually be incremented.
"""
# TODO: right now we'll spin printing bars if the last time
# stamp is before a large period of no market activity.
# Likely the best way to solve this is to make this task
# aware of the instrument's tradable hours?
shm = attach_shm_array(
token=shm_token,
readonly=False,
)
# determine ohlc delay between bars
# to determine time step between datums
times = shm.array['time']
delay_s = times[-1] - times[times != times[-1]][-1]
# TODO: right now we'll spin printing bars if the last time stamp is
# before a large period of no market activity. Likely the best way
# to solve this is to make this task aware of the instrument's
# tradable hours?
# adjust delay to compensate for trio processing time
ad = delay_s - 0.002
ad = min(_shms.keys()) - 0.001
async def sleep():
"""Sleep until next time frames worth has passed from last bar.
"""
# last_ts = shm.array[-1]['time']
# delay = max((last_ts + ad) - time.time(), 0)
# await trio.sleep(delay)
await trio.sleep(ad)
# async def sleep():
# """Sleep until next time frames worth has passed from last bar.
# """
# # last_ts = shm.array[-1]['time']
# # delay = max((last_ts + ad) - time.time(), 0)
# # await trio.sleep(delay)
# await trio.sleep(ad)
total_s = 0 # total seconds counted
lowest = min(_shms.keys())
ad = lowest - 0.001
while True:
# sleep for duration of current bar
await sleep()
# TODO: do we want to support dynamically
# adding a "lower" lowest increment period?
await trio.sleep(ad)
total_s += lowest
# # sleep for duration of current bar
# await sleep()
# increment all subscribed shm arrays
# TODO: this in ``numba``
for delay_s, shms in _shms.items():
if total_s % delay_s != 0:
continue
# TODO: numa this!
for shm in shms:
# TODO: in theory we could make this faster by copying the
# "last" readable value into the underlying larger buffer's
# next value and then incrementing the counter instead of
@ -66,9 +85,15 @@ async def incr_buffer(
# write to the buffer
shm.push(last)
# print('incrementing array')
# print(get_topics())
# broadcast the buffer index step
yield {'index': shm._i.value}
def subscribe_ohlc_for_increment(
shm: ShmArray,
delay: int,
) -> None:
"""Add an OHLC ``ShmArray`` to the increment set.
"""
_shms.setdefault(delay, []).append(shm)

View File

@ -134,7 +134,7 @@ def _make_token(
)
class SharedArray:
class ShmArray:
def __init__(
self,
shmarr: np.ndarray,
@ -216,7 +216,7 @@ def open_shm_array(
size: int = int(2*60*60*10/5),
dtype: np.dtype = base_ohlc_dtype,
readonly: bool = False,
) -> SharedArray:
) -> ShmArray:
"""Open a memory shared ``numpy`` using the standard library.
This call unlinks (aka permanently destroys) the buffer on teardown
@ -245,7 +245,7 @@ def open_shm_array(
)
counter.value = 0
shmarr = SharedArray(
shmarr = ShmArray(
array,
counter,
shm,
@ -268,7 +268,7 @@ def attach_shm_array(
size: int = int(60*60*10/5),
# dtype: np.dtype = base_ohlc_dtype,
readonly: bool = True,
) -> SharedArray:
) -> ShmArray:
"""Load and attach to an existing shared memory array previously
created by another process using ``open_shared_array``.
"""
@ -289,7 +289,7 @@ def attach_shm_array(
# make sure we can read
counter.value
sha = SharedArray(
sha = ShmArray(
shmarr,
counter,
shm,
@ -314,7 +314,7 @@ def maybe_open_shm_array(
key: str,
dtype: np.dtype = base_ohlc_dtype,
**kwargs,
) -> Tuple[SharedArray, bool]:
) -> Tuple[ShmArray, bool]:
"""Attempt to attach to a shared memory block by a
"key" determined by the users overall "system"
(presumes you don't have the block's explicit token).