#!/usr/bin/env python
#2022-03-03: copy from wyget_era5_mn2t_subregion.py
if __name__ == '__main__':
    import sys
    from misc.timer import Timer
    wyt = Timer('start ' + ' '.join(sys.argv)) 
import xarray as xr
import os, os.path, sys, time
import  salem
import pandas as pd
import numpy as np
import geoxarray
from misc.shell import run_shell

cntry_name = 'Brazil'
cntry_sname = cntry_name.replace(' ', '')
data_name = 'mn2t'
dname_long = '2m_temperature_min'
xname = 'longitude'
yname = 'latitude'
shapefile = 'shapefile/gadm40_BRA_1.shp'
shdf = salem.read_shapefile(shapefile)
#name_cols = ['ADM1_EN', 'ADM1_PCODE']
name_col = 'NAME_1'
t = pd.date_range('1999-01', '2020-12', freq='MS')
years, months = t.year.values, t.month.values
print(data_name)
ofiles = []
for year,month in zip(t.year, t.month):
    print(f'{year:04d}-{month:02d}')
    ifile = f'/tigress/wenchang/data/era5/analysis/{dname_long}/daily/era5.{dname_long}.daily.{year:04d}-{month:02d}.nc'
    ofile = f'tmp/era5.{data_name}.daily.{cntry_sname}.subregion.{year:04d}-{month:02d}.nc'
    ofiles.append(ofile)
    if os.path.exists(ofile):
        print('[exists]:', ofile)
        continue
    print(ifile, 'loading...')
    ds = salem.open_xr_dataset(ifile).load()
    if ds.longitude.max() > 180:
        # transform lon [0,360) -> [-180,180)
        ds = ds.roll(longitude=ds.longitude.size//2, roll_coords=True)
        lon_new = ds.longitude.where(ds.longitude<180, other=ds.longitude-360).values
        ds = ds.assign_coords(longitude=lon_new)
    
    das = []
    #region_IDs = []
    region_names = []
    N = len(shdf)
    for i in range(N):
        shape_sub = shdf.iloc[i:i+1, :]
        r = shdf.iloc[i, :]
        #lon_slice = slice(r.min_x, r.max_x)
        #lat_slice = slice(r.min_y, r.max_y)
        #g = r.geometry
        region_name = '_'.join( r[name_col].split() ) #name_col = 'NAME_1'
     
        try:
            #da = ds[data_name].salem.roi(geometry=g).geo.fldmean()
            da = ds[data_name].salem.subset(shape=shape_sub, margin=1) \
                .salem.roi(shape=shape_sub)\
                .geo.fldmean()
            print('[OK]:', i+1,'of', N, region_name)
        except ValueError: #sub shape area is too small to be selected; use value interpolated on centroid instead
            centroid0 = shape_sub.centroid.iloc[0]
            lon0 = centroid0.x
            lat0 = centroid0.y
            da = ds[data_name].interp(latitude=lat0, longitude=lon0).drop(['longitude', 'latitude'])
            #print('\t[Failed]', i+1,'of', N, region_name)
            print('\t[single point used]', i+1,'of', N, region_name)
        das.append(da)
        region_names.append(region_name)
    
    i_record = range(len(das))
    da = xr.concat(das, pd.Index(region_names, name='region_name')).transpose()
    da.attrs = ds[data_name].attrs
    ds_new = da.to_dataset(name=data_name)
    ds_new.attrs = ds.attrs
    ds_new.attrs['region_name'] = '; '.join(region_names)
    
    ds_new.to_netcdf(ofile, unlimited_dims='time', encoding={data_name: {'_FillValue': None}})
    print('[saved]:', ofile)

#ncrcat
ofile = ofiles[0].replace('tmp/', '').replace(f'{years[0]}-{months[0]:02d}', f'{years[0]}-{years[-1]}')
cmd = f'ncrcat {" ".join(ofiles)} {ofile}'
run_shell(cmd)

if __name__ == '__main__':
    wyt.check('**done**')