1
|
|
|
""" |
2
|
|
|
Plots. |
3
|
|
|
""" |
4
|
|
|
import enum |
5
|
|
|
from collections import Mapping |
|
|
|
|
6
|
|
|
from dataclasses import dataclass |
7
|
|
|
from pathlib import Path |
8
|
|
|
from typing import Any, Optional, Tuple, Union |
9
|
|
|
|
10
|
|
|
import numpy as np |
|
|
|
|
11
|
|
|
from typeddfs import TypedDf, AffinityMatrixDf |
|
|
|
|
12
|
|
|
from matplotlib.colors import Colormap |
|
|
|
|
13
|
|
|
|
14
|
|
|
from mandos.analysis._plot_utils import MandosPlotStyling, plt, sns, Figure |
15
|
|
|
from mandos.model.utils import CleverEnum |
16
|
|
|
from mandos.analysis.io_defns import ( |
17
|
|
|
PhiPsiSimilarityDfLongForm, |
18
|
|
|
PsiProjectedDf, |
19
|
|
|
EnrichmentDf, |
20
|
|
|
ConcordanceDf, |
21
|
|
|
SimilarityDfShortForm, |
22
|
|
|
) |
23
|
|
|
|
24
|
|
|
|
25
|
|
|
EN_DASH = "–" |
26
|
|
|
|
27
|
|
|
|
28
|
|
|
@enum.unique |
|
|
|
|
29
|
|
|
class RelPlotType(CleverEnum): |
30
|
|
|
scatter = 1 |
31
|
|
|
line = 2 |
32
|
|
|
regression = 3 |
33
|
|
|
|
34
|
|
|
|
35
|
|
|
@enum.unique |
|
|
|
|
36
|
|
|
class CatPlotType(CleverEnum): |
37
|
|
|
bar = 1 |
|
|
|
|
38
|
|
|
fold = 2 |
39
|
|
|
box = 3 |
40
|
|
|
violin = 4 |
41
|
|
|
strip = 5 |
42
|
|
|
swarm = 6 |
43
|
|
|
|
44
|
|
|
|
45
|
|
|
@dataclass(frozen=True, repr=True) |
|
|
|
|
46
|
|
|
class PlotOptions: |
47
|
|
|
size: Optional[str] |
48
|
|
|
stylesheet: Optional[Path] |
49
|
|
|
rc: Mapping[str, Any] |
|
|
|
|
50
|
|
|
hue: Optional[str] |
51
|
|
|
palette: Union[None, Colormap, Mapping[str, str]] |
52
|
|
|
extra: Mapping[str, Any] |
53
|
|
|
|
54
|
|
|
@property |
55
|
|
|
def width_and_height(self) -> Tuple[float, float]: |
|
|
|
|
56
|
|
|
return MandosPlotStyling.fig_width_and_height(self.size) |
57
|
|
|
|
58
|
|
|
|
59
|
|
|
@dataclass(frozen=True, repr=True) |
|
|
|
|
60
|
|
|
class MandosPlotter: |
61
|
|
|
"""""" |
62
|
|
|
|
63
|
|
|
rc: PlotOptions |
|
|
|
|
64
|
|
|
|
65
|
|
|
def __post_init__(self): |
66
|
|
|
if sns is None or plt is None: |
67
|
|
|
raise ImportError( |
68
|
|
|
"Seaborn and matplotlib required for plotting. Install the 'plots' extra." |
69
|
|
|
) |
70
|
|
|
bad_kwargs = set(self.__dict__.keys()).intersection(self.rc.extra.keys()) |
71
|
|
|
if len(bad_kwargs) > 0: |
72
|
|
|
raise ValueError(f"Overlapping args in extra: {bad_kwargs}") |
73
|
|
|
|
74
|
|
|
def _figure(self): |
75
|
|
|
width, height = self.rc.width_and_height |
76
|
|
|
fig = plt.gca() |
77
|
|
|
fig.set_figwidth(width) |
78
|
|
|
fig.set_figheight(height) |
79
|
|
|
return fig |
80
|
|
|
|
81
|
|
|
|
82
|
|
|
@dataclass(frozen=True, repr=True) |
83
|
|
|
class _CatPlotter(MandosPlotter): |
84
|
|
|
kind: CatPlotType |
85
|
|
|
group: bool |
86
|
|
|
ci: float |
|
|
|
|
87
|
|
|
boot: int |
88
|
|
|
seed: Optional[int] |
89
|
|
|
|
90
|
|
|
def get_kwargs( |
|
|
|
|
91
|
|
|
self, n_rows: int, n_categories: int, more: Mapping[str, Any] |
|
|
|
|
92
|
|
|
) -> Mapping[str, Any]: |
93
|
|
|
kwargs = dict(dropna=False) |
94
|
|
|
kwargs.update(**more) |
95
|
|
|
# the aspect probably doesn't matter much, but it definitely shouldn't be 1 |
96
|
|
|
kwargs["aspect"] = n_categories |
97
|
|
|
if self.kind is CatPlotType.violin: |
98
|
|
|
kwargs.update(inner="quartile") |
99
|
|
|
if self.kind in [CatPlotType.bar, CatPlotType.box, CatPlotType.violin]: |
100
|
|
|
kwargs.update(saturation=1.0) |
101
|
|
|
if self.kind in [CatPlotType.swarm, CatPlotType.strip]: |
102
|
|
|
kwargs.update(edgecolor="black") |
103
|
|
|
if self.kind in [CatPlotType.bar, CatPlotType.fold]: |
104
|
|
|
kwargs.update(errcolor="black") |
105
|
|
|
if self.kind in [ |
106
|
|
|
CatPlotType.bar, |
|
|
|
|
107
|
|
|
CatPlotType.fold, |
|
|
|
|
108
|
|
|
CatPlotType.swarm, |
|
|
|
|
109
|
|
|
CatPlotType.strip, |
|
|
|
|
110
|
|
|
CatPlotType.violin, |
|
|
|
|
111
|
|
|
]: |
112
|
|
|
kwargs.update(dodge=self.group) |
113
|
|
|
if self.group and self.kind is CatPlotType.violin and n_categories == 2: |
114
|
|
|
kwargs.update(dodge=False, split=True) |
115
|
|
|
if self.rc.extra is not None: |
116
|
|
|
kwargs.update(**self.rc.extra) |
117
|
|
|
kwargs.update(seed=self.seed, ci=self.ci, n_boot=self.boot) |
118
|
|
|
return kwargs |
119
|
|
|
|
120
|
|
|
|
121
|
|
|
@dataclass(frozen=True, repr=True) |
122
|
|
|
class _RelPlotter(MandosPlotter): |
123
|
|
|
kind: RelPlotType |
124
|
|
|
ci: float |
|
|
|
|
125
|
|
|
boot: int |
126
|
|
|
seed: Optional[int] |
127
|
|
|
|
128
|
|
|
def get_kwargs(self, n_rows: int, n_cols: int, more: Mapping[str, Any]) -> Mapping[str, Any]: |
|
|
|
|
129
|
|
|
kwargs = dict(dropna=False, dashes=False) |
130
|
|
|
kwargs.update(**more) |
131
|
|
|
if self.rc.extra is not None: |
132
|
|
|
kwargs.update(**self.rc.extra) |
133
|
|
|
kwargs.update(seed=self.seed, ci=self.ci, n_boot=self.boot) |
134
|
|
|
return kwargs |
135
|
|
|
|
136
|
|
|
|
137
|
|
|
@dataclass(frozen=True, repr=True) |
138
|
|
|
class _HeatPlotter(MandosPlotter): |
139
|
|
|
vmin_percentile: float = 0 |
140
|
|
|
vmax_percentile: float = 100 |
141
|
|
|
|
142
|
|
|
def __post_init__(self): |
143
|
|
|
if self.rc.extra.get("mask") is not None: |
144
|
|
|
raise ValueError(f"Cannot set mask in {self.__class__.__name__}") |
145
|
|
|
|
146
|
|
|
def get_kwargs(self, data: AffinityMatrixDf, more: Mapping[str, Any]) -> Mapping[str, Any]: |
|
|
|
|
147
|
|
|
vmin = np.quantile(data.flatten(), self.vmin_percentile / 100) |
148
|
|
|
vmax = np.quantile(data.flatten(), self.vmax_percentile / 100) |
149
|
|
|
mask = data.values == np.nan |
150
|
|
|
kwargs = dict( |
151
|
|
|
vmin=vmin, |
152
|
|
|
vmax=vmax, |
153
|
|
|
square=True, |
154
|
|
|
mask=mask, |
155
|
|
|
hue=self.rc.hue, |
156
|
|
|
palette=self.rc.palette, |
157
|
|
|
) |
158
|
|
|
if self.rc.extra is not None: |
159
|
|
|
kwargs.update(**self.rc.extra) |
160
|
|
|
kwargs.update(**more) |
161
|
|
|
return kwargs |
162
|
|
|
|
163
|
|
|
|
164
|
|
|
@dataclass(frozen=True, repr=True) |
|
|
|
|
165
|
|
|
class ScorePlotter(_CatPlotter): |
166
|
|
|
"""""" |
167
|
|
|
|
168
|
|
|
def plot(self, data: EnrichmentDf) -> Figure: |
|
|
|
|
169
|
|
|
data = data.copy() |
170
|
|
|
data: TypedDf = data |
171
|
|
|
data.only("score_name") # make sure |
172
|
|
|
data[f"object{EN_DASH}predicate"] = data["object"] + " " + data["predicate"] |
173
|
|
|
data[f"predicate{EN_DASH}object"] = data["predicate"] + " " + data["object"] |
174
|
|
|
data = data.sort_natural(f"object{EN_DASH}predicate") |
175
|
|
|
with MandosPlotStyling.context(*self.rc.rc): |
|
|
|
|
176
|
|
|
if self.kind is CatPlotType.fold: |
177
|
|
|
self._plot_fold(data) |
178
|
|
|
else: |
179
|
|
|
self._plot_regular(data) |
180
|
|
|
return self._figure() |
181
|
|
|
|
182
|
|
|
def _plot_fold(self, data: EnrichmentDf): |
183
|
|
|
kwargs = dict( |
184
|
|
|
color="black", |
185
|
|
|
saturation=1, |
186
|
|
|
errcolor="black", |
187
|
|
|
dropna=False, |
188
|
|
|
ci=None, |
189
|
|
|
hue=self.rc.hue, |
190
|
|
|
palette=self.rc.palette, |
191
|
|
|
) |
192
|
|
|
if self.rc.extra is not None: |
193
|
|
|
kwargs.update({k: v for k, v in self.rc.extra if k != "saturation"}) |
194
|
|
|
sns.catplot( |
195
|
|
|
kind="bar", |
196
|
|
|
x=f"predicate{EN_DASH}object", |
197
|
|
|
y="background", |
198
|
|
|
data=data, |
199
|
|
|
row="key", |
200
|
|
|
**kwargs, |
201
|
|
|
) |
202
|
|
|
kwargs = dict( |
203
|
|
|
color="black", |
204
|
|
|
saturation=0.3, |
205
|
|
|
errcolor="black", |
206
|
|
|
hue=self.rc.hue, |
207
|
|
|
palette=self.rc.palette, |
208
|
|
|
) |
209
|
|
|
if self.rc.extra is not None: |
210
|
|
|
kwargs.update(self.rc.extra) |
211
|
|
|
sns.catplot( |
212
|
|
|
kind="bar", |
213
|
|
|
data=data, |
214
|
|
|
x=f"predicate{EN_DASH}object", |
215
|
|
|
y="value", |
216
|
|
|
row="key", |
217
|
|
|
**kwargs, |
218
|
|
|
) |
219
|
|
|
|
220
|
|
|
def _plot_regular(self, data: EnrichmentDf): |
221
|
|
|
keys = data["keys"].unique() |
222
|
|
|
defaults = dict( |
223
|
|
|
saturation=1, |
224
|
|
|
errcolor="black", |
225
|
|
|
hue=self.rc.hue, |
226
|
|
|
palette=self.rc.palette, |
227
|
|
|
) |
228
|
|
|
kwargs = self.get_kwargs(len(keys), 1, defaults) |
229
|
|
|
sns.catplot( |
230
|
|
|
kind=self.kind.name, |
231
|
|
|
data=data, |
232
|
|
|
x=f"predicate{EN_DASH}object", |
233
|
|
|
y="value", |
234
|
|
|
row="key", |
235
|
|
|
**kwargs, |
236
|
|
|
) |
237
|
|
|
|
238
|
|
|
|
239
|
|
|
@dataclass(frozen=True, repr=True) |
|
|
|
|
240
|
|
|
class TauPlotter(_CatPlotter): |
241
|
|
|
""" """ |
242
|
|
|
|
243
|
|
|
def plot(self, data: ConcordanceDf) -> Figure: |
|
|
|
|
244
|
|
|
phis = data["phi"].unique() |
245
|
|
|
# psis = data["psi"].unique() |
246
|
|
|
defaults = dict( |
247
|
|
|
saturation=1, |
248
|
|
|
errcolor="black", |
249
|
|
|
hue=self.rc.hue, |
250
|
|
|
palette=self.rc.palette, |
251
|
|
|
) |
252
|
|
|
kwargs = self.get_kwargs(len(phis), 1, defaults) |
253
|
|
|
with MandosPlotStyling.context(*self.rc.rc): |
|
|
|
|
254
|
|
|
sns.catplot( |
255
|
|
|
kind=self.kind.name, |
256
|
|
|
data=data, |
257
|
|
|
x="psi", |
258
|
|
|
y="tau", |
259
|
|
|
row="phi", |
260
|
|
|
**kwargs, |
261
|
|
|
) |
262
|
|
|
return self._figure() |
263
|
|
|
|
264
|
|
|
|
265
|
|
|
@dataclass(frozen=True, repr=True) |
|
|
|
|
266
|
|
|
class CorrPlotter(_RelPlotter): |
267
|
|
|
"""""" |
268
|
|
|
|
269
|
|
|
def plot(self, data: PhiPsiSimilarityDfLongForm) -> Figure: |
|
|
|
|
270
|
|
|
phis = data["phi"].unique() |
271
|
|
|
psis = data["psi"].unique() |
272
|
|
|
with MandosPlotStyling.context(*self.rc.rc): |
|
|
|
|
273
|
|
|
if self.kind is RelPlotType.regression: |
274
|
|
|
defaults = dict( |
275
|
|
|
truncate=True, |
276
|
|
|
hue=self.rc.hue, |
277
|
|
|
palette=self.rc.palette, |
278
|
|
|
) |
279
|
|
|
kwargs = self.get_kwargs(len(phis), len(psis), defaults) |
280
|
|
|
sns.lmplot( |
281
|
|
|
data=data, |
282
|
|
|
x="phi_value", |
283
|
|
|
y="psi_value", |
284
|
|
|
row="phi", |
285
|
|
|
col="psi", |
286
|
|
|
**kwargs, |
287
|
|
|
) |
288
|
|
|
else: |
289
|
|
|
kwargs = self.get_kwargs(len(phis), len(psis), {}) |
290
|
|
|
sns.relplot( |
291
|
|
|
kind=self.kind.name, |
292
|
|
|
data=data, |
293
|
|
|
x="phi_value", |
294
|
|
|
y="psi_value", |
295
|
|
|
row="phi", |
296
|
|
|
col="psi", |
297
|
|
|
**kwargs, |
298
|
|
|
) |
299
|
|
|
return self._figure() |
300
|
|
|
|
301
|
|
|
|
302
|
|
|
@dataclass(frozen=True, repr=True) |
|
|
|
|
303
|
|
|
class HeatmapPlotter(_HeatPlotter): |
304
|
|
|
def plot(self, data: SimilarityDfShortForm) -> Figure: |
|
|
|
|
305
|
|
|
data = data.triangle() |
306
|
|
|
kwargs = self.get_kwargs(data, {}) |
307
|
|
|
with MandosPlotStyling.context(*self.rc.rc): |
|
|
|
|
308
|
|
|
sns.heatmap( |
309
|
|
|
data, |
310
|
|
|
**kwargs, |
311
|
|
|
) |
312
|
|
|
return self._figure() |
313
|
|
|
|
314
|
|
|
|
315
|
|
|
@dataclass(frozen=True, repr=True) |
|
|
|
|
316
|
|
|
class ProjectionPlotter(MandosPlotter): |
317
|
|
|
def plot(self, data: PsiProjectedDf) -> Figure: |
|
|
|
|
318
|
|
|
psis = set(data["psi"].unique()) |
319
|
|
|
width, height = MandosPlotStyling.fig_width_and_height(self.rc.size) |
320
|
|
|
aspect = width / height |
321
|
|
|
col_wrap = int(np.ceil(np.sqrt(len(psis)) * aspect)) |
322
|
|
|
kwargs = dict( |
323
|
|
|
col_wrap=col_wrap, |
324
|
|
|
hue=self.rc.hue, |
325
|
|
|
palette=self.rc.palette, |
326
|
|
|
) |
327
|
|
|
if self.rc.extra is not None: |
328
|
|
|
kwargs.update(**self.rc.extra) |
329
|
|
|
with MandosPlotStyling.context(*self.rc.rc): |
|
|
|
|
330
|
|
|
sns.relplot( |
331
|
|
|
kind="scatter", |
332
|
|
|
data=data, |
333
|
|
|
x="x", |
334
|
|
|
y="y", |
335
|
|
|
col="psi", |
336
|
|
|
**kwargs, |
337
|
|
|
) |
338
|
|
|
return self._figure() |
339
|
|
|
|