Source code for pycif.plugins.datastreams.fields.netcdf_cams.read

import datetime
import calendar
import os
import pandas as pd
import numpy as np
import xarray as xr
from netCDF4 import Dataset

from .....utils.netcdf import readnc
from .....utils.check.errclass import CifError
from .....utils.hdf5 import _hdf5_lock




def read(
        self,
        name,
        varnames,
        dates,
        files,
        interpol_flx=False,
        comp_type=None,
        tracer=None,
        ddi=None,
        **kwargs
):
    """Get BCs from raw CAMS files and load them into a pyCIF variable.

    For each requested date, computes the time index within the monthly
    file from the file's number of time steps, reads the variable, and
    flips the latitude axis to increasing order (and the vertical axis
    if ``tracer.flip_level`` is set) as needed.

    Args:
        self: the BC Plugin
        name: the name of the component
        varnames: variable name to extract if different from ``name``
        dates: list of ``[start, end]`` date pairs to extract
        files: list of file paths matching ``dates``
        interpol_flx: unused, accepted for interface compatibility
        comp_type: unused, accepted for interface compatibility
        tracer: the fields Plugin; ``tracer.flip_level`` controls whether
            the vertical axis is flipped
        ddi: must not be ``None``; only used to validate the call, not
            otherwise referenced

    Returns:
        xarray.DataArray with dims ``(time, lev, lat, lon)``.

    Raises:
        CifError: if ``ddi`` is ``None``.

    """

    if ddi is None:
        raise CifError("CAMS netCDF read function was called "
                        "without specifying ddi")

    # Variable name to extract
    var2extract = name
    if varnames != "":
        var2extract = varnames

    # Reading fields for periods within the simulation window
    xout = []
    opened_file = ""
    for dd, dd_file in zip(dates, files):
        with _hdf5_lock:
            # Avoid opening the file for all dates
            if dd_file != opened_file:
                ds = xr.open_dataset(dd_file)
                opened_file = dd_file

            ntimes = ds.dims["time"]

            freq = pd.DatetimeIndex([dd[0]]).days_in_month[0] * 24 / ntimes
            date_end = ds["time"].to_pandas().dt.to_pydatetime()[-1]
            date_index = \
                ntimes - 1 - int((date_end - dd[0]) /
                                 datetime.timedelta(hours=freq))

            # bottom of the atmosphere = at the beginning of the table
            lat = ds['latitude']
            conc = ds[var2extract].values[date_index]

        if lat[1] < lat[0] and conc.ndim == 4:
            conc = conc[:, :, ::-1, :]
        elif lat[1] < lat[0] and conc.ndim == 3:
            conc = conc[:, ::-1, :]

        # Swap levels if required to
        if getattr(tracer, "flip_level", False):
            if conc.ndim == 4:
                conc = conc[:, ::-1]
            elif conc.ndim == 3:
                conc = conc[::-1]

        # Append to datastore
        xout.append(conc)

    xmod = xr.DataArray(
        np.array(xout),
        coords={"time": np.array(dates)[:, 0]},
        dims=("time", "lev", "lat", "lon"),
    )

    return xmod