Speedup parser

This commit is contained in:
dvanaken 2020-02-20 15:24:12 -05:00 committed by Dana Van Aken
parent 8768686d17
commit 944a0ccc8d
2 changed files with 31 additions and 24 deletions

View File

@ -213,32 +213,34 @@ class BaseParser:
def extract_valid_variables(self, variables, catalog, default_value=None): def extract_valid_variables(self, variables, catalog, default_value=None):
valid_variables = {} valid_variables = {}
diff_log = [] diff_log = OrderedDict([(k, []) for k in ('miscapitalized', 'extra', 'missing')])
valid_lc_variables = {k.name.lower(): k for k in catalog} lc_catalog = {k.lower(): v for k, v in catalog.items()}
# First check that the names of all variables are valid (i.e., listed # First check that the names of all variables are valid (i.e., listed
# in the official catalog). Invalid variables are logged as 'extras'. # in the official catalog). Invalid variables are logged as 'extras'.
# Variable names that are valid but differ in capitalization are still # Variable names that are valid but differ in capitalization are still
# added to valid_variables but with the proper capitalization. They # added to valid_variables but with the proper capitalization. They
# are also logged as 'miscapitalized'. # are also logged as 'miscapitalized'.
for var_name, var_value in list(variables.items()): for var_name, var_value in variables.items():
if var_name in catalog:
valid_variables[var_name] = var_value
else:
lc_var_name = var_name.lower() lc_var_name = var_name.lower()
if lc_var_name in valid_lc_variables: if lc_var_name in lc_catalog:
valid_name = valid_lc_variables[lc_var_name].name valid_name = lc_catalog[lc_var_name].name
if var_name != valid_name: diff_log['miscapitalized'].append((valid_name, var_name))
diff_log.append(('miscapitalized', valid_name, var_name, var_value))
valid_variables[valid_name] = var_value valid_variables[valid_name] = var_value
else: else:
diff_log.append(('extra', None, var_name, var_value)) diff_log['extra'].append(var_name)
# Next find all item names that are listed in the catalog but missing from # Next find all item names that are listed in the catalog but missing from
# variables. Missing variables are added to valid_variables with the given # variables. Missing variables are added to valid_variables with the given
# default_value if provided (or the item's actual default value if not) and # default_value if provided (or the item's actual default value if not) and
# logged as 'missing'. # logged as 'missing'.
lc_variables = {k.lower(): v for k, v in list(variables.items())} lc_variables = {k.lower() for k in variables.keys()}
for valid_lc_name, metadata in list(valid_lc_variables.items()): for valid_lc_name, metadata in lc_catalog.items():
if valid_lc_name not in lc_variables: if valid_lc_name not in lc_variables:
diff_log.append(('missing', metadata.name, None, None)) diff_log['missing'].append(metadata.name)
valid_variables[metadata.name] = default_value if \ valid_variables[metadata.name] = default_value if \
default_value is not None else metadata.default default_value is not None else metadata.default
assert len(valid_variables) == len(catalog) assert len(valid_variables) == len(catalog)
@ -277,7 +279,7 @@ class BaseParser:
assert len(valid_knobs[k]) == 1 assert len(valid_knobs[k]) == 1
valid_knobs[k] = valid_knobs[k][0] valid_knobs[k] = valid_knobs[k][0]
# Extract all valid knobs # Extract all valid knobs
knob_catalog = KnobCatalog.objects.filter(dbms__id=self.dbms_id) knob_catalog = {k.name: k for k in KnobCatalog.objects.filter(dbms__id=self.dbms_id)}
return self.extract_valid_variables(valid_knobs, knob_catalog) return self.extract_valid_variables(valid_knobs, knob_catalog)
def parse_dbms_metrics(self, metrics): def parse_dbms_metrics(self, metrics):
@ -287,14 +289,15 @@ class BaseParser:
valid_metrics = self.parse_dbms_variables(metrics) valid_metrics = self.parse_dbms_variables(metrics)
# Extract all valid metrics # Extract all valid metrics
metric_catalog = MetricCatalog.objects.filter(dbms__id=self.dbms_id) metric_catalog = {m.name: m for m in MetricCatalog.objects.filter(dbms__id=self.dbms_id)}
valid_metrics, diffs = self.extract_valid_variables( valid_metrics, diffs = self.extract_valid_variables(
valid_metrics, metric_catalog, default_value='0') valid_metrics, metric_catalog, default_value='0')
# Combine values # Combine values
for name, values in list(valid_metrics.items()): for name, values in list(valid_metrics.items()):
metric = metric_catalog.get(name=name) metric = metric_catalog[name]
if metric.metric_type in MetricType.nonnumeric() or len(values) == 1: if len(values) == 1 or metric.metric_type in MetricType.nonnumeric():
valid_metrics[name] = values[0] valid_metrics[name] = values[0]
elif metric.metric_type in MetricType.numeric(): elif metric.metric_type in MetricType.numeric():
conv_fn = int if metric.vartype == VarType.INTEGER else float conv_fn = int if metric.vartype == VarType.INTEGER else float
@ -309,11 +312,12 @@ class BaseParser:
return valid_metrics, diffs return valid_metrics, diffs
def calculate_change_in_metrics(self, metrics_start, metrics_end, fix_metric_type=True): def calculate_change_in_metrics(self, metrics_start, metrics_end, fix_metric_type=True):
metric_catalog = MetricCatalog.objects.filter(dbms__id=self.dbms_id) metric_catalog = {m.name: m for m in MetricCatalog.objects.filter(dbms__id=self.dbms_id)}
adjusted_metrics = {} adjusted_metrics = {}
for met_name, start_val in list(metrics_start.items()):
for met_name, start_val in metrics_start.items():
end_val = metrics_end[met_name] end_val = metrics_end[met_name]
met_info = metric_catalog.get(name=met_name) met_info = metric_catalog[met_name]
if met_info.vartype == VarType.INTEGER or \ if met_info.vartype == VarType.INTEGER or \
met_info.vartype == VarType.REAL: met_info.vartype == VarType.REAL:
conversion_fn = self.convert_integer if \ conversion_fn = self.convert_integer if \

View File

@ -556,7 +556,7 @@ def handle_result_files(session, files, execution_times=None):
else: else:
dbms_type = DBMSType.type(summary['database_type']) dbms_type = DBMSType.type(summary['database_type'])
dbms_version = summary['database_version'] # TODO: fix parse_version_string dbms_version = summary['database_version']
workload_name = summary['workload_name'] workload_name = summary['workload_name']
observation_time = summary['observation_time'] observation_time = summary['observation_time']
start_time = datetime.fromtimestamp( start_time = datetime.fromtimestamp(
@ -604,7 +604,10 @@ def handle_result_files(session, files, execution_times=None):
dbms.pk, JSONUtil.loads(files['metrics_after'])) dbms.pk, JSONUtil.loads(files['metrics_after']))
metric_dict = parser.calculate_change_in_metrics( metric_dict = parser.calculate_change_in_metrics(
dbms.pk, initial_metric_dict, final_metric_dict) dbms.pk, initial_metric_dict, final_metric_dict)
initial_metric_diffs.extend(final_metric_diffs) metric_diffs = OrderedDict([
('metrics_before', initial_metric_diffs),
('metrics_after', final_metric_diffs),
])
numeric_metric_dict = parser.convert_dbms_metrics( numeric_metric_dict = parser.convert_dbms_metrics(
dbms.pk, metric_dict, observation_time, session.target_objective) dbms.pk, metric_dict, observation_time, session.target_objective)
metric_data = MetricData.objects.create_metric_data( metric_data = MetricData.objects.create_metric_data(
@ -631,8 +634,8 @@ def handle_result_files(session, files, execution_times=None):
raw_initial_metrics=files['metrics_before'], raw_initial_metrics=files['metrics_before'],
raw_final_metrics=files['metrics_after'], raw_final_metrics=files['metrics_after'],
raw_summary=files['summary'], raw_summary=files['summary'],
knob_log=knob_diffs, knob_log=JSONUtil.dumps(knob_diffs, pprint=True),
metric_log=initial_metric_diffs) metric_log=JSONUtil.dumps(metric_diffs, pprint=True))
backup_data.save() backup_data.save()
session.project.last_update = now() session.project.last_update = now()