How to use the pygrib.index function in pygrib

try:
            with pygrib.open(self.filename) as grib_file:
                first_msg = grib_file.message(1)
                last_msg = grib_file.message(grib_file.messages)
                start_time = self._convert_datetime(
                    first_msg, 'validityDate', 'validityTime')
                end_time = self._convert_datetime(
                    last_msg, 'validityDate', 'validityTime')
                self._start_time = start_time
                self._end_time = end_time
                if 'keys' not in filetype_info:
                    self._analyze_messages(grib_file)
                    self._idx = None
                else:
                    self._create_dataset_ids(filetype_info['keys'])
                    self._idx = pygrib.index(self.filename,
                                             *filetype_info['keys'].keys())
        except (RuntimeError, KeyError):
            raise IOError("Unknown GRIB file format: {}".format(self.filename))

ts = ts.replace(minute=0)
        fn = None
        for offset in range(0, 24, 4):
            ts2 = ts - datetime.timedelta(hours=offset)
            testfn = ts2.strftime(
                (
                    "/mesonet/ARCHIVE/data/%Y/%m/%d/model/ffg/"
                    "5kmffg_%Y%m%d%H.grib2"
                )
            )
            if os.path.isfile(testfn):
                fn = testfn
                break
        if fn is None:
            raise NoDataFound("No valid grib data found!")
        grbs = pygrib.index(fn, "stepRange")
        grb = grbs.select(stepRange="0-%s" % (hour,))[0]
        lats, lons = grb.latlons()
        data = (
            masked_array(grb.values, data_units=units("mm"))
            .to(units("inch"))
            .m
        )
        plot.pcolormesh(lons, lats, data, bins, cmap=cmap)
        if ilabel:
            plot.drawcounties()
        df = pd.DataFrame()
    return plot.fig, df

#assign dimensions
        times[:]      = self.ndate
        levels[:]     = self.levs
        latitudes[:]  = self.lats
        longitudes[:] = self.lons

        #make actual variables
        variables = []
        for var in self.nams:
            # isolate from [u'Geopotential']
            variables.append(ncd_root.createVariable(var,'f4',
                                                        ('time','level',
                                                         'lat','lon',)))
        #read file, get levels and times
        grbindx = pg.index(self.file_grib,'name','level','dataDate',
                                          'dataTime','step')

        levs = np.array(self.levs)
        for l in self.levs:
            for d in self.jday:
                nd = nc.date2num(d, units = "seconds since 1970-1-1",
                                 calendar = 'standard')
                var_n = 0
                for var in self.nams:
                    #distinguish forecast data to deal with accumulated fields
                    if var in self.accumulated:
                        vpre = 0 #initial for subtraction
                        for s in self.step:
                            sel = grbindx.select(name = var, level = l,
                                             dataDate = int(d.strftime("%Y%m%d")),
                                             dataTime = d.hour * 100, step = s)

#assign dimensions
        times[:]      = self.ndate
        levels[:]     = self.levs
        latitudes[:]  = self.lats
        longitudes[:] = self.lons
        
        #make actual variables
        variables = []
        for var in self.nams:
            # isolate from [u'Geopotential']
            variables.append(ncd_root.createVariable(var,'f4',
                                                        ('time','level',
                                                         'lat','lon',)))
        #read file, get levels and times
        grbindx = pg.index(self.file_grib,'name','level','dataDate',
                                          'dataTime','step')

        levs = np.array(self.levs) 
        for l in self.levs:
            for d in self.jday:
                nd = nc.date2num(d, units = "seconds since 1970-1-1",
                                 calendar = 'standard')
                var_n = 0                 
                for var in self.nams:
                    #distinguish forecast data to deal with accumulated fields
                    if var in self.accumulated:
                        vpre = 0 #initial for subtraction
                        for s in self.step:
                            sel = grbindx.select(name = var, level = l, 
                                             dataDate = int(d.strftime("%Y%m%d")),
                                             dataTime = d.hour * 100, step = s)

#assign dimensions
        times[:]      = self.ndate
        levels[:]     = self.levs
        latitudes[:]  = self.lats
        longitudes[:] = self.lons
        
        #make actual variables
        variables = []
        for var in self.nams:
            # isolate from [u'Geopotential']
            variables.append(ncd_root.createVariable(var,'f4',
                                                        ('time','level',
                                                         'lat','lon',)))
        #read file, get levels and times
        grbindx = pg.index(self.file_grib,'name','level','dataDate',
                                          'dataTime','step')

        levs = np.array(self.levs) 
        for l in self.levs:
            for d in self.jday:
                nd = nc.date2num(d, units = "seconds since 1970-1-1",
                                 calendar = 'standard')
                var_n = 0                 
                for var in self.nams:
                    #distinguish forecast data to deal with accumulated fields
                    if var in self.accumulated:
                        vpre = 0 #initial for subtraction
                        for s in self.step:
                            sel = grbindx.select(name = var, level = l, 
                                             dataDate = int(d.strftime("%Y%m%d")),
                                             dataTime = d.hour * 100, step = s)