def scatter_dist_by_mappings(dataset, x_kdims, y_kdims,
mappings,
selection_dim="Gene",
datashade_=False,
dynspread_=False,
):
data_groups = {name: dataset.sel({selection_dim: genes}) for name, genes in mappings.items()}
data_group_dfs = {k: v[[x_kdims, y_kdims]].to_dataframe() for k, v in data_groups.items()}
points = {k: hv.Points(val, kdims=[x_kdims, y_kdims]) for k, val in data_group_dfs.items()}
dist_x = {k: univariate_kde(hv.Distribution(p, kdims=[y_kdims], group="dist_x"), n_samples=1000)
for k, p in points.items()}
dist_y = {k: univariate_kde(hv.Distribution(p, kdims=[x_kdims], group="dist_y"), n_samples=1000)
for k, p in points.items()}
if datashade_:
points_overlay = datashade(hv.NdOverlay(points))
if dynspread_:
points_overlay = dynspread(points_overlay)
else:
points_overlay = hv.NdOverlay(points)
return points_overlay << hv.NdOverlay(dist_x) << hv.NdOverlay(dist_y)
def test_violin_simple(self):
values = np.random.rand(100)
violin = Violin(values).opts(plot=dict(violin_width=0.7))
qmin, q1, q2, q3, qmax = (np.percentile(values, q=q)
for q in range(0, 125, 25))
iqr = q3 - q1
upper = min(q3 + 1.5 * iqr, np.nanmax(values))
lower = max(q1 - 1.5 * iqr, np.nanmin(values))
r1, r2 = violin.range(0)
kde = univariate_kde(violin, cut=5)
xs, ys = (kde.dimension_values(i) for i in range(2))
ys = (ys / ys.max()) * (0.7 / 2.)
ys = [('', ) + (sign * y, ) for sign, vs in ((-1, ys), (1, ys[::-1]))
for y in vs]
kde = {'x': np.concatenate([xs, xs[::-1]]), 'y': ys}
plot = bokeh_renderer.get_plot(violin)
self.assertEqual(plot.handles['x_range'].factors, [''])
self.assertEqual(plot.handles['y_range'].start, r1)
self.assertEqual(plot.handles['y_range'].end, r2)
self.assertIn('scatter_1_glyph_renderer', plot.handles)
self.assertIn('vbar_1_glyph_renderer', plot.handles)
seg_source = plot.handles['segment_1_source']
self.assertEqual(seg_source.data['x'], [('', 0)])
self.assertEqual(seg_source.data['y0'], [lower])
self.assertEqual(seg_source.data['y1'], [upper])
scatter_source = plot.handles['scatter_1_source']
self.assertEqual(scatter_source.data['x'], [('', 0)])
self.assertEqual(scatter_source.data['y'], [q2])
patch_source = plot.handles['patches_1_source']
self.assertEqual(patch_source.data['xs'], [kde['y']])
self.assertEqual(patch_source.data['ys'], [kde['x']])
def test_violin_simple(self):
values = np.random.rand(100)
violin = Violin(values).opts(plot=dict(violin_width=0.7))
qmin, q1, q2, q3, qmax = (np.percentile(values, q=q)
for q in range(0,125,25))
iqr = q3 - q1
upper = min(q3 + 1.5*iqr, np.nanmax(values))
lower = max(q1 - 1.5*iqr, np.nanmin(values))
r1, r2 = violin.range(0)
kde = univariate_kde(violin, cut=5)
xs, ys = (kde.dimension_values(i) for i in range(2))
ys = (ys/ys.max())*(0.7/2.)
ys = [('',)+(sign*y,) for sign, vs in ((-1, ys), (1, ys[::-1])) for y in vs]
kde = {'x': np.concatenate([xs, xs[::-1]]), 'y': ys}
plot = bokeh_renderer.get_plot(violin)
self.assertEqual(plot.handles['x_range'].factors, [''])
self.assertEqual(plot.handles['y_range'].start, r1)
self.assertEqual(plot.handles['y_range'].end, r2)
self.assertIn('scatter_1_glyph_renderer', plot.handles)
self.assertIn('vbar_1_glyph_renderer', plot.handles)
seg_source = plot.handles['segment_1_source']
self.assertEqual(seg_source.data['x'], [('', 0)])
self.assertEqual(seg_source.data['y0'], [lower])
self.assertEqual(seg_source.data['y1'], [upper])
scatter_source = plot.handles['scatter_1_source']
self.assertEqual(scatter_source.data['x'], [('', 0)])
self.assertEqual(scatter_source.data['y'], [q2])
patch_source = plot.handles['patches_1_source']
self.assertEqual(patch_source.data['xs'], [kde['y']])
self.assertEqual(patch_source.data['ys'], [kde['x']])
def test_violin_inner_stick(self):
values = np.random.rand(100)
violin = Violin(values).opts(plot=dict(inner='stick'))
kde = univariate_kde(violin, cut=5)
xs = kde.dimension_values(0)
plot = bokeh_renderer.get_plot(violin)
self.assertIn('segment_1_glyph_renderer', plot.handles)
segments = np.array([xs[np.argmin(np.abs(xs - v))] for v in values])
self.assertEqual(plot.handles['segment_1_source'].data['x'], segments)
def test_violin_inner_stick(self):
values = np.random.rand(100)
violin = Violin(values).opts(plot=dict(inner='stick'))
kde = univariate_kde(violin, cut=5)
xs = kde.dimension_values(0)
plot = bokeh_renderer.get_plot(violin)
self.assertIn('segment_1_glyph_renderer', plot.handles)
segments = np.array([xs[np.argmin(np.abs(xs-v))] for v in values])
self.assertEqual(plot.handles['segment_1_source'].data['x'],
segments)
def hv_scatter_dist(dataset, x_kdims, y_kdims,
datashade_=False,
dynspread_=False):
if dynspread_ and not datashade_:
warnings.warn("Dynspread can only be used with datashade, setting both to true.")
datashade_ = True
df = dataset[[x_kdims, y_kdims]].to_dataframe()
points = hv.Points(df, kdims=[x_kdims, y_kdims])
dist_x = univariate_kde(hv.Distribution(points, kdims=[y_kdims], group="dist_x"), n_samples=1000)
dist_y = univariate_kde(hv.Distribution(points, kdims=[x_kdims], group="dist_y"), n_samples=1000)
if datashade_:
points = datashade(points)
if dynspread_:
points = dynspread(points)
return points << dist_x << dist_y
def test_violin_inner_quartiles(self):
values = np.random.rand(100)
violin = Violin(values).opts(plot=dict(inner='quartiles'))
kde = univariate_kde(violin, cut=5)
xs = kde.dimension_values(0)
plot = bokeh_renderer.get_plot(violin)
self.assertIn('segment_1_glyph_renderer', plot.handles)
seg_source = plot.handles['segment_1_source']
q1, q2, q3 = (np.percentile(values, q=q) for q in range(25, 100, 25))
y0, y1, y2 = [xs[np.argmin(np.abs(xs - v))] for v in (q1, q2, q3)]
self.assertEqual(seg_source.data['x'], np.array([y0, y1, y2]))
def test_violin_inner_quartiles(self):
values = np.random.rand(100)
violin = Violin(values).opts(plot=dict(inner='quartiles'))
kde = univariate_kde(violin, cut=5)
xs = kde.dimension_values(0)
plot = bokeh_renderer.get_plot(violin)
self.assertIn('segment_1_glyph_renderer', plot.handles)
seg_source = plot.handles['segment_1_source']
q1, q2, q3 = (np.percentile(values, q=q) for q in range(25,100,25))
y0, y1, y2 = [xs[np.argmin(np.abs(xs-v))] for v in (q1, q2, q3)]
self.assertEqual(seg_source.data['x'], np.array([y0, y1, y2]))
def test_univariate_kde_nans(self):
kde = univariate_kde(self.dist_nans, n_samples=5, bin_range=(0, 4))
xs = np.arange(5)
ys = [0, 0, 0, 0, 0]
area = Area((xs, ys), 'Value', ('Value_density', 'Value Density'))
self.assertEqual(kde, area)
def test_univariate_kde_flat_distribution(self):
dist = Distribution([1, 1, 1])
kde = univariate_kde(dist, n_samples=5, bin_range=(0, 4))
area = Area([], 'Value', ('Value_density', 'Value Density'))
self.assertEqual(kde, area)
def test_univariate_kde(self):
kde = univariate_kde(self.dist, n_samples=5, bin_range=(0, 4))
xs = np.arange(5)
ys = [0.17594505, 0.23548218, 0.23548218, 0.17594505, 0.0740306]
area = Area((xs, ys), 'Value', ('Value_density', 'Value Density'))
self.assertEqual(kde, area)
def test_univariate_kde_nans(self):
kde = univariate_kde(self.dist_nans, n_samples=5, bin_range=(0, 4))
xs = np.arange(5)
ys = [0, 0, 0, 0, 0]
area = Area((xs, ys), 'Value', ('Value_density', 'Density'))
self.assertEqual(kde, area)
def test_univariate_kde_flat_distribution(self):
dist = Distribution([1, 1, 1])
kde = univariate_kde(dist, n_samples=5, bin_range=(0, 4))
area = Area([], 'Value', ('Value_density', 'Density'))
self.assertEqual(kde, area)
def test_univariate_kde(self):
kde = univariate_kde(self.dist, n_samples=5, bin_range=(0, 4))
xs = np.arange(5)
ys = [0.17594505, 0.23548218, 0.23548218, 0.17594505, 0.0740306]
area = Area((xs, ys), 'Value', ('Value_density', 'Density'))
self.assertEqual(kde, area)
