Indexing bug

[j6626]

Lines 365 and 372 read response_freqs = (np.squeeze(data.freqs[nchan_kernel/2:-nchan_kernel/2+1]) + data_delfreq/2.0)/1.e6

However, there is no guarantee that nchan_kernel is even, so the code will throw an error when nchan_kernel is odd.

[j6626]

Lines 635 and 641 (the multi-EB case) have the opposite problem. They read

response_freqs = (np.squeeze(data.freqs[int(nchan_kernel/2+1):int(-nchan_kernel/2+1)]) + data_delfreq/2.0)/1.e6

which will cause an issue when nchan_kernel is even because the size of response_freqs will not match the size of curr_xc (they will be off by 1)

[YuRuChou]

Lines 365 and 372 read response_freqs = (np.squeeze(data.freqs[nchan_kernel/2:-nchan_kernel/2+1]) + data_delfreq/2.0)/1.e6

However, there is no guarantee that nchan_kernel is even, so the code will throw an error when nchan_kernel is odd.

Even if the nchan_kernel is even, the code will throw an error as the type of nchan_kernel/2 is float. The solution can be either int(nchan_kernel/2) or nchan_kernel//2 so that the type will be int.

[YuRuChou]

Lines 635 and 641 (the multi-EB case) have the opposite problem. They read

response_freqs = (np.squeeze(data.freqs[int(nchan_kernel/2+1):int(-nchan_kernel/2+1)]) + data_delfreq/2.0)/1.e6

which will cause an issue when nchan_kernel is even because the size of response_freqs will not match the size of curr_xc (they will be off by 1)

This problem can be resolved by response_freqs = (np.squeeze(data.freqs[int(nchan_kernel//2+1):int(-nchan_kernel//2+1)]) + data_delfreq/2.0)/1.e6