forked from JulienPeloton/spark-fits-apps
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathqa_tools.py
335 lines (277 loc) · 10.2 KB
/
qa_tools.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
from pyspark.sql import SparkSession
from pyspark.sql.functions import pandas_udf, PandasUDFType
import numpy as np
import pandas as pd
import healpy as hp
from scipy import stats
from matplotlib import pyplot as plt
#nside=131072
nside=256
nest=False
pixarea=hp.nside2pixarea(nside, degrees=True)*3600
reso= hp.nside2resol(nside,arcmin=True)
#create the ang2pix user-defined-function.
@pandas_udf('long', PandasUDFType.SCALAR)
def Ang2Pix(ra,dec):
return pd.Series(hp.ang2pix(nside,np.radians(90-dec),np.radians(ra),nest=nest))
##################
# req: ipix,tract
def tracts_outline(df):
#list of tracts
tracts=df.select("tract").distinct().toPandas()
df_withpix=df.groupBy(["tract","ipix"]).count().cache()
ipix=np.arange(hp.nside2npix(nside))
pixborder=[]
print("NSIDE={}".format(nside))
for t in tracts['tract'].values :
print("treating tract={}".format(t))
#create a map just for this tract
df_map=df_withpix.filter(df_withpix.tract==int(t))
#create the healpix map
tract_p=df_map.toPandas()
#plt.hist(tract_p['count'].values)
tract_map = np.full(hp.nside2npix(nside),hp.UNSEEN)
tract_map[tract_p['ipix'].values]=1
# for lit pixels compute the neighbours
ipix1=tract_p['ipix'].values
theta,phi=hp.pix2ang(nside,ipix1)
neighbours=hp.get_all_neighbours(nside,theta,phi,0).transpose()
# border if at least one neighbours is UNSEEN
mask=[(hp.UNSEEN in tract_map[neighbours[pixel]]) for pixel in range(len(ipix1))]
pixborder+=list(ipix1[mask])
return pixborder
def get_borders_from_ipix(part):
# Get the pixel ID from the iterator
ipix = [*part]
if len(ipix) == 0:
# Empty partition
yield []
else:
# Get the 8 neighbours of all unique pixels.
theta, phi = hp.pix2ang(nside, ipix)
neighbours = hp.get_all_neighbours(nside, theta, phi, 0).flatten()
# Yield only pixels at the borders
unseen = np.array([i for i in neighbours if i not in ipix])
yield unseen
def tracts_outline2(df):
#remove duplicate tract/ipix pairs
df_pix=df.groupBy(["tract","ipix"]).count().drop('count')
df_pix.show(5)
#reapartition according to tract
df_repart = df_pix.orderBy(df["tract"]).drop("tract").cache()
df_repart.show(5)
pix = df_repart.rdd.mapPartitions(get_borders_from_ipix).collect()
return pix
def projmap_mean(df,col,minmax=None,nsig=2,dohist=True,**kwargs ):
df_map=df.select(col,"ipix").na.drop().groupBy("ipix").avg(col)
#statistics per pixel
var=df_map.columns[-1]
s=df_map.describe([var])
s.show()
r=s.select(var).take(3)
N=int(r[0][0])
mu=float(r[1][0])
sig=float(r[2][0])
map_p=df_map.toPandas()
#now data is reducd create the healpy map
skyMap= np.full(hp.nside2npix(nside),hp.UNSEEN)
skyMap[map_p['ipix'].values]=map_p[var].values
if minmax==None:
minmax=(mu-nsig*sig,mu+nsig*sig)
hp.gnomview(skyMap,reso=reso,min=minmax[0],max=minmax[1],nest=nest,title=var,**kwargs )
if dohist:
plt.figure()
plt.hist(map_p[var].values,bins=80,range=minmax)
plt.xlabel(var)
stat=[r"$N={:d}$".format(N),r"$\mu={:g}$".format(mu),r"$\sigma={:g}$".format(np.sqrt(sig))]
ax=plt.gca()
plt.text(0.7,0.8,"\n".join(stat), horizontalalignment='center',transform=ax.transAxes)
plt.show()
return skyMap
def projmap_max(df,col,minmax=None,dohist=True,**kwargs ):
df_map=df.select(col,"ipix").na.drop().groupBy("ipix").max(col)
#statistics per pixel
var=df_map.columns[-1]
s=df_map.describe([var])
s.show()
r=s.select(var).take(3)
N=int(r[0][0])
mu=float(r[1][0])
sig=float(r[2][0])
map_p=df_map.toPandas()
#now data is reduced create the healpy map
skyMap= np.full(hp.nside2npix(nside),hp.UNSEEN)
skyMap[map_p['ipix'].values]=map_p[var].values
if minmax==None:
minmax=(np.max([0,mu-2*sig]),mu+2*sig)
if dohist:
plt.hist(map_p[var].values,bins=80,range=minmax)
plt.xlabel(var)
hp.gnomview(skyMap,nest=nest,reso=reso,min=minmax[0],max=minmax[1],title=var,**kwargs )
plt.show()
return skyMap
def projmap_stddev(df,col,minmax=None,dohist=True,**kwargs ):
df_map=df.select(col,"ipix").na.drop().groupBy("ipix").agg(F.stddev(col))
#statistics per pixel
var=df_map.columns[-1]
s=df_map.describe([var])
s.show()
r=s.select(var).take(3)
N=int(r[0][0])
mu=float(r[1][0])
sig=float(r[2][0])
map_p=df_map.toPandas()
#now data is reduced create the healpy map
skyMap= np.full(hp.nside2npix(nside),hp.UNSEEN)
skyMap[map_p['ipix'].values]=map_p[var].values
if minmax==None:
minmax=(mu-2*sig,mu+2*sig)
if dohist:
plt.hist(map_p[var].values,bins=80,range=minmax)
plt.xlabel(var)
hp.gnomview(skyMap,nest=nest,reso=reso,min=minmax[0],max=minmax[1],title=var,**kwargs )
plt.show()
return skyMap
def projmap_min(df,col,minmax=None,dohist=True,**kwargs ):
df_map=df.select(col,"ipix").na.drop().groupBy("ipix").min(col)
#statistics per pixel
var=df_map.columns[-1]
s=df_map.describe([var])
s.show()
r=s.select(var).take(3)
N=int(r[0][0])
mu=float(r[1][0])
sig=float(r[2][0])
map_p=df_map.toPandas()
#now data is reduced create the healpy map
skyMap= np.full(hp.nside2npix(nside),hp.UNSEEN)
skyMap[map_p['ipix'].values]=map_p[var].values
if minmax==None:
minmax=(np.max([0,mu-2*sig]),mu+2*sig)
if dohist:
plt.hist(map_p[var].values,bins=80,range=minmax)
plt.xlabel(var)
hp.gnomview(skyMap,nest=nest,reso=reso,min=minmax[0],max=minmax[1],title=var,**kwargs )
plt.show()
return skyMap
def projmap_stddev(df,col,minmax=None,dohist=True,**kwargs ):
df_map=df.select(col,"ipix").na.drop().groupBy("ipix").agg(F.stddev(col))
#statistics per pixel
var=df_map.columns[-1]
s=df_map.describe([var])
s.show()
r=s.select(var).take(3)
N=int(r[0][0])
mu=float(r[1][0])
sig=float(r[2][0])
map_p=df_map.toPandas()
#now data is reduced create the healpy map
skyMap= np.full(hp.nside2npix(nside),hp.UNSEEN)
skyMap[map_p['ipix'].values]=map_p[var].values
if minmax==None:
minmax=(mu-2*sig,mu+2*sig)
if dohist:
plt.hist(map_p[var].values,bins=80,range=minmax)
plt.xlabel(var)
hp.gnomview(skyMap,nest=nest,reso=reso,min=minmax[0],max=minmax[1],title=var,**kwargs )
plt.show()
return skyMap
#from spark v2.4.0 only
@pandas_udf('float', PandasUDFType.GROUPED_AGG) # doctest: +SKIP
def median_udf(v):
return np.median(v)
def projmap_median_udf(df,col,minmax=None,**kwargs ):
df_map=df.select(col,"ipix").na.drop().groupBy("ipix").agg(median_udf(col))
#statistics per pixel
var=df_map.columns[-1]
s=df_map.describe([var])
s.show()
r=s.select(var).take(3)
N=int(r[0][0])
mu=float(r[1][0])
sig=float(r[2][0])
map_p=df_map.toPandas()
#now data is reduced create the healpy map
skyMap= np.full(hp.nside2npix(nside),hp.UNSEEN)
skyMap[map_p['ipix'].values]=map_p[var].values
if minmax==None:
minmax=(np.max([0,mu-2*sig]),mu+2*sig)
hp.gnomview(skyMap,nest=nest,reso=reso,min=minmax[0],max=minmax[1],title=var,**kwargs )
plt.show()
return skyMap
def projmap_median(df,col,minmax=None,**kwargs ):
rdd=df.select("ipix",col).na.drop().rdd.map(lambda r: (r[0],r[1]))
map_rdd=rdd.groupByKey().mapValues(tuple).mapValues(np.median)
#collect renvoie un eliste de tuples
#back to pandas
map_p=map_rdd.map(lambda r: (r[0],float(r[1]))).toDF().toPandas()
val=np.array(map_p._2.values)
mu=val.mean()
sig=val.std()
if minmax==None:
minmax=(mu-2*sig,mu+2*sig)
print("N={} \nmean={} \nsigma={}\nmin={}\nmax={}".format(val.size,mu,sig,min(val),max(val)))
skyMap= np.full(hp.nside2npix(nside),hp.UNSEEN)
skyMap[map_p._1.values]=val
hp.gnomview(skyMap,nest=nest,reso=reso,min=minmax[0],max=minmax[1],title="median("+col+")",**kwargs)
plt.show()
return skyMap
def countsmap(df,minmax=None,doHist=True,density=False,**kwargs):
assert "ipix" in df.columns
df_map=df.select("ipix").groupBy("ipix").count()
#back to python world
map_p=df_map.toPandas()
A=map_p.index.size*pixarea/3600
print("map area={} deg2".format(A))
#statistics per pixel
var='count'
s=df_map.describe([var])
s.show()
r=s.select(var).take(5)
N=int(r[0][0])
mu=float(r[1][0])
sig=float(r[2][0])
imin=int(r[3][0])
imax=int(r[4][0])
if density:
tit=kwargs.pop("title", r"$density/arcmin^2$")
else :
tit=kwargs.pop("title", "counts/pixel (nside={})".format(nside))
if doHist:
plt.figure()
plt.title(tit)
nbins=imax-imin-1
data=map_p[var].values
if density :
data=data/pixarea
plt.hist(data,bins=nbins)
plt.xlabel(var)
t=stats.describe(data)
N=t[0]
m=t[2]
sig2=t[3]
xmin=t[1][0]
xmax=t[1][1]
stat=[r"$N={:d}$".format(N),r"$\mu={:g}$".format(m),r"$\sigma={:g}$".format(np.sqrt(sig2)),r"min={:g}".format(xmin),r"max={:g}".format(xmax)]
ax=plt.gca()
plt.text(0.8,0.7,"\n".join(stat), horizontalalignment='center',transform=ax.transAxes)
#now data is reduced create the healpy map
skyMap= np.full(hp.nside2npix(nside),hp.UNSEEN)
if minmax==None:
minmax=(np.max([0,mu-2*sig]),mu+2*sig)
if density:
skyMap[map_p['ipix'].values]=map_p[var].values/pixarea
hp.gnomview(skyMap,nest=nest,reso=reso,min=minmax[0]/pixarea,max=minmax[1]/pixarea,title=tit,**kwargs)
else :
skyMap[map_p['ipix'].values]=map_p[var].values
hp.gnomview(skyMap,nest=nest,reso=reso,min=minmax[0],max=minmax[1],title=tit,**kwargs)
plt.show()
return skyMap
def add_healpixels(df,ra="ra",dec="dec"):
assert ra in df.columns
assert dec in df.columns
print("NSIDE={}, pixel area={} arcmin^2, resolution={} arcmin".format(nside,pixarea,reso))
spark = SparkSession.builder.getOrCreate()
spark.sparkContext.broadcast(nside)
df=df.withColumn("ipix",Ang2Pix("ra","dec"))
return df