Merge branch 'dev' of hiplab/parser into master
This commit is contained in:
commit
0547746f62
|
@ -192,18 +192,23 @@ class Apex :
|
|||
@staticmethod
|
||||
def scalar(item):
|
||||
_df = item['data']
|
||||
|
||||
name = _df.columns.tolist()[0]
|
||||
value = _df[name].values.round(2)[0]
|
||||
value = '0'
|
||||
unit = ''
|
||||
html = '<div class="scalar"><div class="value">:value</div><div class="label">:label</div></div>'
|
||||
if value > 999 and value < 1000000 :
|
||||
value = " ".join([str(np.divide(value,1000).round(2)),"K"])
|
||||
elif value > 999999 :
|
||||
#@ Think of considering the case of a billion ...
|
||||
value = " ".join([str(np.divide(value,1000000).round(2)),"M"])
|
||||
else:
|
||||
value = str(value)
|
||||
unit = name.replace('_',' ') if 'unit' not in item else item['unit']
|
||||
if _df.shape[0] > 0 :
|
||||
print (_df)
|
||||
print ('_____________________________________')
|
||||
name = _df.columns.tolist()[0]
|
||||
value = _df[name].values[0]
|
||||
|
||||
if value > 999 and value < 1000000 :
|
||||
value = " ".join([str(np.divide(value,1000).round(2)),"K"])
|
||||
elif value > 999999 :
|
||||
#@ Think of considering the case of a billion ...
|
||||
value = " ".join([str(np.divide(value,1000000).round(2)),"M"])
|
||||
else:
|
||||
value = str(value)
|
||||
unit = name.replace('_',' ') if 'unit' not in item else item['unit']
|
||||
return {'html':html.replace(':value',value).replace(":label",unit)}
|
||||
@staticmethod
|
||||
def column(item):
|
||||
|
@ -319,8 +324,9 @@ class Apex :
|
|||
values = df[x_cols].values.round(2).tolist()
|
||||
else:
|
||||
labels = [name.upper().replace('_',' ') for name in df.columns.tolist()]
|
||||
df = df.astype(float)
|
||||
values = df.values.round(2).tolist()[0] if df.shape[1] > 1 else df.values.round(2).tolist()
|
||||
# df = df.astype(float)
|
||||
# values = df.values.round(2).tolist()[0] if df.shape[1] > 1 else df.values.round(2).tolist()
|
||||
values = df[[name for name in df.columns if df[name].dtype in [float,int]] ].values.round(2).tolist()
|
||||
|
||||
colors = COLORS[:len(values)]
|
||||
options = {"series":values,"colors":colors,"labels":labels,"dataLabels":{"enabled":True,"style":{"colors":["#000000"]},"dropShadow":{"enabled":False}},"chart":{"type":"donut","width":200},"plotOptions":{"pie":{"customScale":.9}},"legend":{"position":"right"}}
|
||||
|
@ -343,10 +349,10 @@ class engine :
|
|||
self.store_config = _config['store']
|
||||
self.info = _config['analytics']
|
||||
_args = self.store_config
|
||||
if self.store_config['type'] == 'mongo.MongoWriter' :
|
||||
_args['type'] = 'mongo.MongoReader'
|
||||
else:
|
||||
_args['type'] = 'disk.SQLiteReader'
|
||||
if 'type' not in self.store_config :
|
||||
#
|
||||
# This is the newer version of data-transport
|
||||
self.store_config['context'] = 'read'
|
||||
self.store_config = _args ;
|
||||
|
||||
def filter (self,**args):
|
||||
|
@ -367,8 +373,8 @@ class engine :
|
|||
# conn = lite.connect(self.store_config['args']['path'],isolation_level=None)
|
||||
# conn.create_aggregate("stdev",1,stdev)
|
||||
DB_TYPE = 'mongo' if (type(self.reader) == transport.mongo.MongoReader) else 'sql'
|
||||
if DB_TYPE == 'mongo' :
|
||||
self.store_config['args']['doc'] = args['type']
|
||||
# if DB_TYPE == 'mongo' :
|
||||
# self.store_config['args']['doc'] = args['type']
|
||||
|
||||
self.reader = transport.factory.instance(**self.store_config)
|
||||
r = []
|
||||
|
@ -414,20 +420,8 @@ class engine :
|
|||
_analytics = [_analytics[index]]
|
||||
|
||||
_info = list(_analytics) if 'filter' not in args else [item for item in analytics if args['filter'] == item['id']]
|
||||
# conn = lite.connect(self.store_config['args']['path'],isolation_level=None)
|
||||
# conn.create_aggregate("stdev",1,stdev)
|
||||
#
|
||||
# @TODO: Find a better way to handle database variance
|
||||
#
|
||||
# DB_TYPE = 'mongo' if (type(self.reader) == transport.mongo.MongoReader) else 'sql'
|
||||
|
||||
if 'mongo' in self.store_config['type'] :
|
||||
DB_TYPE='mongo'
|
||||
else:
|
||||
DB_TYPE='sql'
|
||||
self.store_config['args']['table'] = args['type']
|
||||
|
||||
self.reader = transport.factory.instance(**self.store_config)
|
||||
DB_TYPE = 'mongo' if self.store_config ['provider'] in ['mongodb','mongo'] else 'sql'
|
||||
r = []
|
||||
for row in _info :
|
||||
pipeline = row['pipeline']
|
||||
|
@ -440,14 +434,22 @@ class engine :
|
|||
continue
|
||||
if DB_TYPE == 'sql' :
|
||||
query = {"sql":query}
|
||||
else:
|
||||
query = {DB_TYPE:query}
|
||||
|
||||
item['data'] = self.reader.read(**query) #item)
|
||||
_df = self.reader.read(**query) #item)
|
||||
print (query)
|
||||
print (self.reader)
|
||||
if 'serialize' in args :
|
||||
|
||||
# item['data'] = json.dumps(item['data'].to_dict(orient='record')) if type(item['data']) == pd.DataFrame else item['data']
|
||||
item['data'] = json.dumps(item['data'].to_dict('record')) if type(item['data']) == pd.DataFrame else item['data']
|
||||
item['data'] = json.dumps(_df.to_dict(orient='record'))
|
||||
else:
|
||||
item['data'] = (pd.DataFrame(item['data']))
|
||||
# item['data'] = (pd.DataFrame(item['data']))
|
||||
item['data'] = _df
|
||||
pass
|
||||
print (_df.head())
|
||||
break
|
||||
pipeline[index] = item
|
||||
index += 1
|
||||
#
|
||||
|
|
|
@ -15,7 +15,7 @@ RUN ["apt-get","install","-y","mongodb","sqlite3","sqlite3-pcre","libsqlite3-dev
|
|||
#
|
||||
RUN ["pip3","install","--upgrade","pip"]
|
||||
RUN ["pip3","install","numpy","pandas","git+https://dev.the-phi.com/git/steve/data-transport","botocore","matplotlib"]
|
||||
# RUN ["pip3","install","git+https://healthcare.the-phi.com/git/code/parser.git","botocore"]
|
||||
RUN ["pip3","install","git+https://healthcare.the-phi.com/git/code/parser.git"]
|
||||
# RUN ["useradd", "-ms", "/bin/bash", "health-user"]
|
||||
# USER health-user
|
||||
#
|
||||
|
|
|
@ -191,6 +191,7 @@ def init_sql(**_args):
|
|||
_config = jsonmerge.merge(_config,CUSTOM_CONFIG[TYPE])
|
||||
#
|
||||
_info = meta(_config)
|
||||
|
||||
_projectSQLite = [] #-- sqlite projection
|
||||
for field_name in _info['main'] :
|
||||
_projectSQLite += ["json_extract(data,'$."+field_name+"') "+field_name]
|
||||
|
@ -211,12 +212,17 @@ def init_sql(**_args):
|
|||
SQL = "SELECT DISTINCT :fields FROM "+TABLE_NAME+", json_each(data) x, json_each(x.value) i where x.key = ':table'"
|
||||
SQL = SQL.replace(":table",table).replace(":fields",",".join(project))
|
||||
r += [{"table":table,"read":{"sql":SQL},"sql":create(table=table,key='claim_id',fields=fields)}]
|
||||
|
||||
return r
|
||||
def init(**_args):
|
||||
if 'provider' in CONFIG['store'] and CONFIG['store']['provider'] == 'sqlite' :
|
||||
return init_sql(**_args)
|
||||
else:
|
||||
# if 'provider' in CONFIG['store'] and CONFIG['store']['provider'] == 'sqlite' :
|
||||
# return init_sql(**_args)
|
||||
# else:
|
||||
# return init_mongo(**_args)
|
||||
if ('provider' in CONFIG['store'] and CONFIG['store']['provider'] == 'mongo') or ('type' in CONFIG['store'] and 'mongo' in CONFIG['store']['type']):
|
||||
return init_mongo(**_args)
|
||||
else:
|
||||
return init_sql(**_args)
|
||||
def init_mongo (**_args) :
|
||||
"""
|
||||
This function is intended to determine the number of tables to be created, as well as their type.
|
||||
|
@ -330,6 +336,7 @@ class Factory:
|
|||
job_args = init(type=X12_TYPE) #-- getting the queries that will generate the objects we are interested in
|
||||
# print (json.dumps(job_args))
|
||||
_jobs = []
|
||||
|
||||
for row in job_args:
|
||||
# _store = json.loads(json.dumps(wstore))
|
||||
_store = copy.deepcopy(wstore)
|
||||
|
|
|
@ -12,6 +12,7 @@ from multiprocessing import Process, Lock
|
|||
import numpy as np
|
||||
import json
|
||||
import pandas as pd
|
||||
from zmq import has
|
||||
|
||||
class Subject (Process):
|
||||
cache = pd.DataFrame()
|
||||
|
@ -94,8 +95,8 @@ class Worker :
|
|||
except Exception as error:
|
||||
pass
|
||||
finally:
|
||||
|
||||
self.caller.notify()
|
||||
if hasattr(self,'caller') :
|
||||
self.caller.notify()
|
||||
def _apply(self):
|
||||
pass
|
||||
def get(self):
|
||||
|
@ -176,11 +177,16 @@ class Reader(Worker):
|
|||
def init(self,**_args):
|
||||
super().init(**_args)
|
||||
self.rows = []
|
||||
|
||||
|
||||
|
||||
def _apply(self):
|
||||
try:
|
||||
|
||||
if 'type' in self._info :
|
||||
self._info['type'] = self._info['type'].replace('Writer','Reader')
|
||||
if 'fields' in self._info['args'] :
|
||||
del self._info['args']['fields']
|
||||
else:
|
||||
self._info['context'] = 'read'
|
||||
self.reader = transport.factory.instance(**self._info) ;
|
||||
|
||||
# self.rows = self.reader.read(mongo=self.pipeline)
|
||||
|
@ -206,7 +212,7 @@ class Reader(Worker):
|
|||
except Exception as e :
|
||||
_log['status'] = 0
|
||||
_log['info'] = {"error":e.args[0]}
|
||||
print (e)
|
||||
print ([e])
|
||||
|
||||
|
||||
self.log(**_log)
|
||||
|
@ -221,13 +227,13 @@ class Writer(Worker):
|
|||
super().__init__(**_args)
|
||||
if 'provider' in self._info :
|
||||
self._info['context'] = 'write'
|
||||
|
||||
|
||||
def init(self,**_args):
|
||||
"""
|
||||
:param store output data-store needed for writing
|
||||
:param invalues input values with to be written somewhere
|
||||
"""
|
||||
super().init(**_args)
|
||||
|
||||
|
||||
self._invalues = _args['invalues']
|
||||
|
||||
|
@ -259,8 +265,8 @@ class Writer(Worker):
|
|||
# Upon upgrade use the operator "$toString" in export.init function
|
||||
#
|
||||
rows = [dict(item,**{"_id":str(item["_id"])}) for item in rows]
|
||||
|
||||
writer.write(rows)
|
||||
_df = pd.DataFrame(rows)
|
||||
writer.write(_df)
|
||||
index += 1
|
||||
# for _e in rows :
|
||||
# writer.write(_e)
|
||||
|
|
|
@ -122,7 +122,14 @@ def signup (**args) :
|
|||
file = open( filename,'w')
|
||||
file.write( json.dumps(info))
|
||||
file.close()
|
||||
_m = """
|
||||
Thank you for signingup!!
|
||||
Your configuration file is store in :path,
|
||||
- More information visit https://healthcareio.the-phi.com/parser
|
||||
- Access the source https://healthcareio.the-phi.com/git/code/parser
|
||||
|
||||
""".replace(":path",CONFIG_FILE)
|
||||
print (_m)
|
||||
#
|
||||
# Create the sqlite3 database to
|
||||
|
||||
|
@ -155,12 +162,13 @@ def init():
|
|||
# if 'type' in info['store'] :
|
||||
lwriter = None
|
||||
is_sqlite = False
|
||||
if'type' in info['store'] and info['store']['type'] == 'disk.SQLiteWriter' and not os.path.exists(info['store']['args']['path']) :
|
||||
if'type' in info['store'] and info['store']['type'] == 'disk.SQLiteWriter' :
|
||||
lwriter = transport.factory.instance(**info['store'])
|
||||
is_sqlite = True
|
||||
elif 'provider' in info['store'] and info['store']['provider'] == 'sqlite' :
|
||||
lwriter = transport.instance(**info['store']) ;
|
||||
is_sqlite = True
|
||||
|
||||
if lwriter and is_sqlite:
|
||||
for key in info['schema'] :
|
||||
if key != 'logs' :
|
||||
|
@ -168,7 +176,7 @@ def init():
|
|||
else:
|
||||
_id = key
|
||||
|
||||
if not lwriter.has(table=_id) :
|
||||
if not lwriter.has(table=_id) :
|
||||
lwriter.apply(info['schema'][key]['create'])
|
||||
|
||||
# [lwriter.apply( info['schema'][key]['create']) for key in info['schema'] if not lwriter.has(table=key)]
|
||||
|
@ -285,6 +293,24 @@ if __name__ == '__main__' :
|
|||
procs = [proc for proc in procs if proc.is_alive()]
|
||||
time.sleep(2)
|
||||
|
||||
uri = OUTPUT_FOLDER
|
||||
store_config = json.loads( (open(CONFIG_FILE)).read() )['store']
|
||||
if 'type' in store_config :
|
||||
uri = store_config['args']['host'] if 'host' in store_config['args'] else ( store_config['args']['path'] if 'path' in store_config['args'] else store_config['args']['database'])
|
||||
if 'SQLite' in store_config['type']:
|
||||
provider = 'sqlite'
|
||||
elif 'sql' in store_config['type'] :
|
||||
provider = 'SQL'
|
||||
else:
|
||||
provider = 'mongo'
|
||||
|
||||
else:
|
||||
provider = store_config['provider']
|
||||
_msg = """
|
||||
Completed Parsing, The data is available in :provider database at :uri
|
||||
Logs are equally available for errors and summary statistics to be compiled
|
||||
""".replace(":provider",provider).replace(":uri",uri)
|
||||
print (_msg)
|
||||
|
||||
|
||||
pass
|
||||
|
|
|
@ -225,6 +225,7 @@ if __name__ == '__main__' :
|
|||
PATH= SYS_ARGS['config'] if 'config' in SYS_ARGS else os.sep.join([os.environ['HOME'],'.healthcareio','config.json'])
|
||||
#
|
||||
# Adjusting configuration with parameters (batch,folder,resume)
|
||||
SYS_ARGS['config'] = json.loads(open(PATH).read())
|
||||
if 'args' not in SYS_ARGS['config'] :
|
||||
SYS_ARGS['config']["args"] = {"batch":1,"resume":True,"folder":"/data"}
|
||||
|
||||
|
|
|
@ -6,15 +6,16 @@ import json
|
|||
app = Flask(__name__)
|
||||
@app.route("/favicon.ico")
|
||||
def _icon():
|
||||
return send_from_directory(os.path.join([app.root_path, 'static','img','logo.svg']),
|
||||
return send_from_directory(os.path.join([app.root_path, 'static/img/logo.svg']),
|
||||
'favicon.ico', mimetype='image/vnd.microsoft.icon')
|
||||
@app.route("/")
|
||||
def init():
|
||||
e = SYS_ARGS['engine']
|
||||
sections = {"remits":e.info['835'],"claims":e.info['837']}
|
||||
_args = {"sections":sections,"store":SYS_ARGS['config']['store']}
|
||||
print (SYS_ARGS['config']['store'])
|
||||
return render_template("setup.html",**_args)
|
||||
|
||||
_args = {"sections":sections,"store":SYS_ARGS['config']['store'],'args':{'batch':5}}
|
||||
|
||||
return render_template("index.html",**_args)
|
||||
@app.route("/format/<id>/<index>",methods=['POST'])
|
||||
def _format(id,index):
|
||||
|
||||
|
@ -73,13 +74,16 @@ def reload():
|
|||
|
||||
if __name__ == '__main__' :
|
||||
PORT = int(SYS_ARGS['port']) if 'port' in SYS_ARGS else 5500
|
||||
DEBUG= int(SYS_ARGS['debug']) if 'debug' in SYS_ARGS else 0
|
||||
DEBUG= int(SYS_ARGS['debug']) if 'debug' in SYS_ARGS else 1
|
||||
SYS_ARGS['context'] = SYS_ARGS['context'] if 'context' in SYS_ARGS else ''
|
||||
#
|
||||
#
|
||||
PATH= SYS_ARGS['config'] if 'config' in SYS_ARGS else os.sep.join([os.environ['HOME'],'.healthcareio','config.json'])
|
||||
|
||||
#
|
||||
if os.path.exists(PATH) :
|
||||
SYS_ARGS['config'] = json.loads(open(PATH).read())
|
||||
e = healthcareio.analytics.engine(PATH)
|
||||
# e.apply(type='claims',serialize=True)
|
||||
SYS_ARGS['engine'] = e
|
||||
|
||||
app.run(host='0.0.0.0',port=PORT,debug=DEBUG,threaded=True)
|
|
@ -37,14 +37,17 @@ class get :
|
|||
|
||||
@staticmethod
|
||||
def processes(_args):
|
||||
_info = pd.DataFrame(smart.top.read(name='healthcare-io.py'))[['name','cpu','mem']]
|
||||
APP_NAME ='healthcare-io'
|
||||
_info = smart.top.read(name=APP_NAME) #pd.DataFrame(smart.top.read(name='healthcare-io'))[['name','cpu','mem']]
|
||||
|
||||
|
||||
if _info.shape[0] == 0 :
|
||||
_info = pd.DataFrame({"name":["healthcare-io.py"],"cpu":[0],"mem":[0]})
|
||||
_info = pd.DataFrame({"name":[APP_NAME],"cpu":[0],"mem":[0]})
|
||||
# _info = pd.DataFrame(_info.groupby(['name']).sum())
|
||||
# _info['name'] = ['healthcare-io.py']
|
||||
m = {'cpu':'CPU','mem':'RAM','name':'name'}
|
||||
_info.columns = [m[name] for name in _info.columns.tolist()]
|
||||
_info = _info.rename(columns=m)
|
||||
# _info.columns = [m[name] for name in _info.columns.tolist() if name in m]
|
||||
_info.index = np.arange(_info.shape[0])
|
||||
|
||||
charts = []
|
||||
|
@ -56,23 +59,20 @@ class get :
|
|||
{"data":df, "chart":{"type":"radial","axis":{"x":label,"y":"name"}}}
|
||||
)['apex']
|
||||
)
|
||||
#
|
||||
# This will update the counts for the processes, upon subsequent requests so as to show the change
|
||||
#
|
||||
N = 0
|
||||
lprocs = []
|
||||
for proc in get.PROCS :
|
||||
if proc.is_alive() :
|
||||
lprocs.append(proc)
|
||||
N = len(lprocs)
|
||||
get.PROCS = lprocs
|
||||
return {"process":{"chart":charts,"counts":N}}
|
||||
|
||||
return {"process":{"chart":charts,"counts":_info.shape[0]}}
|
||||
@staticmethod
|
||||
def files (_args):
|
||||
_info = smart.folder.read(path='/data')
|
||||
folder = _args['args']['folder']
|
||||
_info = smart.folder.read(path=folder)
|
||||
|
||||
N = _info.files.tolist()[0]
|
||||
if 'mongo' in _args['store']['type'] :
|
||||
store_args = dict(_args['store'].copy(),**{"type":"mongo.MongoReader"})
|
||||
store_args = _args['store'].copy()
|
||||
store_args['context'] = 'read'
|
||||
|
||||
# if 'mongo' in _args['store']['type'] :
|
||||
if _args['store']['provider'] in ['mongo', 'mongodb']:
|
||||
# store_args = dict(_args['store'].copy(),**{"type":"mongo.MongoReader"})
|
||||
# reader = transport.factory.instance(**_args)
|
||||
|
||||
pipeline = [{"$group":{"_id":"$name","count":{"$sum":{"$cond":[{"$eq":["$completed",True]},1,0]}} }},{"$group":{"_id":None,"count":{"$sum":"$count"}}},{"$project":{"_id":0,"status":"completed","count":1}}]
|
||||
|
@ -83,12 +83,15 @@ class get :
|
|||
|
||||
|
||||
else:
|
||||
store_args = dict(_args['store'].copy(),**{"type":"disk.SQLiteReader"})
|
||||
store_args['args']['table'] = 'logs'
|
||||
# store_args = dict(_args['store'].copy(),**{"type":"disk.SQLiteReader"})
|
||||
|
||||
# store_args['args']['table'] = 'logs'
|
||||
store_args['table'] = 'logs'
|
||||
query= {"sql":"select count(distinct json_extract(data,'$.name')) as count, 'completed' status from logs where json_extract(data,'$.completed') = true"}
|
||||
_query={"sql":"select json_extract(data,'$.parse') as type,count(distinct json_extract(data,'$.name')) as count from logs group by type"} #-- distribution claim/remits
|
||||
reader = transport.factory.instance(**store_args)
|
||||
_info = pd.DataFrame(reader.read(**query))
|
||||
|
||||
if not _info.shape[0] :
|
||||
_info = pd.DataFrame({"status":["completed"],"count":[0]})
|
||||
_info['count'] = np.round( (_info['count'] * 100 )/N,2)
|
||||
|
@ -97,11 +100,6 @@ class get :
|
|||
#
|
||||
# Let us classify the files now i.e claims / remits
|
||||
#
|
||||
|
||||
|
||||
# pipeline = [{"$group":{"_id":"$parse","claims":{"$addToSet":"$name"}}},{"$project":{"_id":0,"type":"$_id","count":{"$size":"$claims"}}}]
|
||||
# _args = {"aggregate":"logs","cursor":{},"allowDiskUse":True,"pipeline":pipeline}
|
||||
# r = pd.DataFrame(reader.read(mongo=_args))
|
||||
r = pd.DataFrame(reader.read(**_query)) #-- distribution claims/remits
|
||||
r = Apex.apply({"chart":{"type":"donut","axis":{"x":"count","y":"type"}},"data":r})['apex']
|
||||
r['chart']['height'] = '100%'
|
||||
|
|
|
@ -63,10 +63,10 @@
|
|||
//
|
||||
// We should insure the listeners are enabled
|
||||
if(monitor.listen.handler == null){
|
||||
monitor.listen.handler = setInterval(
|
||||
/*monitor.listen.handler = setInterval(
|
||||
function(){
|
||||
console.log('running ...')
|
||||
monitor.data()},5000)
|
||||
monitor.data()},5000)*/
|
||||
|
||||
}
|
||||
}else{
|
||||
|
|
|
@ -24,6 +24,7 @@ import sys
|
|||
from itertools import islice
|
||||
from multiprocessing import Process
|
||||
import transport
|
||||
from transport import providers
|
||||
import jsonmerge
|
||||
|
||||
import copy
|
||||
|
@ -236,7 +237,54 @@ class Parser (Process):
|
|||
_config[_id] = [_config[_id]]
|
||||
config['parser'] = _config
|
||||
return config
|
||||
|
||||
@staticmethod
|
||||
def init(**_args):
|
||||
"""
|
||||
This function allows to initialize the database that will store the claims if need be
|
||||
:path configuration file
|
||||
"""
|
||||
PATH = os.sep.join([os.environ['HOME'],'.healthcareio'])
|
||||
filename = os.sep.join([PATH,'config.json'])
|
||||
|
||||
filename = _args['path'] if 'path' in _args else filename
|
||||
info = None
|
||||
if os.path.exists(filename):
|
||||
#
|
||||
# Loading the configuration file (JSON format)
|
||||
file = open(filename)
|
||||
info = json.loads(file.read())
|
||||
|
||||
|
||||
OUTPUT_FOLDER = info['out-folder']
|
||||
if 'output-folder' not in info and not os.path.exists(OUTPUT_FOLDER) :
|
||||
os.mkdir(OUTPUT_FOLDER)
|
||||
elif 'output-folder' in info and not os.path.exists(info['out-folder']) :
|
||||
os.mkdir(info['out-folder'])
|
||||
# if 'type' in info['store'] :
|
||||
lwriter = None
|
||||
IS_SQL = False
|
||||
if'type' in info['store'] and info['store']['type'] == 'disk.SQLiteWriter' :
|
||||
lwriter = transport.factory.instance(**info['store'])
|
||||
IS_SQL = True
|
||||
elif 'provider' in info['store'] and info['store']['provider'] == 'sqlite' :
|
||||
lwriter = transport.instance(**info['store']) ;
|
||||
IS_SQL = [providers.SQLITE,providers.POSTGRESQL,providers.NETEZZA,providers.MYSQL,providers.MARIADB]
|
||||
|
||||
if lwriter and IS_SQL:
|
||||
for key in info['schema'] :
|
||||
if key != 'logs' :
|
||||
_id = 'claims' if key == '837' else 'remits'
|
||||
else:
|
||||
_id = key
|
||||
|
||||
if not lwriter.has(table=_id) :
|
||||
lwriter.apply(info['schema'][key]['create'])
|
||||
|
||||
# [lwriter.apply( info['schema'][key]['create']) for key in info['schema'] if not lwriter.has(table=key)]
|
||||
lwriter.close()
|
||||
|
||||
return info
|
||||
|
||||
def __init__(self,path):
|
||||
"""
|
||||
:path path of the configuration file (it can be absolute)
|
||||
|
|
2
setup.py
2
setup.py
|
@ -8,7 +8,7 @@ import sys
|
|||
def read(fname):
|
||||
return open(os.path.join(os.path.dirname(__file__), fname)).read()
|
||||
args = {
|
||||
"name":"healthcareio","version":"1.6.4.6",
|
||||
"name":"healthcareio","version":"1.6.4.8",
|
||||
"author":"Vanderbilt University Medical Center",
|
||||
"author_email":"steve.l.nyemba@vumc.org",
|
||||
"include_package_data":True,
|
||||
|
|
Loading…
Reference in New Issue