Inconsistent sample_shape between VDIF StreamReader and header?

[theXYZT]

Inconsistent sample_shape between VDIF StreamReader and header?

[mhvk]

VDIF has the concept of threads (thread_id in the header), so the sample shape is the number of threads and the number of channels. See https://baseband.readthedocs.io/en/stable/vdif/index.html#file-structure. It is often a thing that makes the stream readers more difficult to generalize...

[theXYZT]

So, does every thread has it's own header?

Edit: I assume I should simply stop relying on header objects for accurate information?

[mhvk]

Indeed, every thread has its own header.

And, indeed, do not rely on just the header (for some Mark 5B headers, the time stamps are even missing the fractional seconds), but instead rely attributes of the stream reader proper (or .info, though for the stream readers that doesn't add much; it is mostly useful for the binary file readers).