feat: reduce centered-master parity-transition churn during new-pane #110

New issue

Closed

opened 2026-04-28 01:57:57 +00:00 by barrettruth · 0 comments

barrettruth commented

2026-04-28 01:57:57 +00:00

Problem

Even after the trivial centered-master first-side-stack birth is fixed, parity transitions still cause nonlocal role shifts. When the side counts rebalance, one or more panes can cross between left stack, center master base, and right stack.

Blocked by: #101, #102, #106

Proposed solution

Investigate and, where possible, reduce visible churn during centered-master parity transitions. Scope includes:

s = 1, where the left stack appears for the first time
later odd-stack transitions where left_n increases and the master base shifts

The goal is not necessarily perfect elimination. It is to reduce avoidable movement and clearly distinguish what remains structurally necessary.

Context

The current matrix classifies these as local role-shift cases rather than simple tail-split cases. The strongest surveyed external analog (main-center in tmux-tiling-revamped) just recomputes and reapplies the full custom layout body after a split.

Acceptance notes

Reduce unnecessary cross-column motion if possible.
If perfect elimination is not possible, make the residual behavior precise enough to test and later document.
Do not regress the simpler first-side-stack case.

Alternatives considered

Accept full reprojection forever.
Change semantic pane-distribution rules instead of improving creation-time behavior.

## Problem Even after the trivial `centered-master` first-side-stack birth is fixed, parity transitions still cause nonlocal role shifts. When the side counts rebalance, one or more panes can cross between left stack, center master base, and right stack. Blocked by: #101, #102, #106 ## Proposed solution Investigate and, where possible, reduce visible churn during `centered-master` parity transitions. Scope includes: - `s = 1`, where the left stack appears for the first time - later odd-stack transitions where `left_n` increases and the master base shifts The goal is not necessarily perfect elimination. It is to reduce avoidable movement and clearly distinguish what remains structurally necessary. ## Context The current matrix classifies these as local role-shift cases rather than simple tail-split cases. The strongest surveyed external analog (`main-center` in `tmux-tiling-revamped`) just recomputes and reapplies the full custom layout body after a split. ## Acceptance notes - Reduce unnecessary cross-column motion if possible. - If perfect elimination is not possible, make the residual behavior precise enough to test and later document. - Do not regress the simpler first-side-stack case. ## Alternatives considered - Accept full reprojection forever. - Change semantic pane-distribution rules instead of improving creation-time behavior.