-
Notifications
You must be signed in to change notification settings - Fork 1
/
toolsHdf5.py
230 lines (196 loc) · 6.53 KB
/
toolsHdf5.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
""" HDF5 UTILS """
from __future__ import print_function
import numpy as np
from toolsLog import logbook
import toolsVarious
import h5py
import os
from multiprocessing import Pool
_datasetsInHdf5File={}
class ixppyHDF5(h5py.File):
def __init__(self,name,mode=None, driver=None, libver=None, userblock_size=None, **kwds):
self.h = h5py.File.__init__(self,name,mode=mode,driver=driver,libver=libver,\
userblock_size=userblock_size,**kwds)
def getNames(self):
if not hasattr(self,"_names"): self._names = getNames(self)
return self._names
def datasetGet(self,name):
if (name in self):
return self[name]
else:
return None
def getObj(self):
if not hasattr(self,"obj"): self.obj = Hdf5ToObj(self)
return self.obj
def datasetRead(self,name,chunksize=1):
dataset = self.dataset(name)
if dataset is None: return None
size = dataset.size
n = dataset.shape[0]
isMultiProcessUseful = size> (1024*1024)
if (nChunkSize > 1) and isMultiProcessUseful:
idx = range(n)
def f(x):
return dataset[x]
p = Pool(); # 16-43 ms overhead
res = p.map_async(f,idx,chunksize=chunksize)
def openOrCreateFile(fname,mode="r",driver=None):
if (os.path.isfile(fname)):
if (mode == "r"):
if not os.access(fname,os.R_OK):
raise IOError("Asked to read %s but it is not possible, check permissions" % fname)
return None
elif (mode=="r+") or (mode=="a") or (mode=="w"):
if not os.access(fname,os.W_OK):
raise IOError("Asked to read/write %s but it is not possible, check permissions" % fname)
return None
h5handle=h5py.File(fname,mode,driver=driver)
logbook("File %s exists already, opening in %s mode" % (fname,mode))
else:
logbook("File %s does not exists, creating it" % (fname))
h5handle=h5py.File(fname,"w")
return h5handle
def getNames(h5handle):
out=[]
h5handle.visit(out.append)
return out
def datasetExists(h5handle,name):
return (name in h5handle)
def datasetWrite(h5handle,name,data):
""" (Over)write a dataset with data, no shape check is done """
if datasetExists(h5handle,name): del h5handle[name]
h5handle[name]=data
def datasetRead(h5handle,name):
""" read name from hdf5, returning the dataset instance, if it does not exist return None """
#print "H5 read",name
if (datasetExists(h5handle,name)):
return h5handle[name]
else:
return None
def getData(h5handle,name,slice=None,chunksize=1):
dataset = getDataset(h5handle,name)
return datasetToNumpy(dataset,slice=slice,chunksize=chunksize)
def getDataset(h5handle,reRead=False):
# caching
if ((h5handle in _datasetsInHdf5File) and (not reRead)):
return _datasetsInHdf5File[h5handle]
out = []
def func(name,obj):
if (isinstance,h5py.Dataset):
out.append(name)
h5handle.visititems(func)
globals()["_datasetsInHdf5File"][h5handle]=out
return out
def getDataset_hack(h5handle,reRead=False):
# caching
if ((h5handle in _datasetsInHdf5File) and (not reRead)):
return _datasetsInHdf5File[h5handle]
out = []
def checkkeys(handle):
for key in handle.keys():
okey = handle[key]
if isinstance(okey,h5py.Dataset):
out.append(okey.name)
elif isinstance(okey,h5py.Group):
checkkeys(okey)
checkkeys(h5handle)
globals()["_datasetsInHdf5File"][h5handle]=out
return out
def Hdf5ToObj(h5handle):
ishdf5group = (isinstance(h5handle,h5py.File)) or (isinstance(h5handle,h5py.Group))
if not ishdf5group:
h5handle = h5py.File(h5handle,'r')
ret = toolsVarious.dropObject()
for h in h5handle:
name = h.replace(":","_")
name = name.replace(".","_")
if not isinstance(h5handle[h],h5py.Dataset):
ret._add(name,Hdf5ToObj(h5handle[h]))
else:
ret._add(name,h5handle[h])
return ret
def f(arg):
dataset,x=arg
return dataset[x]
def f1(dataset,x):
return dataset[x]
def getItemSize(dataset):
# the line below fails for cspad data .... what a pain
#size = dataset.size
return dataset.dtype.itemsize
def _readDataset(args):
""" utility function used by multiprocessing.Pool """
dataset,sliceShot = args
path = dataset.name
filename = dataset.file.filename
h=h5py.File(filename,"r")
return h[path][sliceShot]
def readDataset(dataset,sliceSel=None,chunksize=300):
# the line below fails for cspad data .... what a pain
#size = dataset.size
itemSize = getItemSize(dataset)
n = dataset.shape[0]
if (sliceSel is None): sliceSel = slice(0,n,1)
dataSize = n*itemSize
isMultiProcessUseful = dataSize > (2048*2048)
if isMultiProcessUseful:
# subdivide indices in chunksize
start,stop,step = sliceSel.indices(n)
nC = int(float(stop-start)/step/chunksize+0.5)
print(nC)
args = []
for i in range(nC):
s1 = start+i*(chunksize*step)
s2 = start+(i+1)*(chunksize*step)
print(i,s1,s2)
args.append( (dataset,slice(s1,s2,step) ) )
print(args)
raw_input("Not working yet, use chunksize = 1")
p = Pool(4); # 16-43 ms overhead
res = p.map(_readDataset,args)
data = np.concatenate(res)
else:
data = dataset[sliceSel]
return data
def datasetToNumpy(dataset,sliceSel=None,chunksize=1):
size = dataset.size
n = dataset.shape[0]
if (sliceSel is None): sliceSel = slice(0,n,1)
isMultiProcessUseful = size> (1024*1024)
if (chunksize > 1) and isMultiProcessUseful:
# subdivide indices in chunksize
start,stop,step = sliceSel.indices(n)
nC = int(float(stop-start)/step/chunksize+0.5)
print(nC)
args = []
for i in range(nC):
s1 = start+i*(chunksize*step)
s2 = start+(i+1)*(chunksize*step)
print(i,s1,s2)
args.append( (dataset,slice(s1,s2,step) ) )
print(args)
raw_input("Not working yet, use chunksize = 1")
p = Pool(); # 16-43 ms overhead
res = p.map_async(f,args,chunksize=1)
p.close()
p.join()
data = np.asarray(res.get())
else:
data = dataset[sliceSel]
return data
def getHdf5Format(fileNames):
fileHandles = toolsVarious.iterate(fileNames,openOrCreateFile,"r",driver='sec2')
h5format = ''
for tformat in ['lclsH5','saclaH5']:
try:
if tformat is 'lclsH5':
h5format = int(np.asarray(['Configure:0000' in fh.keys() for fh in fileHandles]).all())*tformat
print(h5format)
elif tformat is 'saclaH5':
#h5format = int(np.asarray([fh['file_info/format_type'].value == 'run_dat_format' for fh in fileHandles]).all())*tformat
h5format = int(np.asarray(['file_info' in fh.keys() for fh in fileHandles]).all())*tformat
if not (h5format==''):
break
except Exception,e:
pass
return h5format