codescout 0.15.0

//! Command type detection and smart output summarization.
//!
//! Detects whether a command is a test runner, a build tool, or something else,
//! then produces a structured summary appropriate for that command type.

use regex::Regex;
use serde_json::{json, Value};
use std::sync::OnceLock;

// ---------------------------------------------------------------------------
// Thresholds
// ---------------------------------------------------------------------------

/// Inline line cap for buffer-only queries (e.g. `grep/sed @cmd_xxx`).
/// Kept separate from the summarization threshold so "when to buffer" and "how much
/// to return from a buffer query" can be tuned independently.
pub(crate) const BUFFER_QUERY_INLINE_CAP: usize = 100;

/// Number of lines to keep from the top in generic summaries.
const HEAD_LINES: usize = 20;

/// Number of lines to keep from the bottom in generic summaries.
const TAIL_LINES: usize = 10;

// ---------------------------------------------------------------------------
// CommandType
// ---------------------------------------------------------------------------

/// Broad category of the command being run.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum CommandType {
    Test,
    Build,
    Generic,
}

// ---------------------------------------------------------------------------
// Regex patterns (compiled once via OnceLock)
// ---------------------------------------------------------------------------

fn test_re() -> &'static Regex {
    static RE: OnceLock<Regex> = OnceLock::new();
    RE.get_or_init(|| {
        Regex::new(
            r"(?x)
            (?:^|\s|/)
            (?:
                cargo\s+test
              | pytest
              | npm\s+test
              | npx\s+jest
              | jest
              | go\s+test
              | mvn\s+test
              | gradle\s+test
            )
            (?:\s|$)",
        )
        .expect("test regex")
    })
}

fn build_re() -> &'static Regex {
    static RE: OnceLock<Regex> = OnceLock::new();
    RE.get_or_init(|| {
        Regex::new(
            r"(?x)
            (?:^|\s|/)
            (?:
                cargo\s+(?:build|clippy|check)
              | npm\s+run\s+build
              | make(?:\s|$)
              | tsc(?:\s|$)
              | gcc(?:\s|$)
              | g\+\+(?:\s|$)
              | clang(?:\s|$)
              | javac(?:\s|$)
              | go\s+build
            )",
        )
        .expect("build regex")
    })
}

fn cargo_test_result_re() -> &'static Regex {
    static RE: OnceLock<Regex> = OnceLock::new();
    RE.get_or_init(|| {
        Regex::new(r"(\d+)\s+passed;\s+(\d+)\s+failed;\s+(\d+)\s+ignored")
            .expect("cargo test regex")
    })
}

fn rust_error_code_re() -> &'static Regex {
    static RE: OnceLock<Regex> = OnceLock::new();
    RE.get_or_init(|| Regex::new(r"^error\[E\d+\]").expect("rust error regex"))
}

fn warning_re() -> &'static Regex {
    static RE: OnceLock<Regex> = OnceLock::new();
    RE.get_or_init(|| Regex::new(r"^warning(\[.+\])?:").expect("warning regex"))
}

// ---------------------------------------------------------------------------
// Public API
// ---------------------------------------------------------------------------

/// Classify a command string into Test, Build, or Generic.
pub fn detect_command_type(command: &str) -> CommandType {
    // Test takes priority over build (e.g. "cargo test" is Test, not Build).
    if test_re().is_match(command) {
        CommandType::Test
    } else if build_re().is_match(command) {
        CommandType::Build
    } else {
        CommandType::Generic
    }
}

/// Returns the byte offset of the `|` pipe character that separates a command
/// from a terminal filter stage (grep, head, tail, sed, awk, cut, wc, sort,
/// uniq, tr, rg, egrep, fgrep).
///
/// Returns `None` if the command has no pipe, or if the last pipe stage is not
/// a known terminal filter. Pipe characters inside quoted strings are ignored.
pub fn detect_terminal_filter(cmd: &str) -> Option<usize> {
    const TERMINAL_FILTERS: &[&str] = &[
        "grep", "egrep", "fgrep", "rg", "head", "tail", "sed", "awk", "cut", "wc", "sort", "uniq",
        "tr",
    ];

    let mut last_pipe: Option<usize> = None;
    let mut in_single = false;
    let mut in_double = false;
    let mut escape_next = false;

    for (i, ch) in cmd.char_indices() {
        if escape_next {
            escape_next = false;
            continue;
        }
        match ch {
            '\\' if !in_single => escape_next = true,
            '\'' if !in_double => in_single = !in_single,
            '"' if !in_single => in_double = !in_double,
            '|' if !in_single && !in_double => {
                // Skip `||` (logical OR) — not a pipeline pipe.
                let next_char = cmd[i + 1..].chars().next();
                let prev_char = if i > 0 { cmd[..i].chars().last() } else { None };
                if next_char != Some('|') && prev_char != Some('|') {
                    last_pipe = Some(i);
                }
            }
            _ => {}
        }
    }

    let pipe_pos = last_pipe?;

    // Extract the first token of the stage after the pipe
    let after_pipe = cmd[pipe_pos + 1..].trim_start();
    let token = after_pipe
        .split(|c: char| c.is_whitespace())
        .next()
        .unwrap_or("");

    // Strip any path prefix so "/usr/bin/grep" matches "grep"
    let name = token.rsplit('/').next().unwrap_or(token);

    if TERMINAL_FILTERS.contains(&name) {
        Some(pipe_pos)
    } else {
        None
    }
}

/// Returns `true` when the combined output is large enough to benefit from
/// summarization rather than raw output.
pub fn needs_summary(stdout: &str, stderr: &str) -> bool {
    (stdout.len() + stderr.len()) / 4 > crate::tools::MAX_INLINE_TOKENS
}

/// Produce a structured summary of test-runner output.
///
/// Parses cargo-test-style result lines, sums across multiple test binaries,
/// and extracts failure details.
pub fn summarize_test_output(stdout: &str, stderr: &str, exit_code: i32) -> Value {
    let mut passed: u64 = 0;
    let mut failed: u64 = 0;
    let mut ignored: u64 = 0;

    let combined = if stderr.is_empty() {
        stdout.to_string()
    } else {
        format!("{stdout}\n{stderr}")
    };

    let re = cargo_test_result_re();
    for line in combined.lines() {
        if let Some(caps) = re.captures(line) {
            passed += caps[1].parse::<u64>().unwrap_or(0);
            failed += caps[2].parse::<u64>().unwrap_or(0);
            ignored += caps[3].parse::<u64>().unwrap_or(0);
        }
    }

    let failures = extract_test_failures(&combined);

    let mut result = json!({
        "type": "test",
        "exit_code": exit_code,
        "passed": passed,
    });

    if failed > 0 {
        result["failed"] = json!(failed);
    }
    if ignored > 0 {
        result["ignored"] = json!(ignored);
    }
    if let Some(f) = failures {
        result["failures"] = Value::String(f);
    }

    result
}

/// Produce a structured summary of compiler / build-tool output.
///
/// Counts errors (with error codes) and warnings, and extracts the first
/// error block for quick diagnosis.
pub fn summarize_build_output(stdout: &str, stderr: &str, exit_code: i32) -> Value {
    let combined = if stderr.is_empty() {
        stdout.to_string()
    } else if stdout.is_empty() {
        stderr.to_string()
    } else {
        format!("{stdout}\n{stderr}")
    };

    let mut errors: u64 = 0;
    let mut warnings: u64 = 0;
    let mut first_error: Option<String> = None;

    let err_re = rust_error_code_re();
    let warn_re = warning_re();
    let lines: Vec<&str> = combined.lines().collect();
    for (i, line) in lines.iter().enumerate() {
        if err_re.is_match(line) {
            errors += 1;
            if first_error.is_none() {
                first_error = Some(extract_error_block(&lines, i));
            }
        } else if warn_re.is_match(line) {
            warnings += 1;
        }
    }

    let mut result = json!({
        "type": "build",
        "exit_code": exit_code,
    });

    if errors > 0 {
        result["errors"] = json!(errors);
    }
    if warnings > 0 {
        result["warnings"] = json!(warnings);
    }
    if let Some(err) = first_error {
        result["first_error"] = Value::String(err);
    }

    result
}

/// Produce a head+tail summary for generic command output.
///
/// If stdout fits within HEAD_LINES + TAIL_LINES, it is returned verbatim.
/// Otherwise the middle is replaced with an "N lines omitted" marker.
pub fn summarize_generic(stdout: &str, stderr: &str, exit_code: i32) -> Value {
    let stdout_lines: Vec<&str> = stdout.lines().collect();
    let total_stdout_lines = stdout_lines.len();

    let summarized_stdout = if total_stdout_lines > HEAD_LINES + TAIL_LINES {
        let head: Vec<&str> = stdout_lines[..HEAD_LINES].to_vec();
        let tail: Vec<&str> = stdout_lines[total_stdout_lines - TAIL_LINES..].to_vec();
        let omitted = total_stdout_lines - HEAD_LINES - TAIL_LINES;
        format!(
            "{}\n--- {} lines omitted ---\n{}",
            head.join("\n"),
            omitted,
            tail.join("\n")
        )
    } else {
        stdout.to_string()
    };

    let mut result = json!({
        "type": "generic",
        "exit_code": exit_code,
    });

    if !summarized_stdout.is_empty() {
        result["stdout"] = Value::String(summarized_stdout);
    }
    if !stderr.is_empty() {
        result["stderr"] = Value::String(stderr.to_string());
    }

    result
}

// ---------------------------------------------------------------------------
// Helpers
// ---------------------------------------------------------------------------

pub fn count_lines(s: &str) -> usize {
    if s.is_empty() {
        0
    } else {
        s.lines().count()
    }
}

/// Truncate `text` to at most `max_lines` lines.
///
/// Returns `(truncated_text, lines_shown, lines_total)`.
/// When `lines_total <= max_lines`, `text` is returned unchanged and
/// `lines_shown == lines_total`.
#[allow(dead_code)]
pub(crate) fn truncate_lines(text: &str, max_lines: usize) -> (String, usize, usize) {
    let total = count_lines(text);
    if total <= max_lines {
        return (text.to_string(), total, total);
    }
    let truncated = text.lines().take(max_lines).collect::<Vec<_>>().join("\n");
    (truncated, max_lines, total)
}

/// Truncate `text` to at most `max_lines` lines **and** at most `max_bytes` bytes.
///
/// Both limits are applied; whichever is more restrictive wins. Truncation
/// always occurs on a line boundary so output is never split mid-line.
///
/// Returns `(content, lines_shown, total_lines)`.
pub(crate) fn truncate_lines_and_bytes(
    text: &str,
    max_lines: usize,
    max_bytes: usize,
) -> (String, usize, usize) {
    let total = count_lines(text);
    let mut result = String::new();
    let mut lines_shown = 0;

    for line in text.lines().take(max_lines) {
        // Each line adds the line itself plus a '\n' separator (except the first).
        let needed = if result.is_empty() {
            line.len()
        } else {
            line.len() + 1 // +1 for '\n'
        };
        if result.len() + needed > max_bytes {
            break;
        }
        if !result.is_empty() {
            result.push('\n');
        }
        result.push_str(line);
        lines_shown += 1;
    }

    (result, lines_shown, total)
}

/// Extract text between the first `failures:` section markers in cargo test output.
fn extract_test_failures(output: &str) -> Option<String> {
    // Cargo test outputs failures between two "failures:" markers.
    // The first "failures:" is followed by stdout of failing tests.
    // The second "failures:" is followed by test names.
    let lines: Vec<&str> = output.lines().collect();
    let mut start = None;
    let mut end = None;

    for (i, line) in lines.iter().enumerate() {
        if line.trim() == "failures:" {
            if start.is_none() {
                start = Some(i);
            } else {
                end = Some(i);
                break;
            }
        }
    }

    // If we found at least one "failures:" marker, collect everything after it
    // up to the second marker (or end of output).
    if let Some(s) = start {
        let e = end.unwrap_or(lines.len());
        // Include from start through the second failures block
        let section: Vec<&str> = if let Some(end_idx) = end {
            // Find the end of the second failures block (next "test result:" line or EOF)
            let block_end = lines[end_idx..]
                .iter()
                .position(|l| l.starts_with("test result:"))
                .map(|p| end_idx + p)
                .unwrap_or(lines.len());
            lines[s..block_end].to_vec()
        } else {
            lines[s..e].to_vec()
        };

        let text = section.join("\n").trim().to_string();
        if text.is_empty() {
            None
        } else {
            Some(text)
        }
    } else {
        None
    }
}

/// Extract an error block: the error line plus continuation lines until
/// the next blank line or next error/warning.
fn extract_error_block(lines: &[&str], start: usize) -> String {
    let err_re = rust_error_code_re();
    let warn_re = warning_re();
    let mut block = vec![lines[start]];
    for line in &lines[start + 1..] {
        // Stop at blank lines or next top-level diagnostic
        if line.is_empty()
            || err_re.is_match(line)
            || warn_re.is_match(line)
            || line.starts_with("error:")
        {
            break;
        }
        block.push(line);
    }
    block.join("\n")
}

/// Strip ANSI CSI escape sequences from `s` (e.g. `\x1b[32m`, `\x1b[0m`).
///
/// Call this on buffered command output before byte-budget calculations. ANSI
/// codes are opaque to LLMs and inflate byte lengths — on ANSI-colored log
/// files a single grep match line can be several KB of escape codes around a
/// few dozen bytes of visible text, silently exhausting the byte budget and
/// causing `stdout_shown=0`.
///
/// Only CSI sequences (`ESC '['` … final ASCII letter) are removed. Other
/// escape types (rare in log files) pass through unchanged.
pub(crate) fn strip_ansi_codes(s: &str) -> String {
    let mut result = String::with_capacity(s.len());
    let mut chars = s.chars().peekable();
    while let Some(ch) = chars.next() {
        if ch == '\x1b' {
            match chars.peek().copied() {
                Some('[') => {
                    chars.next(); // consume '['
                    for c in chars.by_ref() {
                        if c.is_ascii_alphabetic() {
                            break; // final byte consumed
                        }
                    }
                }
                _ => {
                    // Non-CSI escape — preserve; rare in structured log output.
                    result.push(ch);
                }
            }
        } else {
            result.push(ch);
        }
    }
    result
}

// ---------------------------------------------------------------------------
// Tests
// ---------------------------------------------------------------------------

#[cfg(test)]
mod tests {
    use super::*;

    // -- detect_command_type --

    #[test]
    fn strip_ansi_codes_removes_color_sequences() {
        let input = "\x1b[32mINFO\x1b[0m some message";
        assert_eq!(strip_ansi_codes(input), "INFO some message");
    }

    #[test]
    fn strip_ansi_codes_removes_256_color_sequences() {
        // 256-color and bold sequences have multiple parameters separated by ';'
        let input = "\x1b[1;38;5;220mWARN\x1b[0m something";
        assert_eq!(strip_ansi_codes(input), "WARN something");
    }

    #[test]
    fn strip_ansi_codes_plain_text_unchanged() {
        let input = "no escape codes here\nline two";
        assert_eq!(strip_ansi_codes(input), input);
    }

    #[test]
    fn strip_ansi_codes_preserves_newlines() {
        let input = "\x1b[32mline1\x1b[0m\nline2\n\x1b[31mline3\x1b[0m";
        assert_eq!(strip_ansi_codes(input), "line1\nline2\nline3");
    }

    #[test]
    fn strip_ansi_codes_non_csi_escape_preserved() {
        // ESC followed by non-'[' is not a CSI sequence — preserved as-is
        let input = "\x1b=hello";
        assert_eq!(strip_ansi_codes(input), "\x1b=hello");
    }

    #[test]
    fn strip_ansi_codes_empty_string() {
        assert_eq!(strip_ansi_codes(""), "");
    }

    #[test]
    fn detect_test_command() {
        assert_eq!(detect_command_type("cargo test"), CommandType::Test);
        assert_eq!(
            detect_command_type("cargo test --release"),
            CommandType::Test
        );
        assert_eq!(detect_command_type("pytest tests/"), CommandType::Test);
        assert_eq!(detect_command_type("npm test"), CommandType::Test);
        assert_eq!(detect_command_type("npx jest"), CommandType::Test);
        assert_eq!(detect_command_type("go test ./..."), CommandType::Test);
    }

    #[test]
    fn detect_build_command() {
        assert_eq!(detect_command_type("cargo build"), CommandType::Build);
        assert_eq!(
            detect_command_type("cargo clippy -- -D warnings"),
            CommandType::Build
        );
        assert_eq!(detect_command_type("npm run build"), CommandType::Build);
        assert_eq!(detect_command_type("make"), CommandType::Build);
        assert_eq!(detect_command_type("tsc"), CommandType::Build);
        assert_eq!(detect_command_type("gcc main.c"), CommandType::Build);
    }

    #[test]
    fn detect_generic_command() {
        assert_eq!(detect_command_type("echo hello"), CommandType::Generic);
        assert_eq!(detect_command_type("ls -la"), CommandType::Generic);
        assert_eq!(detect_command_type("cat file.txt"), CommandType::Generic);
    }

    // -- needs_summary --

    #[test]
    fn short_output_not_summarized() {
        assert!(!needs_summary("hello\nworld\n", ""));
    }

    #[test]
    fn long_output_needs_summary() {
        // Generate output exceeding MAX_INLINE_TOKENS * 4 bytes (~10KB)
        let stdout: String = (1..=3000).map(|i| format!("line {}\n", i)).collect();
        assert!(needs_summary(&stdout, ""));
    }

    // -- summarize_test_output --

    #[test]
    fn summarize_cargo_test_all_pass() {
        let stdout = "running 5 tests\ntest a ... ok\ntest b ... ok\ntest c ... ok\ntest d ... ok\ntest e ... ok\n\ntest result: ok. 5 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.02s\n";
        let summary = summarize_test_output(stdout, "", 0);
        assert_eq!(summary["passed"], 5);
        assert!(
            summary.get("failed").is_none(),
            "failed:0 should be omitted"
        );
        assert!(
            summary.get("ignored").is_none(),
            "ignored:0 should be omitted"
        );
        assert!(summary.get("failures").is_none() || summary["failures"].is_null());
    }

    #[test]
    fn summarize_cargo_test_with_failures() {
        let stdout = "running 3 tests\ntest ok_test ... ok\ntest failing_test ... FAILED\ntest another ... ok\n\nfailures:\n\n---- failing_test stdout ----\nthread 'failing_test' panicked at 'assertion failed'\n\nfailures:\n    failing_test\n\ntest result: FAILED. 2 passed; 1 failed; 0 ignored; 0 measured; 0 filtered out\n";
        let summary = summarize_test_output(stdout, "", 1);
        assert_eq!(summary["passed"], 2);
        assert_eq!(summary["failed"], 1);
        let failures = summary["failures"].as_str().unwrap();
        assert!(failures.contains("failing_test"));
    }

    #[test]
    fn summarize_cargo_test_multiple_binaries() {
        let stdout = "\
running 3 tests
test a ... ok
test b ... ok
test c ... ok

test result: ok. 3 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out

running 2 tests
test d ... ok
test e ... FAILED

failures:

---- e stdout ----
assertion failed

failures:
    e

test result: FAILED. 1 passed; 1 failed; 0 ignored; 0 measured; 0 filtered out
";
        let summary = summarize_test_output(stdout, "", 1);
        // Sums across both binaries
        assert_eq!(summary["passed"], 4);
        assert_eq!(summary["failed"], 1);
    }

    // -- summarize_build_output --

    #[test]
    fn summarize_build_errors() {
        let stderr = "error[E0308]: mismatched types\n --> src/main.rs:5:20\n  |\n5 |     let x: String = 42;\n  |                     ^^ expected `String`, found integer\n\nwarning: unused variable: `y`\n --> src/main.rs:3:9\n  |\n3 |     let y = 1;\n  |         ^ help: consider prefixing with an underscore: `_y`\n\nerror: aborting due to 1 previous error; 1 warning emitted\n";
        let summary = summarize_build_output("", stderr, 1);
        assert_eq!(summary["errors"], 1); // only error[E...], not "error: aborting"
        assert_eq!(summary["warnings"], 1);
        assert!(summary["first_error"].as_str().unwrap().contains("E0308"));
    }

    #[test]
    fn summarize_build_no_errors() {
        let stderr = "warning: unused variable: `x`\n --> src/main.rs:2:9\n";
        let summary = summarize_build_output("", stderr, 0);
        assert!(
            summary.get("errors").is_none(),
            "errors:0 should be omitted"
        );
        assert_eq!(summary["warnings"], 1);
        assert!(summary.get("first_error").is_none() || summary["first_error"].is_null());
    }

    // -- summarize_generic --

    #[test]
    fn summarize_generic_head_tail() {
        let lines: String = (1..=100).map(|i| format!("line {}\n", i)).collect();
        let summary = summarize_generic(&lines, "", 0);
        let output = summary["stdout"].as_str().unwrap();
        assert!(output.contains("line 1"));
        assert!(output.contains("line 20"));
        assert!(output.contains("lines omitted"));
        assert!(output.contains("line 100"));
    }

    #[test]
    fn summarize_generic_short_output_verbatim() {
        let stdout = "line 1\nline 2\nline 3\n";
        let summary = summarize_generic(stdout, "", 0);
        let output = summary["stdout"].as_str().unwrap();
        assert_eq!(output, stdout);
        assert!(!output.contains("omitted"));
    }

    #[test]
    fn summarize_generic_includes_stderr() {
        let summary = summarize_generic("out\n", "err\n", 1);
        assert!(summary.get("stderr").is_some());
        assert_eq!(summary["stderr"].as_str().unwrap(), "err\n");
    }

    #[test]
    fn summarize_generic_omits_empty_stderr() {
        let summary = summarize_generic("out\n", "", 0);
        assert!(summary.get("stderr").is_none() || summary["stderr"].is_null());
    }

    #[test]
    fn summarize_generic_omits_empty_stdout() {
        let summary = summarize_generic("", "err\n", 1);
        assert!(summary.get("stdout").is_none() || summary["stdout"].is_null());
        assert_eq!(summary["stderr"].as_str().unwrap(), "err\n");
    }

    // -- helpers --

    #[test]
    fn count_lines_empty() {
        assert_eq!(count_lines(""), 0);
    }

    #[test]
    fn count_lines_normal() {
        assert_eq!(count_lines("a\nb\nc"), 3);
    }

    #[test]
    fn truncate_lines_short_returns_unchanged() {
        let text = "a\nb\nc";
        let (out, shown, total) = truncate_lines(text, 10);
        assert_eq!(out, text);
        assert_eq!(shown, 3);
        assert_eq!(total, 3);
    }

    #[test]
    fn truncate_lines_exact_limit_not_truncated() {
        let text: String = (1..=5)
            .map(|i| format!("line {i}"))
            .collect::<Vec<_>>()
            .join("\n");
        let (out, shown, total) = truncate_lines(&text, 5);
        assert_eq!(shown, 5);
        assert_eq!(total, 5);
        assert_eq!(out, text);
    }

    #[test]
    fn truncate_lines_long_truncates_correctly() {
        let text: String = (1..=10)
            .map(|i| format!("line {i}"))
            .collect::<Vec<_>>()
            .join("\n");
        let (out, shown, total) = truncate_lines(&text, 3);
        assert_eq!(shown, 3);
        assert_eq!(total, 10);
        let lines: Vec<&str> = out.lines().collect();
        assert_eq!(lines.len(), 3);
        assert_eq!(lines[0], "line 1");
        assert_eq!(lines[2], "line 3");
    }

    #[test]
    fn truncate_lines_empty_string() {
        let (out, shown, total) = truncate_lines("", 10);
        assert_eq!(out, "");
        assert_eq!(shown, 0);
        assert_eq!(total, 0);
    }

    #[test]
    fn truncate_lines_and_bytes_line_limit_wins() {
        // 3 lines, byte budget is generous — line limit (2) wins.
        let text = "aaa\nbbb\nccc";
        let (out, shown, total) = truncate_lines_and_bytes(text, 2, 1000);
        assert_eq!(out, "aaa\nbbb");
        assert_eq!(shown, 2);
        assert_eq!(total, 3);
    }

    #[test]
    fn truncate_lines_and_bytes_byte_limit_wins() {
        // 5 short lines, byte budget forces truncation after the 2nd line.
        // "aaa\nbbb" = 7 bytes; adding "\nccc" = 11 bytes — over the 8-byte budget.
        let text = "aaa\nbbb\nccc\nddd\neee";
        let (out, shown, total) = truncate_lines_and_bytes(text, 10, 8);
        assert_eq!(out, "aaa\nbbb");
        assert_eq!(shown, 2);
        assert_eq!(total, 5);
    }

    #[test]
    fn truncate_lines_and_bytes_both_fit() {
        let text = "a\nb\nc";
        let (out, shown, total) = truncate_lines_and_bytes(text, 10, 1000);
        assert_eq!(out, text);
        assert_eq!(shown, 3);
        assert_eq!(total, 3);
    }

    #[test]
    fn truncate_lines_and_bytes_long_lines_stay_under_byte_budget() {
        // Simulate log output: 200-char lines, budget = 10_000 bytes (TOOL_OUTPUT_BUFFER_THRESHOLD).
        let long_line = "x".repeat(200);
        let text: String = (0..100).map(|_| format!("{long_line}\n")).collect();
        let (out, shown, _total) = truncate_lines_and_bytes(&text, 100, 10_000);
        // Must fit within budget.
        assert!(
            out.len() <= 10_000,
            "output {} bytes exceeds budget",
            out.len()
        );
        // Should have shown fewer than 100 lines (200-char lines can't all fit in 10KB).
        assert!(shown < 100, "expected byte truncation, shown={shown}");
    }

    // -- detect_terminal_filter --

    #[test]
    fn terminal_filter_grep() {
        let pos = detect_terminal_filter("cargo build 2>&1 | grep error");
        assert!(pos.is_some());
    }

    #[test]
    fn terminal_filter_head() {
        let pos = detect_terminal_filter("cat big_file.log | head -20");
        assert!(pos.is_some());
    }

    #[test]
    fn terminal_filter_tail() {
        let pos = detect_terminal_filter("journalctl | tail -100");
        assert!(pos.is_some());
    }

    #[test]
    fn terminal_filter_no_pipe() {
        assert!(detect_terminal_filter("cargo build").is_none());
    }

    #[test]
    fn terminal_filter_non_filter_pipe() {
        // Second stage is not a known filter
        assert!(detect_terminal_filter("cat file | cargo install").is_none());
    }

    #[test]
    fn terminal_filter_quoted_pipe_ignored() {
        // Pipe inside quotes is not a real pipe
        assert!(detect_terminal_filter("echo 'foo | bar'").is_none());
    }

    #[test]
    fn terminal_filter_nested_filters_last_wins() {
        // cmd | sed | grep  →  finds the grep (last) pipe
        let cmd = "cat file | sed 's/x/y/' | grep foo";
        let pos = detect_terminal_filter(cmd);
        assert!(pos.is_some());
        // The position should be the second pipe (before grep), not the first
        let pipe_pos = pos.unwrap();
        assert!(cmd[pipe_pos + 1..].trim_start().starts_with("grep"));
    }

    #[test]
    fn terminal_filter_returns_pipe_position() {
        let cmd = "cargo build | grep error";
        let pos = detect_terminal_filter(cmd).unwrap();
        // Character at pipe_pos should be '|'
        assert_eq!(&cmd[pos..pos + 1], "|");
    }

    #[test]
    fn terminal_filter_logical_or_not_a_pipe() {
        // || is logical OR, not a pipeline — should not trigger tee injection
        assert!(detect_terminal_filter("ls || grep foo").is_none());
    }

    #[test]
    fn terminal_filter_path_prefixed_filter() {
        // /usr/bin/grep should be recognized as grep
        let pos = detect_terminal_filter("ls | /usr/bin/grep foo");
        assert!(pos.is_some());
    }

    #[test]
    fn bash_stderr_pipe_not_treated_as_filter() {
        // |& is bash's stderr pipe — the & becomes part of the filter name
        // ("&grep"), which doesn't match any known filter, so correctly returns None
        assert!(detect_terminal_filter("cmd |& grep foo").is_none());
    }
}