from __future__ import annotations
import math
import numpy as np
from toolz import memoize
from datashader.enums import AntialiasCombination
from datashader.glyphs.points import _PointLike, _GeometryLike
from datashader.utils import (
isnull, isreal, nanfirst_in_place, nanlast_in_place, nanmax_in_place,
nanmin_in_place, nansum_in_place, ngjit, parallel_fill)
from numba import cuda, literal_unroll
import numba.types as nb_types
try:
import cudf
import cupy as cp
from ..transfer_functions._cuda_utils import cuda_args
except ImportError:
cudf = None
cuda_args = None
try:
import spatialpandas
except Exception:
spatialpandas = None
def _two_stage_agg(antialias_stage_2):
"""Information used to perform the correct stage 2 aggregation."""
if not antialias_stage_2:
# Not using antialiased lines, doesn't matter what is returned.
return False, False
# A single combination in (SUM_2AGG, FIRST, LAST, MIN) means that a 2-stage
# aggregation will be used, otherwise use a 1-stage aggregation that is
# faster.
use_2_stage_agg = False
for comb in antialias_stage_2[0]:
if comb in (AntialiasCombination.SUM_2AGG, AntialiasCombination.MIN,
AntialiasCombination.FIRST, AntialiasCombination.LAST):
use_2_stage_agg = True
break
# Boolean overwrite flag is used in _full_antialias() is True to overwrite
# pixel values (using max of previous and new values) or False for the more
# complicated correction algorithm. Prefer overwrite=True for speed, but
# any SUM_1AGG implies overwrite=False.
overwrite = True
for comb in antialias_stage_2[0]:
if comb == AntialiasCombination.SUM_1AGG:
overwrite = False
break
return overwrite, use_2_stage_agg
class _AntiAliasedLine(object):
""" Methods common to all lines. """
_line_width = 0 # Use antialiasing if > 0.
def set_line_width(self, line_width):
self._line_width = line_width
def _build_extend(self, x_mapper, y_mapper, info, append, antialias_stage_2):
return self._internal_build_extend(
x_mapper, y_mapper, info, append, self._line_width, antialias_stage_2)
[docs]class LineAxis0(_PointLike, _AntiAliasedLine):
"""A line, with vertices defined by ``x`` and ``y``.
Parameters
----------
x, y : str
Column names for the x and y coordinates of each vertex.
"""
@memoize
def _internal_build_extend(
self, x_mapper, y_mapper, info, append, line_width, antialias_stage_2):
antialias = line_width > 0
expand_aggs_and_cols = self.expand_aggs_and_cols(append)
map_onto_pixel = _build_map_onto_pixel_for_line(
x_mapper, y_mapper, antialias)
overwrite, use_2_stage_agg = _two_stage_agg(antialias_stage_2)
draw_segment = _build_draw_segment(
append, map_onto_pixel, expand_aggs_and_cols, line_width, overwrite
)
extend_cpu, extend_cuda = _build_extend_line_axis0(
draw_segment, expand_aggs_and_cols, use_2_stage_agg
)
x_name = self.x
y_name = self.y
def extend(aggs, df, vt, bounds, plot_start=True):
sx, tx, sy, ty = vt
xmin, xmax, ymin, ymax = bounds
aggs_and_cols = aggs + info(df)
if cudf and isinstance(df, cudf.DataFrame):
xs = self.to_cupy_array(df, x_name)
ys = self.to_cupy_array(df, y_name)
do_extend = extend_cuda[cuda_args(xs.shape)]
else:
xs = df.loc[:, x_name].to_numpy()
ys = df.loc[:, y_name].to_numpy()
do_extend = extend_cpu
# line may be clipped, then mapped to pixels
do_extend(
sx, tx, sy, ty, xmin, xmax, ymin, ymax,
xs, ys, plot_start, antialias_stage_2, *aggs_and_cols
)
return extend
[docs]class LineAxis0Multi(_PointLike, _AntiAliasedLine):
"""
"""
def validate(self, in_dshape):
if not all([isreal(in_dshape.measure[str(xcol)]) for xcol in self.x]):
raise ValueError('x columns must be real')
elif not all([isreal(in_dshape.measure[str(ycol)]) for ycol in self.y]):
raise ValueError('y columns must be real')
@property
def x_label(self):
return 'x'
@property
def y_label(self):
return 'y'
def required_columns(self):
return self.x + self.y
def compute_x_bounds(self, df):
bounds_list = [self._compute_bounds(df[x])
for x in self.x]
mins, maxes = zip(*bounds_list)
return self.maybe_expand_bounds((min(mins), max(maxes)))
def compute_y_bounds(self, df):
bounds_list = [self._compute_bounds(df[y])
for y in self.y]
mins, maxes = zip(*bounds_list)
return self.maybe_expand_bounds((min(mins), max(maxes)))
@memoize
def compute_bounds_dask(self, ddf):
r = ddf.map_partitions(lambda df: np.array([[
np.nanmin([np.nanmin(df[c].values).item() for c in self.x]),
np.nanmax([np.nanmax(df[c].values).item() for c in self.x]),
np.nanmin([np.nanmin(df[c].values).item() for c in self.y]),
np.nanmax([np.nanmax(df[c].values).item() for c in self.y])]]
)).compute()
x_extents = np.nanmin(r[:, 0]), np.nanmax(r[:, 1])
y_extents = np.nanmin(r[:, 2]), np.nanmax(r[:, 3])
return (self.maybe_expand_bounds(x_extents),
self.maybe_expand_bounds(y_extents))
@memoize
def _internal_build_extend(
self, x_mapper, y_mapper, info, append, line_width, antialias_stage_2):
antialias = line_width > 0
expand_aggs_and_cols = self.expand_aggs_and_cols(append)
map_onto_pixel = _build_map_onto_pixel_for_line(
x_mapper, y_mapper, antialias)
overwrite, use_2_stage_agg = _two_stage_agg(antialias_stage_2)
draw_segment = _build_draw_segment(
append, map_onto_pixel, expand_aggs_and_cols, line_width, overwrite
)
extend_cpu, extend_cuda = _build_extend_line_axis0_multi(
draw_segment, expand_aggs_and_cols, use_2_stage_agg
)
x_names = self.x
y_names = self.y
def extend(aggs, df, vt, bounds, plot_start=True):
sx, tx, sy, ty = vt
xmin, xmax, ymin, ymax = bounds
aggs_and_cols = aggs + info(df)
if cudf and isinstance(df, cudf.DataFrame):
xs = self.to_cupy_array(df, x_names)
ys = self.to_cupy_array(df, y_names)
do_extend = extend_cuda[cuda_args(xs.shape)]
else:
xs = df.loc[:, list(x_names)].to_numpy()
ys = df.loc[:, list(y_names)].to_numpy()
do_extend = extend_cpu
# line may be clipped, then mapped to pixels
do_extend(
sx, tx, sy, ty, xmin, xmax, ymin, ymax,
xs, ys, plot_start, antialias_stage_2, *aggs_and_cols,
)
return extend
[docs]class LinesAxis1(_PointLike, _AntiAliasedLine):
"""A collection of lines (on line per row) with vertices defined
by the lists of columns in ``x`` and ``y``
Parameters
----------
x, y : list
Lists of column names for the x and y coordinates
"""
def validate(self, in_dshape):
if not all([isreal(in_dshape.measure[str(xcol)])
for xcol in self.x]):
raise ValueError('x columns must be real')
elif not all([isreal(in_dshape.measure[str(ycol)])
for ycol in self.y]):
raise ValueError('y columns must be real')
unique_x_measures = set(in_dshape.measure[str(xcol)]
for xcol in self.x)
if len(unique_x_measures) > 1:
raise ValueError('x columns must have the same data type')
unique_y_measures = set(in_dshape.measure[str(ycol)]
for ycol in self.y)
if len(unique_y_measures) > 1:
raise ValueError('y columns must have the same data type')
def required_columns(self):
return self.x + self.y
@property
def x_label(self):
return 'x'
@property
def y_label(self):
return 'y'
def compute_x_bounds(self, df):
xs = tuple(df[xlabel] for xlabel in self.x)
bounds_list = [self._compute_bounds(xcol) for xcol in xs]
mins, maxes = zip(*bounds_list)
return self.maybe_expand_bounds((min(mins), max(maxes)))
def compute_y_bounds(self, df):
ys = tuple(df[ylabel] for ylabel in self.y)
bounds_list = [self._compute_bounds(ycol) for ycol in ys]
mins, maxes = zip(*bounds_list)
return self.maybe_expand_bounds((min(mins), max(maxes)))
@memoize
def compute_bounds_dask(self, ddf):
r = ddf.map_partitions(lambda df: np.array([[
np.nanmin([np.nanmin(df[c].values).item() for c in self.x]),
np.nanmax([np.nanmax(df[c].values).item() for c in self.x]),
np.nanmin([np.nanmin(df[c].values).item() for c in self.y]),
np.nanmax([np.nanmax(df[c].values).item() for c in self.y])]]
)).compute()
x_extents = np.nanmin(r[:, 0]), np.nanmax(r[:, 1])
y_extents = np.nanmin(r[:, 2]), np.nanmax(r[:, 3])
return (self.maybe_expand_bounds(x_extents),
self.maybe_expand_bounds(y_extents))
@memoize
def _internal_build_extend(
self, x_mapper, y_mapper, info, append, line_width, antialias_stage_2):
antialias = line_width > 0
expand_aggs_and_cols = self.expand_aggs_and_cols(append)
map_onto_pixel = _build_map_onto_pixel_for_line(
x_mapper, y_mapper, antialias)
overwrite, use_2_stage_agg = _two_stage_agg(antialias_stage_2)
draw_segment = _build_draw_segment(
append, map_onto_pixel, expand_aggs_and_cols, line_width, overwrite
)
extend_cpu, extend_cuda = _build_extend_line_axis1_none_constant(
draw_segment, expand_aggs_and_cols, use_2_stage_agg
)
x_names = self.x
y_names = self.y
def extend(aggs, df, vt, bounds, plot_start=True):
sx, tx, sy, ty = vt
xmin, xmax, ymin, ymax = bounds
aggs_and_cols = aggs + info(df)
if cudf and isinstance(df, cudf.DataFrame):
xs = self.to_cupy_array(df, x_names)
ys = self.to_cupy_array(df, y_names)
do_extend = extend_cuda[cuda_args(xs.shape)]
else:
xs = df.loc[:, list(x_names)].to_numpy()
ys = df.loc[:, list(y_names)].to_numpy()
do_extend = extend_cpu
do_extend(
sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys, antialias_stage_2, *aggs_and_cols
)
return extend
[docs]class LinesAxis1XConstant(LinesAxis1):
"""
"""
def validate(self, in_dshape):
if not all([isreal(in_dshape.measure[str(ycol)]) for ycol in self.y]):
raise ValueError('y columns must be real')
unique_y_measures = set(in_dshape.measure[str(ycol)]
for ycol in self.y)
if len(unique_y_measures) > 1:
raise ValueError('y columns must have the same data type')
def required_columns(self):
return self.y
def compute_x_bounds(self, *args):
x_min = np.nanmin(self.x)
x_max = np.nanmax(self.x)
return self.maybe_expand_bounds((x_min, x_max))
@memoize
def compute_bounds_dask(self, ddf):
r = ddf.map_partitions(lambda df: np.array([[
np.nanmin([np.nanmin(df[c].values).item() for c in self.y]),
np.nanmax([np.nanmax(df[c].values).item() for c in self.y])]]
)).compute()
y_extents = np.nanmin(r[:, 0]), np.nanmax(r[:, 1])
return (self.compute_x_bounds(),
self.maybe_expand_bounds(y_extents))
@memoize
def _internal_build_extend(
self, x_mapper, y_mapper, info, append, line_width, antialias_stage_2):
antialias = line_width > 0
expand_aggs_and_cols = self.expand_aggs_and_cols(append)
map_onto_pixel = _build_map_onto_pixel_for_line(
x_mapper, y_mapper, antialias)
overwrite, use_2_stage_agg = _two_stage_agg(antialias_stage_2)
draw_segment = _build_draw_segment(
append, map_onto_pixel, expand_aggs_and_cols, line_width, overwrite
)
extend_cpu, extend_cuda = _build_extend_line_axis1_x_constant(
draw_segment, expand_aggs_and_cols, use_2_stage_agg
)
x_values = self.x
y_names = self.y
def extend(aggs, df, vt, bounds, plot_start=True):
sx, tx, sy, ty = vt
xmin, xmax, ymin, ymax = bounds
aggs_and_cols = aggs + info(df)
if cudf and isinstance(df, cudf.DataFrame):
xs = cp.asarray(x_values)
ys = self.to_cupy_array(df, y_names)
do_extend = extend_cuda[cuda_args(ys.shape)]
else:
xs = x_values
ys = df.loc[:, list(y_names)].to_numpy()
do_extend = extend_cpu
do_extend(
sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys, antialias_stage_2, *aggs_and_cols
)
return extend
[docs]class LinesAxis1YConstant(LinesAxis1):
"""
"""
def validate(self, in_dshape):
if not all([isreal(in_dshape.measure[str(xcol)]) for xcol in self.x]):
raise ValueError('x columns must be real')
unique_x_measures = set(in_dshape.measure[str(xcol)]
for xcol in self.x)
if len(unique_x_measures) > 1:
raise ValueError('x columns must have the same data type')
def required_columns(self):
return self.x
def compute_y_bounds(self, *args):
y_min = np.nanmin(self.y)
y_max = np.nanmax(self.y)
return self.maybe_expand_bounds((y_min, y_max))
@memoize
def compute_bounds_dask(self, ddf):
r = ddf.map_partitions(lambda df: np.array([[
np.nanmin([np.nanmin(df[c].values).item() for c in self.x]),
np.nanmax([np.nanmax(df[c].values).item() for c in self.x])]]
)).compute()
x_extents = np.nanmin(r[:, 0]), np.nanmax(r[:, 1])
return (self.maybe_expand_bounds(x_extents),
self.compute_y_bounds())
@memoize
def _internal_build_extend(
self, x_mapper, y_mapper, info, append, line_width, antialias_stage_2):
antialias = line_width > 0
expand_aggs_and_cols = self.expand_aggs_and_cols(append)
map_onto_pixel = _build_map_onto_pixel_for_line(
x_mapper, y_mapper, antialias)
overwrite, use_2_stage_agg = _two_stage_agg(antialias_stage_2)
draw_segment = _build_draw_segment(
append, map_onto_pixel, expand_aggs_and_cols, line_width, overwrite
)
extend_cpu, extend_cuda = _build_extend_line_axis1_y_constant(
draw_segment, expand_aggs_and_cols, use_2_stage_agg
)
x_names = self.x
y_values = self.y
def extend(aggs, df, vt, bounds, plot_start=True):
sx, tx, sy, ty = vt
xmin, xmax, ymin, ymax = bounds
aggs_and_cols = aggs + info(df)
if cudf and isinstance(df, cudf.DataFrame):
xs = self.to_cupy_array(df, x_names)
ys = cp.asarray(y_values)
do_extend = extend_cuda[cuda_args(xs.shape)]
else:
xs = df.loc[:, list(x_names)].to_numpy()
ys = y_values
do_extend = extend_cpu
do_extend(
sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys, antialias_stage_2, *aggs_and_cols
)
return extend
[docs]class LinesAxis1Ragged(_PointLike, _AntiAliasedLine):
def validate(self, in_dshape):
try:
from datashader.datatypes import RaggedDtype
except ImportError:
RaggedDtype = type(None)
if not isinstance(in_dshape[str(self.x)], RaggedDtype):
raise ValueError('x must be a RaggedArray')
elif not isinstance(in_dshape[str(self.y)], RaggedDtype):
raise ValueError('y must be a RaggedArray')
def required_columns(self):
return (self.x,) + (self.y,)
def compute_x_bounds(self, df):
bounds = self._compute_bounds(df[self.x].array.flat_array)
return self.maybe_expand_bounds(bounds)
def compute_y_bounds(self, df):
bounds = self._compute_bounds(df[self.y].array.flat_array)
return self.maybe_expand_bounds(bounds)
@memoize
def compute_bounds_dask(self, ddf):
r = ddf.map_partitions(lambda df: np.array([[
np.nanmin(df[self.x].array.flat_array).item(),
np.nanmax(df[self.x].array.flat_array).item(),
np.nanmin(df[self.y].array.flat_array).item(),
np.nanmax(df[self.y].array.flat_array).item()]]
)).compute()
x_extents = np.nanmin(r[:, 0]), np.nanmax(r[:, 1])
y_extents = np.nanmin(r[:, 2]), np.nanmax(r[:, 3])
return (self.maybe_expand_bounds(x_extents),
self.maybe_expand_bounds(y_extents))
@memoize
def _internal_build_extend(
self, x_mapper, y_mapper, info, append, line_width, antialias_stage_2):
antialias = line_width > 0
expand_aggs_and_cols = self.expand_aggs_and_cols(append)
map_onto_pixel = _build_map_onto_pixel_for_line(
x_mapper, y_mapper, antialias)
overwrite, use_2_stage_agg = _two_stage_agg(antialias_stage_2)
draw_segment = _build_draw_segment(
append, map_onto_pixel, expand_aggs_and_cols, line_width, overwrite
)
extend_cpu = _build_extend_line_axis1_ragged(
draw_segment, expand_aggs_and_cols, use_2_stage_agg
)
x_name = self.x
y_name = self.y
def extend(aggs, df, vt, bounds, plot_start=True):
sx, tx, sy, ty = vt
xmin, xmax, ymin, ymax = bounds
xs = df[x_name].array
ys = df[y_name].array
aggs_and_cols = aggs + info(df)
# line may be clipped, then mapped to pixels
extend_cpu(
sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys, antialias_stage_2, *aggs_and_cols
)
return extend
[docs]class LineAxis1Geometry(_GeometryLike, _AntiAliasedLine):
# spatialpandas must be available if a LineAxis1Geometry object is created.
@property
def geom_dtypes(self):
from spatialpandas.geometry import (
LineDtype, MultiLineDtype, RingDtype, PolygonDtype,
MultiPolygonDtype
)
return (LineDtype, MultiLineDtype, RingDtype,
PolygonDtype, MultiPolygonDtype)
@memoize
def _internal_build_extend(
self, x_mapper, y_mapper, info, append, line_width, antialias_stage_2):
from spatialpandas.geometry import (
PolygonArray, MultiPolygonArray, RingArray
)
antialias = line_width > 0
expand_aggs_and_cols = self.expand_aggs_and_cols(append)
map_onto_pixel = _build_map_onto_pixel_for_line(
x_mapper, y_mapper, antialias)
overwrite, use_2_stage_agg = _two_stage_agg(antialias_stage_2)
draw_segment = _build_draw_segment(
append, map_onto_pixel, expand_aggs_and_cols, line_width, overwrite
)
perform_extend_cpu = _build_extend_line_axis1_geometry(
draw_segment, expand_aggs_and_cols, use_2_stage_agg
)
geometry_name = self.geometry
def extend(aggs, df, vt, bounds, plot_start=True):
sx, tx, sy, ty = vt
xmin, xmax, ymin, ymax = bounds
aggs_and_cols = aggs + info(df)
geom_array = df[geometry_name].array
# Use type to decide whether geometry represents a closed .
# We skip for closed geometries so as not to double count the first/last
# pixel
if isinstance(geom_array, (PolygonArray, MultiPolygonArray)):
# Convert polygon array to multi line of boundary
geom_array = geom_array.boundary
closed_rings = True
elif isinstance(geom_array, RingArray):
closed_rings = True
else:
closed_rings = False
perform_extend_cpu(
sx, tx, sy, ty, xmin, xmax, ymin, ymax,
geom_array, closed_rings, antialias_stage_2, *aggs_and_cols
)
return extend
def _build_map_onto_pixel_for_line(x_mapper, y_mapper, want_antialias=False):
@ngjit
def map_onto_pixel_snap(sx, tx, sy, ty, xmin, xmax, ymin, ymax, x, y):
"""Map points onto pixel grid.
Points falling on upper bound are mapped into previous bin.
If the line has been clipped, x and y will have been
computed to lie on the bounds; we compare point and bounds
in integer space to avoid fp error. In contrast, with
auto-ranging, a point on the bounds will be the same
floating point number as the bound, so comparison in fp
representation of continuous space or in integer space
doesn't change anything.
"""
xx = int(x_mapper(x) * sx + tx)
yy = int(y_mapper(y) * sy + ty)
# Note that sx and tx were designed so that
# x_mapper(xmax) * sx + tx equals the width of the canvas in pixels
#
# Likewise, sy and ty were designed so that
# y_mapper(ymax) * sy + ty equals the height of the canvas in pixels
#
# We round these results to integers (rather than casting to integers
# with the int constructor) to handle cases where floating-point
# precision errors results in a value just under the integer number
# of pixels.
xxmax = round(x_mapper(xmax) * sx + tx)
yymax = round(y_mapper(ymax) * sy + ty)
return (xx - 1 if xx == xxmax else xx,
yy - 1 if yy == yymax else yy)
@ngjit
def map_onto_pixel_no_snap(sx, tx, sy, ty, xmin, xmax, ymin, ymax, x, y):
xx = x_mapper(x)*sx + tx - 0.5
yy = y_mapper(y)*sy + ty - 0.5
return xx, yy
if want_antialias:
return map_onto_pixel_no_snap
else:
return map_onto_pixel_snap
@ngjit
def _liang_barsky(xmin, xmax, ymin, ymax, x0, x1, y0, y1, skip):
""" An implementation of the Liang-Barsky line clipping algorithm.
https://en.wikipedia.org/wiki/Liang%E2%80%93Barsky_algorithm
"""
# Check if line is fully outside viewport
if x0 < xmin and x1 < xmin:
skip = True
elif x0 > xmax and x1 > xmax:
skip = True
elif y0 < ymin and y1 < ymin:
skip = True
elif y0 > ymax and y1 > ymax:
skip = True
t0, t1 = 0, 1
dx1 = x1 - x0
t0, t1, accept = _clipt(-dx1, x0 - xmin, t0, t1)
if not accept:
skip = True
t0, t1, accept = _clipt(dx1, xmax - x0, t0, t1)
if not accept:
skip = True
dy1 = y1 - y0
t0, t1, accept = _clipt(-dy1, y0 - ymin, t0, t1)
if not accept:
skip = True
t0, t1, accept = _clipt(dy1, ymax - y0, t0, t1)
if not accept:
skip = True
if t1 < 1:
clipped_end = True
x1 = x0 + t1 * dx1
y1 = y0 + t1 * dy1
else:
clipped_end = False
if t0 > 0:
# If x0 is clipped, we need to plot the new start
clipped_start = True
x0 = x0 + t0 * dx1
y0 = y0 + t0 * dy1
else:
clipped_start = False
return x0, x1, y0, y1, skip, clipped_start, clipped_end
@ngjit
def _clipt(p, q, t0, t1):
accept = True
if p < 0 and q < 0:
r = q / p
if r > t1:
accept = False
elif r > t0:
t0 = r
elif p > 0 and q < p:
r = q / p
if r < t0:
accept = False
elif r < t1:
t1 = r
elif q < 0:
accept = False
return t0, t1, accept
@ngjit
def _clamp(x, low, high):
# Clamp ``x`` in the range ``low`` to ``high``.
return max(low, min(x, high))
@ngjit
def _linearstep(edge0, edge1, x):
t = _clamp((x - edge0) / (edge1 - edge0), 0.0, 1.0)
return t
@ngjit
def _x_intercept(y, cx0, cy0, cx1, cy1):
# Return x value of intercept between line at constant y and line
# between corner points.
if cy0 == cy1:
# Line is horizontal, return the "upper", i.e. right-hand, end of it.
return cx1
frac = (y - cy0) / (cy1 - cy0) # In range 0..1
return cx0 + frac*(cx1 - cx0)
def _build_full_antialias(expand_aggs_and_cols):
"""Specialize antialiased line drawing algorithm for a given append/axis combination"""
@ngjit
@expand_aggs_and_cols
def _full_antialias(line_width, overwrite, i, x0, x1, y0, y1,
segment_start, segment_end, xm, ym, append,
nx, ny, buffer, *aggs_and_cols):
"""Draw an antialiased line segment.
If overwrite=True can overwrite each pixel multiple times because
using max for the overwriting. If False can only write each pixel
once per segment and its previous segment.
Argument xm, ym are only valid if overwrite and segment_start are False.
"""
if x0 == x1 and y0 == y1:
return
# Scan occurs in y-direction. But wish to scan in the shortest direction,
# so if |x0-x1| < |y0-y1| then flip (x,y) coords for maths and flip back
# again before setting pixels.
flip_xy = abs(x0-x1) < abs(y0-y1)
if flip_xy:
x0, y0 = y0, x0
x1, y1 = y1, x1
xm, ym = ym, xm
scale = 1.0
# line_width less than 1 is rendered as 1 but with lower intensity.
if line_width < 1.0:
scale *= line_width
line_width = 1.0
aa = 1.0
halfwidth = 0.5*(line_width + aa)
# Want y0 <= y1, so switch vertical direction if this is not so.
flip_order = y1 < y0 or (y1 == y0 and x1 < x0)
# Start (x0, y0), end (y0, y1)
# c1 +-------------+ c2 along | right
# (x0, y0) | o o | (x1, y1) vector | vector
# c0 +-------------+ c3 ----> v
alongx = float(x1 - x0)
alongy = float(y1 - y0) # Always +ve
length = math.sqrt(alongx**2 + alongy**2)
alongx /= length
alongy /= length
rightx = alongy
righty = -alongx
# 4 corners, x and y. Uses buffer, which must have length 8. Order of coords is
# (x0, x1, x2, x3, y0, y1, y2, y3). Each CPU/GPU thread has its own local buffer
# so there is no cross-talk. Contents of buffer are written and read within the
# lifetime of this function, so it doesn't matter what they are before this
# function is called or after it returns.
if flip_order:
buffer[0] = x1 - halfwidth*( rightx - alongx)
buffer[1] = x1 - halfwidth*(-rightx - alongx)
buffer[2] = x0 - halfwidth*(-rightx + alongx)
buffer[3] = x0 - halfwidth*( rightx + alongx)
buffer[4] = y1 - halfwidth*( righty - alongy)
buffer[5] = y1 - halfwidth*(-righty - alongy)
buffer[6] = y0 - halfwidth*(-righty + alongy)
buffer[7] = y0 - halfwidth*( righty + alongy)
else:
buffer[0] = x0 + halfwidth*( rightx - alongx)
buffer[1] = x0 + halfwidth*(-rightx - alongx)
buffer[2] = x1 + halfwidth*(-rightx + alongx)
buffer[3] = x1 + halfwidth*( rightx + alongx)
buffer[4] = y0 + halfwidth*( righty - alongy)
buffer[5] = y0 + halfwidth*(-righty - alongy)
buffer[6] = y1 + halfwidth*(-righty + alongy)
buffer[7] = y1 + halfwidth*( righty + alongy)
xmax = nx-1
ymax = ny-1
if flip_xy:
xmax, ymax = ymax, xmax
# Index of lowest-y point.
if flip_order:
lowindex = 0 if x0 > x1 else 1
else:
lowindex = 0 if x1 > x0 else 1
if not overwrite and not segment_start:
prev_alongx = x0 - xm
prev_alongy = y0 - ym
prev_length = math.sqrt(prev_alongx**2 + prev_alongy**2)
if prev_length > 0.0:
prev_alongx /= prev_length
prev_alongy /= prev_length
prev_rightx = prev_alongy
prev_righty = -prev_alongx
else:
overwrite = True
# y limits of scan.
ystart = _clamp(math.ceil(buffer[4 + lowindex]), 0, ymax)
yend = _clamp(math.floor(buffer[4 + (lowindex+2) % 4]), 0, ymax)
# Need to know which edges are to left and right; both will change.
ll = lowindex # Index of lower point of left edge.
lu = (ll + 1) % 4 # Index of upper point of left edge.
rl = lowindex # Index of lower point of right edge.
ru = (rl + 3) % 4 # Index of upper point of right edge.
for y in range(ystart, yend+1):
if ll == lowindex and y > buffer[4 + lu]:
ll = lu
lu = (ll + 1) % 4
if rl == lowindex and y > buffer[4 + ru]:
rl = ru
ru = (rl + 3) % 4
# Find x limits of scan at this y.
xleft = _clamp(math.ceil(_x_intercept(
y, buffer[ll], buffer[4+ll], buffer[lu], buffer[4+lu])), 0, xmax)
xright = _clamp(math.floor(_x_intercept(
y, buffer[rl], buffer[4+rl], buffer[ru], buffer[4+ru])), 0, xmax)
for x in range(xleft, xright+1):
along = (x-x0)*alongx + (y-y0)*alongy # dot product
prev_correction = False
if along < 0.0:
# Before start of segment
if overwrite or segment_start or (x-x0)*prev_alongx + (y-y0)*prev_alongy > 0.0:
distance = math.sqrt((x-x0)**2 + (y-y0)**2) # round join/end cap
else:
continue
elif along > length:
# After end of segment
if overwrite or segment_end:
distance = math.sqrt((x-x1)**2 + (y-y1)**2) # round join/end cap
else:
continue
else:
# Within segment
distance = abs((x-x0)*rightx + (y-y0)*righty)
if not overwrite and not segment_start and \
-prev_length <= (x-x0)*prev_alongx + (y-y0)*prev_alongy <= 0.0 and \
abs((x-x0)*prev_rightx + (y-y0)*prev_righty) <= halfwidth:
prev_correction = True
value = 1.0 - _linearstep(0.5*(line_width - aa), halfwidth, distance)
value *= scale
if prev_correction:
# Already set pixel from previous segment, need to correct it
prev_distance = abs((x-x0)*prev_rightx + (y-y0)*prev_righty)
prev_value = 1.0 - _linearstep(0.5*(line_width - aa), halfwidth, prev_distance)
prev_value *= scale
value = value - prev_value # Correction from previous segment.
if value > 0.0:
xx, yy = (y, x) if flip_xy else (x, y)
append(i, xx, yy, value, *aggs_and_cols)
return _full_antialias
def _build_bresenham(expand_aggs_and_cols):
"""Specialize a bresenham kernel for a given append/axis combination"""
@ngjit
@expand_aggs_and_cols
def _bresenham(i, sx, tx, sy, ty, xmin, xmax, ymin, ymax, segment_start,
x0, x1, y0, y1, clipped, append, *aggs_and_cols):
"""Draw a line segment using Bresenham's algorithm
This method plots a line segment with integer coordinates onto a pixel
grid.
"""
dx = x1 - x0
ix = (dx > 0) - (dx < 0)
dx = abs(dx) * 2
dy = y1 - y0
iy = (dy > 0) - (dy < 0)
dy = abs(dy) * 2
# If vertices weren't clipped and are concurrent in integer space,
# call append and return, so that the second vertex won't be hit below.
if not clipped and not (dx | dy):
append(i, x0, y0, *aggs_and_cols)
return
if segment_start:
append(i, x0, y0, *aggs_and_cols)
if dx >= dy:
error = 2 * dy - dx
while x0 != x1:
if error >= 0 and (error or ix > 0):
error -= 2 * dx
y0 += iy
error += 2 * dy
x0 += ix
append(i, x0, y0, *aggs_and_cols)
else:
error = 2 * dx - dy
while y0 != y1:
if error >= 0 and (error or iy > 0):
error -= 2 * dy
x0 += ix
error += 2 * dx
y0 += iy
append(i, x0, y0, *aggs_and_cols)
return _bresenham
def _build_draw_segment(append, map_onto_pixel, expand_aggs_and_cols, line_width, overwrite):
"""Specialize a line plotting kernel for a given append/axis combination"""
if line_width > 0.0:
_bresenham = None
_full_antialias = _build_full_antialias(expand_aggs_and_cols)
else:
_bresenham = _build_bresenham(expand_aggs_and_cols)
_full_antialias = None
@ngjit
@expand_aggs_and_cols
def draw_segment(
i, sx, tx, sy, ty, xmin, xmax, ymin, ymax, segment_start, segment_end,
x0, x1, y0, y1, xm, ym, buffer, *aggs_and_cols
):
# xm, ym are only valid if segment_start is True.
# buffer is a length-8 float64 array if antialiasing is to be used,
# or None otherwise. It is allocated in the appropriate extend_cpu or
# extend_cuda function so that it is of the correct type (numpy or
# cupy) and that there is one per CPU/GPU thread.
# NOTE: The slightly bizarre variable versioning herein for variables
# x0, y0, y0, y1 is to deal with Numba not having SSA form prior to
# version 0.49.0. The result of lack of SSA is that the type inference
# algorithms would widen types that are multiply defined as would be the
# case in code such as `x, y = function(x, y)` if the function returned
# a wider type for x, y then the input x, y.
skip = False
# If any of the coordinates are NaN, there's a discontinuity.
# Skip the entire segment.
if isnull(x0) or isnull(y0) or isnull(x1) or isnull(y1):
skip = True
# Use Liang-Barsky to clip the segment to a bounding box
x0_1, x1_1, y0_1, y1_1, skip, clipped_start, clipped_end = \
_liang_barsky(xmin, xmax, ymin, ymax, x0, x1, y0, y1, skip)
if not skip:
clipped = clipped_start or clipped_end
segment_start = segment_start or clipped_start
x0_2, y0_2 = map_onto_pixel(
sx, tx, sy, ty, xmin, xmax, ymin, ymax, x0_1, y0_1
)
x1_2, y1_2 = map_onto_pixel(
sx, tx, sy, ty, xmin, xmax, ymin, ymax, x1_1, y1_1
)
if line_width > 0.0:
if segment_start:
xm_2 = ym_2 = 0.0
else:
xm_2, ym_2 = map_onto_pixel(
sx, tx, sy, ty, xmin, xmax, ymin, ymax, xm, ym)
nx = round((xmax - xmin)*sx)
ny = round((ymax - ymin)*sy)
_full_antialias(line_width, overwrite, i, x0_2, x1_2, y0_2, y1_2,
segment_start, segment_end, xm_2, ym_2, append,
nx, ny, buffer, *aggs_and_cols)
else:
_bresenham(i, sx, tx, sy, ty, xmin, xmax, ymin, ymax,
segment_start, x0_2, x1_2, y0_2, y1_2,
clipped, append, *aggs_and_cols)
return draw_segment
def _build_extend_line_axis0(draw_segment, expand_aggs_and_cols, use_2_stage_agg):
@ngjit
@expand_aggs_and_cols
def perform_extend_line(i, sx, tx, sy, ty, xmin, xmax, ymin, ymax,
plot_start, xs, ys, buffer, *aggs_and_cols):
x0 = xs[i]
y0 = ys[i]
x1 = xs[i + 1]
y1 = ys[i + 1]
segment_start = (plot_start if i == 0 else
(isnull(xs[i - 1]) or isnull(ys[i - 1])))
segment_end = (i == len(xs)-2) or isnull(xs[i+2]) or isnull(ys[i+2])
if segment_start or use_2_stage_agg:
xm = 0.0
ym = 0.0
else:
xm = xs[i-1]
ym = ys[i-1]
draw_segment(i, sx, tx, sy, ty, xmin, xmax, ymin, ymax,
segment_start, segment_end, x0, x1, y0, y1,
xm, ym, buffer, *aggs_and_cols)
@ngjit
@expand_aggs_and_cols
def extend_cpu(sx, tx, sy, ty, xmin, xmax, ymin, ymax,
xs, ys, plot_start, antialias_stage_2, *aggs_and_cols):
"""Aggregate along a line formed by ``xs`` and ``ys``"""
antialias = antialias_stage_2 is not None
buffer = np.empty(8) if antialias else None
nrows = xs.shape[0]
for i in range(nrows - 1):
perform_extend_line(i, sx, tx, sy, ty, xmin, xmax, ymin, ymax,
plot_start, xs, ys, buffer, *aggs_and_cols)
@cuda.jit
@expand_aggs_and_cols
def extend_cuda(sx, tx, sy, ty, xmin, xmax, ymin, ymax,
xs, ys, plot_start, antialias_stage_2, *aggs_and_cols):
antialias = antialias_stage_2 is not None
buffer = cuda.local.array(8, nb_types.float64) if antialias else None
i = cuda.grid(1)
if i < xs.shape[0] - 1:
perform_extend_line(i, sx, tx, sy, ty, xmin, xmax, ymin, ymax,
plot_start, xs, ys, buffer, *aggs_and_cols)
return extend_cpu, extend_cuda
@ngjit
def _combine_in_place(accum_agg, other_agg, antialias_combination):
if antialias_combination == AntialiasCombination.MAX:
nanmax_in_place(accum_agg, other_agg)
elif antialias_combination == AntialiasCombination.MIN:
nanmin_in_place(accum_agg, other_agg)
elif antialias_combination == AntialiasCombination.FIRST:
nanfirst_in_place(accum_agg, other_agg)
elif antialias_combination == AntialiasCombination.LAST:
nanlast_in_place(accum_agg, other_agg)
else:
nansum_in_place(accum_agg, other_agg)
@ngjit
def _aa_stage_2_accumulate(aggs_and_copies, antialias_combinations):
k = 0
# Numba access to heterogeneous tuples is only permitted using literal_unroll.
for agg_and_copy in literal_unroll(aggs_and_copies):
_combine_in_place(agg_and_copy[1], agg_and_copy[0], antialias_combinations[k])
k += 1
@ngjit
def _aa_stage_2_clear(aggs_and_copies, antialias_zeroes):
k = 0
# Numba access to heterogeneous tuples is only permitted using literal_unroll.
for agg_and_copy in literal_unroll(aggs_and_copies):
parallel_fill(agg_and_copy[0], antialias_zeroes[k])
k += 1
@ngjit
def _aa_stage_2_copy_back(aggs_and_copies):
k = 0
# Numba access to heterogeneous tuples is only permitted using literal_unroll.
for agg_and_copy in literal_unroll(aggs_and_copies):
agg_and_copy[0][:] = agg_and_copy[1][:]
k += 1
def _build_extend_line_axis0_multi(draw_segment, expand_aggs_and_cols, use_2_stage_agg):
@ngjit
@expand_aggs_and_cols
def perform_extend_line(i, j, sx, tx, sy, ty, xmin, xmax, ymin, ymax,
plot_start, xs, ys, buffer, *aggs_and_cols):
x0 = xs[i, j]
y0 = ys[i, j]
x1 = xs[i + 1, j]
y1 = ys[i + 1, j]
segment_start = (plot_start if i == 0 else
(isnull(xs[i - 1, j]) or isnull(ys[i - 1, j])))
segment_end = (i == len(xs)-2) or isnull(xs[i+2, j]) or isnull(ys[i+2, j])
if segment_start or use_2_stage_agg:
xm = 0.0
ym = 0.0
else:
xm = xs[i-1, j]
ym = ys[i-1, j]
draw_segment(i, sx, tx, sy, ty, xmin, xmax, ymin, ymax,
segment_start, segment_end, x0, x1, y0, y1,
xm, ym, buffer, *aggs_and_cols)
@ngjit
@expand_aggs_and_cols
def extend_cpu(sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys,
plot_start, antialias_stage_2, *aggs_and_cols):
"""Aggregate along a line formed by ``xs`` and ``ys``"""
antialias = antialias_stage_2 is not None
buffer = np.empty(8) if antialias else None
nrows, ncols = xs.shape
for j in range(ncols):
for i in range(nrows - 1):
perform_extend_line(i, j, sx, tx, sy, ty, xmin, xmax, ymin, ymax,
plot_start, xs, ys, buffer, *aggs_and_cols)
def extend_cpu_antialias_2agg(sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys,
plot_start, antialias_stage_2, *aggs_and_cols):
"""Aggregate along a line formed by ``xs`` and ``ys``"""
n_aggs = len(antialias_stage_2[0])
aggs_and_accums = tuple((agg, agg.copy()) for agg in aggs_and_cols[:n_aggs])
cpu_antialias_2agg_impl(sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys,
plot_start, antialias_stage_2, aggs_and_accums, *aggs_and_cols)
@ngjit
@expand_aggs_and_cols
def cpu_antialias_2agg_impl(sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys,
plot_start, antialias_stage_2, aggs_and_accums, *aggs_and_cols):
antialias = antialias_stage_2 is not None
buffer = np.empty(8) if antialias else None
antialias_combinations, antialias_zeroes = antialias_stage_2
nrows, ncols = xs.shape
for j in range(ncols):
for i in range(nrows - 1):
perform_extend_line(i, j, sx, tx, sy, ty, xmin, xmax, ymin, ymax,
plot_start, xs, ys, buffer, *aggs_and_cols)
if ncols == 1:
return
_aa_stage_2_accumulate(aggs_and_accums, antialias_combinations)
if j < ncols - 1:
_aa_stage_2_clear(aggs_and_accums, antialias_zeroes)
_aa_stage_2_copy_back(aggs_and_accums)
@cuda.jit
@expand_aggs_and_cols
def extend_cuda(sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys,
plot_start, antialias_stage_2, *aggs_and_cols):
antialias = antialias_stage_2 is not None
buffer = cuda.local.array(8, nb_types.float64) if antialias else None
i, j = cuda.grid(2)
if i < xs.shape[0] - 1 and j < xs.shape[1]:
perform_extend_line(i, j, sx, tx, sy, ty, xmin, xmax, ymin, ymax,
plot_start, xs, ys, buffer, *aggs_and_cols)
if use_2_stage_agg:
return extend_cpu_antialias_2agg, extend_cuda
else:
return extend_cpu, extend_cuda
def _build_extend_line_axis1_none_constant(draw_segment, expand_aggs_and_cols, use_2_stage_agg):
@ngjit
@expand_aggs_and_cols
def perform_extend_line(
i, j, sx, tx, sy, ty, xmin, xmax, ymin, ymax,
xs, ys, buffer, *aggs_and_cols
):
x0 = xs[i, j]
y0 = ys[i, j]
x1 = xs[i, j + 1]
y1 = ys[i, j + 1]
segment_start = (
(j == 0) or isnull(xs[i, j - 1]) or isnull(ys[i, j - 1])
)
segment_end = (j == xs.shape[1]-2) or isnull(xs[i, j+2]) or isnull(ys[i, j+2])
if segment_start or use_2_stage_agg:
xm = 0.0
ym = 0.0
else:
xm = xs[i, j-1]
ym = ys[i, j-1]
draw_segment(i, sx, tx, sy, ty, xmin, xmax, ymin, ymax,
segment_start, segment_end, x0, x1, y0, y1,
xm, ym, buffer, *aggs_and_cols)
@ngjit
@expand_aggs_and_cols
def extend_cpu(sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys, antialias_stage_2, *aggs_and_cols):
antialias = antialias_stage_2 is not None
buffer = np.empty(8) if antialias else None
ncols = xs.shape[1]
for i in range(xs.shape[0]):
for j in range(ncols - 1):
perform_extend_line(
i, j, sx, tx, sy, ty, xmin, xmax, ymin, ymax,
xs, ys, buffer, *aggs_and_cols
)
def extend_cpu_antialias_2agg(sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys,
antialias_stage_2, *aggs_and_cols):
n_aggs = len(antialias_stage_2[0])
aggs_and_accums = tuple((agg, agg.copy()) for agg in aggs_and_cols[:n_aggs])
cpu_antialias_2agg_impl(sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys,
antialias_stage_2, aggs_and_accums, *aggs_and_cols)
@ngjit
@expand_aggs_and_cols
def cpu_antialias_2agg_impl(sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys,
antialias_stage_2, aggs_and_accums, *aggs_and_cols):
antialias = antialias_stage_2 is not None
buffer = np.empty(8) if antialias else None
antialias_combinations, antialias_zeroes = antialias_stage_2
ncols = xs.shape[1]
for i in range(xs.shape[0]):
for j in range(ncols - 1):
perform_extend_line(i, j, sx, tx, sy, ty, xmin, xmax, ymin, ymax,
xs, ys, buffer, *aggs_and_cols)
if xs.shape[0] == 1:
return
_aa_stage_2_accumulate(aggs_and_accums, antialias_combinations)
if i < xs.shape[0] - 1:
_aa_stage_2_clear(aggs_and_accums, antialias_zeroes)
_aa_stage_2_copy_back(aggs_and_accums)
@cuda.jit
@expand_aggs_and_cols
def extend_cuda(sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys, antialias_stage_2, *aggs_and_cols):
antialias = antialias_stage_2 is not None
buffer = cuda.local.array(8, nb_types.float64) if antialias else None
i, j = cuda.grid(2)
if i < xs.shape[0] and j < xs.shape[1] - 1:
perform_extend_line(
i, j, sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys,
buffer, *aggs_and_cols
)
if use_2_stage_agg:
return extend_cpu_antialias_2agg, extend_cuda
else:
return extend_cpu, extend_cuda
def _build_extend_line_axis1_x_constant(draw_segment, expand_aggs_and_cols, use_2_stage_agg):
@ngjit
@expand_aggs_and_cols
def perform_extend_line(
i, j, sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys, buffer, *aggs_and_cols
):
x0 = xs[j]
y0 = ys[i, j]
x1 = xs[j + 1]
y1 = ys[i, j + 1]
segment_start = (
(j == 0) or isnull(xs[j - 1]) or isnull(ys[i, j - 1])
)
segment_end = (j == len(xs)-2) or isnull(xs[j+2]) or isnull(ys[i, j+2])
if segment_start or use_2_stage_agg:
xm = 0.0
ym = 0.0
else:
xm = xs[j-1]
ym = ys[i, j-1]
draw_segment(i, sx, tx, sy, ty, xmin, xmax, ymin, ymax,
segment_start, segment_end, x0, x1, y0, y1,
xm, ym, buffer, *aggs_and_cols)
@ngjit
@expand_aggs_and_cols
def extend_cpu(sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys, antialias_stage_2, *aggs_and_cols):
antialias = antialias_stage_2 is not None
buffer = np.empty(8) if antialias else None
ncols = ys.shape[1]
for i in range(ys.shape[0]):
for j in range(ncols - 1):
perform_extend_line(
i, j, sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys, buffer, *aggs_and_cols
)
def extend_cpu_antialias_2agg(sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys,
antialias_stage_2, *aggs_and_cols):
n_aggs = len(antialias_stage_2[0])
aggs_and_accums = tuple((agg, agg.copy()) for agg in aggs_and_cols[:n_aggs])
cpu_antialias_2agg_impl(sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys,
antialias_stage_2, aggs_and_accums, *aggs_and_cols)
@ngjit
@expand_aggs_and_cols
def cpu_antialias_2agg_impl(sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys,
antialias_stage_2, aggs_and_accums, *aggs_and_cols):
antialias = antialias_stage_2 is not None
buffer = np.empty(8) if antialias else None
antialias_combinations, antialias_zeroes = antialias_stage_2
ncols = ys.shape[1]
for i in range(ys.shape[0]):
for j in range(ncols - 1):
perform_extend_line(
i, j, sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys,
buffer, *aggs_and_cols
)
if ys.shape[0] == 1:
return
_aa_stage_2_accumulate(aggs_and_accums, antialias_combinations)
if i < ys.shape[0] - 1:
_aa_stage_2_clear(aggs_and_accums, antialias_zeroes)
_aa_stage_2_copy_back(aggs_and_accums)
@cuda.jit
@expand_aggs_and_cols
def extend_cuda(sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys, antialias_stage_2, *aggs_and_cols):
antialias = antialias_stage_2 is not None
buffer = cuda.local.array(8, nb_types.float64) if antialias else None
i, j = cuda.grid(2)
if i < ys.shape[0] and j < ys.shape[1] - 1:
perform_extend_line(
i, j, sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys,
buffer, *aggs_and_cols
)
if use_2_stage_agg:
return extend_cpu_antialias_2agg, extend_cuda
else:
return extend_cpu, extend_cuda
def _build_extend_line_axis1_y_constant(draw_segment, expand_aggs_and_cols, use_2_stage_agg):
@ngjit
@expand_aggs_and_cols
def perform_extend_line(
i, j, sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys, buffer, *aggs_and_cols
):
x0 = xs[i, j]
y0 = ys[j]
x1 = xs[i, j + 1]
y1 = ys[j + 1]
segment_start = (
(j == 0) or isnull(xs[i, j - 1]) or isnull(ys[j - 1])
)
segment_end = (j == len(ys)-2) or isnull(xs[i, j+2]) or isnull(ys[j+2])
if segment_start or use_2_stage_agg:
xm = 0.0
ym = 0.0
else:
xm = xs[i, j-1]
ym = ys[j-1]
draw_segment(i, sx, tx, sy, ty, xmin, xmax, ymin, ymax,
segment_start, segment_end, x0, x1, y0, y1,
xm, ym, buffer, *aggs_and_cols)
@ngjit
@expand_aggs_and_cols
def extend_cpu(sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys, antialias_stage_2, *aggs_and_cols):
antialias = antialias_stage_2 is not None
buffer = np.empty(8) if antialias else None
ncols = xs.shape[1]
for i in range(xs.shape[0]):
for j in range(ncols - 1):
perform_extend_line(
i, j, sx, tx, sy, ty, xmin, xmax, ymin, ymax,
xs, ys, buffer, *aggs_and_cols
)
def extend_cpu_antialias_2agg(sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys,
antialias_stage_2, *aggs_and_cols):
n_aggs = len(antialias_stage_2[0])
aggs_and_accums = tuple((agg, agg.copy()) for agg in aggs_and_cols[:n_aggs])
cpu_antialias_2agg_impl(sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys,
antialias_stage_2, aggs_and_accums, *aggs_and_cols)
@ngjit
@expand_aggs_and_cols
def cpu_antialias_2agg_impl(sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys,
antialias_stage_2, aggs_and_accums, *aggs_and_cols):
antialias = antialias_stage_2 is not None
buffer = np.empty(8) if antialias else None
antialias_combinations, antialias_zeroes = antialias_stage_2
ncols = xs.shape[1]
for i in range(xs.shape[0]):
for j in range(ncols - 1):
perform_extend_line(
i, j, sx, tx, sy, ty, xmin, xmax, ymin, ymax,
xs, ys, buffer, *aggs_and_cols
)
if xs.shape[0] == 1:
return
_aa_stage_2_accumulate(aggs_and_accums, antialias_combinations)
if i < xs.shape[0] - 1:
_aa_stage_2_clear(aggs_and_accums, antialias_zeroes)
_aa_stage_2_copy_back(aggs_and_accums)
@cuda.jit
@expand_aggs_and_cols
def extend_cuda(sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys, antialias_stage_2, *aggs_and_cols):
antialias = antialias_stage_2 is not None
buffer = cuda.local.array(8, nb_types.float64) if antialias else None
i, j = cuda.grid(2)
if i < xs.shape[0] and j < xs.shape[1] - 1:
perform_extend_line(
i, j, sx, tx, sy, ty, xmin, xmax, ymin, ymax,
xs, ys, buffer, *aggs_and_cols
)
if use_2_stage_agg:
return extend_cpu_antialias_2agg, extend_cuda
else:
return extend_cpu, extend_cuda
def _build_extend_line_axis1_ragged(draw_segment, expand_aggs_and_cols, use_2_stage_agg):
def extend_cpu(
sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys, antialias_stage_2, *aggs_and_cols
):
x_start_i = xs.start_indices
x_flat = xs.flat_array
y_start_i = ys.start_indices
y_flat = ys.flat_array
extend_cpu_numba(
sx, tx, sy, ty, xmin, xmax, ymin, ymax,
x_start_i, x_flat, y_start_i, y_flat, antialias_stage_2, *aggs_and_cols
)
@ngjit
@expand_aggs_and_cols
def extend_cpu_numba(
sx, tx, sy, ty, xmin, xmax, ymin, ymax,
x_start_i, x_flat, y_start_i, y_flat, antialias_stage_2, *aggs_and_cols
):
antialias = antialias_stage_2 is not None
buffer = np.empty(8) if antialias else None
nrows = len(x_start_i)
x_flat_len = len(x_flat)
y_flat_len = len(y_flat)
for i in range(nrows):
# Get x index range
x_start_index = x_start_i[i]
x_stop_index = (x_start_i[i + 1]
if i < nrows - 1
else x_flat_len)
# Get y index range
y_start_index = y_start_i[i]
y_stop_index = (y_start_i[i + 1]
if i < nrows - 1
else y_flat_len)
# Find line segment length as shorter of the two segments
segment_len = min(x_stop_index - x_start_index,
y_stop_index - y_start_index)
for j in range(segment_len - 1):
x0 = x_flat[x_start_index + j]
y0 = y_flat[y_start_index + j]
x1 = x_flat[x_start_index + j + 1]
y1 = y_flat[y_start_index + j + 1]
segment_start = (
(j == 0) or
isnull(x_flat[x_start_index + j - 1]) or
isnull(y_flat[y_start_index + j - 1])
)
segment_end = (
(j == segment_len-2) or
isnull(x_flat[x_start_index + j + 2]) or
isnull(y_flat[y_start_index + j + 2])
)
if segment_start or use_2_stage_agg:
xm = 0.0
ym = 0.0
else:
xm = x_flat[x_start_index + j - 1]
ym = y_flat[y_start_index + j - 1]
draw_segment(i, sx, tx, sy, ty, xmin, xmax, ymin, ymax,
segment_start, segment_end, x0, x1, y0, y1,
xm, ym, buffer, *aggs_and_cols)
def extend_cpu_antialias_2agg(
sx, tx, sy, ty, xmin, xmax, ymin, ymax, xs, ys, antialias_stage_2, *aggs_and_cols
):
x_start_i = xs.start_indices
x_flat = xs.flat_array
y_start_i = ys.start_indices
y_flat = ys.flat_array
n_aggs = len(antialias_stage_2[0])
aggs_and_accums = tuple((agg, agg.copy()) for agg in aggs_and_cols[:n_aggs])
extend_cpu_numba_antialias_2agg(
sx, tx, sy, ty, xmin, xmax, ymin, ymax, x_start_i, x_flat,
y_start_i, y_flat, antialias_stage_2, aggs_and_accums, *aggs_and_cols
)
@ngjit
@expand_aggs_and_cols
def extend_cpu_numba_antialias_2agg(
sx, tx, sy, ty, xmin, xmax, ymin, ymax, x_start_i, x_flat,
y_start_i, y_flat, antialias_stage_2, aggs_and_accums, *aggs_and_cols
):
antialias = antialias_stage_2 is not None
buffer = np.empty(8) if antialias else None
antialias_combinations, antialias_zeroes = antialias_stage_2
nrows = len(x_start_i)
x_flat_len = len(x_flat)
y_flat_len = len(y_flat)
for i in range(nrows):
# Get x index range
x_start_index = x_start_i[i]
x_stop_index = (x_start_i[i + 1]
if i < nrows - 1
else x_flat_len)
# Get y index range
y_start_index = y_start_i[i]
y_stop_index = (y_start_i[i + 1]
if i < nrows - 1
else y_flat_len)
# Find line segment length as shorter of the two segments
segment_len = min(x_stop_index - x_start_index,
y_stop_index - y_start_index)
for j in range(segment_len - 1):
x0 = x_flat[x_start_index + j]
y0 = y_flat[y_start_index + j]
x1 = x_flat[x_start_index + j + 1]
y1 = y_flat[y_start_index + j + 1]
segment_start = (
(j == 0) or
isnull(x_flat[x_start_index + j - 1]) or
isnull(y_flat[y_start_index + j - 1])
)
segment_end = (
(j == segment_len-2) or
isnull(x_flat[x_start_index + j + 2]) or
isnull(y_flat[y_start_index + j + 2])
)
draw_segment(i, sx, tx, sy, ty, xmin, xmax, ymin, ymax,
segment_start, segment_end, x0, x1, y0, y1,
0.0, 0.0, buffer, *aggs_and_cols)
if nrows == 1:
return
_aa_stage_2_accumulate(aggs_and_accums, antialias_combinations)
if i < nrows - 1:
_aa_stage_2_clear(aggs_and_accums, antialias_zeroes)
_aa_stage_2_copy_back(aggs_and_accums)
if use_2_stage_agg:
return extend_cpu_antialias_2agg
else:
return extend_cpu
def _build_extend_line_axis1_geometry(draw_segment, expand_aggs_and_cols, use_2_stage_agg):
def extend_cpu(
sx, tx, sy, ty, xmin, xmax, ymin, ymax,
geometry, closed_rings, antialias_stage_2, *aggs_and_cols
):
values = geometry.buffer_values
missing = geometry.isna()
offsets = geometry.buffer_offsets
if len(offsets) == 2:
# MultiLineArray
offsets0, offsets1 = offsets
else:
# LineArray
offsets1 = offsets[0]
offsets0 = np.arange(len(offsets1))
if geometry._sindex is not None:
# Compute indices of potentially intersecting polygons using
# geometry's R-tree if there is one
eligible_inds = geometry.sindex.intersects((xmin, ymin, xmax, ymax))
else:
# Otherwise, process all indices
eligible_inds = np.arange(0, len(geometry), dtype='uint32')
extend_cpu_numba(
sx, tx, sy, ty, xmin, xmax, ymin, ymax,
values, missing, offsets0, offsets1, eligible_inds,
closed_rings, antialias_stage_2, *aggs_and_cols
)
@ngjit
@expand_aggs_and_cols
def extend_cpu_numba(
sx, tx, sy, ty, xmin, xmax, ymin, ymax,
values, missing, offsets0, offsets1, eligible_inds,
closed_rings, antialias_stage_2, *aggs_and_cols
):
antialias = antialias_stage_2 is not None
buffer = np.empty(8) if antialias else None
for i in eligible_inds:
if missing[i]:
continue
start0 = offsets0[i]
stop0 = offsets0[i + 1]
for j in range(start0, stop0):
start1 = offsets1[j]
stop1 = offsets1[j + 1]
for k in range(start1, stop1 - 2, 2):
x0 = values[k]
if not np.isfinite(x0):
continue
y0 = values[k + 1]
if not np.isfinite(y0):
continue
x1 = values[k + 2]
if not np.isfinite(x1):
continue
y1 = values[k + 3]
if not np.isfinite(y1):
continue
segment_start = (
(k == start1 and not closed_rings) or
(k > start1 and
(not np.isfinite(values[k - 2]) or not np.isfinite(values[k - 1])))
)
segment_end = (
(not closed_rings and k == stop1-4) or
(k < stop1-4 and
(not np.isfinite(values[k + 4]) or not np.isfinite(values[k + 5])))
)
if segment_start or use_2_stage_agg:
xm = 0.0
ym = 0.0
elif k == start1 and closed_rings:
xm = values[stop1-4]
ym = values[stop1-3]
else:
xm = values[k-2]
ym = values[k-1]
draw_segment(i, sx, tx, sy, ty, xmin, xmax, ymin, ymax,
segment_start, segment_end, x0, x1, y0, y1,
xm, ym, buffer, *aggs_and_cols)
def extend_cpu_antialias_2agg(
sx, tx, sy, ty, xmin, xmax, ymin, ymax,
geometry, closed_rings, antialias_stage_2, *aggs_and_cols
):
values = geometry.buffer_values
missing = geometry.isna()
offsets = geometry.buffer_offsets
if len(offsets) == 2:
# MultiLineArray
offsets0, offsets1 = offsets
else:
# LineArray
offsets1 = offsets[0]
offsets0 = np.arange(len(offsets1))
if geometry._sindex is not None:
# Compute indices of potentially intersecting polygons using
# geometry's R-tree if there is one
eligible_inds = geometry.sindex.intersects((xmin, ymin, xmax, ymax))
else:
# Otherwise, process all indices
eligible_inds = np.arange(0, len(geometry), dtype='uint32')
n_aggs = len(antialias_stage_2[0])
aggs_and_accums = tuple((agg, agg.copy()) for agg in aggs_and_cols[:n_aggs])
extend_cpu_numba_antialias_2agg(
sx, tx, sy, ty, xmin, xmax, ymin, ymax,
values, missing, offsets0, offsets1, eligible_inds,
closed_rings, antialias_stage_2, aggs_and_accums, *aggs_and_cols
)
@ngjit
@expand_aggs_and_cols
def extend_cpu_numba_antialias_2agg(
sx, tx, sy, ty, xmin, xmax, ymin, ymax,
values, missing, offsets0, offsets1, eligible_inds,
closed_rings, antialias_stage_2, aggs_and_accums, *aggs_and_cols
):
antialias = antialias_stage_2 is not None
buffer = np.empty(8) if antialias else None
antialias_combinations, antialias_zeroes = antialias_stage_2
for i in eligible_inds:
if missing[i]:
continue
start0 = offsets0[i]
stop0 = offsets0[i + 1]
for j in range(start0, stop0):
start1 = offsets1[j]
stop1 = offsets1[j + 1]
for k in range(start1, stop1 - 2, 2):
x0 = values[k]
if not np.isfinite(x0):
continue
y0 = values[k + 1]
if not np.isfinite(y0):
continue
x1 = values[k + 2]
if not np.isfinite(x1):
continue
y1 = values[k + 3]
if not np.isfinite(y1):
continue
segment_start = (
(k == start1 and not closed_rings) or
(k > start1 and
(not np.isfinite(values[k - 2]) or not np.isfinite(values[k - 1])))
)
segment_end = (
(not closed_rings and k == stop1-4) or
(k < stop1-4 and
(not np.isfinite(values[k + 4]) or not np.isfinite(values[k + 5])))
)
draw_segment(i, sx, tx, sy, ty, xmin, xmax, ymin, ymax,
segment_start, segment_end, x0, x1, y0, y1,
0.0, 0.0, buffer, *aggs_and_cols)
_aa_stage_2_accumulate(aggs_and_accums, antialias_combinations)
_aa_stage_2_clear(aggs_and_accums, antialias_zeroes)
_aa_stage_2_copy_back(aggs_and_accums)
if use_2_stage_agg:
return extend_cpu_antialias_2agg
else:
return extend_cpu