feat: define and mitigate unavoidable global retiles in grid new-pane #115

New issue

Closed

opened 2026-04-28 01:58:02 +00:00 by barrettruth · 0 comments

barrettruth commented

2026-04-28 01:58:02 +00:00

Problem

grid has pane counts where the final new-pane region is a true global repartition rather than a child split of any current pane. It also has smaller role-shift cases like 2 -> 3. Without an explicit issue, those cases get mixed up with avoidable wrong-branch births.

Blocked by: #101, #102

Proposed solution

Define the exact unavoidable global-retile counts and mitigate the rest. At a minimum, this issue should:

enumerate the true global-reshape counts in tests and docs
separate them from avoidable wrong-branch births in non-global counts
decide what “best effort” means for grid when exact first-split birth is impossible

Context

The current matrix identifies true global-reshape counts when current n is k^2 or k(k + 1) for k >= 2, with examples like 4 -> 5, 6 -> 7, and 9 -> 10. Surveyed external plugins (tmux-tiling-revamped, tmux-tilit) simply call built-in tiled after split.

Acceptance notes

Do not pretend grid can always be solved with one exact birth split.
Make the unavoidable cases explicit and testable.
Improve the avoidable cases where possible.

Alternatives considered

Keep generic split -> tiled for every grid case.
Demand exact first-split births even when the final region is not a refinement of one current pane.

## Problem `grid` has pane counts where the final new-pane region is a true global repartition rather than a child split of any current pane. It also has smaller role-shift cases like `2 -> 3`. Without an explicit issue, those cases get mixed up with avoidable wrong-branch births. Blocked by: #101, #102 ## Proposed solution Define the exact unavoidable global-retile counts and mitigate the rest. At a minimum, this issue should: - enumerate the true global-reshape counts in tests and docs - separate them from avoidable wrong-branch births in non-global counts - decide what “best effort” means for grid when exact first-split birth is impossible ## Context The current matrix identifies true global-reshape counts when current `n` is `k^2` or `k(k + 1)` for `k >= 2`, with examples like `4 -> 5`, `6 -> 7`, and `9 -> 10`. Surveyed external plugins (`tmux-tiling-revamped`, `tmux-tilit`) simply call built-in `tiled` after split. ## Acceptance notes - Do not pretend grid can always be solved with one exact birth split. - Make the unavoidable cases explicit and testable. - Improve the avoidable cases where possible. ## Alternatives considered - Keep generic split -> tiled for every grid case. - Demand exact first-split births even when the final region is not a refinement of one current pane.