|
1
|
|
|
""" |
|
2
|
|
|
Processor base class and helper functions |
|
3
|
|
|
""" |
|
4
|
|
|
|
|
5
|
|
|
__all__ = [ |
|
6
|
|
|
'Processor', |
|
7
|
|
|
'generate_processor_help', |
|
8
|
|
|
'run_cli', |
|
9
|
|
|
'run_processor' |
|
10
|
|
|
] |
|
11
|
|
|
|
|
12
|
|
|
from os import makedirs |
|
13
|
|
|
from os.path import exists, isdir, join |
|
14
|
|
|
from shutil import copyfileobj |
|
15
|
|
|
import json |
|
16
|
|
|
import os |
|
17
|
|
|
import re |
|
18
|
|
|
from pkg_resources import resource_filename |
|
19
|
|
|
|
|
20
|
|
|
import requests |
|
21
|
|
|
|
|
22
|
|
|
from ocrd_utils import ( |
|
23
|
|
|
VERSION as OCRD_VERSION, |
|
24
|
|
|
MIMETYPE_PAGE, |
|
25
|
|
|
list_resource_candidates, |
|
26
|
|
|
list_all_resources, |
|
27
|
|
|
XDG_CACHE_HOME |
|
28
|
|
|
) |
|
29
|
|
|
from ocrd_validators import ParameterValidator |
|
30
|
|
|
from ocrd_models.ocrd_page import MetadataItemType, LabelType, LabelsType |
|
31
|
|
|
|
|
32
|
|
|
# XXX imports must remain for backwards-compatibilty |
|
33
|
|
|
from .helpers import run_cli, run_processor, generate_processor_help # pylint: disable=unused-import |
|
34
|
|
|
|
|
35
|
|
|
class Processor(): |
|
36
|
|
|
""" |
|
37
|
|
|
A processor runs an algorithm based on the workspace, the mets.xml in the |
|
38
|
|
|
workspace (and the input files defined therein) as well as optional |
|
39
|
|
|
parameter. |
|
40
|
|
|
""" |
|
41
|
|
|
|
|
42
|
|
|
def __init__( |
|
43
|
|
|
self, |
|
44
|
|
|
workspace, |
|
45
|
|
|
ocrd_tool=None, |
|
46
|
|
|
parameter=None, |
|
47
|
|
|
# TODO OCR-D/core#274 |
|
48
|
|
|
# input_file_grp=None, |
|
49
|
|
|
# output_file_grp=None, |
|
50
|
|
|
input_file_grp="INPUT", |
|
51
|
|
|
output_file_grp="OUTPUT", |
|
52
|
|
|
page_id=None, |
|
53
|
|
|
show_help=False, |
|
54
|
|
|
show_version=False, |
|
55
|
|
|
dump_json=False, |
|
56
|
|
|
version=None |
|
57
|
|
|
): |
|
58
|
|
|
if parameter is None: |
|
59
|
|
|
parameter = {} |
|
60
|
|
|
if dump_json: |
|
61
|
|
|
print(json.dumps(ocrd_tool, indent=True)) |
|
62
|
|
|
return |
|
63
|
|
|
self.ocrd_tool = ocrd_tool |
|
64
|
|
|
if show_help: |
|
65
|
|
|
self.show_help() |
|
66
|
|
|
return |
|
67
|
|
|
self.version = version |
|
68
|
|
|
if show_version: |
|
69
|
|
|
self.show_version() |
|
70
|
|
|
return |
|
71
|
|
|
self.workspace = workspace |
|
72
|
|
|
# FIXME HACK would be better to use pushd_popd(self.workspace.directory) |
|
73
|
|
|
# but there is no way to do that in process here since it's an |
|
74
|
|
|
# overridden method. chdir is almost always an anti-pattern. |
|
75
|
|
|
if self.workspace: |
|
76
|
|
|
os.chdir(self.workspace.directory) |
|
77
|
|
|
self.input_file_grp = input_file_grp |
|
78
|
|
|
self.output_file_grp = output_file_grp |
|
79
|
|
|
self.page_id = None if page_id == [] or page_id is None else page_id |
|
80
|
|
|
parameterValidator = ParameterValidator(ocrd_tool) |
|
81
|
|
|
report = parameterValidator.validate(parameter) |
|
82
|
|
|
if not report.is_valid: |
|
83
|
|
|
raise Exception("Invalid parameters %s" % report.errors) |
|
84
|
|
|
self.parameter = parameter |
|
85
|
|
|
|
|
86
|
|
|
def show_help(self): |
|
87
|
|
|
print(generate_processor_help(self.ocrd_tool, processor_instance=self)) |
|
88
|
|
|
|
|
89
|
|
|
def show_version(self): |
|
90
|
|
|
print("Version %s, ocrd/core %s" % (self.version, OCRD_VERSION)) |
|
91
|
|
|
|
|
92
|
|
|
def verify(self): |
|
93
|
|
|
""" |
|
94
|
|
|
Verify that the input fulfills the processor's requirements. |
|
95
|
|
|
""" |
|
96
|
|
|
return True |
|
97
|
|
|
|
|
98
|
|
|
def process(self): |
|
99
|
|
|
""" |
|
100
|
|
|
Process the workspace |
|
101
|
|
|
""" |
|
102
|
|
|
raise Exception("Must be implemented") |
|
103
|
|
|
|
|
104
|
|
|
|
|
105
|
|
|
def add_metadata(self, pcgts): |
|
106
|
|
|
""" |
|
107
|
|
|
Adds PAGE-XML MetadataItem describing the processing step |
|
108
|
|
|
""" |
|
109
|
|
|
pcgts.get_Metadata().add_MetadataItem( |
|
110
|
|
|
MetadataItemType(type_="processingStep", |
|
111
|
|
|
name=self.ocrd_tool['steps'][0], |
|
112
|
|
|
value=self.ocrd_tool['executable'], |
|
113
|
|
|
Labels=[LabelsType( |
|
114
|
|
|
externalModel="ocrd-tool", |
|
115
|
|
|
externalId="parameters", |
|
116
|
|
|
Label=[LabelType(type_=name, |
|
117
|
|
|
value=self.parameter[name]) |
|
118
|
|
|
for name in self.parameter.keys()])])) |
|
119
|
|
|
|
|
120
|
|
|
def resolve_resource(self, parameter_name, val): |
|
121
|
|
|
""" |
|
122
|
|
|
Resolve a resource name with the algorithm in |
|
123
|
|
|
https://ocr-d.de/en/spec/ocrd_tool#file-parameters |
|
124
|
|
|
|
|
125
|
|
|
Args: |
|
126
|
|
|
parameter_name (string): name of parameter to resolve resource for |
|
127
|
|
|
val (string): resource value to resolve |
|
128
|
|
|
""" |
|
129
|
|
|
executable = self.ocrd_tool['executable'] |
|
130
|
|
|
try: |
|
131
|
|
|
param = self.ocrd_tool['parameter'][parameter_name] |
|
132
|
|
|
except KeyError: |
|
133
|
|
|
raise ValueError("Parameter '%s' not defined in ocrd-tool.json" % parameter_name) |
|
134
|
|
|
if not param['mimetype']: |
|
135
|
|
|
raise ValueError("Parameter '%s' is not a file parameter (has no 'mimetype' field)" % |
|
136
|
|
|
parameter_name) |
|
137
|
|
|
if val.startswith('http:') or val.startswith('https:'): |
|
138
|
|
|
cache_dir = join(XDG_CACHE_HOME, executable) |
|
139
|
|
|
cache_key = re.sub('[^A-Za-z0-9]', '', val) |
|
140
|
|
|
cache_fpath = join(cache_dir, cache_key) |
|
141
|
|
|
# TODO Proper caching (make head request for size, If-Modified etc) |
|
142
|
|
|
if not exists(cache_fpath): |
|
143
|
|
|
if not isdir(cache_dir): |
|
144
|
|
|
makedirs(cache_dir) |
|
145
|
|
|
with requests.get(val, stream=True) as r: |
|
146
|
|
|
with open(cache_fpath, 'wb') as f: |
|
147
|
|
|
copyfileobj(r.raw, f) |
|
148
|
|
|
return cache_fpath |
|
149
|
|
|
ret = next([cand for cand in list_resource_candidates(executable, val) if exists(cand)]) |
|
150
|
|
|
if ret: |
|
151
|
|
|
return ret |
|
152
|
|
|
bundled_fpath = resource_filename(__name__, val) |
|
153
|
|
|
if exists(bundled_fpath): |
|
154
|
|
|
return bundled_fpath |
|
155
|
|
|
raise FileNotFoundError("Could not resolve '%s' file parameter value '%s'" % |
|
156
|
|
|
(parameter_name, val)) |
|
157
|
|
|
|
|
158
|
|
|
def list_all_resources(self): |
|
159
|
|
|
""" |
|
160
|
|
|
List all resources found in the filesystem |
|
161
|
|
|
""" |
|
162
|
|
|
return list_all_resources(self.ocrd_tool['executable']) |
|
163
|
|
|
|
|
164
|
|
|
@property |
|
165
|
|
|
def input_files(self): |
|
166
|
|
|
""" |
|
167
|
|
|
List the input files. |
|
168
|
|
|
|
|
169
|
|
|
- If there's a PAGE-XML for the page, take it (and forget about all |
|
170
|
|
|
other files for that page) |
|
171
|
|
|
- Else if there's only one image, take it (and forget about all other |
|
172
|
|
|
files for that page) |
|
173
|
|
|
- Otherwise raise an error (complaining that only PAGE-XML warrants |
|
174
|
|
|
|
|
175
|
|
|
having multiple images for a single page) |
|
176
|
|
|
(https://github.com/cisocrgroup/ocrd_cis/pull/57#issuecomment-656336593) |
|
177
|
|
|
""" |
|
178
|
|
|
ret = self.workspace.mets.find_all_files( |
|
179
|
|
|
fileGrp=self.input_file_grp, pageId=self.page_id, mimetype=MIMETYPE_PAGE) |
|
180
|
|
|
if ret: |
|
181
|
|
|
return ret |
|
182
|
|
|
ret = self.workspace.mets.find_all_files( |
|
183
|
|
|
fileGrp=self.input_file_grp, pageId=self.page_id, mimetype="//image/.*") |
|
184
|
|
|
if self.page_id and len(ret) > 1: |
|
185
|
|
|
raise ValueError("No PAGE-XML %s in fileGrp '%s' but multiple images." % ( |
|
186
|
|
|
"for page '%s'" % self.page_id if self.page_id else '', |
|
187
|
|
|
self.input_file_grp |
|
188
|
|
|
)) |
|
189
|
|
|
return ret |
|
190
|
|
|
|
OCR-D /
core
