def work(self):
"""
Look for Documents in the given session Return for which no Analysis exists and
creates one with Status.New. Returns True iff some Analyses were created
"""
# start a new session for each job
session = Session()
try:
# Get a Document
# ... for which no Analysis exists
# ... and lock it for updates
# ... sort by created date
# ... pick the first (oldest)
gkgs = session.query(Gkg) \
.filter(~session.query(Analysis).filter(Gkg.id == Analysis.gkg_id).exists()) \
.with_for_update() \
.order_by(Gkg.date) \
.limit(1000).all()
if len(gkgs) == 0:
return False # no work to be done
for gkg in gkgs:
analysis = Analysis(gkg=gkg, status=Status.NEW)
session.add(analysis)
session.commit()
logger.info(
"Worker {} created Analysis {} in status {}".format(
os.getpid(), analysis.gkg_id, analysis.status))
finally:
# make sure to release a FOR UPDATE lock, if we got one
if session is not None:
session.rollback()
session.close()
return True
class TestScraper(TestCase):
def setUp(self):
db_host = os.environ.get('DB_HOST')
db_url = 'postgresql://{user}:{passwd}@{db_host}/{db}'.format(
user='******', passwd='tester', db_host=db_host, db='idetect_test')
engine = create_engine(db_url)
Session.configure(bind=engine)
Base.metadata.drop_all(engine)
Base.metadata.create_all(engine)
self.session = Session()
def tearDown(self):
self.session.rollback()
for gkg in self.session.query(Gkg).all():
self.session.delete(gkg)
self.session.commit()
def test_scrape_html(self):
gkg = Gkg(
document_identifier="http://www.cnn.com/2013/08/23/us/hurricane-katrina-statistics-fast-facts/index.html")
analysis = Analysis(gkg=gkg, status=Status.NEW)
self.session.add(analysis)
self.session.commit()
scrape(analysis)
content = analysis.content
self.assertEqual("text", content.content_type)
self.assertTrue("Katrina" in content.content_clean)
self.assertTrue("Louisiana" in content.content_clean)
self.assertTrue("\n" not in content.content_clean)
self.assertTrue(content.content_ts is not None)
matches = (
self.session.query(DocumentContent)
.filter(DocumentContent.content_ts.match('Katrina & Louisiana')).all()
)
self.assertIn(content, matches)
def test_scrape_pdf(self):
gkg = Gkg(
document_identifier="https://www1.ncdc.noaa.gov/pub/data/extremeevents/specialreports/Hurricane-Katrina.pdf")
analysis = Analysis(gkg=gkg, status=Status.NEW)
self.session.add(analysis)
self.session.commit()
scrape(analysis)
content = analysis.content
self.assertEqual("pdf", content.content_type)
self.assertTrue("Katrina" in content.content)
self.assertTrue("Louisiana" in content.content)
self.assertTrue("\n" not in content.content)
def work(self):
"""
Look for analyses in the given session and run function on them
if any are found, managing status appropriately. Return True iff some Analyses were processed (successfully or not)
"""
# start a new session for each job
session = Session()
try:
# Get an analysis
# ... and lock it for updates
# ... that meets the conditions specified in the filter function
# ... sort by updated date
# ... pick the first (oldest)
analysis = self.filter_function(session.query(Analysis)) \
.with_for_update() \
.order_by(Analysis.updated) \
.first()
if analysis is None:
return False # no work to be done
analysis_status = analysis.status
analysis.create_new_version(self.working_status)
logger.info("Worker {} claimed Analysis {} in status {}".format(
os.getpid(), analysis.gkg_id, analysis_status))
finally:
# make sure to release a FOR UPDATE lock, if we got one
session.rollback()
start = time.time()
try:
# set a timeout so if this worker stalls, we recover
signal.alarm(self.timeout_seconds)
# actually run the work function on this analysis
self.function(analysis)
delta = time.time() - start
logger.info("Worker {} processed Analysis {} {} -> {} {}s".format(
os.getpid(), analysis.gkg_id, analysis_status,
self.success_status, delta))
analysis.error_msg = None
analysis.processing_time = delta
analysis.create_new_version(self.success_status)
except Exception as e:
delta = time.time() - start
logger.warning(
"Worker {} failed to process Analysis {} {} -> {}".format(
os.getpid(), analysis.gkg_id, analysis_status,
self.failure_status),
exc_info=e)
analysis.error_msg = str(e)
analysis.processing_time = delta
analysis.create_new_version(self.failure_status)
session.commit()
finally:
# clear the timeout
signal.alarm(0)
if session is not None:
session.rollback()
session.close()
return True
def homepage():
session = Session()
try:
articles = session.query(Analysis).order_by(desc(
Analysis.updated)).limit(10).all()
counts = Analysis.status_counts(session)
cat_counts = Analysis.category_counts(session)
return render_template('index.html',
articles=articles,
counts=counts,
cat_counts=cat_counts)
finally:
session.close()
def article(doc_id):
session = Session()
try:
analysis = session.query(Analysis) \
.filter(Analysis.gkg_id == doc_id).one()
coords = {
tuple(l.latlong.split(","))
for f in analysis.facts for l in f.locations
if l.latlong is not None
}
return render_template('article.html',
article=analysis,
coords=list(coords))
finally:
session.close()
def test_status_update(self):
gkg = self.session.query(Gkg).first()
analysis = Analysis(gkg=gkg, status=Status.NEW)
self.session.add(analysis)
self.session.commit()
analysis.create_new_version(Status.SCRAPING)
self.assertEqual(analysis.status, Status.SCRAPING)
# meanwhile, some other process changed the status of this...
session2 = Session()
try:
other = session2.query(Analysis).get(analysis.gkg_id)
other.create_new_version(Status.SCRAPING_FAILED)
finally:
session2.rollback()
with self.assertRaises(NotLatestException):
analysis.create_new_version(Status.SCRAPED)
def search_url():
url = request.args.get('url')
if url is None:
return json.dumps({'success': False}), 422, {
'ContentType': 'application/json'
}
session = Session()
try:
gkg = session.query(Gkg).filter(
Gkg.document_identifier.like("%" + url + "%")).order_by(
Gkg.date.desc()).first()
if gkg:
resp = jsonify({'doc_id': gkg.id})
resp.status_code = 200
return resp
else:
return json.dumps({'success': False}), 422, {
'ContentType': 'application/json'
}
finally:
session.close()
class TestSyriaYear(TestCase):
syria_location_ids = [
127, 270, 281, 284, 307, 332, 372, 412, 429, 431, 531, 591, 612, 618,
644, 671, 764, 807, 905, 958, 996, 1018, 1019, 1188, 1190, 1212, 1352,
1357, 1524, 1678, 1898, 1981, 1990, 2058, 2060, 2272, 2378, 2735, 2933,
3262, 3323, 3327, 3372, 3391, 3404, 3660, 3708, 3725, 3834, 3915, 3924,
4069, 4172, 4399, 4509, 4648, 4824, 4890, 5017, 5285, 5833, 6053, 6070,
6270, 6760, 6832, 7121, 7122, 7151, 7222, 7244, 7248, 7641, 7723, 7749,
7757, 7827, 7919, 7970, 8078, 8107, 8131, 8166, 8176, 8210, 8222, 8240,
8254, 8367, 8442, 8659, 8660, 8730, 8788, 8793, 8941, 9045, 9167, 9285,
9370, 9531, 9606, 9775, 9909, 9913, 9916, 9917, 9933, 10136, 10312,
10464, 10532, 10795, 10971, 11052, 11076, 11174, 11194, 11216, 11250,
11311, 11501, 11703, 11727, 11916, 11933, 12242, 12387, 12990, 13126,
13130, 13142, 13171, 13348, 13531, 13659, 13722, 14225, 14718, 14732,
14737, 14917, 14930, 14988, 15215, 15257, 15984, 15993, 16188, 17034,
17090, 17373, 17404, 17873, 18019, 18131, 18267, 18396, 18403, 18578,
19550, 19641, 19721, 20180, 21339, 21894, 22003, 22022, 22162, 22201,
22850, 23189, 23414, 23532, 23875, 24851, 25171, 25415, 25894, 25927,
26024, 26283, 26458, 26545, 26909, 27027, 27393, 27507, 28185, 28626,
28628, 29703, 29704, 29754, 29942, 30210, 30286, 30302, 30442, 30993,
31492, 31743
]
start_date = '2017-01-01'
plus_1_yr = '2018-01-01'
plus_6_mo = '2017-07-01'
plus_3_mo = '2017-04-01'
plus_1_mo = '2017-02-01'
def setUp(self):
logger.debug("setUp")
worker_logger = logging.getLogger("idetect.worker")
worker_logger.setLevel(logging.INFO)
logger.debug("Connecting to DB")
db_host = os.environ.get('DB_HOST')
db_port = os.environ.get('DB_PORT', 5432)
db_user = os.environ.get('DB_USER', 'tester')
db_pass = os.environ.get('DB_PASSWORD', 'tester')
db_url = 'postgresql://{user}:{passwd}@{db_host}:{db_port}/{db}'.format(
user=db_user,
passwd=db_pass,
db_host=db_host,
db_port=db_port,
db='idetect')
self.engine = create_engine(db_url, echo=False)
Session.configure(bind=self.engine)
self.session = Session()
self.session.query(FactApi).count()
logger.debug("setUp complete")
def tearDown(self):
logger.debug("tearDown")
self.session.rollback()
logger.debug("sessions rolled back")
logger.debug("tearDown complete")
def test_categories(self):
syr_year_by_category = add_filters(
self.session.query(func.count(FactApi.fact), FactApi.category),
fromdate=self.start_date,
todate=self.plus_1_yr,
location_ids=self.syria_location_ids).group_by(FactApi.category)
t0 = time.time()
result = {
category: count
for count, category in syr_year_by_category.all()
}
t1 = time.time()
# print(result)
self.assertEqual(set(result.keys()), {'Conflict', 'Disaster', 'Other'})
# print(explain_text(self.session, syr_year_by_category))
print(t1 - t0)
self.assertLess(t1 - t0, 1.0)
def test_filter_counts(self):
f_c = get_filter_counts(self.session,
fromdate=self.start_date,
todate=self.plus_1_yr,
location_ids=self.syria_location_ids)
# print(f_c)
self.assertGreater(len(f_c), 1000)
def test_filter_counts_speed(self):
for end_date in (self.plus_1_mo, self.plus_3_mo, self.plus_6_mo):
# Adding this usually fails: , self.plus_1_yr):
t0 = time.time()
f_c = get_filter_counts(self.session,
fromdate=self.start_date,
todate=end_date,
location_ids=self.syria_location_ids)
t1 = time.time()
print('{} - {}: {}s'.format(self.start_date, end_date, t1 - t0))
self.assertLess(
t1 - t0, 1.0,
'Calculating filter counts {} - {} took too long'.format(
self.start_date, end_date))
def test_timeline(self):
t0 = time.time()
counts = get_timeline_counts(self.session,
fromdate=self.start_date,
todate=self.plus_1_yr,
location_ids=self.syria_location_ids)
t1 = time.time()
days = {t['gdelt_day'] for t in counts}
self.assertGreater(len(days), 180)
categories = {t['category'] for t in counts}
self.assertEqual(categories, {'Conflict', 'Disaster', 'Other'})
print(t1 - t0)
self.assertLess(
t1 - t0, 1.0,
'Calculating timeline counts {} - {} took too long'.format(
self.start_date, self.plus_1_yr))
def test_histogram(self):
t0 = time.time()
counts = get_histogram_counts(self.session,
fromdate=self.start_date,
todate=self.plus_1_yr,
location_ids=self.syria_location_ids)
t1 = time.time()
print(len(counts))
figures = {
t['specific_reported_figure']
for t in counts if t['specific_reported_figure']
}
self.assertLess(min(figures), 10)
self.assertGreater(max(figures), 1000000)
units = {t['unit'] for t in counts}
self.assertEqual(units, {'Household', 'Person'})
print(t1 - t0)
self.assertLess(
t1 - t0, 1.0,
'Calculating histogram counts {} - {} took too long'.format(
self.start_date, self.plus_1_yr))
def test_wordcloud(self):
t0 = time.time()
terms = get_wordcloud(self.session,
self.engine,
fromdate=self.start_date,
todate=self.plus_1_yr,
location_ids=self.syria_location_ids)
t1 = time.time()
print(t1 - t0)
print(len(terms))
print(tabulate(terms))
self.assertLess(
t1 - t0, 5.0, 'Calculating wordcloud {} - {} took too long'.format(
self.start_date, self.plus_1_yr))
def test_none_location(self):
# TODO this isn't about Syria, move it somewhere else
counts = get_filter_counts(self.session, location_ids=['NULL'])
self.assertGreater(len(counts), 1000)
self.assertEqual(
counts, get_filter_counts(self.session, location_ids=['null']))
self.assertEqual(counts,
get_filter_counts(self.session, location_ids=[None]))
counts2 = get_filter_counts(self.session, location_ids=['NULL', 1])
self.assertGreater(len(counts2), len(counts))
def test_none_specific_reported_figure_1(self):
# TODO this isn't about Syria, move it somewhere else
counts = get_filter_counts(self.session,
specific_reported_figures=['NULL'])
srf_counts = [
c for c in counts if c['filter_type'] == 'specific_reported_figure'
]
self.assertEqual(len(srf_counts), 1)
self.assertTrue([c for c in srf_counts if c['value'] is None])
self.assertEqual(
counts,
get_filter_counts(self.session,
specific_reported_figures=['null']))
self.assertEqual(
counts,
get_filter_counts(self.session, specific_reported_figures=[None]))
def test_none_specific_reported_figure_2(self):
# TODO this isn't about Syria, move it somewhere else
counts = get_filter_counts(self.session,
specific_reported_figures=['NULL', 1])
srf_counts = [
c for c in counts if c['filter_type'] == 'specific_reported_figure'
]
self.assertEqual(len(srf_counts), 2)
self.assertTrue([c for c in srf_counts if c['value'] is None])
self.assertTrue([c for c in srf_counts if c['value'] == 1])
def test_none_specific_reported_figure_3(self):
# TODO this isn't about Syria, move it somewhere else
counts = get_filter_counts(self.session,
specific_reported_figures=['NULL', 1, 1000])
srf_counts = [
c for c in counts if c['filter_type'] == 'specific_reported_figure'
]
self.assertGreater(len(srf_counts), 2)
self.assertTrue([c for c in srf_counts if c['value'] is None])
self.assertTrue([c for c in srf_counts if c['value'] == 1])
def test_specific_reported_figure(self):
# TODO this isn't about Syria, move it somewhere else
counts = get_filter_counts(self.session,
specific_reported_figures=[1, 1000])
srf_counts = [
c for c in counts if c['filter_type'] == 'specific_reported_figure'
]
self.assertGreater(len(srf_counts), 2)
self.assertFalse([c for c in srf_counts if c['value'] is None])
self.assertTrue([c for c in srf_counts if c['value'] == 1])
def test_filter_ts(self):
t0 = time.time()
query = add_filters(self.session.query(FactApi.content_id,
DocumentContent.content_clean),
fromdate=self.start_date,
todate=self.plus_1_yr,
location_ids=self.syria_location_ids,
ts='Jordan').order_by(FactApi.gdelt_day).limit(32)
results = query.all()
t1 = time.time()
print(t1 - t0)
for id, content_clean in results:
self.assertTrue('jordan' in content_clean.lower())
@skip("Too slow in practice")
def test_filter_ts_exhaustive(self):
# make sure that the query found everything that it was supposed to
t0 = time.time()
query = add_filters(self.session.query(FactApi.content_id,
DocumentContent.content_clean,
FactApi.gdelt_day),
fromdate=self.start_date,
todate=self.plus_1_yr,
location_ids=self.syria_location_ids,
ts='Jordan').order_by(FactApi.gdelt_day).limit(32)
results = query.all()
t1 = time.time()
print(t1 - t0)
matched = set()
max_day = date.min
for id, content_clean, gdelt_day in results:
self.assertTrue('jordan' in content_clean.lower())
matched.add(id)
max_day = max(max_day, gdelt_day)
t2 = time.time()
query = add_filters(self.session.query(FactApi.content_id,
DocumentContent.content_clean),
fromdate=self.start_date,
todate=self.plus_1_yr,
location_ids=self.syria_location_ids).filter(
FactApi.gdelt_day <= max_day)
results = query.all()
t3 = time.time()
print(t2 - t3)
print(len(results))
for id, content_clean in results:
self.assertEqual(id in matched, 'jordan' in content_clean.lower())
def test_urllist(self):
t0 = time.time()
result1 = get_urllist(self.session,
fromdate=self.start_date,
todate=self.plus_1_yr,
location_ids=self.syria_location_ids)
t1 = time.time()
print(t1 - t0)
self.assertEqual(32, len(list(result1)))
t2 = time.time()
result2 = get_urllist(self.session,
offset=32,
limit=100,
fromdate=self.start_date,
todate=self.plus_1_yr,
location_ids=self.syria_location_ids)
t3 = time.time()
print(t3 - t2)
self.assertEqual(100, len(list(result2)))
for r1 in result1:
# print(r1)
for r2 in result2:
self.assertLessEqual(
(r1['gdelt_day'], r1['gkg_id']),
(r2['gdelt_day'], r2['gkg_id']),
)
self.assertIn('display_color', result1[0])
def test_urllist_unique_fact_id(self):
t0 = time.time()
result1 = get_urllist(self.session,
fromdate=self.start_date,
todate=self.plus_1_yr,
limit=1000000,
location_ids=self.syria_location_ids)
t1 = time.time()
print(t1 - t0)
ids = [f['fact_id'] for f in result1]
self.assertGreater(len(ids), 1000)
self.assertEqual(len(ids), len(set(ids)))
def test_urllist_ts(self):
t0 = time.time()
result1 = get_urllist(self.session,
fromdate=self.start_date,
todate=self.plus_1_yr,
location_ids=self.syria_location_ids,
ts='Jordan')
t1 = time.time()
print(t1 - t0)
self.assertEqual(32, len(list(result1)))
t2 = time.time()
result2 = get_urllist(self.session,
offset=32,
limit=100,
fromdate=self.start_date,
todate=self.plus_1_yr,
location_ids=self.syria_location_ids,
ts='Jordan')
t3 = time.time()
print(t3 - t2)
self.assertEqual(100, len(list(result2)))
for r1 in result1:
# print(r1)
for r2 in result2:
self.assertLessEqual(
(r1['gdelt_day'], r1['gkg_id']),
(r2['gdelt_day'], r2['gkg_id']),
)
def test_count_ts(self):
t0 = time.time()
c = get_count(self.session,
fromdate=self.start_date,
todate=self.plus_1_yr,
location_ids=self.syria_location_ids,
ts='Jordan')
t1 = time.time()
print(t1 - t0)
self.assertGreater(c, 5000)
self.assertLess(c, 10000)
def test_urllist_grouped(self):
t0 = time.time()
result = get_urllist_grouped(self.session,
fromdate=self.start_date,
todate=self.plus_1_yr,
location_ids=self.syria_location_ids,
limit=100)
t1 = time.time()
print(t1 - t0)
self.assertEqual(100, len(result))
for entry in result:
self.assertEqual(entry['nfacts'], len(entry['entry']))
for fact in entry['entry']:
self.assertEqual(entry['specific_reported_figure'],
fact['specific_reported_figure'])
self.assertEqual(entry['unit'], fact['unit'])
self.assertEqual(entry['term'], fact['term'])
self.assertGreater(len(fact['content_clean']), 0)
fact_ids = [f['fact'] for f in entry['entry']]
self.assertEqual(len(fact_ids), len(set(fact_ids)),
"fact repeated")
class TestGeoTagger(TestCase):
def setUp(self):
db_host = os.environ.get('DB_HOST')
db_url = 'postgresql://{user}:{passwd}@{db_host}/{db}'.format(
user='******', passwd='tester', db_host=db_host, db='idetect_test')
engine = create_engine(db_url)
Session.configure(bind=engine)
Base.metadata.drop_all(engine)
Base.metadata.create_all(engine)
self.session = Session()
load_countries(self.session)
def tearDown(self):
self.session.rollback()
def test_sets_no_results_flag(self):
"""Sets no-results flag if nothing found"""
results = get_geo_info("xghijdshfkljdes")
self.assertEqual(results['flag'], "no-results")
def test_returns_detail_for_places(self):
"""Returns sufficient level of detail for results"""
results = get_geo_info("Paris")
self.assertNotEqual(results['country_code'], '')
self.assertNotEqual(results['coordinates'], '')
self.assertNotEqual(results['type'], '')
def test_accuracy(self):
"""Returns sufficient level of detail for results"""
results = get_geo_info("Beijing")
self.assertEqual(results['country_code'], 'CHN')
self.assertEqual(results['coordinates'], "39.9059631,116.391248")
def test_country_code(self):
"""Returns sufficient level of detail for results"""
results = get_geo_info("Bidibidi")
self.assertEqual(results['country_code'], 'UGA')
results = get_geo_info("Marrakech")
self.assertEqual(results['country_code'], 'MAR')
results = get_geo_info("Fairfax County")
self.assertEqual(results['country_code'], 'USA')
def test_location_types(self):
"""Corectly distinguishes between Countries, Cities and Subdivisions"""
results = get_geo_info("London")
self.assertEqual(results['type'], LocationType.CITY)
results = get_geo_info("India")
self.assertEqual(results['type'], LocationType.COUNTRY)
results = get_geo_info("Alaska")
self.assertEqual(results['type'], LocationType.SUBDIVISION)
# DONT RUN geotagging if detail already exists
@mock.patch('idetect.geotagger.nominatim_coordinates')
def dont_geotag_if_detail_exists(self, nominatim):
gkg = Gkg(
id=3771256,
gkgrecordid="20170215174500-2503",
date=20170215174500,
document_identifier=
"http://www.philstar.com/headlines/2017/02/16/1672746/yasay-harris-affirm-stronger-phl-us-ties"
)
self.session.add(gkg)
analysis = Analysis(gkg=gkg, status=Status.NEW)
self.session.add(analysis)
content = DocumentContent(
content_clean=
"It was early Saturday when a flash flood hit large parts of India and Pakistan and washed away more than 500 houses"
)
self.session.add(content)
self.session.commit()
analysis.content_id = content.id
self.session.commit()
fact = Fact(unit='person', term='displaced')
self.session.add(fact)
self.session.commit()
loc1 = self.session.query(Location).filter(
Location.location_name == 'India').one_or_none()
fact.locations.append(loc1)
analysis.facts.append(fact)
self.session.commit()
process_locations(analysis)
assert not nominatim.called
def test_create_duplicate_fact(self):
"""Creates duplicate fact if locations from multiple countries exist"""
gkg = Gkg(
id=3771256,
gkgrecordid="20170215174500-2503",
date=20170215174500,
document_identifier=
"http://www.philstar.com/headlines/2017/02/16/1672746/yasay-harris-affirm-stronger-phl-us-ties"
)
self.session.add(gkg)
analysis = Analysis(gkg=gkg, status=Status.NEW)
self.session.add(analysis)
self.session.commit()
fact = Fact(unit='person', term='displaced')
self.session.add(fact)
self.session.commit()
loc1 = self.session.query(Location).filter(
Location.location_name == 'India').one_or_none()
loc2 = self.session.query(Location).filter(
Location.location_name == 'Pakistan').one_or_none()
fact.locations.append(loc1)
fact.locations.append(loc2)
analysis.facts.append(fact)
self.session.commit()
self.assertEqual(1, len(analysis.facts))
process_locations(analysis)
self.assertEqual(2, len(analysis.facts))
fact_countries = [f.iso3 for f in analysis.facts]
self.assertIn('IND', fact_countries)
self.assertIn('PAK', fact_countries)
self.assertEqual(1, len(analysis.facts[0].locations))
self.assertEqual(1, len(analysis.facts[1].locations))
@mock.patch('idetect.geotagger.nominatim_coordinates')
def test_fail_if_geotagging_fails(self, nominatim):
"""Location processing should fail if geotagging fails"""
nominatim.side_effect = GeotagException()
gkg = Gkg(
id=3771256,
gkgrecordid="20170215174500-2503",
date=20170215174500,
document_identifier=
"http://www.philstar.com/headlines/2017/02/16/1672746/yasay-harris-affirm-stronger-phl-us-ties"
)
self.session.add(gkg)
analysis = Analysis(gkg=gkg, status=Status.NEW)
self.session.add(analysis)
self.session.commit()
fact = Fact(unit='person', term='displaced')
self.session.add(fact)
self.session.commit()
loc1 = Location(location_name="Ruislip")
fact.locations.append(loc1)
analysis.facts.append(fact)
self.session.commit()
with self.assertRaises(GeotagException):
process_locations(analysis)
class TestFactExtractor(TestCase):
def setUp(self):
db_host = os.environ.get('DB_HOST')
db_url = 'postgresql://{user}:{passwd}@{db_host}/{db}'.format(
user='******', passwd='tester', db_host=db_host, db='idetect_test')
engine = create_engine(db_url)
Session.configure(bind=engine)
Base.metadata.drop_all(engine)
Base.metadata.create_all(engine)
self.session = Session()
load_countries(self.session)
load_terms(self.session)
def tearDown(self):
self.session.rollback()
for article in self.session.query(Gkg).all():
self.session.delete(article)
self.session.commit()
def test_extract_facts_simple(self):
"""Extracts simple facts when present and saves to DB"""
gkg = Gkg()
analysis = Analysis(gkg=gkg, status=Status.NEW)
self.session.add(analysis)
content = DocumentContent(
content_clean=
"It was early Saturday when a flash flood hit the area and washed away more than 500 houses"
)
self.session.add(content)
self.session.commit()
analysis.content_id = content.id
self.session.commit()
extract_facts(analysis)
self.assertEqual(1, len(analysis.facts))
def test_extract_refugee_facts(self):
"""Extracts refugee-related facts with Refugee Term"""
gkg = Gkg()
analysis = Analysis(gkg=gkg, status=Status.NEW)
self.session.add(analysis)
content = DocumentContent(
content_clean=
"It was early Saturday when government troops entered the area and forced more than 20000 refugees to flee."
)
self.session.add(content)
self.session.commit()
analysis.content_id = content.id
self.session.commit()
extract_facts(analysis)
self.assertEqual(FactTerm.REFUGEE, analysis.facts[0].term)
def test_extract_evicted_facts(self):
"""Extracts eviction-related facts with eviction Term"""
gkg = Gkg()
analysis = Analysis(gkg=gkg, status=Status.NEW)
self.session.add(analysis)
content = DocumentContent(
content_clean=
"2000 people have been evicted from their homes in Bosnia")
self.session.add(content)
self.session.commit()
analysis.content_id = content.id
self.session.commit()
extract_facts(analysis)
self.assertEqual(FactTerm.EVICTED, analysis.facts[0].term)
def test_extract_eviction_facts(self):
"""Extracts eviction-related facts with eviction Term"""
gkg = Gkg()
analysis = Analysis(gkg=gkg, status=Status.NEW)
self.session.add(analysis)
content = DocumentContent(
content_clean=
"ordered eviction for 2000 people from their homes in Bosnia")
self.session.add(content)
self.session.commit()
analysis.content_id = content.id
self.session.commit()
extract_facts(analysis)
self.assertEqual(FactTerm.EVICTED, analysis.facts[0].term)
def test_extract_forced_eviction_facts(self):
"""Extracts eviction-related facts with eviction Term"""
gkg = Gkg()
analysis = Analysis(gkg=gkg, status=Status.NEW)
self.session.add(analysis)
content = DocumentContent(
content_clean=
"ordered forced eviction for 2000 people from their homes in Bosnia"
)
self.session.add(content)
self.session.commit()
analysis.content_id = content.id
self.session.commit()
extract_facts(analysis)
self.assertEqual(FactTerm.EVICTED, analysis.facts[0].term)
def test_extract_forcibly_evicted_facts(self):
"""Extracts eviction-related facts with eviction Term"""
gkg = Gkg()
analysis = Analysis(gkg=gkg, status=Status.NEW)
self.session.add(analysis)
content = DocumentContent(
content_clean=
"2000 people were forcibly evicted from their homes in Bosnia")
self.session.add(content)
self.session.commit()
analysis.content_id = content.id
self.session.commit()
extract_facts(analysis)
self.assertEqual(FactTerm.EVICTED, analysis.facts[0].term)
def test_extract_sacked_facts(self):
"""Extracts sacked-related facts with eviction Term"""
gkg = Gkg()
analysis = Analysis(gkg=gkg, status=Status.NEW)
self.session.add(analysis)
content = DocumentContent(
content_clean=
"last week 2000 people have been sacked from their homes in Nigeria"
)
self.session.add(content)
self.session.commit()
analysis.content_id = content.id
self.session.commit()
extract_facts(analysis)
self.assertEqual(FactTerm.SACKED, analysis.facts[0].term)
def test_create_locations_with_names(self):
"""Creates locations for facts only with location names"""
gkg = Gkg()
analysis = Analysis(gkg=gkg, status=Status.NEW)
self.session.add(analysis)
content = DocumentContent(
content_clean=
"It was early Saturday when a flash flood hit large parts of London and Middlesex and washed away more than 500 houses"
)
self.session.add(content)
self.session.commit()
analysis.content_id = content.id
self.session.commit()
extract_facts(analysis)
facts = analysis.facts
self.assertEqual(1, len(facts))
fact = facts[0]
self.assertEqual(2, len(fact.locations))
loc_names = [loc.location_name for loc in fact.locations]
self.assertIn('London', loc_names)
self.assertIn('Middlesex', loc_names)
self.assertEqual([None, None], [loc.country for loc in fact.locations])
def test_use_existing_location(self):
"""Uses existing locations when they exist"""
gkg = Gkg()
analysis = Analysis(gkg=gkg, status=Status.NEW)
self.session.add(analysis)
content = DocumentContent(
content_clean=
"It was early Saturday when a flash flood hit large parts of Bosnia and washed away more than 500 houses"
)
self.session.add(content)
location = Location(location_name='Bosnia')
self.session.add(location)
self.session.commit()
analysis.content_id = content.id
self.session.commit()
extract_facts(analysis)
fact = analysis.facts[0]
extracted_location = fact.locations[0]
self.assertEqual(location.id, extracted_location.id)
import string
import numpy as np
import pandas as pd
from idetect.nlp_models.category import *
from idetect.nlp_models.relevance import *
from idetect.nlp_models.base_model import CustomSklLsiModel
if __name__ == "__main__":
# Create the Database
engine = create_engine(db_url())
Session.configure(bind=engine)
Base.metadata.create_all(engine)
session = Session()
# Load the Countries data if necessary
countries = session.query(Country).all()
if len(countries) == 0:
load_countries(session)
# Load the Keywords if neccessary
keywords = session.query(FactKeyword).all()
if len(keywords) == 0:
load_terms(session)
session.close()
# Load the Classifier models once to ensure they are downloaded
CategoryModel()
RelevanceModel()
class TestManager(TestCase):
start_date = '2017-01-01'
plus_1_yr = '2018-01-01'
plus_6_mo = '2017-07-01'
plus_3_mo = '2017-04-01'
plus_1_mo = '2017-02-01'
def setUp(self):
logger.debug("setUp")
worker_logger = logging.getLogger("idetect.worker")
worker_logger.setLevel(logging.INFO)
logger.debug("Connecting to DB")
db_host = os.environ.get('DB_HOST')
db_port = os.environ.get('DB_PORT', 5432)
db_user = os.environ.get('DB_USER', 'tester')
db_pass = os.environ.get('DB_PASSWORD', 'tester')
db_url = 'postgresql://{user}:{passwd}@{db_host}:{db_port}/{db}'.format(
user=db_user,
passwd=db_pass,
db_host=db_host,
db_port=db_port,
db='idetect')
self.engine = create_engine(db_url, echo=False)
Session.configure(bind=self.engine)
self.session = Session()
self.session.query(FactApi).count()
logger.debug("setUp complete")
def tearDown(self):
logger.debug("tearDown")
self.session.rollback()
logger.debug("sessions rolled back")
logger.debug("tearDown complete")
def test_timeline(self):
t0 = time.time()
counts = get_timeline_counts(self.session)
t1 = time.time()
days = {t['gdelt_day'] for t in counts}
self.assertGreater(len(days), 180)
categories = {t['category'] for t in counts}
self.assertEqual(categories, {'Conflict', 'Disaster', 'Other'})
print(t1 - t0)
self.assertLess(
t1 - t0, 1.0,
'Calculating timeline counts {} - {} took too long'.format(
self.start_date, self.plus_1_yr))
def test_histogram(self):
t0 = time.time()
counts = get_histogram_counts(self.session)
t1 = time.time()
print(len(counts))
figures = {
t['specific_reported_figure']
for t in counts if t['specific_reported_figure']
}
self.assertLess(min(figures), 10)
self.assertGreater(max(figures), 1000000)
units = {t['unit'] for t in counts}
self.assertEqual(units, {'Household', 'Person'})
print(t1 - t0)
self.assertLess(
t1 - t0, 1.0,
'Calculating histogram counts {} - {} took too long'.format(
self.start_date, self.plus_1_yr))
def test_wordcloud(self):
t0 = time.time()
terms = get_wordcloud(self.session, self.engine)
t1 = time.time()
print(t1 - t0)
print(len(terms))
print(tabulate(terms))
self.assertLess(
t1 - t0, 5.0, 'Calculating wordcloud {} - {} took too long'.format(
self.start_date, self.plus_1_yr))
def test_timeline_year(self):
t0 = time.time()
counts = get_timeline_counts(self.session,
fromdate=self.start_date,
todate=self.plus_1_yr)
t1 = time.time()
days = {t['gdelt_day'] for t in counts}
self.assertGreater(len(days), 180)
categories = {t['category'] for t in counts}
self.assertEqual(categories, {'Conflict', 'Disaster', 'Other'})
print(t1 - t0)
self.assertLess(
t1 - t0, 1.0,
'Calculating timeline counts {} - {} took too long'.format(
self.start_date, self.plus_1_yr))
def test_histogram_year(self):
t0 = time.time()
counts = get_histogram_counts(self.session,
fromdate=self.start_date,
todate=self.plus_1_yr)
t1 = time.time()
print(len(counts))
figures = {
t['specific_reported_figure']
for t in counts if t['specific_reported_figure']
}
self.assertLess(min(figures), 10)
self.assertGreater(max(figures), 1000000)
units = {t['unit'] for t in counts}
self.assertEqual(units, {'Household', 'Person'})
print(t1 - t0)
self.assertLess(
t1 - t0, 1.0,
'Calculating histogram counts {} - {} took too long'.format(
self.start_date, self.plus_1_yr))
def test_wordcloud_year(self):
t0 = time.time()
terms = get_wordcloud(self.session,
self.engine,
fromdate=self.start_date,
todate=self.plus_1_yr)
t1 = time.time()
print(t1 - t0)
print(len(terms))
print(tabulate(terms))
self.assertLess(
t1 - t0, 5.0, 'Calculating wordcloud {} - {} took too long'.format(
self.start_date, self.plus_1_yr))
def test_map_week(self):
print("hello")
t0 = time.time()
entries = get_map_week(self.session)
t1 = time.time()
print(t1 - t0)
# print(json.dumps(entries, indent=2))
self.assertEqual(len(entries), 1)
self.assertIsNotNone(entries[0].get('entries'))
class TestModel(TestCase):
def setUp(self):
db_host = os.environ.get('DB_HOST')
db_url = 'postgresql://{user}:{passwd}@{db_host}/{db}'.format(
user='******', passwd='tester', db_host=db_host, db='idetect_test')
engine = create_engine(db_url)
Session.configure(bind=engine)
Base.metadata.drop_all(engine)
Base.metadata.create_all(engine)
self.session = Session()
self.sample_data()
def sample_data(self):
gkg1 = Gkg(
id=3771256,
gkgrecordid="20170215174500-2503",
date=20170215174500,
document_identifier=
"http://www.philstar.com/headlines/2017/02/16/1672746/yasay-harris-affirm-stronger-phl-us-ties"
)
self.session.add(gkg1)
gkg2 = Gkg(
id=3771257,
gkgrecordid="20170215174500-1536",
date=20170215174500,
document_identifier=
"http://wynkcountry.iheart.com/onair/cmt-cody-alan-54719/thomas-rhett-and-lauren-akins-are-15565244/"
)
self.session.add(gkg2)
self.session.commit()
def tearDown(self):
self.session.rollback()
for gkg in self.session.query(Gkg).all():
self.session.delete(gkg)
self.session.commit()
def test_status_update(self):
gkg = self.session.query(Gkg).first()
analysis = Analysis(gkg=gkg, status=Status.NEW)
self.session.add(analysis)
self.session.commit()
analysis.create_new_version(Status.SCRAPING)
self.assertEqual(analysis.status, Status.SCRAPING)
# meanwhile, some other process changed the status of this...
session2 = Session()
try:
other = session2.query(Analysis).get(analysis.gkg_id)
other.create_new_version(Status.SCRAPING_FAILED)
finally:
session2.rollback()
with self.assertRaises(NotLatestException):
analysis.create_new_version(Status.SCRAPED)
def test_version_lifecycle(self):
gkg = self.session.query(Gkg).first()
analysis = Analysis(gkg=gkg, status=Status.NEW)
self.session.add(analysis)
self.session.commit()
analysis.create_new_version(Status.SCRAPING)
history = self.session.query(AnalysisHistory).filter(
AnalysisHistory.gkg == gkg)
self.assertEqual(1, history.count())
self.assertEqual(
1,
history.filter(AnalysisHistory.status == Status.NEW).count())
content = DocumentContent(content_type="text/html",
content="Lorem ipsum")
analysis.content = content
analysis.create_new_version(Status.SCRAPED)
self.assertEqual(2, history.count())
self.assertEqual(
1,
history.filter(AnalysisHistory.status == Status.NEW).count())
self.assertEqual(
1,
history.filter(AnalysisHistory.status == Status.SCRAPING).count())
analysis.create_new_version(Status.EXTRACTING)
self.assertEqual(3, history.count())
self.assertEqual(
1,
history.filter(AnalysisHistory.status == Status.NEW).count())
self.assertEqual(
1,
history.filter(AnalysisHistory.status == Status.SCRAPING).count())
self.assertEqual(
1,
history.filter(AnalysisHistory.status == Status.SCRAPED).count())
# content is preserved
scraped = history.filter(
AnalysisHistory.status == Status.SCRAPED).one_or_none()
self.assertEqual(analysis.content, scraped.content)
fact = Fact(analysis_date=datetime.now())
analysis.facts = [fact]
analysis.create_new_version(Status.EXTRACTED)
self.assertEqual(4, history.count())
self.assertEqual(
1,
history.filter(AnalysisHistory.status == Status.NEW).count())
self.assertEqual(
1,
history.filter(AnalysisHistory.status == Status.SCRAPING).count())
self.assertEqual(
1,
history.filter(AnalysisHistory.status == Status.SCRAPED).count())
self.assertEqual(
1,
history.filter(
AnalysisHistory.status == Status.EXTRACTING).count())
# content still preserved
extracting = history.filter(
AnalysisHistory.status == Status.EXTRACTING).one_or_none()
self.assertEqual(analysis.content, extracting.content)
analysis.create_new_version(Status.EDITING)
analysis.content = DocumentContent(content_type="text/html",
content="Lorem edited")
analysis.create_new_version(Status.EDITED)
self.assertEqual(6, history.count())
self.assertEqual(
1,
history.filter(AnalysisHistory.status == Status.NEW).count())
self.assertEqual(
1,
history.filter(AnalysisHistory.status == Status.SCRAPING).count())
self.assertEqual(
1,
history.filter(AnalysisHistory.status == Status.SCRAPED).count())
self.assertEqual(
1,
history.filter(
AnalysisHistory.status == Status.EXTRACTING).count())
self.assertEqual(
1,
history.filter(AnalysisHistory.status == Status.EXTRACTED).count())
self.assertEqual(
1,
history.filter(AnalysisHistory.status == Status.EDITING).count())
# content has changed, but reports are preserved
extracted = history.filter(
AnalysisHistory.status == Status.EXTRACTED).one_or_none()
self.assertNotEqual(analysis.content.id, extracted.content.id)
self.assertCountEqual([f.id for f in analysis.facts],
[f.id for f in extracted.facts])
analysis.create_new_version(Status.EDITING)
fact2 = Fact(analysis_date=datetime.now())
analysis.facts.append(fact2)
analysis.create_new_version(Status.EDITED)
self.assertEqual(8, history.count())
self.assertEqual(
1,
history.filter(AnalysisHistory.status == Status.NEW).count())
self.assertEqual(
1,
history.filter(AnalysisHistory.status == Status.SCRAPING).count())
self.assertEqual(
1,
history.filter(AnalysisHistory.status == Status.SCRAPED).count())
self.assertEqual(
1,
history.filter(
AnalysisHistory.status == Status.EXTRACTING).count())
self.assertEqual(
1,
history.filter(AnalysisHistory.status == Status.EXTRACTED).count())
self.assertEqual(
2,
history.filter(AnalysisHistory.status == Status.EDITING).count())
self.assertEqual(
1,
history.filter(AnalysisHistory.status == Status.EDITED).count())
edited = history.filter(
AnalysisHistory.status == Status.EDITED).one_or_none()
self.assertCountEqual([f.id for f in analysis.facts],
[fact.id, fact2.id])
self.assertCountEqual([f.id for f in edited.facts], [fact.id])
def test_status_counts(self):
gkgs = self.session.query(Gkg).all()[:2]
analysis1 = Analysis(gkg=gkgs[0], status=Status.NEW)
self.session.add(analysis1)
self.session.commit()
self.assertEqual(Analysis.status_counts(self.session), {Status.NEW: 1})
analysis1.create_new_version(Status.SCRAPING)
self.assertEqual(Analysis.status_counts(self.session),
{Status.SCRAPING: 1})
analysis2 = Analysis(gkg=gkgs[1], status=Status.NEW)
self.session.add(analysis2)
self.session.commit()
self.assertEqual(Analysis.status_counts(self.session), {
Status.NEW: 1,
Status.SCRAPING: 1
})
analysis2.create_new_version(Status.SCRAPING)
self.assertEqual(Analysis.status_counts(self.session),
{Status.SCRAPING: 2})
analysis2.create_new_version(Status.SCRAPED)
self.assertEqual(Analysis.status_counts(self.session), {
Status.SCRAPED: 1,
Status.SCRAPING: 1
})
def test_country_term(self):
mmr = Country(iso3="MMR", preferred_term="Myanmar")
myanmar = CountryTerm(term="Myanmar", country=mmr)
burma = CountryTerm(term="Burma", country=mmr)
yangon = Location(location_name="Yangon",
location_type=LocationType.CITY,
country=mmr,
latlong="16°51′N 96°11′E")
self.assertEqual(yangon.country, myanmar.country)
self.assertEqual(yangon.country, burma.country)
def analyse_url():
session = Session()
status = None
gkg_id = None
try:
url = request.get_json(silent=True)['url'] or request.form['url']
except Exception as e:
return json.dumps({
'success': False,
'Exception': str(e),
'status': 'missing or null url parameter'
}), 422, {
'ContentType': 'application/json'
}
if url is None:
return json.dumps({
'success': False,
'status': 'null url parameter'
}), 422, {
'ContentType': 'application/json'
}
gkg = session.query(Gkg.id).filter(
Gkg.document_identifier.like("%" + url + "%")).order_by(
Gkg.date.asc()).first()
if gkg:
gkg_id = gkg.id
status = 'url already in IDETECT DB'
else:
analysis = create_new_analysis_from_url(session, url)
gkg_id = analysis.gkg_id
status = 'url added to IDETECT DB'
try:
work(session, analysis, Status.SCRAPING, Status.SCRAPED,
Status.SCRAPING_FAILED, scrape)
# TODO add classification, missing modules
# work(session,analysis,Status.CLASSIFYING,Status.CLASSIFIED,Status.CLASSIFYING_FAILED,lambda article: classify(article, get_c_m(), get_r_m()))
work(session, analysis, Status.EXTRACTING, Status.EXTRACTED,
Status.EXTRACTING_FAILED, extract_facts)
work(session, analysis, Status.GEOTAGGING, Status.GEOTAGGED,
Status.GEOTAGGING_FAILED, process_locations)
except Exception as e:
return json.dumps({
'success': False,
'Exception': str(e)
}), 422, {
'ContentType': 'application/json'
}
finally:
session.close()
try:
document = get_document(session, gkg_id)
entries = get_facts_for_document(session, gkg_id)
resp = jsonify({
'document': document,
'facts': entries,
'status': status
})
resp.status_code = 200
return resp
finally:
session.close()