5. Views¶

Views in bifrost are conceptually similar to the numpy.view(), in that it changes the interpretation of the data in a ring, without copying or changing the underlying data. These are useful for preparing your data for later processing. The views provided are:

• views.split_axis: reshape an array to split an axis into two: [freq, time] becomes [freq, time, window] to chunk up the data in preparation for an FFT.

• views.merge_axis: flatten two axes of an array: [channel, freq] becomes [freq].

• views.rename_axis: change the label on an axis: [one_over_wavelength] becomes [freq]

• views.expand_dims: add an extra dimension: shape[-1, 10, 100] becomes [-1, 10, 1, 100]

• views.custom: apply a custom view

As an example, this reads two “time steps” from a GUPPI RAW file, and squishes it into one longer time step:

# Read from guppi raw file
b_guppi   = blocks.read_guppi_raw(filelist, core=1, buffer_nframe=4)
b_gup2    = views.rename_axis(b_guppi, 'freq', 'channel')

# Buffer up two time steps & reshape to allow longer FFT
b_gup2    = views.split_axis(b_gup2, axis='time', n=n_chunks, label='time_chunk')
b_gup2    = blocks.transpose(b_gup2, axes=['time', 'channel', 'time_chunk', 'fine_time', 'pol'], buffer_nframe=1)
b_gup2    = views.merge_axes(b_gup2, 'time_chunk', 'fine_time', label='fine_time')