labkeyInterface/importXLSX.py

import os
import pandas 
import sys
import importlib
import re
import datetime
import chardet
import json
import math

#same directory!
import labkeyInterface
import labkeyDatabaseBrowser
import labkeyFileBrowser

def connectDB(server):
   net=labkeyInterface.labkeyInterface()
   qfile='{}.json'.format(server)
   fconfig=os.path.join(os.path.expanduser('~'),'.labkey',qfile)
   net.init(fconfig)
   net.getCSRF()
   return labkeyDatabaseBrowser.labkeyDB(net)

def getDB(pars):
   try:
      return pars['db']
   except KeyError:
      pass
   server=pars.get('server','onko-nix')
   db=connectDB(server)
   pars['db']=db
   return db

def getFields(pars):
   project=pars.get('project','DCIS/Study')
   schema=pars.get('schema','demographics')
   query=pars.get('query','demographics')
   db=getDB(pars)
    
#data on query structure are both in queryDesign and selectRows/metadata content
    
   dsgn=db.getQueryDesign(project,schema,query)
   dsgnFields={f['name']:f for f in dsgn['fields']} 

   ds=db.selectRows(project,schema,query,[])
   mdFields={x['name']:x for x in ds['metaData']['fields']}
    
#in principle, any property from mdFields could get copied 
#to content reported from dsgnFields
#try/except to robustify against missing components
   copyFields=['lookup']
   for f in dsgnFields:
      try:
         dsgnFields[f].update({x:mdFields[f][x] for x in copyFields})
      except KeyError:
         pass
   return dsgnFields

def getAlias(fields):
   fieldMap={}
   for f in fields.values():
      aliasList=getAliasList(f['importAliases'])
      fieldMap.update({x:f['name'] for x in aliasList})
   return fieldMap

def invertMap(qmap):
   return {qmap[x]:x for x in qmap} 

def getVariables(fields,fieldType='LOOKUP',fieldName=None):
#get list of variables of particular type to help manage import
#if type is LOOKUP also return data on lookup query
   if fieldName:
      return {fieldName:fields[fieldName]['lookupQuery']}
   if fieldType=='LOOKUP':
      return {f['name']:f['lookupQuery'] for f in fields.values() \
         if f['lookupQuery']}
   if fieldType=='DATE':
      return {f['name']:fieldType for f in fields.values() \
         if f['rangeURI'].find('dateTime')>-1}
   if fieldType=='DOUBLE':
      return {f['name']:fieldType for f in fields.values() \
         if f['rangeURI'].find('double')>-1}
   return {}


def getLookupMap(pars,fields,fieldName):
#get possible values of categorical variables/factors from labkey
   try:
      lookup=fields[fieldName]['lookup']
   except KeyError:
      print(fields[fieldName])
      raise KeyError(f'Could not find lookup for {fieldName}')
   schema=lookup['schemaName']
   query=lookup['queryName']

   key=lookup['keyColumn']
   val=lookup['displayColumn']

   project=pars['project']
   db=getDB(pars)
   ds=db.selectRows(project,schema,query,[])
   cMap={r[val]:r[key] for r in ds['rows']}
   return cMap
    
def parseLookup(lookup,qv):
#parse/convert lookup 
#is it key?
   try:
      return lookup[qv]
        
   except KeyError:
      pass
    
   if pandas.isna(qv):
      return qv
    
   try:
      qv=qv.item()
      qv=str(qv)
      return lookup[qv]
    
   except AttributeError:
      pass

   qv=qv.replace('Č','C')
   return lookup[qv]

def asKey(qv):
   if not qv:
      return qv
   try:
      return int(qv)
   except (TypeError,ValueError):
      print(f'Failed to parse {qv} as key')
      return None

def parseDate(qv):

   if not qv:
      return qv

#from xls format to native python format
   fmts=['datetime','pandas','%d.%m.%y','%Y-%m-%d %H:%M:%S.%f',\
      '%d/%m/%Y','%Y-%m-%d %H:%M:%S']

   usedFmt=None
   for fmt in fmts:
      usedFmt=fmt
      try:

         if fmt=='pandas':
#print(f'Trying {qv} as pandas.Timestamp')
            date=pandas.Timestamp.to_pydatetime(qv)
            break
         elif fmt=='datetime':
#print(f'Trying {qv} as datetime.datetime')
            if not isinstance(qv,datetime.datetime):
               raise TypeError('Not a datetime object')
            date=qv
            break
         else:
#print(f'Trying {qv} with {fmt}')
            date=datetime.datetime.strptime(qv,fmt)
            break

      except TypeError:
#print('Failed (type): {}'.format(type(qv)))
         continue
      except ValueError:
#print('Failed (value)')
         continue
      
#sometimes parsing fails
   try:
      return date.isoformat()
   except UnboundLocalError:
      print (f'Failed to parse {qv} as date, used {usedFmt}')
      return None

def parseDouble(qv):
   try:
      return float(qv)
   except (ValueError,TypeError):
      return None
    
    #m
def setMissingLookupValues(filename,xlsFieldName,project,lookup,\
   labkeyFieldName=None,dryRun=True):
    
#list all possible values for a field
#perhaps needs to be updated
   df=pandas.read_excel(filename)
   vars=df.columns
    
   vals=set([df.at[r,xlsFieldName] for r in df.index if not pandas.isna(df.at[r,xlsFieldName])])
   try:
      vals={v.item() for v in vals}
   except AttributeError:
      pass
   print(vals)
    
    
   if not labkeyFieldName:
      labkeyFieldName=xlsFieldName
   db=connectDB('onko-nix')
   ds=db.selectRows(project,'lists',lookup,[])
#only record values from labkey (should be unique anyhow)
   setVals=set([r[labkeyFieldName] for r in ds['rows']])
   off=len(list(setVals))
   missing=sorted(list(vals-setVals))
#print('Missing {}'.format(missing))
   n=len(missing)
   entries=[{'Key':'{}'.format(i+1+off),columnName:missing[i]} \
      for i in range(n)]
   print(entries)
   if dryRun:
      return
   db.modifyRows('insert',project,'lists',lookup,entries)

def getAliasList(x):
#split aliases by comma, taking into account quotation marks, 
#where commas are ignored
   if not x:
      return []
#sophisticated spliting that ignores commas in double (and single!) quotes
   ar=re.split(''',(?=(?:[^'"]|'[^']*'|"[^"]*")*$)''', x)
#remove the quotes afterwards
   ar=[s.replace('"','') for s in ar]
   ar=[s.strip() for s in ar]
   return ar

def printErr(resp):
#print error from server response
   try:
      print(resp['exception'])
   except KeyError:
      pass
     
def findAlias(v,alias):
#find matchng alias for field v from XLS
   try:
      return alias[v.strip()]
   except KeyError:
      pass
#remove .N qualifiers, attach to the end by pandas.read_excel
   v=re.sub(r'(.*)\.[0-9]*$',r'\1',v.strip())
   try:
      return alias[v]
   except KeyError:
      pass
    
   return None

def getSequenceNum(pMap,id):
    #updates pMap and return seqnum for this entry
   try:
      n=len(pMap[id])
   except KeyError:
      n=0
   seqNum=n+1
   if n==0:
      pMap[id]=[seqNum]
   else:
      pMap[id].append(seqNum)
   return seqNum


def entryInList(r,entries):
#is entry in list
   candidates=[x for x in entries if entriesMatch(x,r)]
   return len(candidates)>0

def entriesMatch(x,r):
#do a pair of entries match? Should we update data from one over the other?
   matchFields=['ParticipantId','SequenceNum']
   for f in matchFields:
      if x[f]!=r[f]:
         return False
   return True

def validate(qv):
#rough checks on value read from XLSX

#NAN
   try:
      if math.isnan(qv):
         return None
   except TypeError:
      pass

#NAD
   if pandas.isnull(qv):
      return None
   return qv

#defaults for modify, getId and convertLookup
def modify(qv,columnName):
   return qv

def getId(df,r):
   try:
      rawId=str(df.at[r,'ID'])
   except KeyError:
      msg='Getting id from field ID field. ' 
      msg+=' Overload getId function with getId=getIdFnc in importData'
      print(msg)
      raise KeyError
   return rawId.replace(' ','')

def convertLookup(xlsColumnName):
   return True

def importData(pars,filename,getId=getId,modify=modify,\
   convertLookup=convertLookup,dryRun=True,debug=True):
#master routine that imports data based on pars, 
#applies user supplied functions modify, convertLookup and get Id and 
#updates relevant database

#some useful fields from pars (d is for default value)
# - skiprows removes irelevant rows (number, d: 0)
# - usecols is EXCEL like notations for cols taken (identifer, d: None for all)
# - sheet_name selects sheet for import (number, d:0 for first sheet)
# - seqNumOffset specify visit/sequenceNum offset (number, d:0 will result in 1)
# - project - labkey project
# - schema - labkey schema (list/study, d: study)
# - query - labkey query

   skiprows=pars.get('skiprows',0)
   usecols=pars.get('usecols',None)
   sheet_name=pars.get('sheet_name',0)
#set this is as sequenceNum for entries, or initial seqNum if more than a single entry is in the dataset
   seqNumOffset=pars.get('seqNumOffset',0)
   fields=getFields(pars)
   lookupVars=getVariables(fields,fieldType='LOOKUP')
   dateVars=getVariables(fields,fieldType='DATE')
   doubleVars=getVariables(fields,fieldType='DOUBLE')
   usecols=pars.get('usecols',None)
#convert dates to list
   dateVars=list(dateVars.keys())
   print(f'dateVars: {dateVars}')
   lookupMap={f:getLookupMap(pars,fields,f) for f in lookupVars}
   if debug:
      print(lookupMap)
   alias=getAlias(fields)
   print(f'aliases: {alias}')

   df=pandas.read_excel(filename,sheet_name=sheet_name,skiprows=skiprows,\
      usecols=usecols)
   vars=df.columns
   print(vars)

   pMap={}
   print('Index: {}'.format(len(df.index)))
   idx=df.index #for all
   if debug:
      idx=df.index[0:10] #for debug

   entries=[]
   for r in idx:
      id=getId(df,r)    
      entry={}
      entry['ParticipantId']=id

      for v in vars:
         qv=validate(df.at[r,v])
         qv=modify(qv,v)
         f=findAlias(v,alias)
         if not f:
            continue
         if f in lookupMap:
            if convertLookup(v):
               qv=parseLookup(lookupMap[f],qv)
            else:
               qv=asKey(qv)
                
            
         if f in dateVars:
            qv=parseDate(qv)
                
         if f in doubleVars:
            qv=parseDouble(qv)
         try:
            numpyType=qv.dtype
            qv=qv.item()
         except AttributeError:
            pass
            
         entry[f]=qv
            
#print('{}:{}/{}'.format(f,qv,type(qv)))
      seqNum=getSequenceNum(pMap,id)
      entry['SequenceNum']=seqNum+seqNumOffset
      entries.append(entry)
            
#for p in pMap:
#    print('{}: {}'.format(p,len(pMap[p])))
   print(entries)
#delete previous incarnations
   loadSafely(pars,entries,dryRun=dryRun)

def loadSafely(pars,entries,keyColumn=None,dryRun=True):
#allow additional keys in keyColumn

   db=getDB(pars)
   project=pars.get('project','DCIS/Study')
   schema=pars.get('schema','demographics')
   query=pars.get('query','demographics')
   
   updateRows=[]
   insertRows=[]


   selVal=['ParticipantId','SequenceNum']
   if keyColumn:
      selVal.append(keyColumn)

   for entry in entries:

      qFilter=[{'variable':v,'value':'{}'.format(entry[v]),'oper':'eq'} for v in selVal]
      ds=db.selectRows(project,schema,query,qFilter)

      if len(ds['rows'])>0:
         r=ds['rows'][0]
         r.update(entry)
         updateRows.append(r)
      else:
         insertRows.append(entry)
        
   n=len(updateRows)
   print(f'Updating {n} entries')
   if n and not dryRun:
      printErr(db.modifyRows('update',project,schema,query,updateRows))
   n=len(insertRows)
   print(f'Inserting {n} entries')
   if n and not dryRun:
      printErr(db.modifyRows('insert',project,schema,query,insertRows))