Entropy coding wants the coefficients in zig-zag order, so the encoder gathers the natural-order grid into the diagonal sequence. Today you build that forward zig-zag, the mirror of the decoder's scatter.
Gather an 8-by-8 natural-order coefficient grid into a 64-entry zig-zag sequence.
The entropy coder processes coefficients in zig-zag order so that low frequencies come first and the quantized-to-zero high frequencies bunch at the end, ready for run-length coding. So before coding, the encoder gathers its natural-order grid into the zig-zag sequence - the exact inverse of the scatter the decoder did with the same ZigZag map. Where the decoder wrote grid[ZigZag[k]] = seq[k], the encoder reads seq[k] = grid[ZigZag[k]].
The map guarantees the two directions agree: sequence position 1 comes from natural index 1 (just right of DC), and position 2 comes from natural index 8 (just below DC), following the diagonal sweep. Reusing the single shared map for both encode and decode is what keeps them consistent - any mismatch here would scramble every block. With the coefficients in sequence, the encoder can split off the DC and run-length code the ACs, which are the next lessons.
// gather is the inverse of the decoder's scatter:// seq[k] = grid[ZigZag[k]]func zigzagOrder(grid [64]int) (seq [64]int) { }