Describe the bug
See here.
We see that initially rectpacking makes correct rows (see placement of n5).
However, during compaction it seems that n1 and n2 steal n3 and n4 and increase the height of the row (somehow).
Hence, n5 can fit all nodes to n10 on top of it, potentially making it harder to visually spot the row.
This does also not seem to be an optimization for the last row as seen at n11.
Expected behavior
I would expect a row with n1 and n2, then a row with n3 and n4 and n5, n6, and n7 stacked next to it and the other nodes in a new row.
Screenshots
What I get:
What I want:
ELK Version
All versions that include rectpacking.
Describe the bug See here. We see that initially rectpacking makes correct rows (see placement of n5). However, during compaction it seems that n1 and n2 steal n3 and n4 and increase the height of the row (somehow). Hence, n5 can fit all nodes to n10 on top of it, potentially making it harder to visually spot the row. This does also not seem to be an optimization for the last row as seen at n11.
Expected behavior I would expect a row with n1 and n2, then a row with n3 and n4 and n5, n6, and n7 stacked next to it and the other nodes in a new row.
Screenshots What I get:
What I want:
ELK Version All versions that include rectpacking.