global: removal of workflows dependency

.travis.invenio.cfg

@@ -31,7 +31,6 @@ ASSETS_AUTO_BUILD = False
 
 PACKAGES = [
     'invenio_oaiharvester',
-    'invenio_workflows',
     'invenio_accounts',
     'invenio_records',
     'invenio_formatter',

README.rst

@@ -1,6 +1,6 @@
 ..
     This file is part of Invenio.
-    Copyright (C) 2015 CERN.
+    Copyright (C) 2015, 2016 CERN.
 
     Invenio is free software; you can redistribute it
     and/or modify it under the terms of the GNU General Public License as
@@ -51,9 +51,10 @@ Invenio module for OAI-PMH metadata harvesting between repositories.
 Features
 ========
 
-This module allows you to easily harvest OAI-PMH repositories, thanks to the `Sickle`_ module, and feed the
-output into your ingestion workflows, or simply to files. You can configure
-your OAI-PMH sources via a web-interface and run or schedule immediate harvesting jobs
+This module allows you to easily harvest OAI-PMH repositories, thanks to the `Sickle`_ module, and via signals
+you can hook the output into your application, or simply to files.
+
+You keep configurations of your OAI-PMH sources via SQLAlchemy models and run or schedule immediate harvesting jobs
 via command-line or regularly via `Celery beat`_.
 
 .. _Celery beat: http://celery.readthedocs.org/en/latest/userguide/periodic-tasks.html
@@ -73,43 +74,40 @@ If you want to have your harvested records saved in a directory automatically, i
 
 .. code-block:: shell
 
-    inveniomanage oaiharvester get -u http://export.arxiv.org/oai2 -i oai:arXiv.org:1507.07286 -o dir
-
+    inveniomanage oaiharvester get -u http://export.arxiv.org/oai2 -i oai:arXiv.org:1507.07286 -d /tmp
 
-Note the  output ``-o`` parameter that specifies how to output the harvested records. The three options are:
 
-   * Sent to a workflow (E.g. `-o workflow`)
-   * Saved files in a folder (E.g. `-o dir`)
-   * Printed to stdout (default)
+Note the directory ``-d`` parameter that specifies a directory to save harvested XML files.
 
 
-Harvesting with workflows
-=========================
+Integration with your application
+=================================
 
-.. code-block:: shell
+If you want to integrate ``invenio-oaiharvester`` into your application, you should hook into
+the signals sent by the harvester upon completed harvesting.
 
-    inveniomanage oaiharvester get -u http://export.arxiv.org/oai2 -i oai:arXiv.org:1507.07286 -o workflow
+See ``invenio_oaiharvester.signals:oaiharvest_finished``.
 
-When you send an harvested record to a workflow you can process the harvested
-files however you'd like and then even upload it automatically into your own repository.
-
-This module already provides some
+Check also the defined Celery tasks under ``invenio_oaiharvester.tasks``.
 
 
 Managing OAI-PMH sources
 ========================
 
 If you want to store configuration for an OAI repository, you can use the
-administration interface available via the admin panel. This is useful if you regularly need to query a server.
+SQLAlchemy model ``invenio_oaiharvester.models:OaiHARVEST``.
 
-Here you can add information about the server URL, metadataPrefix to use etc. This information is also available when scheduling and running tasks:
+This is useful if you regularly need to query a server.
 
+Here you can add information about the server URL, metadataPrefix to use etc.
+This information is also available when scheduling and running tasks:
 
 .. code-block:: shell
 
     inveniomanage oaiharvester get -n somerepo -i oai:example.org:1234
 
-Here we are using the `-n, --name` parameter to specify which stored OAI-PMH source to query, by name.
+Here we are using the `-n, --name` parameter to specify which configured
+OAI-PMH source to query, using the ``name`` property.
 
 
 API
@@ -120,6 +118,8 @@ If you need to schedule or run harvests via Python, you can use our API:
 .. code-block:: python
 
     from invenio_oaiharvester.api import get_records
-    for rec in get_records(identifiers=["oai:arXiv.org:1207.7214"],
-                           url="http://export.arxiv.org/oai2"):
+
+    request, records = get_records(identifiers=["oai:arXiv.org:1207.7214"],
+                                   url="http://export.arxiv.org/oai2")
+    for record in records:
         print rec.raw

invenio_oaiharvester/__init__.py

@@ -1,7 +1,7 @@
 # -*- coding: utf-8 -*-
 #
 # This file is part of Invenio.
-# Copyright (C) 2013 CERN.
+# Copyright (C) 2013, 2015, 2016 CERN.
 #
 # Invenio is free software; you can redistribute it and/or
 # modify it under the terms of the GNU General Public License as
@@ -18,3 +18,9 @@
 # 59 Temple Place, Suite 330, Boston, MA 02111-1307, USA.
 
 from __future__ import absolute_import, print_function, unicode_literals
+
+from .version import __version__
+
+from .api import get_records, list_records
+
+__all__ = ('__version__', 'get_records', 'list_records')

invenio_oaiharvester/api.py

@@ -1,7 +1,7 @@
 # -*- coding: utf-8 -*-
 #
 # This file is part of Invenio.
-# Copyright (C) 2015 CERN.
+# Copyright (C) 2015, 2016 CERN.
 #
 # Invenio is free software; you can redistribute it and/or
 # modify it under the terms of the GNU General Public License as
@@ -19,36 +19,40 @@
 
 from __future__ import absolute_import, print_function, unicode_literals
 
+import datetime
+
 from sickle import Sickle
 
 from .errors import NameOrUrlMissing, WrongDateCombination
 from .utils import get_oaiharvest_object
 
 
 def list_records(metadata_prefix=None, from_date=None, until_date=None,
-                 url=None, name=None, setSpec=None):
+                 url=None, name=None, setspecs=None):
     """Harvest records from an OAI repo, based on datestamp and/or set parameters.
 
     :param metadata_prefix: The prefix for the metadata return (defaults to 'oai_dc').
     :param from_date: The lower bound date for the harvesting (optional).
     :param until_date: The upper bound date for the harvesting (optional).
     :param url: The The url to be used to create the endpoint.
     :param name: The name of the OaiHARVEST object that we want to use to create the endpoint.
-    :param setSpec: The 'set' criteria for the harvesting (optional).
+    :param setspecs: The 'set' criteria for the harvesting (optional).
     :return: An iterator of harvested records.
     """
-    if url:
-        request = Sickle(url)
-    elif name:
-        request, _metadata_prefix, lastrun = get_from_oai_name(name)
+    if name:
+        url, _metadata_prefix, lastrun, _setspecs = get_info_by_oai_name(name)
 
         # In case we provide a prefix, we don't want it to be
         # overwritten by the one we get from the name variable.
         if metadata_prefix is None:
             metadata_prefix = _metadata_prefix
-    else:
+        if setspecs is None:
+            setspecs = _setspecs
+    elif not url:
         raise NameOrUrlMissing("Retry using the parameters -n <name> or -u <url>.")
 
+    request = Sickle(url)
+
     # By convention, when we have a url we have no lastrun, and when we use
     # the name we can either have from_date (if provided) or lastrun.
     dates = {
@@ -60,12 +64,19 @@ def list_records(metadata_prefix=None, from_date=None, until_date=None,
     if (dates['until'] is not None) and (dates['from'] > dates['until']):
         raise WrongDateCombination("'Until' date larger than 'from' date.")
 
-    if metadata_prefix is None:
-        metadata_prefix = "oai_dc"
-
-    return request.ListRecords(metadataPrefix=metadata_prefix,
-                               set=setSpec,
-                               **dates)
+    lastrun_date = datetime.datetime.now()
+    records = []
+    for spec in setspecs.split():
+        for record in request.ListRecords(metadataPrefix=metadata_prefix or "oai_dc",
+                                          set=spec,
+                                          **dates):
+            records.append(record)
+    # Update lastrun?
+    if from_date is None and until_date is None and name is not None:
+        oai_source = get_oaiharvest_object(name)
+        oai_source.update_lastrun(lastrun_date)
+        oai_source.save()
+    return request, records
 
 
 def get_records(identifiers, metadata_prefix=None, url=None, name=None):
@@ -77,39 +88,34 @@ def get_records(identifiers, metadata_prefix=None, url=None, name=None):
     :param name: The name of the OaiHARVEST object that we want to use to create the endpoint.
     :return: An iterator of harvested records.
     """
-    if url:
-        request = Sickle(url)
-    elif name:
-        request, _metadata_prefix, _ = get_from_oai_name(name)
+    if name:
+        url, _metadata_prefix, _, __ = get_info_by_oai_name(name)
 
         # In case we provide a prefix, we don't want it to be
         # overwritten by the one we get from the name variable.
         if metadata_prefix is None:
             metadata_prefix = _metadata_prefix
-    else:
+    elif not url:
         raise NameOrUrlMissing("Retry using the parameters -n <name> or -u <url>.")
 
-    if metadata_prefix is None:
-        metadata_prefix = "oai_dc"
-
+    request = Sickle(url)
+    records = []
     for identifier in identifiers:
         arguments = {
             'identifier': identifier,
-            'metadataPrefix': metadata_prefix
+            'metadataPrefix': metadata_prefix or "oai_dc"
         }
-        yield request.GetRecord(**arguments)
+        records.append(request.GetRecord(**arguments))
+    return request, records
 
 
-def get_from_oai_name(name):
+def get_info_by_oai_name(name):
     """Get basic OAI request data from the OaiHARVEST model.
 
     :param name: name of the source (OaiHARVEST.name)
 
-    :return: (Sickle obj, metadataprefix, lastrun)
+    :return: (url, metadataprefix, lastrun as YYYY-MM-DD, setspecs)
     """
     obj = get_oaiharvest_object(name)
-
-    req = Sickle(obj.baseurl)
-    metadata_prefix = obj.metadataprefix
-    lastrun = obj.lastrun
-    return req, metadata_prefix, lastrun
+    lastrun = obj.lastrun.strftime("%Y-%m-%d")
+    return obj.baseurl, obj.metadataprefix, lastrun, obj.setspecs

invenio_oaiharvester/config.py

@@ -1,7 +1,7 @@
 # -*- coding: utf-8 -*-
 #
 # This file is part of Invenio.
-# Copyright (C) 2014, 2015 CERN.
+# Copyright (C) 2014, 2015, 2016 CERN.
 #
 # Invenio is free software; you can redistribute it and/or
 # modify it under the terms of the GNU General Public License as
@@ -34,6 +34,3 @@
 
 OAIHARVESTER_STORAGEDIR = os.path.join(CFG_DATADIR, "oaiharvester", "storage")
 """Path to a storage directory where the oaiharvester may put files."""
-
-OAIHARVESTER_RECORD_ARXIV_ID_LOOKUP = "system_control_number.value"
-"""Path to the arXiv ID value used by sample post-process tasks."""

invenio_oaiharvester/errors.py

@@ -1,7 +1,7 @@
 # -*- coding: utf-8; -*-
 #
 # This file is part of Invenio.
-# Copyright (C) 2015 CERN.
+# Copyright (C) 2015, 2016 CERN.
 #
 # Invenio is free software; you can redistribute it and/or
 # modify it under the terms of the GNU General Public License as
@@ -36,11 +36,3 @@ class WrongDateCombination(Exception):
 
 class IdentifiersOrDates(Exception):
     """Identifiers cannot be used in combination with dates."""
-
-
-class WrongOutputIdentifier(Exception):
-    """Output type not recognized. Try 'workflow', directory' or omit for stdout."""
-
-
-class WorkflowNotFound(Exception):
-    """Workflow not found. Try '-o workflow -w <workflow name> or provide a name (-n <name>)."""