Datasets:

ArXiv:
License:
Wave-Gauss / assemble_data.py
maximilianherde's picture
Update assemble_data.py
d0eadc4 verified
import argparse
import os
from netCDF4 import Dataset
def assemble_data(input_dir, output_file):
nc_files = [f for f in os.listdir(input_dir) if f.endswith(".nc") and not f.startswith("c")]
c_files = [f for f in os.listdir(input_dir) if f.endswith(".nc") and f.startswith("c")]
nc_files.sort()
c_files.sort()
samples = [0]
with Dataset(os.path.join(input_dir, nc_files[0]), "r") as first_nc:
samples.append(first_nc.dimensions["sample"].size)
num_times = first_nc.dimensions["time"].size
try:
num_channels = first_nc.dimensions["channel"].size
except:
num_channels = None
x_size = first_nc.dimensions["x"].size
y_size = first_nc.dimensions["y"].size
dtype = first_nc.variables[nc_files[0].split("_")[0]].dtype
for nc_file in nc_files[1:]:
with Dataset(os.path.join(input_dir, nc_file), "r") as nc:
samples.append(nc.dimensions["sample"].size)
num_samples = sum(samples)
for i in range(1, len(samples)):
samples[i] += samples[i - 1]
with Dataset(output_file, "w") as out_nc:
out_nc.createDimension("sample", num_samples)
out_nc.createDimension("time", num_times)
if num_channels is not None:
out_nc.createDimension("channel", num_channels)
out_nc.createDimension("x", x_size)
out_nc.createDimension("y", y_size)
if num_channels is not None:
out_nc.createVariable(
nc_files[0].split("_")[0], dtype, ("sample", "time", "channel", "x", "y"), chunksizes=(1, 1, num_channels, x_size, y_size)
)
else:
out_nc.createVariable(
nc_files[0].split("_")[0], dtype, ("sample", "time", "x", "y"), chunksizes=(1, 1, x_size, y_size)
)
out_nc.createVariable(
"c", dtype, ("sample", "x", "y"), chunksizes=(1, x_size, y_size)
)
for i, nc_file in enumerate(nc_files):
with Dataset(os.path.join(input_dir, nc_file), "r") as nc:
print(f"Processing {os.path.join(input_dir, nc_file)}")
variable = nc.variables[nc_file.split("_")[0]]
out_nc[nc_file.split("_")[0]][samples[i] : samples[i + 1]] = variable[:]
samples = [0]
for nc_file in c_files:
with Dataset(os.path.join(input_dir, nc_file), "r") as nc:
samples.append(nc.dimensions["sample"].size)
for i in range(1, len(samples)):
samples[i] += samples[i - 1]
for i, c_file in enumerate(c_files):
with Dataset(os.path.join(input_dir, c_file), "r") as nc:
print(f"Processing {os.path.join(input_dir, c_file)}")
variable = nc.variables[c_file.split("_")[0]]
out_nc[c_file.split("_")[0]][samples[i] : samples[i + 1]] = variable[:]
print(f"Saved data to {output_file}")
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument("--input_dir", type=str, required=True)
parser.add_argument("--output_file", type=str, required=True)
args = parser.parse_args()
assemble_data(args.input_dir, args.output_file)