import numpy as np  # this module handles arrays, but here we need it for its NaN value
import pandas as pd # this module contains a lot of tools for handling tabular data
import re

# define paths to the source files and eventual output file
pathBottle='/ocean/eolson/MEOPAR/obs/NemcekHPLC/All 2015 SoG_bottle.xlsx'
pathPhyto='/ocean/eolson/MEOPAR/obs/NemcekHPLC/2015-2019 SoG abs values New ALLO.xlsx'
pathOut='/ocean/eolson/MEOPAR/obs/NemcekHPLC/bottlePhytoMerged2015_NewALLO.csv'

def fmtCruise(istr):
    if re.fullmatch('[0-9]{4}-[0-9]{2}',istr): 
        sp=re.split('-',istr)
        rstr=sp[0]+'-0'+sp[1]
    elif re.fullmatch('[0-9]{4}-[0-9]{3}',istr):
        rstr=istr
    else:
        raise ValueError('Input had unexpected format:',istr)
    return rstr

def subval(idf,colList):
    # first value in colList should be the column you are going to keep
    # follow with other columns that will be used to fill in when that column is NaN
    # in order of precedence
    if len(colList)==2:
        idf[colList[0]]=[r[colList[0]] if not pd.isna(r[colList[0]]) \
                         else  r[colList[1]] for i,r in idf.iterrows()]
    elif len(colList)==3:
        idf[colList[0]]=[r[colList[0]] if not pd.isna(r[colList[0]]) \
                         else  r[colList[1]] if not pd.isna(r[colList[1]]) \
                         else r[colList[2]] for i,r in idf.iterrows()]
    else:
        raise NotImplementedError('Add to code to handle this case')
    idf.drop(columns=list(colList[1:]),inplace=True)
    return idf

dfbot=subval(dfbot,('Oxygen:Dissolved','Oxygen:Dissolved.1'))
dfbot=subval(dfbot,('Temperature:Primary','Temperature:Secondary'))
dfbot=subval(dfbot,('Salinity:T0:C0','Salinity:T1:C1'))
dfbot=subval(dfbot,('Oxygen:Dissolved:SBE','Oxygen:Dissolved:SBE.1'))
dfbot.rename(columns={'LOC:LATITUDE':'Lat',
                      'LOC:LONGITUDE':'Lon',
                      'Temperature:Primary':'Temperature',
                      'Oxygen:Dissolved:SBE':'Oxygen:Dissolved:CTD',
                      'Salinity:T0:C0':'Salinity',},inplace=True)

# define a function that will be applied to the values in the index column;
# this makes it easier to drop non-data rows later
def convertIndex(val):
    try:
        x =int(val)
    except ValueError:
        x=-999
    return x

# now, drop any rows from the table that have NaN values in either of the columns
#  'Index' or 'TchlA (ug/L)'
# This is why we applied a function to the Index column to make sure all 
#  non-numeric Index values would have a consistent NaN entry, making them easy to identify
#  and remove
dfPhyto.dropna(subset=['TchlA'],how='any',inplace=True)

dfPhyto.drop(dfPhyto[dfPhyto['Sample#']<0].index,inplace=True)

# pandas creates its own index, and after dropping rows I like to reset it -
# this is just for convenience
dfPhyto.reset_index(drop=True,inplace=True)

# due to repeated column names in the original spreadsheet, '.1' was appended to the names
# of the phytoplankton columns; 
# these lines correct the column names, removing the '.1':
renameDict=dict()
for colName in dfPhyto.keys():
    if colName.endswith('.1'):
        renameDict[colName]=colName.split('.1')[0]
dfPhyto.rename(columns=renameDict,inplace=True)

# This is the important step- join the two tables ('left' and 'right'), 
#  matching the cruise IDs and sample numbers
#   how='outer'  -> all rows from both the left and the right tables will be included, 
#                   even if they cannot be matched; this makes it easy to check for 
#                   unmatched data later
#   left_on  specifies the name of the column to match in the left table (dfbot) 
#   right_on specifies the name of the column to match in the right table (dfPhyto)
dfout = pd.merge(dfbot, dfPhyto,  how='outer', 
                 left_on=['Cruise','Sample_Number'], right_on = ['Cruise','Sample#'])

# Identify cases where phytoplankton data were matched to multiple samples in bottle data:
dftest=pd.merge(dfbot, dfPhyto,how='right', left_on=['Cruise','Sample_Number'],right_on = ['Cruise','Sample#'])

temp=dftest.groupby(['Cruise','Sample#']).agg({'Cruise':['count']})
temp.columns = ['icount']

# drop repetetive/unecessary columns: 'Index','Bin #', 'Subgroup', 'Month',
dfout.drop(labels=['Date', 'Station', 'Sample#', 'rep',
                   'depth',],axis=1,inplace=True)

# truncate phyto group values to 3 decimal places:
for col in ('Cyanobacteria', 'Prasinophytes', 'Cryptophytes', 'Diatoms-1',
       'Diatoms-2', 'Dinoflagellates-1', 'Haptophytes', 'Dictyo', 'Raphido',
       'TchlA'):
    dfout[col]=[np.round(ii,decimals=3) for ii in dfout[col]] # use list comprehension to set values for entire column

# now write the output table to a .csv file:
dfout.to_csv(pathOut, index=False)