pikers
/
piker
3
0
Fork 0
Code Issues 2 Pull Requests 4 Packages Projects Releases Wiki Activity

Move path ops routines to top of mod 

Planning to put the formatters into a new mod and aggregate all path
gen/op helpers into this module.

Further tweak include:
- moving `path_arrays_from_ohlc()` back to module level
- slice out the last xy datum for `OHLCBarsAsCurveFmtr` 1d formatting
- always copy the new x-value from the source to `.x_nd`
epoch_indexing_and_dataviz_layer
Tyler Goodlet 2022-11-30 15:47:06 -05:00
parent 6ea04f850d
commit 7f3f6f871a
1 changed files with 156 additions and 168 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

330
piker/data/_pathops.py
View File

@ -50,6 +50,129 @@ if TYPE_CHECKING:
from .._profile import Profiler from .._profile import Profiler
def xy_downsample(
x,
y,
uppx,
x_spacer: float = 0.5,
) -> tuple[
np.ndarray,
np.ndarray,
float,
float,
]:
'''
Downsample 1D (flat ``numpy.ndarray``) arrays using M4 given an input
``uppx`` (units-per-pixel) and add space between discreet datums.
'''
# downsample whenever more then 1 pixels per datum can be shown.
# always refresh data bounds until we get diffing
# working properly, see above..
bins, x, y, ymn, ymx = ds_m4(
x,
y,
uppx,
)
# flatten output to 1d arrays suitable for path-graphics generation.
x = np.broadcast_to(x[:, None], y.shape)
x = (x + np.array(
[-x_spacer, 0, 0, x_spacer]
)).flatten()
y = y.flatten()
return x, y, ymn, ymx
@njit(
# NOTE: need to construct this manually for readonly
# arrays, see https://github.com/numba/numba/issues/4511
# (
# types.Array(
# numba_ohlc_dtype,
# 1,
# 'C',
# readonly=True,
# ),
# int64,
# types.unicode_type,
# optional(float64),
# ),
nogil=True
)
def path_arrays_from_ohlc(
data: np.ndarray,
start: int64,
bar_gap: float64 = 0.43,
# index_field: str,
) -> tuple[
np.ndarray,
np.ndarray,
np.ndarray,
]:
'''
Generate an array of lines objects from input ohlc data.
'''
size = int(data.shape[0] * 6)
# XXX: see this for why the dtype might have to be defined outside
# the routine.
# https://github.com/numba/numba/issues/4098#issuecomment-493914533
x = np.zeros(
shape=size,
dtype=float64,
)
y, c = x.copy(), x.copy()
# TODO: report bug for assert @
# /home/goodboy/repos/piker/env/lib/python3.8/site-packages/numba/core/typing/builtins.py:991
for i, q in enumerate(data[start:], start):
# TODO: ask numba why this doesn't work..
# open, high, low, close, index = q[
# ['open', 'high', 'low', 'close', 'index']]
open = q['open']
high = q['high']
low = q['low']
close = q['close']
# index = float64(q[index_field])
index = float64(q['index'])
istart = i * 6
istop = istart + 6
# x,y detail the 6 points which connect all vertexes of a ohlc bar
x[istart:istop] = (
index - bar_gap,
index,
index,
index,
index,
index + bar_gap,
)
y[istart:istop] = (
open,
open,
low,
high,
close,
close,
)
# specifies that the first edge is never connected to the
# prior bars last edge thus providing a small "gap"/"space"
# between bars determined by ``bar_gap``.
c[istart:istop] = (1, 1, 1, 1, 1, 0)
return x, y, c
class IncrementalFormatter(msgspec.Struct): class IncrementalFormatter(msgspec.Struct):
''' '''
Incrementally updating, pre-path-graphics tracking, formatter. Incrementally updating, pre-path-graphics tracking, formatter.
@ -132,7 +255,6 @@ class IncrementalFormatter(msgspec.Struct):
np.ndarray, np.ndarray,
np.ndarray, np.ndarray,
]: ]:
# TODO: # TODO:
# - can the renderer just call ``Viz.read()`` directly? unpack # - can the renderer just call ``Viz.read()`` directly? unpack
# latest source data read # latest source data read
@ -423,18 +545,11 @@ class IncrementalFormatter(msgspec.Struct):
) -> None: ) -> None:
# write pushed data to flattened copy # write pushed data to flattened copy
new_y_nd = new_from_src[data_field] new_y_nd = new_from_src[data_field]
# XXX
# TODO: this should be returned and written by caller!
# XXX
# generate same-valued-per-row x support with Nx1 shape
index_field = self.index_field
if index_field != 'index':
x_nd_new = self.x_nd[read_slc]
x_nd_new[:] = new_from_src[index_field]
self.y_nd[read_slc] = new_y_nd self.y_nd[read_slc] = new_y_nd
x_nd_new = self.x_nd[read_slc]
x_nd_new[:] = new_from_src[self.index_field]
# XXX: was ``.format_xy()`` # XXX: was ``.format_xy()``
def format_xy_nd_to_1d( def format_xy_nd_to_1d(
self, self,
@ -454,6 +569,8 @@ class IncrementalFormatter(msgspec.Struct):
Return single field column data verbatim Return single field column data verbatim
''' '''
# NOTE: we don't include the very last datum which is filled in
# normally by another graphics object.
return ( return (
array[self.index_field][:-1], array[self.index_field][:-1],
array[array_key][:-1], array[array_key][:-1],
@ -507,92 +624,37 @@ class OHLCBarsFmtr(IncrementalFormatter):
y_nd, y_nd,
) )
@staticmethod def incr_update_xy_nd(
@njit( self,
# NOTE: need to construct this manually for readonly
# arrays, see https://github.com/numba/numba/issues/4511 src_shm: ShmArray,
# ( data_field: str,
# types.Array(
# numba_ohlc_dtype, new_from_src: np.ndarray, # portion of source that was updated
# 1,
# 'C', read_slc: slice,
# readonly=True, ln: int, # len of updated
# ),
# int64, nd_start: int,
# types.unicode_type, nd_stop: int,
# optional(float64),
# ), is_append: bool,
nogil=True
) -> None:
# write newly pushed data to flattened copy
# a struct-arr is always passed in.
new_y_nd = rfn.structured_to_unstructured(
new_from_src[self.fields]
) )
def path_arrays_from_ohlc( self.y_nd[read_slc] = new_y_nd
data: np.ndarray,
start: int64,
bar_gap: float64 = 0.43,
# index_field: str,
) -> tuple[ # generate same-valued-per-row x support based on y shape
np.ndarray, x_nd_new = self.x_nd[read_slc]
np.ndarray, x_nd_new[:] = np.broadcast_to(
np.ndarray, new_from_src[self.index_field][:, None],
]: new_y_nd.shape,
''' ) + np.array([-0.5, 0, 0, 0.5])
Generate an array of lines objects from input ohlc data.
'''
size = int(data.shape[0] * 6)
# XXX: see this for why the dtype might have to be defined outside
# the routine.
# https://github.com/numba/numba/issues/4098#issuecomment-493914533
x = np.zeros(
shape=size,
dtype=float64,
)
y, c = x.copy(), x.copy()
# TODO: report bug for assert @
# /home/goodboy/repos/piker/env/lib/python3.8/site-packages/numba/core/typing/builtins.py:991
for i, q in enumerate(data[start:], start):
# TODO: ask numba why this doesn't work..
# open, high, low, close, index = q[
# ['open', 'high', 'low', 'close', 'index']]
open = q['open']
high = q['high']
low = q['low']
close = q['close']
# index = float64(q[index_field])
# index = float64(q['time'])
index = float64(q['index'])
istart = i * 6
istop = istart + 6
# x,y detail the 6 points which connect all vertexes of a ohlc bar
x[istart:istop] = (
index - bar_gap,
index,
index,
index,
index,
index + bar_gap,
)
y[istart:istop] = (
open,
open,
low,
high,
close,
close,
)
# specifies that the first edge is never connected to the
# prior bars last edge thus providing a small "gap"/"space"
# between bars determined by ``bar_gap``.
c[istart:istop] = (1, 1, 1, 1, 1, 0)
return x, y, c
# TODO: can we drop this frame and just use the above? # TODO: can we drop this frame and just use the above?
def format_xy_nd_to_1d( def format_xy_nd_to_1d(
@ -617,7 +679,7 @@ class OHLCBarsFmtr(IncrementalFormatter):
for line spacing. for line spacing.
''' '''
x, y, c = self.path_arrays_from_ohlc( x, y, c = path_arrays_from_ohlc(
array, array,
start, start,
# self.index_field, # self.index_field,
@ -625,43 +687,6 @@ class OHLCBarsFmtr(IncrementalFormatter):
) )
return x, y, c return x, y, c
def incr_update_xy_nd(
self,
src_shm: ShmArray,
data_field: str,
new_from_src: np.ndarray, # portion of source that was updated
read_slc: slice,
ln: int, # len of updated
nd_start: int,
nd_stop: int,
is_append: bool,
) -> None:
# write newly pushed data to flattened copy
# a struct-arr is always passed in.
new_y_nd = rfn.structured_to_unstructured(
new_from_src[self.fields]
)
# XXX
# TODO: this should be returned and written by caller!
# XXX
# generate same-valued-per-row x support based on y shape
index_field: str = self.index_field
if index_field != 'index':
x_nd_new = self.x_nd[read_slc]
x_nd_new[:] = new_from_src[index_field][:, np.newaxis]
if (self.x_nd[self.xy_slice] == 0.5).any():
breakpoint()
self.y_nd[read_slc] = new_y_nd
class OHLCBarsAsCurveFmtr(OHLCBarsFmtr): class OHLCBarsAsCurveFmtr(OHLCBarsFmtr):
@ -681,8 +706,8 @@ class OHLCBarsAsCurveFmtr(OHLCBarsFmtr):
# should we be passing in array as an xy arrays tuple? # should we be passing in array as an xy arrays tuple?
# 2 more datum-indexes to capture zero at end # 2 more datum-indexes to capture zero at end
x_flat = self.x_nd[self.xy_nd_start:self.xy_nd_stop] x_flat = self.x_nd[self.xy_nd_start:self.xy_nd_stop-1]
y_flat = self.y_nd[self.xy_nd_start:self.xy_nd_stop] y_flat = self.y_nd[self.xy_nd_start:self.xy_nd_stop-1]
# slice to view # slice to view
ivl, ivr = vr ivl, ivr = vr
@ -871,40 +896,3 @@ class StepCurveFmtr(IncrementalFormatter):
# ) # )
return x_1d, y_1d, 'all' return x_1d, y_1d, 'all'
def xy_downsample(
x,
y,
uppx,
x_spacer: float = 0.5,
) -> tuple[
np.ndarray,
np.ndarray,
float,
float,
]:
'''
Downsample 1D (flat ``numpy.ndarray``) arrays using M4 given an input
``uppx`` (units-per-pixel) and add space between discreet datums.
'''
# downsample whenever more then 1 pixels per datum can be shown.
# always refresh data bounds until we get diffing
# working properly, see above..
bins, x, y, ymn, ymx = ds_m4(
x,
y,
uppx,
)
# flatten output to 1d arrays suitable for path-graphics generation.
x = np.broadcast_to(x[:, None], y.shape)
x = (x + np.array(
[-x_spacer, 0, 0, x_spacer]
)).flatten()
y = y.flatten()
return x, y, ymn, ymx