First, the TraceHeaderRev[0,1] classes use more specific field types to constrain their values. This is analogous to the recent changes to BinaryReelHeader.
Second, the trace headers we generate for round-trip testing are more sensible, and our InMemoryDataset more correctly calculates things like num_traces, etc. As a result, we can actually do round-trip comparisons between traces, trace headers, and samples now.
Third, I cleaned up a lot of incorrectly specified multi-line docstrings. They were pasted from the SEG Y spec as multiple strings, not multiline strings. As a result, the docstrings for many header fields were just the first line of text from the spec.
Coverage increased (+0.4%) to 85.94% when pulling 548d49ca228f1e02def7eb2c7be8d49641efe0dd on abingham:reader-tests into d0a6f57cff53718a27e13902275921d50d3d2a94 on sixty-north:master.
This includes a few related changes.
First, the
TraceHeaderRev[0,1]
classes use more specific field types to constrain their values. This is analogous to the recent changes toBinaryReelHeader
.Second, the trace headers we generate for round-trip testing are more sensible, and our
InMemoryDataset
more correctly calculates things like num_traces, etc. As a result, we can actually do round-trip comparisons between traces, trace headers, and samples now.Third, I cleaned up a lot of incorrectly specified multi-line docstrings. They were pasted from the SEG Y spec as multiple strings, not multiline strings. As a result, the docstrings for many header fields were just the first line of text from the spec.