Skip to content

Arkindex Workers

arkindex_worker.worker

Base classes to implement Arkindex workers.

Classes

ElementsWorker

ElementsWorker(
    description: str = "Arkindex Elements Worker",
    support_cache: bool = False,
)

Bases: ElementMixin, DatasetMixin, BaseWorker, ClassificationMixin, CorpusMixin, TranscriptionMixin, WorkerVersionMixin, EntityMixin, MetaDataMixin, ImageMixin, ProcessMixin

Base class for ML workers that operate on Arkindex elements.

This class inherits from numerous mixin classes found in other modules of arkindex.worker, which provide helpers to read and write to the Arkindex API.

Parameters:

Name Type Description Default
description str

The worker’s description

'Arkindex Elements Worker'
support_cache bool

Whether the worker supports cache

False
Source code in arkindex_worker/worker/__init__.py
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
def __init__(
    self, description: str = "Arkindex Elements Worker", support_cache: bool = False
):
    """
    :param description: The worker's description
    :param support_cache: Whether the worker supports cache
    """
    super().__init__(description, support_cache)

    self.classes = {}

    self.entity_types = {}
    """Known and available entity types in processed corpus
    """

    self.corpus_types = {}
    """Known and available element types in processed corpus
    """

    self._worker_version_cache = {}
Attributes
entity_types instance-attribute
entity_types = {}

Known and available entity types in processed corpus

corpus_types instance-attribute
corpus_types = {}

Known and available element types in processed corpus

store_activity property
store_activity: bool

Whether or not WorkerActivity support has been enabled on the DataImport used to run this worker.

Functions
get_elements
get_elements() -> (
    Iterable[CachedElement] | list[str] | list[Element]
)

List the elements to be processed, either from the CLI arguments or the cache database when enabled.

Returns:

Type Description
Iterable[CachedElement] | list[str] | list[Element]

An iterable of CachedElement when cache support is enabled, or a list of strings representing element IDs otherwise.

Source code in arkindex_worker/worker/__init__.py
 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
def get_elements(self) -> Iterable[CachedElement] | list[str] | list[Element]:
    """
    List the elements to be processed, either from the CLI arguments or
    the cache database when enabled.

    :return: An iterable of [CachedElement][arkindex_worker.cache.CachedElement] when cache support is enabled,
       or a list of strings representing element IDs otherwise.
    """
    assert not (
        self.args.elements_list and self.args.element
    ), "elements-list and element CLI args shouldn't be both set"

    def invalid_element_id(value: str) -> bool:
        """
        Return whether the ID of an element is a valid UUID or not
        """
        try:
            uuid.UUID(value)
        except Exception:
            return True

        return False

    out = []

    # Load from the cache when available
    # Flake8 wants us to use 'is True', but Peewee only supports '== True'
    cache_query = CachedElement.select().where(
        CachedElement.initial == True  # noqa: E712
    )
    if self.use_cache and cache_query.exists():
        return cache_query
    elif self.args.elements_list:
        # Process elements from JSON file
        data = json.load(self.args.elements_list)
        assert isinstance(data, list), "Elements list must be a list"
        assert len(data), "No elements in elements list"
        out += list(filter(None, [element.get("id") for element in data]))
    elif self.args.element:
        # Add any extra element from CLI
        out += self.args.element
    elif self.process_mode == ProcessMode.Dataset or self.args.set:
        # Elements from datasets
        return list(
            chain.from_iterable(map(self.list_set_elements, self.list_sets()))
        )
    elif self.process_mode == ProcessMode.Export:
        # For export mode processes, use list_process_elements and return element IDs
        return {item["id"] for item in self.list_process_elements()}

    invalid_element_ids = list(filter(invalid_element_id, out))
    assert (
        not invalid_element_ids
    ), f"These element IDs are invalid: {', '.join(invalid_element_ids)}"

    return out
run
run()

Implements an Arkindex worker that goes through each element returned by get_elements. It calls process_element, catching exceptions, and handles saving WorkerActivity updates when enabled.

Source code in arkindex_worker/worker/__init__.py
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
def run(self):
    """
    Implements an Arkindex worker that goes through each element returned by
    [get_elements][arkindex_worker.worker.ElementsWorker.get_elements].
    It calls [process_element][arkindex_worker.worker.ElementsWorker.process_element],
    catching exceptions, and handles saving WorkerActivity updates when enabled.
    """
    self.configure()

    # List all elements either from JSON file
    # or direct list of elements on CLI
    elements = self.get_elements()
    if not elements:
        logger.warning("No elements to process, stopping.")
        sys.exit(1)

    if not self.store_activity:
        logger.info(
            "No worker activity will be stored as it is disabled for this process"
        )

    # Process every element
    count = len(elements)
    failed = 0
    for i, item in enumerate(elements, start=1):
        element = None
        try:
            if isinstance(item, CachedElement | Element):
                # Just use the result of get_elements as the element
                element = item
            else:
                # Load element using the Arkindex API
                element = Element(
                    **self.api_client.request("RetrieveElement", id=item)
                )

            logger.info(f"Processing {element} ({i}/{count})")

            # Process the element and report its progress if activities are enabled
            if self.update_activity(element.id, ActivityState.Started):
                self.process_element(element)
                self.update_activity(element.id, ActivityState.Processed)
            else:
                logger.info(
                    f"Skipping element {element.id} as it was already processed"
                )
                continue
        except Exception as e:
            # Handle errors occurring while retrieving, processing or patching the activity for this element.
            # Count the element as failed in case the activity update to "started" failed with no conflict.
            # This prevent from processing the element
            failed += 1

            # Handle the case where we failed retrieving the element
            element_id = element.id if element else item

            if isinstance(e, ErrorResponse):
                message = f"An API error occurred while processing element {element_id}: {e.title} - {e.content}"
            else:
                message = (
                    f"Failed running worker on element {element_id}: {repr(e)}"
                )

            logger.warning(
                message,
                exc_info=e if self.args.verbose else None,
            )
            if element:
                # Try to update the activity to error state regardless of the response
                with contextlib.suppress(Exception):
                    self.update_activity(element.id, ActivityState.Error)

    message = f'Ran on {count} {pluralize("element", count)}: {count - failed} completed, {failed} failed'
    if failed:
        logger.error(message)
        if failed >= count:  # Everything failed!
            sys.exit(1)
    else:
        logger.info(message)
process_element
process_element(element: Element | CachedElement)

Override this method to implement your worker and process a single Arkindex element at once.

Parameters:

Name Type Description Default
element Element | CachedElement

The element to process. Will be a CachedElement instance if cache support is enabled, and an Element instance otherwise.

required
Source code in arkindex_worker/worker/__init__.py
232
233
234
235
236
237
238
239
def process_element(self, element: Element | CachedElement):
    """
    Override this method to implement your worker and process a single Arkindex element at once.

    :param element: The element to process.
       Will be a CachedElement instance if cache support is enabled,
       and an Element instance otherwise.
    """
update_activity
update_activity(
    element_id: str | uuid.UUID, state: ActivityState
) -> bool

Update the WorkerActivity for this element and worker.

Parameters:

Name Type Description Default
element_id str | UUID

ID of the element.

required
state ActivityState

New WorkerActivity state for this element.

required

Returns:

Type Description
bool

True if the update has been successful or WorkerActivity support is disabled. False if the update has failed due to a conflict; this worker might have already processed this element.

Source code in arkindex_worker/worker/__init__.py
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
def update_activity(
    self, element_id: str | uuid.UUID, state: ActivityState
) -> bool:
    """
    Update the WorkerActivity for this element and worker.

    :param element_id: ID of the element.
    :param state: New WorkerActivity state for this element.
    :returns: True if the update has been successful or WorkerActivity support is disabled.
       False if the update has failed due to a conflict; this worker might have already processed
       this element.
    """
    if not self.store_activity:
        logger.debug(
            "Activity is not stored as the feature is disabled on this process"
        )
        return True

    assert element_id and isinstance(
        element_id, uuid.UUID | str
    ), "element_id shouldn't be null and should be an UUID or str"
    assert isinstance(state, ActivityState), "state should be an ActivityState"

    try:
        self.api_client.request(
            "UpdateWorkerActivity",
            id=self.worker_run_id,
            body={
                "element_id": str(element_id),
                "process_id": self.process_information["id"],
                "state": state.value,
            },
        )
    except ErrorResponse as e:
        if state == ActivityState.Started and e.status_code == 409:
            # 409 conflict error when updating the state of an activity to "started" mean that we
            # cannot process this element. We assume that the reason is that the state transition
            # was forbidden i.e. that the activity was already in a started or processed state.
            # This allow concurrent access to an element activity between multiple processes.
            # Element should not be counted as failed as it is probably handled somewhere else.
            logger.debug(
                f"Cannot start processing element {element_id} due to a conflict. "
                f"Another process could have processed it with the same version already."
            )
            return False
        logger.warning(
            f"Failed to update activity of element {element_id} to {state.value} due to an API error: {e.content}"
        )
        raise e

    logger.debug(f"Updated activity of element {element_id} to {state}")
    return True

DatasetWorker

DatasetWorker(
    description: str = "Arkindex Dataset Worker",
    support_cache: bool = False,
)

Bases: DatasetMixin, BaseWorker, TaskMixin

Base class for ML workers that operate on Arkindex dataset sets.

This class inherits from numerous mixin classes found in other modules of arkindex.worker, which provide helpers to read and write to the Arkindex API.

Parameters:

Name Type Description Default
description str

The worker’s description.

'Arkindex Dataset Worker'
support_cache bool

Whether the worker supports cache.

False
Source code in arkindex_worker/worker/__init__.py
303
304
305
306
307
308
309
310
311
312
313
314
315
316
def __init__(
    self,
    description: str = "Arkindex Dataset Worker",
    support_cache: bool = False,
):
    """
    :param description: The worker's description.
    :param support_cache: Whether the worker supports cache.
    """
    super().__init__(description, support_cache)

    # Path to the dataset compressed archive (containing images and a SQLite database)
    # Set as an instance variable as dataset workers might use it to easily extract its content
    self.downloaded_dataset_artifact: Path | None = None
Functions
cleanup_downloaded_artifact
cleanup_downloaded_artifact() -> None

Cleanup the downloaded dataset artifact if any

Source code in arkindex_worker/worker/__init__.py
318
319
320
321
322
323
324
325
def cleanup_downloaded_artifact(self) -> None:
    """
    Cleanup the downloaded dataset artifact if any
    """
    if not self.downloaded_dataset_artifact:
        return

    self.downloaded_dataset_artifact.unlink(missing_ok=True)
download_dataset_artifact
download_dataset_artifact(dataset: Dataset) -> None

Find and download the compressed archive artifact describing a dataset using the list_artifacts and download_artifact methods.

Parameters:

Name Type Description Default
dataset Dataset

The dataset to retrieve the compressed archive artifact for.

required

Raises:

Type Description
MissingDatasetArchive

When the dataset artifact is not found.

Source code in arkindex_worker/worker/__init__.py
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
def download_dataset_artifact(self, dataset: Dataset) -> None:
    """
    Find and download the compressed archive artifact describing a dataset using
    the [list_artifacts][arkindex_worker.worker.task.TaskMixin.list_artifacts] and
    [download_artifact][arkindex_worker.worker.task.TaskMixin.download_artifact] methods.

    :param dataset: The dataset to retrieve the compressed archive artifact for.
    :raises MissingDatasetArchive: When the dataset artifact is not found.
    """
    extra_dir = self.find_extras_directory()
    archive = extra_dir / dataset.filepath
    if archive.exists():
        return

    # Cleanup the dataset artifact that was downloaded previously
    self.cleanup_downloaded_artifact()

    logger.info(f"Downloading artifact for {dataset}")
    task_id = uuid.UUID(dataset.task_id)
    for artifact in self.list_artifacts(task_id):
        if artifact.path != dataset.filepath:
            continue

        archive.write_bytes(self.download_artifact(task_id, artifact).read())
        self.downloaded_dataset_artifact = archive
        return

    raise MissingDatasetArchive(
        "The dataset compressed archive artifact was not found."
    )
process_set
process_set(set: Set)

Override this method to implement your worker and process a single Arkindex dataset set at once.

Parameters:

Name Type Description Default
set Set

The set to process.

required
Source code in arkindex_worker/worker/__init__.py
358
359
360
361
362
363
def process_set(self, set: Set):
    """
    Override this method to implement your worker and process a single Arkindex dataset set at once.

    :param set: The set to process.
    """
run
run()

Implements an Arkindex worker that goes through each dataset set returned by list_sets.

It calls process_set, catching exceptions.

Source code in arkindex_worker/worker/__init__.py
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
def run(self):
    """
    Implements an Arkindex worker that goes through each dataset set returned by
    [list_sets][arkindex_worker.worker.dataset.DatasetMixin.list_sets].

    It calls [process_set][arkindex_worker.worker.DatasetWorker.process_set],
    catching exceptions.
    """
    self.configure()

    dataset_sets: list[Set] = list(self.list_sets())
    if not dataset_sets:
        logger.warning("No sets to process, stopping.")
        sys.exit(1)

    # Process every set
    count = len(dataset_sets)
    failed = 0
    for i, dataset_set in enumerate(dataset_sets, start=1):
        try:
            assert (
                dataset_set.dataset.state == DatasetState.Complete.value
            ), "When processing a set, its dataset state should be Complete."

            logger.info(f"Retrieving data for {dataset_set} ({i}/{count})")
            self.download_dataset_artifact(dataset_set.dataset)

            logger.info(f"Processing {dataset_set} ({i}/{count})")
            self.process_set(dataset_set)
        except Exception as e:
            # Handle errors occurring while retrieving or processing this dataset set
            failed += 1

            if isinstance(e, ErrorResponse):
                message = f"An API error occurred while processing {dataset_set}: {e.title} - {e.content}"
            else:
                message = f"Failed running worker on {dataset_set}: {repr(e)}"

            logger.warning(message, exc_info=e if self.args.verbose else None)

    # Cleanup the latest downloaded dataset artifact
    self.cleanup_downloaded_artifact()

    message = f'Ran on {count} {pluralize("set", count)}: {count - failed} completed, {failed} failed'
    if failed:
        logger.error(message)
        if failed >= count:  # Everything failed!
            sys.exit(1)
    else:
        logger.info(message)

Functions