class UpsertIngester(Utils):
def __init__(self, db_connection):
self.db = Database(helpers_connection=db_connection)
def add_target(self, target_data, table_name, output_schema, primary_keys,
date_format):
self.target_data = target_data
self.table_name = table_name
self.schema = output_schema
self.primary_keys = primary_keys
date_cols = [
x['column_name'] for x in self.db.execute_query(
f""" SELECT column_name FROM information_schema.columns WHERE table_schema = '{output_schema}' AND table_name = '{table_name}' AND ( data_type ILIKE '%timestamp%' OR data_type ILIKE '%date%' ); """,
return_object=True)
]
if not self.target_data:
raise IndexError('Target data for table ' + self.table_name +
' is empty.')
else:
self.target_data, self.target_columns = Utils.add_missing_keys(
Utils.keys_to_snakecase(self, self.target_data))
self.column_names = ', '.join([x for x in self.target_columns])
self.columns_insert = ', '.join([
'%(' + x +
')s' if x not in [k
for k in date_cols] else f"TO_TIMESTAMP(%(" +
x + ")s, '" + date_format + "')" for x in self.target_columns
])
self._ingest_target()
def _ingest_target(self):
self.db.execute_query(
f""" INSERT INTO {self.schema}.{self.table_name}
({self.column_names})
VALUES ({self.columns_insert})
ON CONFLICT (""" +
', '.join([x for x in self.primary_keys]) + """)
DO NOTHING
;
""",
multiple=True,
args=self.target_data)
def close(self):
self.db.close()
def main():
# clean output folder
if CLEAN:
clean_folder(cfg.PATH_2D)
clean_folder(cfg.PATH_3D)
database = Database(name=cfg.DATABASE)
# LEA
if LEA_IN:
winjulea_in(os.path.join(cfg.PATH_LEA, 'winjulea.lea'))
if LEA_OUT:
evals, depths = winjulea_out(os.path.join(cfg.PATH_LEA, 'winjulea.rpt'))
np.save(os.path.join(cfg.PATH_LEA, 'evals'+cfg.FILE_SUFFIX_LEA+'.npy'), evals)
np.save(os.path.join(cfg.PATH_LEA, 'depths'+cfg.FILE_SUFFIX_LEA+'.npy'), depths)
# 2D
if D2:
d2 = Superposition(database)
d2.run(cfg.VEHICLE, cfg.NUM_TIRES)
# 3D
if D3:
d3 = FEM3D(database)
d3.run(cfg.VEHICLE, cfg.NUM_TIRES)
# plot
if PLOT:
"""
# LEA vs superposition 2D
evals_lea = np.load(os.path.join(cfg.PATH_LEA, 'evals'+cfg.FILE_SUFFIX_LEA+'.npy'))
depths_lea = np.load(os.path.join(cfg.PATH_LEA, 'depths'+cfg.FILE_SUFFIX_LEA+'.npy'))
evals_2d = np.load(os.path.join(cfg.PATH_2D, 'evals'+cfg.FILE_SUFFIX_2D+'.npy'))
depths_2d = np.load(os.path.join(cfg.PATH_2D, 'depths'+cfg.FILE_SUFFIX_2D+'.npy'))
plot_eval_depth_lea_2d(evals_lea, depths_lea, evals_2d, depths_2d, cfg.PATH_LEA)
"""
# LEA vs superposition 2D vs 3D
evals_lea = np.load(os.path.join(cfg.PATH_LEA, 'evals'+cfg.FILE_SUFFIX_LEA+'.npy'))
depths_lea = np.load(os.path.join(cfg.PATH_LEA, 'depths'+cfg.FILE_SUFFIX_LEA+'.npy'))
evals_2d = np.load(os.path.join(cfg.PATH_2D, 'evals'+cfg.FILE_SUFFIX_2D+'.npy'))
depths_2d = np.load(os.path.join(cfg.PATH_2D, 'depths'+cfg.FILE_SUFFIX_2D+'.npy'))
evals_3d = np.load(os.path.join(cfg.PATH_3D, 'evals'+cfg.FILE_SUFFIX_3D+'.npy'))
depths_3d = np.load(os.path.join(cfg.PATH_3D, 'depths'+cfg.FILE_SUFFIX_3D+'.npy'))
plot_eval_depth_all(evals_lea, depths_lea, evals_2d, depths_2d, evals_3d, depths_3d, cfg.PATH_3D)
database.close()
"""Old version
dataArray = []
for i in fields:
array = vtk.vtkDoubleArray()
array.SetName(i)
array.SetNumberOfTuples(grid.GetNumberOfPoints()) # axis 0
array.SetNumberOfComponents(1) # axis 1
dataArray.append(array)
grid.GetPointData().AddArray(array)
for i in range(results.shape[0]): # points
for j in range(results.shape[1]): # fields
dataArray[j].SetValue(i, results[i,j])
"""
if __name__ == "__main__":
database = Database(name=cfg.DATABASE)
task = Superposition(database)
mode = 'single'
if mode == 'single':
# run one vehicle (single mode)
task.run(cfg.VEHICLE, cfg.NUM_TIRES)
elif mode == 'batch':
# run all vehicles (batch mode)
for vehicle in database.vehicle_all():
task.run(vehicle, cfg.NUM_TIRES)
database.close()
from models.intent.intent_model import IntentModel
intent = IntentModel(model_name='../models/intent/intent_model.h5', proprocess=p)
predict = intent.predict_class(query)
intent_name = intent.labels[predict]
from models.ner.nermodel import NerModel
ner = NerModel(model_name='../models/ner/ner_model.h5', proprocess=p)
predicts = ner.predict(query)
ner_tags = ner.predict_tags(query)
print('질문 : ', query)
print('=' * 40)
print('의도 파악 : ', intent_name)
print('개체명 인식 : ', predicts)
print('답변 검색에 필요한 NER 태그 : ', ner_tags)
print('=' * 40)
from utils.findanswer import FindAnswer
try:
f = FindAnswer(db)
answer_text, answer_image = f.search(intent_name, ner_tags)
answer = f.tag_to_word(predicts, answer_text)
except:
answer = '죄송해요, 무슨 말인지 모르겠어요.'
print('답변 : ', answer)
db.close()
