fix: preserve indentation when wrapping multiline task detail text #153

noe.trevino · 2026-06-23T02:39:05-05:00

noe.trevino commented

2026-06-23 02:39:05 -05:00

Closes: #151

Summary

Fixes #151 — paso task show rendered multiline descriptions, comments, and relationship bullets as a "stream of text" instead of preserving their structure.

Root cause

The detail card builds content with manual indents, bullets, and newlines, then hands the whole string to a fixed-Width lipgloss card. lipgloss preserves hard newlines but re-wraps any over-long line and drops the continuation to the card margin, discarding the manual indent. There was also an off-by: in this lipgloss version Width includes the border, so the usable content width is 80 − border(2) − padding(4) = 74, not 76.

Fix

Wrap the text ourselves to the card's real content width with a hanging indent, so continuation lines stay aligned and lipgloss never needs to re-wrap.

styles.go: add CardContentWidth(), WrapText() (word-wrap + hard-break long words), and renderBullet() (2-space first indent, 4-space hanging indent). RenderTaskReference/…WithLabel route through it.
task_helpers.go: WriteIndentedLines (descriptions + comments) now wraps each line to the indent-aware width.
show_logic.go: dropped the now-redundant manual " " prefix on the four relationship bullet call sites.

Before / after (wrapped blocker)

Before:                                      After:
  • …-10 - Create OrderRepository (session     • …-10 - Create OrderRepository (session
lookup, token lookup, atomic email guard)        lookup, token lookup, atomic email guard)

Testing

Added tests for CardContentWidth, WrapText, the bullet hanging indent, and WriteIndentedLines wrapping.
Verified manually via go run main.go task show 233.
go build, go vet, gofmt, and the styles/task suites pass.

Note: two pre-existing failures in internal/tui/activity_view_test.go are unrelated (confirmed failing on a clean tree).

Closes: #151 ## Summary Fixes #151 — `paso task show` rendered multiline descriptions, comments, and relationship bullets as a "stream of text" instead of preserving their structure. ### Root cause The detail card builds content with manual indents, bullets, and newlines, then hands the whole string to a fixed-`Width` lipgloss card. lipgloss preserves hard newlines but **re-wraps any over-long line and drops the continuation to the card margin**, discarding the manual indent. There was also an off-by: in this lipgloss version `Width` *includes* the border, so the usable content width is `80 − border(2) − padding(4) = 74`, not 76. ### Fix Wrap the text ourselves to the card's real content width with a **hanging indent**, so continuation lines stay aligned and lipgloss never needs to re-wrap. - `styles.go`: add `CardContentWidth()`, `WrapText()` (word-wrap + hard-break long words), and `renderBullet()` (2-space first indent, 4-space hanging indent). `RenderTaskReference`/`…WithLabel` route through it. - `task_helpers.go`: `WriteIndentedLines` (descriptions + comments) now wraps each line to the indent-aware width. - `show_logic.go`: dropped the now-redundant manual `" "` prefix on the four relationship bullet call sites. ### Before / after (wrapped blocker) ``` Before: After: • …-10 - Create OrderRepository (session • …-10 - Create OrderRepository (session lookup, token lookup, atomic email guard) lookup, token lookup, atomic email guard) ``` ## Testing - Added tests for `CardContentWidth`, `WrapText`, the bullet hanging indent, and `WriteIndentedLines` wrapping. - Verified manually via `go run main.go task show 233`. - `go build`, `go vet`, `gofmt`, and the styles/task suites pass. Note: two pre-existing failures in `internal/tui/activity_view_test.go` are unrelated (confirmed failing on a clean tree).

claude[bot] (Migrated from github.com) reviewed

2026-06-23 02:58:08 -05:00

internal/cli/task/task_helpers.go

The new wrapping introduces a hard-break regression for long tokens in task descriptions and comments. WrapText chains wordwrap.String (soft-break) then wrap.String (hard-break at limit). When a task description contains a URL longer than CardContentWidth() - len(indent) (~72 chars), wrap.String cuts it mid-string, producing unclickable fragments. The old implementation passed lines through to lipgloss intact; lipgloss would reflow at the card margin (losing indentation) but never cut inside a token. Consider preserving whole lines that contain no whitespace (e.g., URLs that are a single token), or at minimum document that hard-breaks are expected.

The new wrapping introduces a hard-break regression for long tokens in task descriptions and comments. `WrapText` chains `wordwrap.String` (soft-break) then `wrap.String` (hard-break at limit). When a task description contains a URL longer than `CardContentWidth() - len(indent)` (~72 chars), `wrap.String` cuts it mid-string, producing unclickable fragments. The old implementation passed lines through to lipgloss intact; lipgloss would reflow at the card margin (losing indentation) but never cut inside a token. Consider preserving whole lines that contain no whitespace (e.g., URLs that are a single token), or at minimum document that hard-breaks are expected.

claude[bot] (Migrated from github.com) reviewed

2026-06-23 02:58:16 -05:00

internal/cli/task/task_helpers_test.go

					
				@ -0,0 +1,46 @@

				package task

len(l) counts bytes; CardContentWidth() is a visual terminal column count. muesli/reflow wraps by visual width (via go-runewidth), not bytes. For any multi-byte Unicode character in a task title (accented Latin = 2 bytes/1 col, CJK = 3 bytes/1-2 cols), a visually-correct line can have len(l) > CardContentWidth(), causing this assertion to fire spuriously. Use runewidth.StringWidth(l) or strip to ASCII before comparing, consistent with how the library itself measures width.

`len(l)` counts bytes; `CardContentWidth()` is a visual terminal column count. `muesli/reflow` wraps by visual width (via `go-runewidth`), not bytes. For any multi-byte Unicode character in a task title (accented Latin = 2 bytes/1 col, CJK = 3 bytes/1-2 cols), a visually-correct line can have `len(l) > CardContentWidth()`, causing this assertion to fire spuriously. Use `runewidth.StringWidth(l)` or strip to ASCII before comparing, consistent with how the library itself measures width.

claude[bot] (Migrated from github.com) reviewed

2026-06-23 02:58:22 -05:00

internal/cli/task/task_helpers_test.go

					
				@ -0,0 +1,46 @@

				package task

internal/testing/fixtures/cli.go already exports fixtures.StripANSI (same regex, same logic), and it is imported by 17+ other test files in this repo. The styles test package has its own copy as well (success_test.go). This new copy makes three. Use the existing shared helper to keep the regex in one place.

`internal/testing/fixtures/cli.go` already exports `fixtures.StripANSI` (same regex, same logic), and it is imported by 17+ other test files in this repo. The `styles` test package has its own copy as well (`success_test.go`). This new copy makes three. Use the existing shared helper to keep the regex in one place.

claude[bot] (Migrated from github.com) reviewed

2026-06-23 02:58:44 -05:00

internal/cli/styles/styles_test.go

Same byte-vs-visual-width mismatch as in task_helpers_test.go: len(l) is bytes, CardContentWidth() is terminal columns. The muesli/reflow wrapping library uses go-runewidth for visual measurement. A task title with multi-byte Unicode characters will produce lines where len(l) > CardContentWidth() while the visual rendering is correct, causing this assertion to spuriously fail.

Same byte-vs-visual-width mismatch as in `task_helpers_test.go`: `len(l)` is bytes, `CardContentWidth()` is terminal columns. The `muesli/reflow` wrapping library uses `go-runewidth` for visual measurement. A task title with multi-byte Unicode characters will produce lines where `len(l) > CardContentWidth()` while the visual rendering is correct, causing this assertion to spuriously fail.

claude[bot] commented

2026-06-23 02:59:33 -05:00

(Migrated from github.com)

Review

This PR fixes the multiline text wrapping regression in paso task show by implementing manual word-wrap with hanging indent before lipgloss renders the card border, and simultaneously renames the ticket_number field to task_number throughout the model, DB, and CLI layers.

Issues found

1. Hard-break regression for long tokens in descriptions/comments (task_helpers.go line 43)
The new WriteIndentedLines calls WrapText, which chains wordwrap.String (soft-break) then wrap.String (hard-break at limit). Any long single token in a task description—a URL, a file path, a hash—that exceeds ~72 characters will be cut mid-string. The old code did not hard-break: it passed lines through to lipgloss, which reflowed at the card margin (losing indentation) but never split inside a token. This is a new UX regression for any user whose descriptions contain long URLs.

2. len(l) (bytes) compared against CardContentWidth() (terminal columns) in tests (styles_test.go ~line 168, task_helpers_test.go line 46)
muesli/reflow wraps by visual cell width via go-runewidth. For multi-byte Unicode characters (accented Latin = 2 bytes/1 col, CJK = 3 bytes/1-2 cols), a visually-correct wrapped line will have len(l) > CardContentWidth() and the assertion fires spuriously. The tests pass today only because the fixture uses ASCII-only input.

3. stripANSI duplicated when a shared helper already exists (task_helpers_test.go line 14)
internal/testing/fixtures/cli.go exports StripANSI (identical regex, same logic) and is already imported by 17+ test files in this repo. The new task_helpers_test.go adds a third independent copy (the styles package has its own as well). Use fixtures.StripANSI.

Score: 7/10

## Review This PR fixes the multiline text wrapping regression in `paso task show` by implementing manual word-wrap with hanging indent before lipgloss renders the card border, and simultaneously renames the `ticket_number` field to `task_number` throughout the model, DB, and CLI layers. ### Issues found **1. Hard-break regression for long tokens in descriptions/comments** (`task_helpers.go` line 43) The new `WriteIndentedLines` calls `WrapText`, which chains `wordwrap.String` (soft-break) then `wrap.String` (hard-break at limit). Any long single token in a task description—a URL, a file path, a hash—that exceeds ~72 characters will be cut mid-string. The old code did not hard-break: it passed lines through to lipgloss, which reflowed at the card margin (losing indentation) but never split inside a token. This is a new UX regression for any user whose descriptions contain long URLs. **2. `len(l)` (bytes) compared against `CardContentWidth()` (terminal columns) in tests** (`styles_test.go` ~line 168, `task_helpers_test.go` line 46) `muesli/reflow` wraps by visual cell width via `go-runewidth`. For multi-byte Unicode characters (accented Latin = 2 bytes/1 col, CJK = 3 bytes/1-2 cols), a visually-correct wrapped line will have `len(l) > CardContentWidth()` and the assertion fires spuriously. The tests pass today only because the fixture uses ASCII-only input. **3. `stripANSI` duplicated when a shared helper already exists** (`task_helpers_test.go` line 14) `internal/testing/fixtures/cli.go` exports `StripANSI` (identical regex, same logic) and is already imported by 17+ test files in this repo. The new `task_helpers_test.go` adds a third independent copy (the `styles` package has its own as well). Use `fixtures.StripANSI`. --- **Score: 7/10**

Sign in to join this conversation.

No reviewers