[max-sixty]

The wt list entry point carried three independently-resolved booleans (format, show_progress, render_table) plus a RenderMode enum to encode just three real modes: JSON, buffered table, progressive table. This replaces the boolean soup with a RenderTarget enum (Json | Table { progressive: bool }) resolved once from the CLI flags + stdout TTY at the top of handle_list and threaded through to collect::collect.

TableRenderPlan no longer needs its is_tty fallback — RenderTarget factors TTY detection up front, so a progressive table only exists when it's actually going to be used. The Some(t) → progressive finalize, None → buffered render split is now the only path.

RowCache shrinks to its load-bearing role and is then inlined: the has_data bitmap that tells the reveal-tick renderer which rows still need the skeleton vs the full row. The set_result dedup is gone — ProgressiveTable::update_row already dedups internally and the picker handler writes through an idempotent Mutex<String>, so the second layer was wasted work. The remaining helper inlines as a free render_reveal function.

Follow-up commit (in response to auto-review): PickerProgressHandler::on_reveal tightens from Vec<Option<String>> to Vec<String>. After the RowCache removal, render_reveal always returns a real line for every row, so the Option wrapper was vestigial — the trait now describes the actual post-refactor behavior and the call sites drop their unwrap branches.

Net: −85 lines across the touched list/picker files. No observable behavior change — all wt list snapshots are unchanged.

This was written by Claude Code on behalf of @max-sixty

[worktrunk-bot]

Nice cleanup — the RenderTarget enum reads much better than the boolean trio, and the RowCache removal is a straightforward consequence of the dedup already happening in update_row and the picker slot.

One follow-up that fell out of the refactor — the PickerProgressHandler::on_reveal trait at src/commands/list/collect/mod.rs:351-354 is now stale:

/// Fired at the 200ms reveal deadline. Entry per row: `Some(line)` for
/// rows still at skeleton state (placeholder needs promoting to `·`),
/// `None` for rows that already received real data via `on_update`.
fn on_reveal(&self, rendered: Vec<Option<String>>);

After this refactor, render_reveal always returns Some(line) for every row (skeleton for rows without data, format_list_item_line for rows with data), so the Some/None semantics described here no longer hold and the Option wrapper carries no information.

Worth tightening to fn on_reveal(&self, rendered: Vec<String>); — the picker handler in progressive_handler.rs drops its if let Some(line) = line and writes every slot (idempotent through Mutex<String>), the test at progressive_handler.rs:242 updates to plain strings, and the doc here describes the actual post-refactor behavior. Otherwise the doc should at least be updated — it now describes a None case that production never produces.