def process_database(species = None, commit = False):
session = get_session()
if species == None:
taxa = [taxon_id for (taxon_id,) in session.execute("SELECT DISTINCT taxon_id FROM processing_method WHERE data_type = 2").fetchall()]
else:
taxa = [taxon_id for (taxon_id,) in session.execute(
"SELECT DISTINCT taxon_id FROM processing_method, taxon WHERE taxon.id = taxon_id AND data_type = 2 AND spno IN %s" % sql_list_placeholder('species', species),
sql_list_argument('species', species)
).fetchall()]
create_region_lookup_table(session)
log.info("Step 1/2: Monthly aggregation")
fn = functools.partial(aggregate_by_month, commit = commit)
for result, error in tqdm(run_parallel(fn, taxa), total = len(taxa)):
pass
log.info("Step 2/2: Yearly aggregation")
fn = functools.partial(aggregate_by_year, commit = commit)
for result, error in tqdm(run_parallel(fn, taxa), total = len(taxa)):
pass
cleanup_region_lookup_table(session)
def process_database(species=None, commit=False):
session = get_session()
if species == None:
taxa = [
taxon_id for (taxon_id, ) in session.execute(
"SELECT DISTINCT taxon_id FROM t1_sighting").fetchall()
]
else:
taxa = [
taxon_id for (taxon_id, ) in session.execute(
"SELECT DISTINCT taxon_id FROM t1_sighting, taxon WHERE taxon.id = taxon_id AND spno IN (%s)"
% sql_list_placeholder('species', species),
sql_list_argument('species', species)).fetchall()
]
create_region_lookup_table(session)
# Process in parallel
tasks = [(taxon_id, commit) for taxon_id in taxa]
log.info("Step 1/2: Monthly aggregation")
for result, error in tqdm(run_parallel(aggregate_monthly, tasks),
total=len(tasks)):
if error:
print error
log.info("Step 1/2: Yearly aggregation")
for result, error in tqdm(run_parallel(aggregate_yearly, tasks),
total=len(tasks)):
if error:
print error
cleanup_region_lookup_table(session)
def process_database(species=None, commit=False):
session = get_session()
# Just get taxa that have range polygons
if species == None:
taxa = session.execute(
"SELECT DISTINCT taxon_id FROM taxon_range").fetchall()
else:
taxa = session.execute(
"SELECT DISTINCT taxon_id FROM taxon_range, taxon WHERE taxon_id = taxon.id AND spno IN (%s)"
% sql_list_placeholder('species', species),
sql_list_argument('species', species)).fetchall()
# Unwrap tuple
taxa = [taxon_id for (taxon_id, ) in taxa]
# Delete old results
if commit and len(taxa) > 0:
# session.execute("DELETE FROM t2_ultrataxon_sighting WHERE taxon_id IN (:taxa)", { 'taxa': taxa })
session.execute(
"DELETE FROM t2_ultrataxon_sighting WHERE taxon_id IN (%s)" %
sql_list_placeholder('taxa', taxa),
sql_list_argument('taxa', taxa))
session.commit()
# Process in parallel
tasks = [(taxon_id, commit) for taxon_id in taxa]
# This is important because we are about to spawn child processes, and this stops them attempting to share the
# same database connection pool
session.close()
for result, error in tqdm(run_parallel(process_taxon, tasks),
total=len(tasks)):
if error:
print(error)
def process_database(species=None, commit=False):
"""
Calculates spatial representativeness using alpha hulls
Generates alpha hulls from each source x taxon combination
Intersects alpha hulls with range layers, and then calculates percentage of range covered
"""
session = get_session()
if commit:
if species == None:
session.execute("DELETE FROM taxon_source_alpha_hull")
else:
session.execute(
"""DELETE FROM taxon_source_alpha_hull
WHERE taxon_id IN (SELECT id FROM taxon WHERE spno IN (%s))"""
% sql_list_placeholder('species', species),
sql_list_argument('species', species))
session.commit()
# Load coastal shapefile
coastal_shape_filename = tsx.config.config.get("processing.alpha_hull",
"coastal_shp")
with fiona.Env(OSR_WKT_FORMAT="WKT2_2018"), fiona.open(
coastal_shape_filename, 'r') as coastal_shape:
# Convert from fiona dictionary to shapely geometry and reproject
shp_to_working_transformer = pyproj.Transformer.from_proj(
pyproj.CRS.from_wkt(coastal_shape.crs_wkt),
working_proj,
always_xy=True)
coastal_shape = reproject(shape(coastal_shape[0]['geometry']),
shp_to_working_transformer)
# Simplify coastal boundary - makes things run ~20X faster
log.info("Simplifying coastal boundary")
coastal_shape = coastal_shape.buffer(10000).simplify(10000)
log.info("Generating alpha shapes")
for data_type in 1, 2:
log.info("Processing type %s data" % data_type)
taxa = get_taxa(session, data_type, species)
tasks = [(taxon_id, coastal_shape, data_type, commit)
for taxon_id in taxa]
# This is important because we are about to spawn child processes, and this stops them attempting to share the
# same database connection pool
session.close() # TODO: not sure if this is needed now
# Process all the species in parallel
for result, error in tqdm(run_parallel(process, tasks),
total=len(tasks)):
if error:
print(error)
def process_database(species = None, commit = False):
"""
Generates alpha hulls from raw sighting data in the database
Intersects alpha hulls with range layers, and inserts the result back into the database
"""
session = get_session()
if species is None:
species = get_all_spno(session)
if commit:
for spno in species:
session.execute("""
DELETE FROM taxon_presence_alpha_hull
WHERE taxon_id IN (SELECT id FROM taxon WHERE spno = :spno)
""", { 'spno': spno })
session.execute("""
DELETE FROM taxon_presence_alpha_hull_subdiv
WHERE taxon_id IN (SELECT id FROM taxon WHERE spno = :spno)
""", { 'spno': spno })
session.commit()
db_proj = pyproj.Proj('+init=EPSG:4326') # Database always uses WGS84
working_proj = pyproj.Proj('+init=EPSG:3112') # GDA94 / Geoscience Australia Lambert - so that we can buffer in metres
# Load coastal shapefile
coastal_shape_filename = tsx.config.config.get("processing.alpha_hull", "coastal_shp")
with fiona.open(coastal_shape_filename, 'r') as coastal_shape:
# Convert from fiona dictionary to shapely geometry and reproject
coastal_shape = reproject(shape(coastal_shape[0]['geometry']), pyproj.Proj(coastal_shape.crs), working_proj)
# Simplify coastal boundary - makes things run ~20X faster
log.info("Simplifying coastal boundary")
coastal_shape = coastal_shape.buffer(10000).simplify(10000)
log.info("Generating alpha shapes")
# Process a single species.
# This gets run off the main thread.
def process_spno(spno):
session = get_session()
try:
# Get raw points from DB
raw_points = get_species_points(session, spno)
if len(raw_points) < 4:
# Not enough points to create an alpha hull
return
# Read points from database
points = [reproject(p, db_proj, working_proj) for p in raw_points]
# Generate alpha shape
alpha_shp = make_alpha_hull(
points = points,
coastal_shape = coastal_shape,
thinning_distance = tsx.config.config.getfloat('processing.alpha_hull', 'thinning_distance'),
alpha = tsx.config.config.getfloat('processing.alpha_hull', 'alpha'),
hullbuffer_distance = tsx.config.config.getfloat('processing.alpha_hull', 'hullbuffer_distance'),
isolatedbuffer_distance = tsx.config.config.getfloat('processing.alpha_hull', 'isolatedbuffer_distance'))
# Convert back to DB projection
alpha_shp = reproject(alpha_shp, working_proj, db_proj)
# Clean up geometry
alpha_shp = alpha_shp.buffer(0)
# Get range polygons to intersect with alpha shape
for taxon_id, range_id, breeding_range_id, geom_wkb in get_species_range_polygons(session, spno):
# Intersect and insert into DB
geom = shapely.wkb.loads(binascii.unhexlify(geom_wkb)).buffer(0)
geom = to_multipolygon(geom.intersection(alpha_shp)) # slow
if len(geom) > 0:
session.execute("""INSERT INTO taxon_presence_alpha_hull (taxon_id, range_id, breeding_range_id, geometry)
VALUES (:taxon_id, :range_id, :breeding_range_id, ST_GeomFromWKB(_BINARY :geom_wkb))""", {
'taxon_id': taxon_id,
'range_id': range_id,
'breeding_range_id': breeding_range_id,
'geom_wkb': shapely.wkb.dumps(geom)
}
)
# We also subdivide the geometries into small pieces and insert this into the database. This allows for much faster
# spatial queries in the database.
for subgeom in subdivide_geometry(geom, max_points = 100):
session.execute("""INSERT INTO taxon_presence_alpha_hull_subdiv (taxon_id, range_id, breeding_range_id, geometry)
VALUES (:taxon_id, :range_id, :breeding_range_id, ST_GeomFromWKB(_BINARY :geom_wkb))""", {
'taxon_id': taxon_id,
'range_id': range_id,
'breeding_range_id': breeding_range_id,
'geom_wkb': shapely.wkb.dumps(subgeom)
}
)
if commit:
session.commit()
except:
log.exception("Exception processing alpha hull")
raise
finally:
session.close()
# This is important because we are about to spawn child processes, and this stops them attempting to share the
# same database connection pool
session.close()
tsx.db.connect.engine.dispose()
# Process all the species in parallel
for result, error in tqdm(run_parallel(process_spno, species, use_processes = True), total = len(species)):
if error:
print error
def process_sites(session):
run_sql(
session, "Populate t2_survey_site",
"""INSERT INTO t2_survey_site (survey_id, site_id)
SELECT t2_survey.id, t2_site.id
FROM t2_survey, t2_site
WHERE site_id = t2_site.id""")
# Query OK, 14073 rows affected (0.68 sec)
# Records: 14073 Duplicates: 0 Warnings: 0
run_sql(
session, "Populate t2_survey_site (spatial)",
"""INSERT INTO t2_survey_site
SELECT t2_survey.id, t2_site.id
FROM t2_site STRAIGHT_JOIN t2_survey USE INDEX (coords)
WHERE site_id IS NULL
AND t2_survey.search_type_id = t2_site.search_type_id
AND ST_Intersects(geometry, coords)""")
# Query OK, 278678 rows affected (15.77 sec)
# Records: 278678 Duplicates: 0 Warnings: 0
run_sql(
session, "Populate standardised site surveys",
"""INSERT INTO t2_processed_survey (raw_survey_id, site_id, search_type_id, start_date_y, start_date_m, source_id, experimental_design_type_id)
SELECT t2_survey.id, t2_survey_site.site_id, search_type_id, start_date_y, start_date_m, source_id, 1
FROM t2_survey, t2_survey_site
WHERE t2_survey.id = t2_survey_site.survey_id""")
# Query OK, 292751 rows affected (5.00 sec)
# Records: 292751 Duplicates: 0 Warnings: 0
run_sql(
session, "Populate presences / non-pseudo-absences",
"""INSERT INTO t2_processed_sighting (survey_id, taxon_id, count, unit_id, pseudo_absence)
SELECT t2_processed_survey.id, t2_ultrataxon_sighting.taxon_id, count, unit_id, 0
FROM t2_ultrataxon_sighting, t2_sighting, t2_processed_survey
WHERE t2_ultrataxon_sighting.sighting_id = t2_sighting.id
AND t2_sighting.survey_id = t2_processed_survey.raw_survey_id
AND t2_processed_survey.experimental_design_type_id = 1""")
# Query OK, 5775718 rows affected (2 min 35.55 sec)
# Records: 5775718 Duplicates: 0 Warnings: 0
log.info("Identify taxa for each site based on alpha hulls")
taxa = [
taxon_id for (taxon_id, ) in session.execute(
"SELECT DISTINCT taxon_id FROM taxon_presence_alpha_hull_subdiv").
fetchall()
]
session.execute("""DROP TABLE IF EXISTS tmp_taxon_site""")
# Note for some reason CREATE TEMPORARY TABLE doesn't work as expected, the table seems to be empty by the time we get
# to the next step... I'm guessing after some kind of timeout the transaction gets rolled back(?)
session.execute("""CREATE TABLE tmp_taxon_site (
site_id INT NOT NULL,
taxon_id CHAR(6) NOT NULL,
INDEX (taxon_id, site_id)
)""")
# The next step was originally a very slow query, directly populating the tmp_taxon_site table.
# Instead, I've broken the query down to process one taxon at a time
# Furthermore, we process taxa in parallel threads to fully utilise the CPU, and then the results of the query are
# inserted in bulk on the main thread.
# Time taken: 17 min 4 s
for result, error in tqdm(run_parallel(get_taxon_sites, taxa),
total=len(taxa)):
# Perform bulk insert on main thread
if len(result) > 0:
insert_data = [{
'site_id': site_id,
'taxon_id': taxon_id
} for site_id, taxon_id in result]
session.execute(
"""INSERT INTO tmp_taxon_site (site_id, taxon_id) VALUES (:site_id, :taxon_id)""",
insert_data)
log.info("Insert pseudo absences")
# This next step originally ran in about 15 minutes on my laptop as a single query, but took forever on the server
# (I gave up after a couple of hours) so I've split it up by taxon with is a bit slower overall but at least you can
# see progress
# Create temporary table so we aren't inserting and selecting from the same table
session.execute(
"""CREATE TEMPORARY TABLE tmp_processed_sighting SELECT survey_id, taxon_id, count, unit_id, pseudo_absence FROM t2_processed_sighting WHERE FALSE"""
)
# Running this in parallel originally resulted in MySQL deadlock errors, but then we were inserting and selecting from the same table.
# It might be worth trying a parallel approach again.
for taxon_id in tqdm(taxa):
# This query is a bit tricky. We do a left join to find taxa that are not present for a survey, and match on t2_processed_sighting.id = NULL to generate the pseudo-absences
session.execute(
"""INSERT INTO tmp_processed_sighting (survey_id, taxon_id, count, unit_id, pseudo_absence)
SELECT t2_processed_survey.id, tmp_taxon_site.taxon_id, 0, 2, 1
FROM t2_survey
INNER JOIN t2_survey_site ON t2_survey.id = t2_survey_site.survey_id
INNER JOIN t2_processed_survey ON t2_survey.id = t2_processed_survey.raw_survey_id AND t2_processed_survey.experimental_design_type_id = 1
INNER JOIN tmp_taxon_site ON t2_survey_site.site_id = tmp_taxon_site.site_id AND taxon_id = :taxon_id
LEFT JOIN t2_processed_sighting ON t2_processed_sighting.survey_id = t2_processed_survey.id AND t2_processed_sighting.taxon_id = tmp_taxon_site.taxon_id
WHERE t2_processed_sighting.id IS NULL""", {'taxon_id': taxon_id})
session.execute(
"""INSERT INTO t2_processed_sighting(survey_id, taxon_id, count, unit_id, pseudo_absence) SELECT survey_id, taxon_id, count, unit_id, pseudo_absence FROM tmp_processed_sighting"""
)
session.execute("""DROP TABLE tmp_taxon_site""")
def process_database(species = None, commit = False):
"""
Calculates spatial representativeness using alpha hulls
Generates alpha hulls from each source x taxon combination
Intersects alpha hulls with range layers, and then calculates percentage of range covered
"""
session = get_session()
if commit:
if species == None:
session.execute("DELETE FROM taxon_source_alpha_hull")
else:
session.execute("""DELETE FROM taxon_source_alpha_hull
WHERE taxon_id IN (SELECT id FROM taxon WHERE spno IN (%s))""" % sql_list_placeholder('species', species),
sql_list_argument('species', species))
session.commit()
db_proj = pyproj.Proj('+init=EPSG:4326') # Database always uses WGS84
working_proj = pyproj.Proj('+init=EPSG:3112') # GDA94 / Geoscience Australia Lambert - so that we can buffer in metres
# Load coastal shapefile
coastal_shape_filename = tsx.config.config.get("processing.alpha_hull", "coastal_shp")
with fiona.open(coastal_shape_filename, 'r') as coastal_shape:
# Convert from fiona dictionary to shapely geometry and reproject
coastal_shape = reproject(shape(coastal_shape[0]['geometry']), pyproj.Proj(coastal_shape.crs), working_proj)
# Simplify coastal boundary - makes things run ~20X faster
log.info("Simplifying coastal boundary")
coastal_shape = coastal_shape.buffer(10000).simplify(10000)
log.info("Generating alpha shapes")
for data_type in 1,2:
log.info("Processing type %s data" % data_type)
# Process a single species.
# This gets run off the main thread.
def process(taxon_id):
session = get_session()
try:
# Load core range geometry
core_range_geom = reproject(get_core_range_geometry(session, taxon_id), db_proj, working_proj).buffer(0).intersection(coastal_shape)
for source_id in get_source_ids(session, data_type, taxon_id):
log.info("Processing taxon_id: %s, source_id: %s" % (taxon_id, source_id))
# Get raw points from DB
raw_points = get_raw_points(session, data_type, taxon_id, source_id)
empty = len(raw_points) < 4
if empty:
log.info("Taxon %s: not enough points to create alpha hull (%s)" % (taxon_id, len(raw_points)))
if not empty:
# Read points from database
points = [reproject(p, db_proj, working_proj) for p in raw_points]
# Generate alpha shape
alpha_shp = make_alpha_hull(
points = points,
coastal_shape = None,
thinning_distance = tsx.config.config.getfloat('processing.alpha_hull', 'thinning_distance'),
alpha = tsx.config.config.getfloat('processing.alpha_hull', 'alpha'),
hullbuffer_distance = tsx.config.config.getfloat('processing.alpha_hull', 'hullbuffer_distance'),
isolatedbuffer_distance = tsx.config.config.getfloat('processing.alpha_hull', 'isolatedbuffer_distance'))
# Clean up geometry
alpha_shp = alpha_shp.buffer(0)
if core_range_geom.area == 0:
log.info("Core range geometry area is zero")
empty = True
else:
# Intersect alpha hull with core range
intersected_alpha = to_multipolygon(core_range_geom.intersection(alpha_shp))
empty = intersected_alpha.is_empty
if empty:
session.execute("""INSERT INTO taxon_source_alpha_hull (source_id, taxon_id, data_type, core_range_area_in_m2, alpha_hull_area_in_m2)
VALUES (:source_id, :taxon_id, :data_type, 0, 0)""", {
'source_id': source_id,
'taxon_id': taxon_id,
'data_type': data_type
})
else:
session.execute("""INSERT INTO taxon_source_alpha_hull (source_id, taxon_id, data_type, geometry, core_range_area_in_m2, alpha_hull_area_in_m2)
VALUES (:source_id, :taxon_id, :data_type, ST_GeomFromWKB(_BINARY :geom_wkb), :core_range_area, :alpha_hull_area)""", {
'source_id': source_id,
'taxon_id': taxon_id,
'data_type': data_type,
'geom_wkb': shapely.wkb.dumps(reproject(intersected_alpha, working_proj, db_proj)),
'core_range_area': core_range_geom.area,
'alpha_hull_area': intersected_alpha.area
})
if commit:
session.commit()
except:
log.exception("Exception processing alpha hull")
raise
finally:
session.close()
taxa = get_taxa(session, data_type, species)
# This is important because we are about to spawn child processes, and this stops them attempting to share the
# same database connection pool
session.close()
tsx.db.connect.engine.dispose()
# Process all the species in parallel
for result, error in tqdm(run_parallel(process, taxa, use_processes = True), total = len(taxa)):
if error:
print error
