def latest_run(model, valid_time):
with tracer.span(name="download_content"):
download_content()
pattern = re.compile("./%s" \
"/grib" \
"\\/(\d{2})" \
"/t" \
"/icon-eu_europe_regular-lat-lon_model-level_(\d{10})_(\d{3})_1_T.grib2.bz2" % model)
max_t: int = 0
result = None
with tracer.span(name="parse_content"):
for i, line in enumerate(open('tmp/content.log')):
for match in re.finditer(pattern, line):
matches = match.groups()
match_valid_at = datetime.datetime.strptime(matches[1], "%Y%m%d%H")
match_valid_at = pytz.timezone('UTC').localize(match_valid_at)
match_valid_at = match_valid_at + datetime.timedelta(hours=int(matches[2]))
delta_t = abs((match_valid_at - valid_time).total_seconds())
if delta_t <= 30 * 60 and int(matches[1]) > max_t:
result = matches
max_t = int(matches[1])
return result
def hello():
if request.method == 'POST':
n = int(request.get_json().get("n"))
else:
n = 10
tracer = initialize_tracer(PROJECT_ID)
with tracer.span(name="gae_gapic") as span:
latencies = []
temp = time.time()
response = client.publish(topic_path, [{'data': b'first'}])
assert response.message_ids
logger.log_text(f"{int((time.time()- temp)*1000)}")
for i in range(n):
messages = [{'data': str(i).encode('utf-8')}]
temp = time.time()
response = client.publish(topic_path, messages)
assert response.message_ids
latencies.append(time.time() - temp)
logger.log_text(f"{','.join([str(int(i*1000)) for i in latencies])}")
return ("Published messages.\n")
def hello_world(request):
request_json = request.get_json(silent=True)
if 'n' in request_json:
n = int(request_json['n'])
else:
n = 10
tracer = initialize_tracer(PROJECT_ID)
with tracer.span(name="cf_gapic") as span:
latencies = []
temp = time.time()
response = client.publish(topic_path, [{'data': b'first'}])
assert response.message_ids
logger.log_text(f"{int((time.time()- temp)*1000)}")
for i in range(n):
messages = [{'data': str(i).encode('utf-8')}]
temp = time.time()
response = client.publish(topic_path, messages)
assert response.message_ids
latencies.append(time.time() - temp)
logger.log_text(f"{','.join([str(int(i*1000)) for i in latencies])}")
return ("Published messages.")
def hello():
if request.method == 'POST':
n = int(request.get_json().get("n"))
else:
n = 10
tracer = initialize_tracer(PROJECT_ID)
with tracer.span(name="gae_apiary") as span:
latencies = []
message = base64.b64encode(b'first')
body = {'messages': [{'data': message.decode('utf-8')}]}
temp = time.time()
response = service.projects().topics().publish(
topic=topic_path,
body=body,
).execute()
assert response['messageIds']
logger.log_text(f"{int((time.time()- temp)*1000)}")
for i in range(n):
message = base64.b64encode(str(i).encode('utf-8'))
body = {'messages': [{'data': message.decode('utf-8')}]}
temp = time.time()
response = service.projects().topics().publish(
topic=topic_path,
body=body,
).execute()
assert response['messageIds']
latencies.append(time.time() - temp)
logger.log_text(f"{','.join([str(int(i*1000)) for i in latencies])}")
return("Published messages.\n")
def hello():
if request.method == 'POST':
n = int(request.get_json().get("n"))
else:
n = 10
tracer = initialize_tracer(PROJECT_ID)
def get_callback(f, data):
def callback(f):
try:
f.result()
logger.log_text(f"{int((time.time()- futures[data])*1000)}")
except Exception:
raise Exception
return callback
with tracer.span(name="gae_cl_mixed") as span:
futures = dict()
latencies = []
futures.update({b"first": time.time()})
future = client.publish(topic_path, data=b"first")
future.add_done_callback(get_callback(future, b"first"))
for i in range(n):
data = str(i)
temp = time.time()
future = client.publish(topic_path, data=data.encode('utf-8'))
future.result()
latencies.append(time.time() - temp)
logger.log_text(f"{','.join([str(int(i*1000)) for i in latencies])}")
return("Published messages.\n")
def skewt(latitude, longitude, valid_at):
with tracer.span(name="skewt") as span:
span.add_attribute("latitude", str(latitude))
span.add_attribute("longitude", str(longitude))
span.add_attribute("valid_at", str(valid_at))
valid_at_parsed = datetime.datetime.strptime(valid_at, "%Y%m%d%H")
valid_at_parsed = pytz.timezone('UTC').localize(valid_at_parsed)
with tracer.span(name="sounding"):
sounding = load_weather_model_sounding(latitude, longitude, valid_at_parsed)
with tracer.span(name="plotting"):
model_time = str(np.datetime_as_string(sounding.metadata.model_time))
valid_time = str(np.datetime_as_string(sounding.metadata.valid_time))
model_time_for_file_name = str(np.datetime_as_string(sounding.metadata.model_time, unit='m')).replace(":", "_")
valid_time_for_file_name = str(np.datetime_as_string(sounding.metadata.valid_time, unit='m')).replace(":", "_")
full_plot = plot_skewt_icon(sounding=sounding, parcel="surface-based")
full_plot_filename = f"plot_{sounding.latitude_pretty}_{sounding.longitude_pretty}_" \
f"{model_time_for_file_name}_{valid_time_for_file_name}_full.png"
full_plot.savefig(full_plot_filename)
detail_plot = plot_skewt_icon(sounding=sounding, parcel="surface-based", base=1000, top=500, skew=15)
detail_plot_filename = f"plot_{sounding.latitude_pretty}_{sounding.longitude_pretty}_" \
f"{model_time_for_file_name}_{valid_time_for_file_name}_detail.png"
detail_plot.savefig(detail_plot_filename)
with tracer.span(name="cloud_upload"):
# Google Cloud Upload
storage_client = storage.Client()
bucket = storage_client.bucket(config.bucket_name)
blob_full = bucket.blob(full_plot_filename)
blob_detail = bucket.blob(detail_plot_filename)
with ThreadPoolExecutor(max_workers=2) as executor:
executor.submit(blob_full.upload_from_filename, full_plot_filename)
executor.submit(blob_detail.upload_from_filename, detail_plot_filename)
result = json.dumps(SkewTResult(model_time, valid_time,
config.bucket_public_url + full_plot_filename,
config.bucket_public_url + detail_plot_filename).__dict__)
response = make_response(result)
response.mimetype = 'application/json'
return response
def parameter_all_levels(model, latitude, longitude, run_hour, run_datetime, timestep,
parameter, level_type="model_level", base_level=60, top_level=1):
with tracer.span(name="download"):
levels = list(range(base_level, top_level - 1, -1))
paths = [level_path(model, run_hour, run_datetime, timestep, parameter, level, level_type) for level in levels]
session = requests.sessions.Session()
with concurrent.futures.ThreadPoolExecutor(max_workers=config.http_download_pool) as executor:
futures = list(executor.submit(download_file(path, session)) for path in paths)
concurrent.futures.wait(futures, timeout=None, return_when=ALL_COMPLETED)
with tracer.span(name="parsing"):
data_set = xr.open_mfdataset(paths, engine="cfgrib", concat_dim="generalVerticalLayer",
combine='nested', parallel=config.cfgrib_parallel)
interpolated = data_set.to_array()[0].interp(latitude=latitude, longitude=longitude)
data = AllLevelDataResult(interpolated.values, interpolated.time.values, interpolated.valid_time.values)
data_set.close()
return data
def parameter_all_levels(model, run, parameter, latitude, longitude, level_type="model_level", session=requests.Session()):
with tracer.span(name="load_parameters") as span:
run_hour = run[0]
run_datetime = run[1]
timestep = int(run[2])
logging.info(f"Loading sounding for latitude={latitude} longitude={longitude} with "
f"run_hour={run_hour} run_datetime={run_datetime} timestep={timestep} "
f"level_type={level_type} and parameter={parameter}")
span.add_attribute("parameter", parameter)
span.add_attribute("run_hour", str(run_hour))
span.add_attribute("run_datetime", str(run_datetime))
span.add_attribute("timestep", str(timestep))
levels = np.floor(np.linspace(60, 0, config.level_workers)).astype(int).tolist()
urls = list()
for i in range(0, len(levels) - 1):
base = levels[i]
top = levels[i + 1] + 1
# example URL:
# https://nwp-sounding-mw5zsrftba-ew.a.run.app/48.21/16.37/06/2020030406/4/p
url = f"{config.sounding_api}" \
f"/{latitude}" \
f"/{longitude}" \
f"/{run_hour}" \
f"/{run_datetime}" \
f"/{timestep}" \
f"/{parameter}" \
f"?level_type={level_type}" \
f"&base={base}" \
f"&top={top}"
urls.append(url)
result = AllLevelData(data=np.empty(0), model_time=None, valid_time=None)
with ThreadPoolExecutor(max_workers=config.level_workers) as executor:
responses = list(executor.map(session.get, urls))
for response in responses:
response.raise_for_status()
json_result = json.loads(response.content)
result.data = np.append(result.data, np.array(json_result["data"]))
json_first = json.loads(responses[0].content)
result.model_time = np.datetime64(json_result["model_time"])
result.valid_time = np.datetime64(json_result["valid_time"])
return result
def handleHttp(request: 'flask.Request') -> str:
tracer = initialize_tracer(request)
req = WebhookRequest()
res = WebhookResponse()
try:
json_format.Parse(request.data, req, ignore_unknown_fields=True)
if req.query_result.action == "roll":
with tracer.span(name='roll'):
handleRoll(req, res)
except UnfulfillableRequestError as e:
logging.exception(e)
if STACKDRIVER_ERROR_REPORTING:
try:
client = error_reporting.Client()
client.report_exception(
http_context=error_reporting.build_flask_context(request))
except Exception:
logging.exception("Failed to send error report to Google")
add_fulfillment_messages(res, str(e))
return json_format.MessageToJson(res)
def sounding(latitude, longitude, run_hour, run_datetime, timestep, parameter):
with tracer.span(name="sounding") as span:
span.add_attribute("latitude", str(latitude))
span.add_attribute("longitude", str(longitude))
span.add_attribute("run_hour", str(run_hour))
span.add_attribute("run_datetime", str(run_datetime))
span.add_attribute("timestep", str(timestep))
span.add_attribute("parameter", str(parameter))
level_type = request.args.get("level_type", "model_level")
base_level = int(request.args.get("base", "60"))
top_level = int(request.args.get("top", "1"))
sounding = parameter_all_levels(config.model, latitude, longitude,
run_hour, run_datetime, int(timestep),
parameter, level_type, base_level, top_level)
response = make_response(json.dumps(sounding.__dict__))
response.mimetype = 'application/json'
return response
def trace_requests(request):
"""Responds to any HTTP request.
Args:
request (flask.Request): HTTP request object.
Returns:
The response text or any set of values that can be turned into a
Response object using
`make_response <http://flask.pocoo.org/docs/1.0/api/#flask.Flask.make_response>`.
"""
tracer = initTracer()
base_name = os.environ.get("FUNCTION_NAME") + "-trace-"
with tracer.span(name=base_name + "metadata-call"):
import requests
r = requests.get(
"http://metadata.google.internal/computeMetadata/v1/project/numeric-project-id",
headers={'Metadata-Flavor': 'Google'})
return r.content
def load_weather_model_sounding(latitude, longitude, valid_time):
with tracer.span(name="latest_run"):
model = "icon-eu"
run = latest_run(model, valid_time)
http_session = session()
with ThreadPoolExecutor(max_workers=config.parameter_all_levels_workers) as executor:
p_future = executor.submit(parameter_all_levels, model, run, "p", latitude, longitude, session=http_session)
T_future = executor.submit(parameter_all_levels, model, run, "T", latitude, longitude, session=http_session)
QV_future = executor.submit(parameter_all_levels, model, run, "QV", latitude, longitude, session=http_session)
U_future = executor.submit(parameter_all_levels, model, run, "U", latitude, longitude, session=http_session)
V_future = executor.submit(parameter_all_levels, model, run, "V", latitude, longitude, session=http_session)
HHL_future = executor.submit(parameter_all_levels, model, run, "HHL", latitude, longitude, "time_invariant", session=http_session)
# Pressure Pa
p_raw = p_future.result()
p = p_raw.data
# Temperature K
T = T_future.result().data
# Specific Humidty kg/kg
QV = QV_future.result().data
# Dewpoint K
Td = mpcalc.dewpoint_from_specific_humidity(QV * units("kg/kg"), T * units.K, p * units.Pa)
# Wind m/s
U = U_future.result().data
V = V_future.result().data
# Height above MSL for model level
HHL = HHL_future.result().data
meta_data = WeatherModelSoundingMetaData(p_raw.model_time, p_raw.valid_time)
return WeatherModelSounding(latitude, longitude, p, T, QV, Td, U, V, HHL, meta_data)
