Ch26 Discussion
Discussion Guide
Prompts
1. "A visualization file contains no picture." Defend or refute, and say what it contains instead. Why is
that separation (data now, picture later) powerful for debugging and for a paper?
2. The chapter insists POINT_DATA be computed as field%nx * field%ny, never hard-coded. What class of bug
does that discipline prevent, and how does it echo the general "single source of truth" idea from the
modules chapter?
3. VTK wants x fastest; Fortran stores the first index fastest. Is it luck that our write loop is both correct
and cache-friendly, or a rule you can state and reuse? Phrase the rule.
4. When would you reach for ASCII VTK, and when for the binary formats of Chapter 25? Give the crossover in
concrete terms (grid size, run length, precision needs).
5. Why is choosing viridis over jet a matter of scientific honesty and not taste? What can a rainbow
colormap make a reader see that is not in the data?
Mini group activity (≈ 25 min). In pairs: hand each pair a deliberately broken VTK file (rotate among:
wrong magic-string case; POINT_DATA off by one; DIMENSIONS with nz=2 for a 2-D field; values in i-outer
transposed order; unpadded filenames in a series). Each pair must (a) predict what ParaView will do — reject,
error, or show a wrong picture — before opening it, (b) open it and confirm, and (c) fix the file and
re-open. Then pairs swap files. The key learning outcome: a file that compiles/opens can still be wrong,
and the "shows a plausible but wrong picture" failures (transpose, off-by-one) are the dangerous ones. Close
by collecting the taxonomy of failure modes on the board.