Feature/nicke adpsfc prepbufr

dump/mapping/bufr_adpsfc_prepbufr.py

@@ -0,0 +1,255 @@
+#!/usr/bin/env python3
+import os
+import sys
+import bufr
+import argparse
+import copy
+import numpy as np
+import numpy.ma as ma
+import math
+import calendar
+import time
+from datetime import datetime
+from pyioda.ioda.Engines.Bufr import Encoder as iodaEncoder
+from bufr.encoders.netcdf import Encoder as netcdfEncoder
+from wxflow import Logger
+
+# Initialize Logger
+# Get log level from the environment variable, default to 'INFO it not set
+log_level = os.getenv('LOG_LEVEL', 'INFO')
+logger = Logger('bufr_adpsfc_prepbufr.py', level=log_level, colored_log=False)
+
+def logging(comm, level, message):
+    """
+    Logs a message to the console or log file, based on the specified logging level.
+
+    This function ensures that logging is only performed by the root process (`rank 0`)
+    in a distributed computing environment. The function maps the logging level to
+    appropriate logger methods and defaults to the 'INFO' level if an invalid level is provided.
+
+    Parameters:
+        comm: object
+            The communicator object, typically from a distributed computing framework
+            (e.g., MPI). It must have a `rank()` method to determine the process rank.
+        level: str
+            The logging level as a string. Supported levels are:
+                - 'DEBUG'
+                - 'INFO'
+                - 'WARNING'
+                - 'ERROR'
+                - 'CRITICAL'
+            If an invalid level is provided, a warning will be logged, and the level
+            will default to 'INFO'.
+        message: str
+            The message to be logged.
+
+    Behavior:
+        - Logs messages only on the root process (`comm.rank() == 0`).
+        - Maps the provided logging level to a method of the logger object.
+        - Defaults to 'INFO' and logs a warning if an invalid logging level is given.
+        - Supports standard logging levels for granular control over log verbosity.
+
+    Example:
+        >>> logging(comm, 'DEBUG', 'This is a debug message.')
+        >>> logging(comm, 'ERROR', 'An error occurred!')
+
+    Notes:
+        - Ensure that a global `logger` object is configured before using this function.
+        - The `comm` object should conform to MPI-like conventions (e.g., `rank()` method).
+    """
+
+    if comm.rank() == 0:
+        # Define a dictionary to map levels to logger methods
+        log_methods = {
+            'DEBUG': logger.debug,
+            'INFO': logger.info,
+            'WARNING': logger.warning,
+            'ERROR': logger.error,
+            'CRITICAL': logger.critical,
+        }
+
+        # Get the appropriate logging method, default to 'INFO'
+        log_method = log_methods.get(level.upper(), logger.info)
+
+        if log_method == logger.info and level.upper() not in log_methods:
+            # Log a warning if the level is invalid
+            logger.warning(f'log level = {level}: not a valid level --> set to INFO')
+
+        # Call the logging method
+        log_method(message)
+
+
+def _compute_datetime(cycleTimeSinceEpoch, dhr):
+    """
+    Compute dateTime using the cycleTimeSinceEpoch and Cycle Time
+        minus Cycle Time
+
+    Parameters:
+        cycleTimeSinceEpoch: Time of cycle in Epoch Time
+        dhr: Observation Time Minus Cycle Time
+
+    Returns:
+        Masked array of dateTime values
+    """
+
+    int64_fill_value = np.int64(0)
+
+    dateTime = np.zeros(dhr.shape, dtype=np.int64)
+    for i in range(len(dateTime)):
+        if ma.is_masked(dhr[i]):
+            continue
+        else:
+            dateTime[i] = np.int64(dhr[i]*3600) + cycleTimeSinceEpoch
+
+    dateTime = ma.array(dateTime)
+    dateTime = ma.masked_values(dateTime, int64_fill_value)
+
+    return dateTime
+
+
+def _make_description(mapping_path, cycle_time, update=False):
+    description = bufr.encoders.Description(mapping_path)
+
+    ReferenceTime = np.int64(calendar.timegm(time.strptime(str(int(cycle_time)), '%Y%m%d%H')))
+
+    if update:
+        # Define the variables to be added in a list of dictionaries
+        variables = [
+            {
+                'name': 'MetaData/sequenceNumber',
+                'source': 'variables/sequenceNumber',
+                'units': '1',
+                'longName': 'Sequence Number (Obs Subtype)',
+            }
+        ]
+
+        # Loop through each variable and add it to the description
+        for var in variables:
+            description.add_variable(
+                name=var['name'],
+                source=var['source'],
+                units=var['units'],
+                longName=var['longName']
+            )
+
+        #description.add_global(name='datetimeReference', value=str(ReferenceTime))
+
+    return description
+
+
+def _make_obs(comm, input_path, mapping_path, cycle_time):
+    """
+    Create the ioda adpsfc prepbufr observations:
+    - reads values
+    - adds sequenceNum
+
+    Parameters
+    ----------
+    comm: object
+            The communicator object (e.g., MPI)
+    input_path: str
+            The input bufr file
+    mapping_path: str
+            The input bufr2ioda mapping file
+    cycle_time: str
+            The cycle in YYYYMMDDHH format
+    """
+
+    # Get container from mapping file first
+    logging(comm, 'INFO', 'Get container from bufr')
+    container = bufr.Parser(input_path, mapping_path).parse(comm)
+
+    logging(comm, 'DEBUG', f'container list (original): {container.list()}')
+    logging(comm, 'DEBUG', f'Change longitude range from [0,360] to [-180,180]')
+    lon = container.get('variables/longitude')
+    lon_paths = container.get_paths('variables/longitude')
+    lon[lon>180] -= 360
+    lon = ma.round(lon, decimals=2)
+    logging(comm, 'DEBUG', f'longitude max and min are {lon.max()}, {lon.min()}')
+
+    logging(comm, 'DEBUG', f'Do DateTime calculation')
+    otmct = container.get('variables/obsTimeMinusCycleTime')
+    otmct_paths = container.get_paths('variables/obsTimeMinusCycleTime')
+    otmct2 = np.array(otmct)
+    cycleTimeSinceEpoch = np.int64(calendar.timegm(time.strptime(str(int(cycle_time)), '%Y%m%d%H')))
+    dateTime = _compute_datetime(cycleTimeSinceEpoch, otmct2)
+    min_dateTime_ge_zero = min(x for x in dateTime if x > -1)
+    logging(comm, 'DEBUG', f'dateTime min/max = {min_dateTime_ge_zero} {dateTime.max()}')
+
+    logging(comm, 'DEBUG', f'Make an array of 0s for MetaData/sequenceNumber')
+    sequenceNum = np.zeros(lon.shape, dtype=np.int32)
+    logging(comm, 'DEBUG', f' sequenceNummin/max =  {sequenceNum.min()} {sequenceNum.max()}')
+
+    logging(comm, 'DEBUG', f'Update variables in container')
+    container.replace('variables/longitude', lon)
+    container.replace('variables/timestamp', dateTime)
+
+    logging(comm, 'DEBUG', f'Add variables to container')
+    container.add('variables/sequenceNumber', sequenceNum, lon_paths)
+
+    # Check
+    logging(comm, 'DEBUG', f'container list (updated): {container.list()}')
+
+    return container
+
+
+def create_obs_group(input_path, mapping_path, cycle_time, env):
+
+    comm = bufr.mpi.Comm(env["comm_name"])
+
+    logging(comm, 'INFO', f'Make description and make obs')
+
+    container = _make_obs(comm, input_path, mapping_path, cycle_time)
+    description = _make_description(mapping_path, cycle_time, update=True)
+
+    # Gather data from all tasks into all tasks. Each task will have the complete record
+    logging(comm, 'INFO', f'Gather data from all tasks into all tasks')
+    container.all_gather(comm)
+
+    logging(comm, 'INFO', f'Encode the data')
+    data = next(iter(iodaEncoder(description).encode(container).values()))
+
+    logging(comm, 'INFO', f'Return the encoded data.')
+
+    return data
+
+
+def create_obs_file(input_path, mapping_path, output_path, cycle_time):
+
+    comm = bufr.mpi.Comm("world")
+    container = _make_obs(comm, input_path, mapping_path, cycle_time)
+    container.gather(comm)
+
+    description = _make_description(mapping_path, cycle_time, update=True)
+
+    # Encode the data
+    if comm.rank() == 0:
+        netcdfEncoder(description).encode(container, output_path)
+
+    logging(comm, 'INFO', f'Return the encoded data')
+
+
+if __name__ == '__main__':
+    start_time = time.time()
+
+    bufr.mpi.App(sys.argv)
+    comm = bufr.mpi.Comm("world")
+
+    # Required input arguments as positional arguments
+    parser = argparse.ArgumentParser(description="Convert BUFR to NetCDF using a mapping file.")
+    parser.add_argument('input', type=str, help='Input BUFR file')
+    parser.add_argument('mapping', type=str, help='BUFR2IODA Mapping File')
+    parser.add_argument('output', type=str, help='Output NetCDF file')
+    parser.add_argument('cycle_time', type=str, help='cycle time in YYYYMMDDHH format')
+
+    args = parser.parse_args()
+    infile = args.input
+    mapping = args.mapping
+    output = args.output
+    cycle_time = args.cycle_time
+
+    create_obs_file(infile, mapping, output, cycle_time)
+
+    end_time = time.time()
+    running_time = end_time - start_time
+    logging(comm, 'INFO', f'Total running time: {running_time}')