|
1
|
|
|
''' |
|
2
|
|
|
run_savu |
|
3
|
|
|
This is a refactor of the code that used to be contained in dawn. |
|
4
|
|
|
It's used to mock up a runner for individual savu plugins from a python shell. |
|
5
|
|
|
It is currently very early in development and will be subject to massive refactor in the future. |
|
6
|
|
|
''' |
|
7
|
|
|
from savu.data.experiment_collection import Experiment |
|
8
|
|
|
from savu.data.meta_data import MetaData |
|
9
|
|
|
from savu.plugins.utils import get_plugin |
|
10
|
|
|
import savu.plugins.loaders.utils.yaml_utils as yaml |
|
11
|
|
|
import os, sys |
|
12
|
|
|
import numpy as np |
|
13
|
|
|
from copy import deepcopy as copy |
|
14
|
|
|
import time |
|
15
|
|
|
from collections import OrderedDict |
|
16
|
|
|
|
|
17
|
|
|
|
|
18
|
|
|
def get_output_rank(path2plugin, inputs, params, persistence): |
|
19
|
|
|
sys_path_0_lock = persistence['sys_path_0_lock'] |
|
20
|
|
|
sys_path_0_lock.acquire() |
|
21
|
|
|
try: |
|
22
|
|
|
parameters = {} |
|
23
|
|
|
# slight repack here |
|
24
|
|
|
for key in list(params.keys()): |
|
25
|
|
|
val = params[key]["value"] |
|
26
|
|
|
if type(val)==type(''): |
|
27
|
|
|
val = val.replace('\n','').strip() |
|
28
|
|
|
parameters[key] = val |
|
29
|
|
|
plugin = _savu_setup(path2plugin, inputs, parameters) |
|
30
|
|
|
persistence['plugin_object'] = plugin |
|
31
|
|
|
finally: |
|
32
|
|
|
sys_path_0_lock.release() |
|
33
|
|
|
|
|
34
|
|
|
return len(plugin.get_plugin_out_datasets()[0].get_core_dimensions()) |
|
35
|
|
|
|
|
36
|
|
|
|
|
37
|
|
|
def runSavu(path2plugin, params, metaOnly, inputs, persistence): |
|
38
|
|
|
''' |
|
39
|
|
|
path2plugin - is the path to the user script that should be run |
|
40
|
|
|
params - are the savu parameters |
|
41
|
|
|
metaOnly - a boolean for whether the data is kept in metadata or is passed as data |
|
42
|
|
|
inputs - is a dictionary of input objects |
|
43
|
|
|
''' |
|
44
|
|
|
t1 = time.time() |
|
45
|
|
|
sys_path_0_lock = persistence['sys_path_0_lock'] |
|
46
|
|
|
sys_path_0_set = persistence['sys_path_0_set'] |
|
47
|
|
|
plugin_object = persistence['plugin_object'] |
|
48
|
|
|
axis_labels = persistence['axis_labels'] |
|
49
|
|
|
axis_values = persistence['axis_values'] |
|
50
|
|
|
string_key = persistence['string_key'] |
|
51
|
|
|
parameters = persistence['parameters'] |
|
52
|
|
|
aux = persistence['aux'] |
|
53
|
|
|
sys_path_0_lock.acquire() |
|
54
|
|
|
try: |
|
55
|
|
|
result = copy(inputs) |
|
56
|
|
|
|
|
57
|
|
|
scriptDir = os.path.dirname(path2plugin) |
|
58
|
|
|
sys_path_0 = sys.path[0] |
|
59
|
|
|
if sys_path_0_set and scriptDir != sys_path_0: |
|
60
|
|
|
raise Exception("runSavu attempted to change sys.path[0] in a way that " |
|
61
|
|
|
"could cause a race condition. Current sys.path[0] is {!r}, " |
|
62
|
|
|
"trying to set to {!r}".format(sys_path_0, scriptDir)) |
|
63
|
|
|
else: |
|
64
|
|
|
sys.path[0] = scriptDir |
|
65
|
|
|
sys_path_0_set = True |
|
66
|
|
|
|
|
67
|
|
|
if not plugin_object: |
|
68
|
|
|
parameters = {} |
|
69
|
|
|
# slight repack here |
|
70
|
|
|
for key in list(params.keys()): |
|
71
|
|
|
# print "here" |
|
72
|
|
|
val = params[key]["value"] |
|
73
|
|
|
if type(val)==type(''): |
|
74
|
|
|
val = val.replace('\n','').strip() |
|
75
|
|
|
# print val |
|
76
|
|
|
parameters[key] = val |
|
77
|
|
|
print(("val: {}".format(val))) |
|
78
|
|
|
# print "initialising the object" |
|
79
|
|
|
plugin_object = _savu_setup(path2plugin, inputs, parameters) |
|
80
|
|
|
persistence['plugin_object'] = plugin_object |
|
81
|
|
|
axis_labels, axis_values = process_init(plugin_object) |
|
82
|
|
|
# print "I did the initialisation" |
|
83
|
|
|
# print "axis labels",axis_labels |
|
84
|
|
|
# print "axis_values", axis_values |
|
85
|
|
|
# print plugin_object |
|
86
|
|
|
chkstring = [any(isinstance(ix, str) for ix in axis_values[label]) for label in axis_labels] |
|
87
|
|
|
if any(chkstring): # are any axis values strings we instead make this an aux out |
|
88
|
|
|
metaOnly = True |
|
89
|
|
|
# print "AXIS LABELS"+str(axis_values) |
|
90
|
|
|
string_key = axis_labels[chkstring.index(True)] |
|
91
|
|
|
aux = OrderedDict.fromkeys(axis_values[string_key]) |
|
92
|
|
|
# print aux.keys() |
|
93
|
|
|
else: |
|
94
|
|
|
string_key = axis_labels[0]# will it always be the first one? |
|
95
|
|
|
if not metaOnly: |
|
96
|
|
|
if len(axis_labels) == 1: |
|
97
|
|
|
result['xaxis']=axis_values[axis_labels[0]] |
|
98
|
|
|
result['xaxis_title']=axis_labels[0] |
|
99
|
|
|
if len(axis_labels) == 2: |
|
100
|
|
|
# print "set the output axes" |
|
101
|
|
|
x = axis_labels[0] |
|
102
|
|
|
result['xaxis_title']=x |
|
103
|
|
|
y = axis_labels[1] |
|
104
|
|
|
result['yaxis_title']=y |
|
105
|
|
|
result['yaxis']=axis_values[y] |
|
106
|
|
|
result['xaxis']=axis_values[x] |
|
107
|
|
|
else: |
|
108
|
|
|
pass |
|
109
|
|
|
finally: |
|
110
|
|
|
sys_path_0_lock.release() |
|
111
|
|
|
|
|
112
|
|
|
if plugin_object.get_max_frames()>1: # we need to get round this since we are frame independant |
|
113
|
|
|
data = np.expand_dims(inputs['data'], 0) |
|
114
|
|
|
else: |
|
115
|
|
|
data = inputs['data'] |
|
116
|
|
|
|
|
117
|
|
|
print(("metaOnly: {}".format(metaOnly))) |
|
118
|
|
|
|
|
119
|
|
|
if not metaOnly: |
|
120
|
|
|
|
|
121
|
|
|
out = plugin_object.process_frames([data]) |
|
122
|
|
|
# print "ran the plugin" |
|
123
|
|
|
|
|
124
|
|
|
result['data'] = out |
|
125
|
|
|
elif metaOnly: |
|
126
|
|
|
result['data'] = inputs['data'] |
|
127
|
|
|
# print type(result['data']) |
|
128
|
|
|
out_array = plugin_object.process_frames([inputs['data']]) |
|
129
|
|
|
|
|
130
|
|
|
# print aux.keys() |
|
131
|
|
|
|
|
132
|
|
|
for k,key in enumerate(aux.keys()): |
|
133
|
|
|
aux[key]=np.array([out_array[k]])# wow really |
|
134
|
|
|
|
|
135
|
|
|
result['auxiliary'] = aux |
|
136
|
|
|
t2 = time.time() |
|
137
|
|
|
print("time to runSavu = "+str((t2-t1))) |
|
138
|
|
|
return result |
|
139
|
|
|
|
|
140
|
|
|
|
|
141
|
|
|
def _savu_setup(path2plugin, inputs, parameters): |
|
142
|
|
|
print("running _savu_setup") |
|
143
|
|
|
parameters['in_datasets'] = [inputs['dataset_name']] |
|
144
|
|
|
parameters['out_datasets'] = [inputs['dataset_name']] |
|
145
|
|
|
plugin = get_plugin(path2plugin.split('.py')[0]+'.py') |
|
146
|
|
|
plugin.exp = setup_exp_and_data(inputs, inputs['data'], plugin) |
|
147
|
|
|
plugin._set_parameters(parameters) |
|
148
|
|
|
plugin._set_plugin_datasets() |
|
149
|
|
|
plugin.setup() |
|
150
|
|
|
return plugin |
|
151
|
|
|
|
|
152
|
|
|
|
|
153
|
|
|
def process_init(plugin): |
|
154
|
|
|
axis_labels = plugin.get_out_datasets()[0].get_axis_label_keys() |
|
155
|
|
|
axis_labels.remove('idx') # get the labels |
|
156
|
|
|
axis_values = {} |
|
157
|
|
|
plugin._clean_up() # this copies the metadata! |
|
158
|
|
|
for label in axis_labels: |
|
159
|
|
|
axis_values[label] = plugin.get_out_datasets()[0].meta_data.get(label) |
|
160
|
|
|
plugin.base_pre_process() |
|
161
|
|
|
plugin.pre_process() |
|
162
|
|
|
return axis_labels, axis_values |
|
163
|
|
|
|
|
164
|
|
|
|
|
165
|
|
|
def setup_exp_and_data(inputs, data, plugin): |
|
166
|
|
|
exp = DawnExperiment(get_options()) |
|
167
|
|
|
data_obj = exp.create_data_object('in_data', inputs['dataset_name']) |
|
168
|
|
|
data_obj.data = None |
|
169
|
|
|
if len(inputs['data'].shape)==1: |
|
170
|
|
|
# print data.shape |
|
171
|
|
|
if inputs['xaxis_title'] is None or inputs['xaxis_title'].isspace(): |
|
172
|
|
|
inputs['xaxis_title']='x' |
|
173
|
|
|
inputs['xaxis'] = np.arange(inputs['data'].shape[0]) |
|
174
|
|
|
data_obj.set_axis_labels('idx.units', inputs['xaxis_title'] + '.units') |
|
175
|
|
|
data_obj.meta_data.set('idx', np.array([1])) |
|
176
|
|
|
data_obj.meta_data.set(str(inputs['xaxis_title']), inputs['xaxis']) |
|
177
|
|
|
data_obj.add_pattern(plugin.get_plugin_pattern(), core_dims=(1,), slice_dims=(0, )) |
|
178
|
|
|
data_obj.add_pattern('SINOGRAM', core_dims=(1,), slice_dims=(0, )) # good to add these two on too |
|
179
|
|
|
data_obj.add_pattern('PROJECTION', core_dims=(1,), slice_dims=(0, )) |
|
180
|
|
|
if len(inputs['data'].shape)==2: |
|
181
|
|
|
if inputs['xaxis_title'] is None or inputs['xaxis_title'].isspace(): |
|
182
|
|
|
print("set x") |
|
183
|
|
|
inputs['xaxis_title']='x' |
|
184
|
|
|
inputs['xaxis'] = np.arange(inputs['data'].shape[0]) |
|
185
|
|
|
if inputs['yaxis_title'] is None or inputs['yaxis_title'].isspace(): |
|
186
|
|
|
print("set y") |
|
187
|
|
|
inputs['yaxis_title']='y' |
|
188
|
|
|
size_y_axis = inputs['data'].shape[1] |
|
189
|
|
|
inputs['yaxis'] = np.arange(size_y_axis) |
|
190
|
|
|
|
|
191
|
|
|
data_obj.set_axis_labels('idx.units', inputs['xaxis_title'] + '.units', inputs['yaxis_title'] + '.units') |
|
192
|
|
|
data_obj.meta_data.set('idx', np.array([1])) |
|
193
|
|
|
data_obj.meta_data.set(str(inputs['xaxis_title']), inputs['xaxis']) |
|
194
|
|
|
data_obj.meta_data.set(str(inputs['yaxis_title']), inputs['yaxis']) |
|
195
|
|
|
data_obj.add_pattern(plugin.get_plugin_pattern(), core_dims=(1,2,), slice_dims=(0, )) |
|
196
|
|
|
data_obj.add_pattern('SINOGRAM', core_dims=(1,2,), slice_dims=(0, )) # good to add these two on too |
|
197
|
|
|
data_obj.add_pattern('PROJECTION', core_dims=(1,2,), slice_dims=(0, )) |
|
198
|
|
|
|
|
199
|
|
|
data_obj.set_shape((1, ) + data.shape) # need to add for now for slicing... |
|
200
|
|
|
data_obj.get_preview().set_preview([]) |
|
201
|
|
|
return exp |
|
202
|
|
|
|
|
203
|
|
|
class DawnExperiment(Experiment): |
|
204
|
|
|
def __init__(self, options): |
|
205
|
|
|
self.index={"in_data": {}, "out_data": {}, "mapping": {}} |
|
206
|
|
|
self.meta_data = MetaData(get_options()) |
|
207
|
|
|
self.nxs_file = None |
|
208
|
|
|
|
|
209
|
|
|
def get_options(): |
|
210
|
|
|
options = {} |
|
211
|
|
|
options['dawn_runner'] = True |
|
212
|
|
|
options['transport'] = 'hdf5' |
|
213
|
|
|
options['process_names'] = 'CPU0' |
|
214
|
|
|
options['processes'] = 'CPU0' |
|
215
|
|
|
options['data_file'] = '' |
|
216
|
|
|
options['process_file'] = '' |
|
217
|
|
|
options['out_path'] = '' |
|
218
|
|
|
options['inter_path'] = '' |
|
219
|
|
|
options['log_path'] = '' |
|
220
|
|
|
options['run_type'] = '' |
|
221
|
|
|
options['verbose'] = 'True' |
|
222
|
|
|
options['system_params'] = _set_system_params() |
|
223
|
|
|
return options |
|
224
|
|
|
|
|
225
|
|
|
def _set_system_params(): |
|
226
|
|
|
# look in conda environment to see which version is being used |
|
227
|
|
|
savu_path = sys.modules['savu'].__path__[0] |
|
228
|
|
|
sys_files = os.path.join( |
|
229
|
|
|
os.path.dirname(savu_path), 'system_files') |
|
230
|
|
|
subdirs = os.listdir(sys_files) |
|
231
|
|
|
sys_folder = 'dls' if len(subdirs) > 1 else subdirs[0] |
|
232
|
|
|
fname = 'system_parameters.yml' |
|
233
|
|
|
sys_file = os.path.join(sys_files, sys_folder, fname) |
|
234
|
|
|
return yaml.read_yaml(sys_file) |
|
235
|
|
|
|
DiamondLightSource /
Savu
