tui: @ file search drops filename text after # even without a line suffix

Description

TUI @ file autocomplete treats the last # as the start of a line-number suffix even when what follows is not a line number. For a valid filename such as report#2026.md, it searches only for report; for #notes.md, it searches for an empty base. The suffix of the filename cannot narrow the results, and with the 20-result cap the intended file may be absent.

Reproduction

On dev commit 82666b6612cdb2ba13a7a26e150b21e897a80373, run from packages/tui:

bun -e 'import {parseFileLineRange} from "./src/prompt/parse.ts"; for (const q of ["report#2026.md", "#notes.md", "report#2026.md#3"]) console.log(q,JSON.stringify(parseFileLineRange(q)))'

Actual:

report#2026.md {"base":"report"}
#notes.md {"base":""}
report#2026.md#3 {"base":"report#2026.md","lineRange":{"startLine":3}}

Expected: the first two inputs keep the full filename as the search base; a valid #3 line suffix is still recognized. The autocomplete implementation passes this base to file search. To confirm the user-visible consequence, I used the actual ripgrep-backed FileSystemSearch with 40 reportNN.txt files plus report#2026.md and a limit of 20: searching report returned 20 other files, while searching report#2026.md returned the target. Ranking may vary by filesystem. This verifies the parser and real search separately, not a full interactive TUI run. The parser blob matches current upstream dev.

Thanks for the clear report. I reproduced this on v2.2.1 and on the latest dev (ccd406b).

Steps: a project with 40 reportNN.txt files plus report#2026.md, then type @report#2026.md in the prompt. The autocomplete shows report00.txt, report01.txt and so on, and report#2026.md is not among the visible entries. As you found, parseFileLineRange in packages/tui/src/prompt/parse.ts returns the text before the last # even when what follows isn't a line range, so the search runs on report only.

The recording and screenshots below show it, and the script below reproduces it.

Related: #58033 is a different bug in the same #N line-range handling.

recording.mp4

Reproduction script (drive.ts)

// Repro for #60304: `@` file autocomplete drops the text after '#' even when it is not a line suffix.
import { Effect } from "effect"
import { OpenCodeDriver } from "opencode-drive"

const files: Record<string, string> = { "report#2026.md": "# target\n" }
for (let i = 0; i < 40; i++) files[`report${String(i).padStart(2, "0")}.txt`] = `filler ${i}\n`

export default OpenCodeDriver.use(
  {
    ...(process.env.OC_DEV ? { opencode: { dev: process.env.OC_DEV } } : {}),
    project: { git: true, files },
    tui: { recording: true, keypressOverlay: true },
  },
  ({ ui }) =>
    Effect.gen(function* () {
      yield* Effect.sleep("2 seconds")
      // Type the full, valid filename after '@'
      yield* ui.type("@report#2026.md")
      yield* Effect.sleep("2 seconds")
      yield* ui.screenshot("full-filename-query")
      // Bug: the list shows report00.txt, report01.txt, ... (search base "report") instead of report#2026.md
      yield* Effect.sleep("1 second")
    }),
)

Run it with opencode-drive: opencode-drive run ./drive.ts. It drives opencode from your PATH; set OC_DEV to the path of an OpenCode checkout to run that from source instead.

Navigation Menu

Repository navigation

tui: @ file search drops filename text after # even without a line suffix #60304

Description

Reproduction

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

tui: @ file search drops filename text after # even without a line suffix #60304

Description

Description

Reproduction

Activity

opencodejr commented on Oct 29, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions