warpui_core(tui): build TUI element primitives on ratatui-core

[kevinyang372]

Description

Swap the prototype TUI element library's hand-rolled primitive layer for ratatui 0.30 primitives, behind the existing additive tui cargo feature. The default (GUI) and wasm builds are unaffected — tui is off by default and ratatui is an optional dependency.

This is the first step of building a TUI rendering backend for WarpUI (to support a TUI version of the agent harness). It is stacked on the prototype element library (zb/tui-backend/03b-tui-elements).

What changed:

• TuiBuffer / TuiStyle / Cell now re-export ratatui::buffer::Buffer / ratatui::style::Style / Cell, plus a TuiBufferExt::to_lines headless test helper.
• TuiSize / TuiRect re-export ratatui::layout::{Size, Rect}. A small local TuiConstraint is kept (ratatui's Constraint is the layout-solver enum, not a min/max measure box), and a TuiRectExt keeps the inset / split_top / split_left helpers the hand-rolled column/container rely on.
• TuiText wraps and measures via ratatui::widgets::Paragraph (line_count / line_width) instead of the naive splitter.
• TuiContainer paints through ratatui's Buffer writers (cell_mut / set_style).
• Depends on the ratatui umbrella crate (crossterm 0.29, matching the prototype) with unstable-rendered-line-info + unstable-widget-ref.

Rationale and the full design (why hybrid / ratatui-core over hand-rolling) are in the plan linked below.

Linked Issue

No linked issue — internal, WIP infrastructure gated behind the tui feature (not user-visible).

Testing

• cargo nextest run -p warpui_core --features tui — all pass (geometry / buffer / text / element tests ported to ratatui).
• cargo test -p warpui_core --features tui --doc — pass.
• cargo build -p warpui_core (default, no tui) — unaffected.
• ./script/format and cargo clippy -p warpui_core --features tui --all-targets -- -D warnings — clean.

Agent Mode

• Warp Agent Mode - This PR was created via Warp's AI Agent Mode

---

Plan: https://staging.warp.dev/drive/notebook/RQNlhRWhQK3Djnr50YbC4G
Conversation: https://staging.warp.dev/conversation/9293e035-3939-42c7-aeee-110a5a131011

CHANGELOG-NONE

[kevinyang372]

Warning

This pull request is not mergeable via GitHub because a downstack PR is open. Once all requirements are satisfied, merge this PR as a stack on Graphite.
Learn more

• get tool calls working #12849
• string together input editor + input / output + scrolling #12827
• add tech spec #12794
• Prototype TUI #12786
• warpui_core(tui): add ratatui-backed TUI presenter and runtime #12759
• warpui_core(tui): build TUI element primitives on ratatui-core #12758 👈 (View in Graphite)
• warpui_core: TUI element library (text, column, container, child_view, event_handler) #12634 : 1 other dependent PR (#12601 )
• Add TUI API surface to AppContext under tui #12633
• warpui_core: backend-neutral view hierarchy for a shared responder chain #12413 : 1 other dependent PR (#12414 )
• master

This stack of pull requests is managed by Graphite. Learn more about stacking.

[warp-dev-github-integration]

Overview

Swaps the prototype TUI element library's hand-rolled primitive layer (TuiBuffer/Cell/TuiStyle, TuiSize/TuiRect, and the naive text wrapper) for ratatui 0.30 re-exports, behind the additive tui cargo feature. ratatui is an optional dep and tui is off by default, so the GUI/wasm builds are untouched. Geometry keeps a local TuiConstraint (correctly distinguished from ratatui's layout-solver Constraint) and moves the inset/split_top/split_left helpers onto a TuiRectExt trait; TuiText now measures/paints through Paragraph; TuiContainer paints via Buffer writers; to_lines becomes a TuiBufferExt test helper.

Assessment

Nice, tightly-scoped refactor — the module docs are excellent and the design rationale (hybrid: ratatui primitives + local constraint/slicing helpers) is sound.

Things I checked:

• contains_position removal is safe — it was only referenced by its own (now-removed) test; no production caller depends on it (ratatui's Rect::contains is available if hit-testing needs it later).
• Empty-text layout path is consistent — constraint.clamp(TuiSize::ZERO) produces the same result as the non-empty (constrain_width(0), constrain_height(0)) path, and layout/render/desired_height now all defer to Paragraph, so they agree.
• TuiBufferExt::to_lines wide-grapheme skip is correct — the leading cell sets skip = cell_width-1, trailing continuation columns are skipped, so every glyph is emitted exactly once.
• Intentional behavior change worth calling out for downstream PRs: wrap policy now preserves whitespace (Wrap { trim: false }) instead of collapsing space runs, and wrapping only breaks once the row width is reached. Tests were updated accordingly.

Concerns

None blocking. Two non-blocking suggestions below (one inline). As an initial, feature-gated, non-user-visible infra PR, I'm treating these as optional.

Note: I reviewed the diff statically and did not independently compile/run the suite (the crate pulls a heavy dep graph); relying on the author's stated cargo nextest/clippy/doc runs.

Verdict

Found: 0 critical, 0 important, 2 suggestions

Approve

[warp-dev-github-integration]

💡 [SUGGESTION] Of the four enabled ratatui features, only unstable-rendered-line-info looks exercised by this PR's code (Paragraph::line_count/line_width in text.rs); Widget::render, Buffer, and cell_mut/set_style are all stable. If scrolling-regions, unstable-backend-writer, and unstable-widget-ref are intended for the later stacked runtime/backend PRs, consider enabling each in the PR that first uses it so the feature set tracks the code. As-is this also drifts from the PR description, which lists only two unstable features. Non-blocking.