Numpy 数组到列表非常慢

Question

我正在阅读 NetCDF 文件并将其转换为 GeoJSON。 功能已完成，但完成执行大约需要 40 秒。 任何优化此代码的方法。 这可能是由于三个不同位置上的 numpy 阵列的三个 for 循环和 tolist 转换。

import time
import geobuf
from netCDF4 import Dataset
from collections import OrderedDict

def main():
    '''
    The main method that holds the logic for conversion
    '''
    dataset = Dataset('outfile.nc', 'r')

    data_dict = OrderedDict({
        'type': 'FeatureCollection',
        'features': []
    })

    for traj in range(dataset.variables['trajectory'].size):
        for t in range(dataset.variables['time'].size):
            lat = dataset.variables['lat'][traj, t].tolist()
            lon = dataset.variables['lon'][traj, t].tolist()

            if lat is None or lon is None:
                continue
            feature_dict = OrderedDict({
                'type': 'Feature',
                'geometry': {
                    'type': 'Point',
                    'coordinates': [lat, lon]
                },
                'properties': {}
            })
            for var in dataset.variables.keys():
                if var not in ['lon', 'lat', 'time', 'trajectory']:
                    dvar = dataset.variables[var]
                    feature_dict['properties'][var] = dvar[traj, t].tolist()
            data_dict['features'].append(feature_dict)
    
    # This encoding is not slow
    pbf = geobuf.encode(data_dict)
    print(pbf)

    dataset.close()


if __name__ == '__main__':
    '''
    Entrypoint to this script
    '''
    start_time = time.time()
    main()
    print("--- %s seconds ---" % (time.time() - start_time))