本文整理汇总了Python中mi.dataset.parser.dosta_ln_wfp.DostaLnWfpParser类的典型用法代码示例。如果您正苦于以下问题:Python DostaLnWfpParser类的具体用法?Python DostaLnWfpParser怎么用?Python DostaLnWfpParser使用的例子?那么恭喜您, 这里精选的类代码示例或许可以为您提供帮助。
在下文中一共展示了DostaLnWfpParser类的10个代码示例,这些例子默认根据受欢迎程度排序。您可以为喜欢或者感觉有用的代码点赞,您的评价将有助于系统推荐出更棒的Python代码示例。
示例1: test_bad_header
def test_bad_header(self):
"""
Ensure that bad data is skipped when it exists.
"""
# This case tests against a header that does not match
# 0000 0000 0000 0100 0000 0000 0000 0151
file_path = os.path.join(RESOURCE_PATH, 'E0000001-BAD-HEADER1.DAT')
self.stream_handle = open(file_path, 'rb')
with self.assertRaises(SampleException):
self.parser = DostaLnWfpParser(self.config, self.start_state, self.stream_handle,
self.state_callback, self.pub_callback, self.exception_callback)
self.stream_handle.close()
# This case tests against a header that does not match global, but matches coastal
# 0001 0000 0000 0000 0001 0001 0000 0000
file_path = os.path.join(RESOURCE_PATH, 'E0000001-BAD-HEADER2.DAT')
self.stream_handle = open(file_path, 'rb')
with self.assertRaises(SampleException):
self.parser = DostaLnWfpParser(self.config, self.start_state, self.stream_handle,
self.state_callback, self.pub_callback, self.exception_callback)
self.stream_handle.close()
示例2: test_get_many
def test_get_many(self):
"""
Read test data and pull out multiple data particles at one time.
Assert that the results are those we expected.
"""
file_path = os.path.join(RESOURCE_PATH, 'E0000001.DAT')
self.stream_handle = open(file_path, 'rb')
self.parser = DostaLnWfpParser(self.config, self.start_state, self.stream_handle,
self.state_callback, self.pub_callback, self.exception_callback)
particles = self.parser.get_records(20)
# Should end up with 20 particles
self.assertTrue(len(particles) == 20)
self.assert_result(self.test_particle3, particles[19])
particles = self.parser.get_records(30)
# Should end up with 30 particles
self.assertTrue(len(particles) == 30)
self.assert_result(self.test_particle4, particles[29])
self.stream_handle.close()
示例3: test_simple
def test_simple(self):
"""
Read test data and pull out data particles one at a time.
Assert that the results are those we expected.
"""
file_path = os.path.join(RESOURCE_PATH, 'small.DAT')
self.stream_handle = open(file_path, 'rb')
self.parser = DostaLnWfpParser(self.config, self.start_state, self.stream_handle,
self.state_callback, self.pub_callback, self.exception_callback)
particles = self.parser.get_records(6)
for particle in particles:
log.info(particle.generate_dict())
# Make sure the fifth particle has the correct values
self.assert_result(self.test_particle1, particles[5])
test_data = self.get_dict_from_yml('good.yml')
for i in range(0,6):
self.assert_result(test_data['data'][i], particles[i])
self.stream_handle.close()
示例4: test_set_state
def test_set_state(self):
"""
Test changing to a new state after initializing the parser and
reading data, as if new data has been found and the state has
changed
"""
filepath = os.path.join(RESOURCE_PATH, "E0000001.DAT")
self.stream_handle = open(filepath, "rb")
# Moving the file position past the header and two records
new_state = {StateKey.POSITION: HEADER_BYTES + (WFP_E_GLOBAL_RECOVERED_ENG_DATA_SAMPLE_BYTES * 2)}
self.parser = DostaLnWfpParser(
self.config, new_state, self.stream_handle, self.state_callback, self.pub_callback, self.exception_callback
)
particles = self.parser.get_records(4)
# Should end up with 4 particles
self.assertTrue(len(particles) == 4)
self.assert_result(self.test_particle1, particles[3])
# Moving the file position past the header and three records
new_state = {StateKey.POSITION: HEADER_BYTES + (WFP_E_GLOBAL_RECOVERED_ENG_DATA_SAMPLE_BYTES * 3)}
self.parser = DostaLnWfpParser(
self.config, new_state, self.stream_handle, self.state_callback, self.pub_callback, self.exception_callback
)
particles = self.parser.get_records(10)
# Should end up with 10 particles
self.assertTrue(len(particles) == 10)
self.assert_result(self.test_particle2, particles[9])
self.stream_handle.close()
示例5: test_bad_data
def test_bad_data(self):
"""
Ensure that bad data is skipped when it exists.
"""
file_path = os.path.join(RESOURCE_PATH, 'E0000001-BAD-DATA.DAT')
self.stream_handle = open(file_path, 'rb')
self.parser = DostaLnWfpParser(self.config, self.start_state, self.stream_handle,
self.state_callback, self.pub_callback, self.exception_callback)
with self.assertRaises(SampleException):
self.parser.get_records(1)
self.stream_handle.close()
示例6: create_large_yml
def create_large_yml(self):
"""
Create a large yml file corresponding to an actual recovered dataset. This is not an actual test - it allows
us to create what we need for integration testing, i.e. a yml file.
"""
file_path = os.path.join(RESOURCE_PATH, 'E0000001.DAT')
self.stream_handle = open(file_path, 'rb')
self.parser = DostaLnWfpParser(self.config, self.start_state, self.stream_handle,
self.state_callback, self.pub_callback, self.exception_callback)
# In a single read, get all particles in this file.
result = self.parser.get_records(1000)
self.particle_to_yml(result, 'E0000001.yml')
示例7: test_long_stream
def test_long_stream(self):
"""
Test a long stream
"""
file_path = os.path.join(RESOURCE_PATH, 'E0000002.DAT')
self.stream_handle = open(file_path, 'rb')
self.parser = DostaLnWfpParser(self.config, self.start_state, self.stream_handle,
self.state_callback, self.pub_callback, self.exception_callback)
particles = self.parser.get_records(1000)
# Should end up with 683 particles
self.assertTrue(len(particles) == 683)
self.stream_handle.close()
示例8: test_verify_record_against_yaml
def test_verify_record_against_yaml(self):
"""
Read data from a file and pull out data particles
one at a time. Verify that the results are those we expected.
"""
file_path = os.path.join(RESOURCE_PATH, 'E0000001.DAT')
self.stream_handle = open(file_path, 'rb')
self.parser = DostaLnWfpParser(self.config, self.start_state, self.stream_handle,
self.state_callback, self.pub_callback, self.exception_callback)
# In a single read, get all particles in this file.
result = self.parser.get_records(1000)
self.assert_particles(result, 'E0000001.yml', RESOURCE_PATH)
self.stream_handle.close()
示例9: test_mid_state_start
def test_mid_state_start(self):
"""
Test starting the parser in a state in the middle of processing
"""
file_path = os.path.join(RESOURCE_PATH, 'E0000001.DAT')
self.stream_handle = open(file_path, 'rb')
# Moving the file position past the header and two records
new_state = {StateKey.POSITION: HEADER_BYTES+(WFP_E_GLOBAL_RECOVERED_ENG_DATA_SAMPLE_BYTES*2)}
self.parser = DostaLnWfpParser(self.config, new_state, self.stream_handle,
self.state_callback, self.pub_callback, self.exception_callback)
particles = self.parser.get_records(4)
# Should end up with 4 particles
self.assertTrue(len(particles) == 4)
self.assert_result(self.test_particle1, particles[3])
self.stream_handle.close()
示例10: DostaLnWfpParserUnitTestCase
class DostaLnWfpParserUnitTestCase(ParserUnitTestCase):
"""
dosta_ln_wfp Parser unit test suite
"""
def state_callback(self, state, file_ingested):
""" Call back method to watch what comes in via the position callback """
self.state_callback_value = state
self.file_ingested_value = file_ingested
def pub_callback(self, pub):
""" Call back method to watch what comes in via the publish callback """
self.publish_callback_value = pub
def exception_callback(self, exception):
""" Call back method to match what comes in via the exception callback """
self.exception_callback_value = exception
def setUp(self):
ParserUnitTestCase.setUp(self)
self.config = {
DataSetDriverConfigKeys.PARTICLE_MODULE: 'mi.dataset.parser.dosta_ln_wfp',
DataSetDriverConfigKeys.PARTICLE_CLASS: 'DostaLnWfpInstrumentParserDataParticle'
}
# Define test data particles and their associated timestamps which will be
# compared with returned results
self.start_state = {StateKey.POSITION: 0}
self.file_ingested_value = None
self.state_callback_value = None
self.publish_callback_value = None
self.test_particle1 = {}
self.test_particle1['internal_timestamp'] = 3583638177
self.test_particle1[StateKey.POSITION] = 204
self.test_particle1[DostaLnWfpInstrumentParserDataParticleKey.ESTIMATED_OXYGEN_CONCENTRATION] = \
154.23699951171875
self.test_particle1[DostaLnWfpInstrumentParserDataParticleKey.OPTODE_TEMPERATURE] = 1.4819999933242798
self.test_particle1[DostaLnWfpInstrumentParserDataParticleKey.WFP_TIMESTAMP] = 1374649377
self.test_particle2 = {}
self.test_particle2['internal_timestamp'] = 3583638247
self.test_particle2[StateKey.POSITION] = 414
self.test_particle2[DostaLnWfpInstrumentParserDataParticleKey.ESTIMATED_OXYGEN_CONCENTRATION] = \
153.7899932861328
self.test_particle2[DostaLnWfpInstrumentParserDataParticleKey.OPTODE_TEMPERATURE] = 1.4950000047683716
self.test_particle2[DostaLnWfpInstrumentParserDataParticleKey.WFP_TIMESTAMP] = 1374649447
self.test_particle3 = {}
self.test_particle3['internal_timestamp'] = 3583638317
self.test_particle3[StateKey.POSITION] = 624
self.test_particle3[DostaLnWfpInstrumentParserDataParticleKey.ESTIMATED_OXYGEN_CONCENTRATION] = \
153.41099548339844
self.test_particle3[DostaLnWfpInstrumentParserDataParticleKey.OPTODE_TEMPERATURE] = 1.5
self.test_particle3[DostaLnWfpInstrumentParserDataParticleKey.WFP_TIMESTAMP] = 1374649517
self.test_particle4 = {}
self.test_particle4['internal_timestamp'] = 3583638617
self.test_particle4[StateKey.POSITION] = 1524
self.test_particle4[DostaLnWfpInstrumentParserDataParticleKey.ESTIMATED_OXYGEN_CONCENTRATION] = \
152.13600158691406
self.test_particle4[DostaLnWfpInstrumentParserDataParticleKey.OPTODE_TEMPERATURE] = 1.5019999742507935
self.test_particle4[DostaLnWfpInstrumentParserDataParticleKey.WFP_TIMESTAMP] = 1374649817
def test_simple(self):
"""
Read test data and pull out data particles one at a time.
Assert that the results are those we expected.
"""
file_path = os.path.join(RESOURCE_PATH, 'small.DAT')
self.stream_handle = open(file_path, 'rb')
self.parser = DostaLnWfpParser(self.config, self.start_state, self.stream_handle,
self.state_callback, self.pub_callback, self.exception_callback)
particles = self.parser.get_records(6)
# Make sure the fifth particle has the correct values
self.assert_result(self.test_particle1, particles[5])
test_data = self.get_dict_from_yml('good.yml')
for i in range(0,6):
self.assert_result(test_data['data'][i], particles[i])
self.stream_handle.close()
def test_get_many(self):
"""
Read test data and pull out multiple data particles at one time.
Assert that the results are those we expected.
"""
file_path = os.path.join(RESOURCE_PATH, 'E0000001.DAT')
self.stream_handle = open(file_path, 'rb')
self.parser = DostaLnWfpParser(self.config, self.start_state, self.stream_handle,
self.state_callback, self.pub_callback, self.exception_callback)
#.........这里部分代码省略.........