dsc/

utils.rs

use anyhow::{Context, Result};
use std::collections::HashMap;
use std::fs;
use std::io::IsTerminal;
use std::path::{Path, PathBuf};

/// Trim trailing slashes from a base URL.
pub fn normalize_baseurl(baseurl: &str) -> String {
    baseurl.trim_end_matches('/').to_string()
}

/// Create a URL-safe slug from arbitrary input.
///
/// Wraps the [`slug`] crate, which transliterates Unicode (so `"Café"`
/// becomes `"cafe"`, Cyrillic and CJK get sensible romanisations) and
/// emits the standard kebab-case shape used across most slug-generating
/// tooling. Returns `"untitled"` when the slug would otherwise be empty
/// (the `slug` crate itself returns an empty string for input that has
/// no transliterable characters).
pub fn slugify(input: &str) -> String {
    let s = slug::slugify(input);
    if s.is_empty() {
        "untitled".to_string()
    } else {
        s
    }
}

/// Ensure a directory exists.
pub fn ensure_dir(path: &Path) -> Result<()> {
    fs::create_dir_all(path).with_context(|| format!("creating {}", path.display()))?;
    Ok(())
}

/// Resolve a topic path from a user-provided path and a topic title.
pub fn resolve_topic_path(
    provided: Option<&Path>,
    title: &str,
    default_dir: &Path,
) -> Result<PathBuf> {
    let filename = format!("{}.md", slugify(title));
    match provided {
        Some(path) if path.exists() && path.is_dir() => Ok(path.join(filename)),
        Some(path) if path.extension().is_some() => Ok(path.to_path_buf()),
        Some(path) => Ok(path.join(filename)),
        None => Ok(default_dir.join(filename)),
    }
}

/// Read a Markdown file.
pub fn read_markdown(path: &Path) -> Result<String> {
    let raw = fs::read_to_string(path).with_context(|| format!("reading {}", path.display()))?;
    Ok(raw)
}

/// Write a Markdown file, creating parent directories if needed.
pub fn write_markdown(path: &Path, content: &str) -> Result<()> {
    if let Some(parent) = path.parent() {
        ensure_dir(parent)?;
    }
    fs::write(path, content).with_context(|| format!("writing {}", path.display()))?;
    Ok(())
}

/// Quote a YAML scalar if it contains characters that would confuse the
/// parser. Keeps simple values unquoted. Shared by every command that
/// writes YAML front matter (`topic pull --full`, `category pull`).
pub fn yaml_scalar(value: &str) -> String {
    let needs_quoting = value.is_empty()
        || value.contains(':')
        || value.contains('#')
        || value.contains('\n')
        || value.starts_with(['-', '?', '!', '&', '*', '|', '>', '@', '`', '%', '\'', '"', '['])
        || value.starts_with("  ");
    if needs_quoting {
        let escaped = value.replace('\\', "\\\\").replace('"', "\\\"");
        format!("\"{}\"", escaped)
    } else {
        value.to_string()
    }
}

/// Split a Markdown document into its leading YAML front matter (if any) and
/// the body that follows.
///
/// Front matter is recognised only when the file's very first line is exactly
/// `---` (an optional leading BOM is tolerated), terminated by a later line
/// that is exactly `---`. The fenced block is parsed shallowly into a flat
/// `key → value` map (one `key: value` pair per line; lines without a colon
/// are ignored) — `dsc` front matter is intentionally shallow (`title`,
/// `topic_id`, `url`, `pulled_at`), so a full YAML parse is unnecessary and a
/// flat scan keeps the body intact.
///
/// Returns `(map, body)`. When there is no recognisable front matter the map
/// is empty and the body is the original content unchanged, so callers can
/// treat "no front matter" and "empty front matter" identically. One blank
/// line separating the closing fence from the body is consumed (it mirrors
/// what the `pull` side writes), giving a stable pull → push round-trip.
///
/// Note the inherent ambiguity shared with Jekyll/Hugo: a file with no front
/// matter whose body genuinely opens with a `---` thematic break followed by
/// another `---` will be misread as front matter. This is accepted; real
/// snapshots written by `dsc` always carry proper front matter.
pub fn strip_frontmatter(raw: &str) -> (HashMap<String, String>, String) {
    let mut map = HashMap::new();
    let text = raw.strip_prefix('\u{feff}').unwrap_or(raw);

    let mut lines = text.lines();
    if lines.next().map(str::trim_end) != Some("---") {
        return (map, raw.to_string());
    }

    let mut body_lines: Vec<&str> = Vec::new();
    let mut closed = false;
    for line in &mut lines {
        if line.trim_end() == "---" {
            closed = true;
            break;
        }
        if let Some((key, value)) = line.split_once(':') {
            map.insert(key.trim().to_string(), unquote_yaml_scalar(value.trim()));
        }
    }

    if !closed {
        // Opening fence with no matching close: not front matter after all.
        return (HashMap::new(), raw.to_string());
    }

    body_lines.extend(lines);
    // Consume a single conventional blank line between fence and body.
    if body_lines.first() == Some(&"") {
        body_lines.remove(0);
    }
    let mut body = body_lines.join("\n");
    if raw.ends_with('\n') && !body.is_empty() {
        body.push('\n');
    }
    (map, body)
}

/// Inverse of [`yaml_scalar`]'s quoting: if `value` is wrapped in double
/// quotes, strip them and unescape `\"` and `\\`. Bare values pass through
/// unchanged, so a value Discourse never sees as quoted (an integer, a URL)
/// is untouched.
fn unquote_yaml_scalar(value: &str) -> String {
    let bytes = value.as_bytes();
    if bytes.len() < 2 || bytes[0] != b'"' || bytes[bytes.len() - 1] != b'"' {
        return value.to_string();
    }
    let inner = &value[1..value.len() - 1];
    let mut out = String::with_capacity(inner.len());
    let mut chars = inner.chars();
    while let Some(c) = chars.next() {
        if c == '\\' {
            match chars.next() {
                Some('"') => out.push('"'),
                Some('\\') => out.push('\\'),
                Some(other) => {
                    out.push('\\');
                    out.push(other);
                }
                None => out.push('\\'),
            }
        } else {
            out.push(c);
        }
    }
    out
}

/// Current time in `YYYY-MM-DDTHH:MM:SSZ` form, derived directly from
/// `SystemTime` to avoid a chrono dependency where one is not otherwise
/// needed. Used for the `pulled_at` front-matter stamp.
pub fn current_utc_iso8601() -> String {
    use std::time::{SystemTime, UNIX_EPOCH};
    let secs = SystemTime::now()
        .duration_since(UNIX_EPOCH)
        .map(|d| d.as_secs())
        .unwrap_or(0);
    // Days-from-epoch arithmetic (proleptic Gregorian via the standard
    // 1970-01-01 epoch). Good for any year `dsc` will plausibly run in.
    let days = (secs / 86_400) as i64;
    let secs_of_day = secs % 86_400;
    let hh = secs_of_day / 3600;
    let mm = (secs_of_day % 3600) / 60;
    let ss = secs_of_day % 60;
    let (y, m, d) = civil_from_days(days);
    format!("{:04}-{:02}-{:02}T{:02}:{:02}:{:02}Z", y, m, d, hh, mm, ss)
}

/// Convert days-from-1970-01-01 to (year, month, day).
/// Reference: Howard Hinnant, "chrono-Compatible Low-Level Date Algorithms".
fn civil_from_days(z: i64) -> (i32, u32, u32) {
    let z = z + 719_468;
    let era = if z >= 0 { z } else { z - 146_096 } / 146_097;
    let doe = (z - era * 146_097) as u64; // [0, 146096]
    let yoe = (doe - doe / 1460 + doe / 36_524 - doe / 146_096) / 365; // [0, 399]
    let y = yoe as i64 + era * 400;
    let doy = doe - (365 * yoe + yoe / 4 - yoe / 100); // [0, 365]
    let mp = (5 * doy + 2) / 153; // [0, 11]
    let d = doy - (153 * mp + 2) / 5 + 1; // [1, 31]
    let m = if mp < 10 { mp + 3 } else { mp - 9 }; // [1, 12]
    let y = y + if m <= 2 { 1 } else { 0 };
    (y as i32, m as u32, d as u32)
}

fn color_mode() -> &'static str {
    match std::env::var("DSC_COLOR") {
        Ok(value) => match value.trim().to_ascii_lowercase().as_str() {
            "always" => "always",
            "never" => "never",
            _ => "auto",
        },
        Err(_) => "auto",
    }
}

fn color_allowed_for_stdout() -> bool {
    if std::env::var_os("NO_COLOR").is_some() {
        return false;
    }
    match color_mode() {
        "always" => true,
        "never" => false,
        _ => std::io::stdout().is_terminal(),
    }
}

fn discourse_color_code(key: &str) -> u8 {
    const COLORS: [u8; 12] = [31, 32, 33, 34, 35, 36, 91, 92, 93, 94, 95, 96];
    let hash = key.bytes().fold(0usize, |acc, b| {
        acc.wrapping_mul(31).wrapping_add(b as usize)
    });
    COLORS[hash % COLORS.len()]
}

pub fn color_discourse_label(label: &str, key: &str) -> String {
    if !color_allowed_for_stdout() {
        return label.to_string();
    }
    let code = discourse_color_code(key);
    format!("\x1b[1;{}m{}\x1b[0m", code, label)
}

/// Parse a `--since`-style value. Accepts either a relative duration
/// (`7d`, `24h`, `30m`, `1w`, `90s`) or an ISO-8601 absolute timestamp
/// (`2026-04-01`, `2026-04-01T12:00:00Z`). Returns the resulting cutoff
/// instant (now - duration, or the ISO value itself).
pub fn parse_since_cutoff(input: &str) -> anyhow::Result<chrono::DateTime<chrono::Utc>> {
    use anyhow::anyhow;
    let trimmed = input.trim();
    if trimmed.is_empty() {
        return Err(anyhow!("empty --since value"));
    }

    if let Some(duration) = parse_relative_duration(trimmed) {
        return Ok(chrono::Utc::now() - duration);
    }

    // Try RFC3339 (full timestamp).
    if let Ok(dt) = chrono::DateTime::parse_from_rfc3339(trimmed) {
        return Ok(dt.with_timezone(&chrono::Utc));
    }
    // Try date-only — treat as midnight UTC.
    if let Ok(d) = chrono::NaiveDate::parse_from_str(trimmed, "%Y-%m-%d") {
        return Ok(
            chrono::NaiveDateTime::new(d, chrono::NaiveTime::from_hms_opt(0, 0, 0).unwrap())
                .and_utc(),
        );
    }

    Err(anyhow!(
        "unrecognised --since value: {:?} (expected e.g. `7d`, `24h`, `30m`, `1w`, or an ISO-8601 timestamp)",
        input
    ))
}

/// Parse a relative duration like `7d`, `24h`, `1w`, `1m`, `90s`, `1y`.
///
/// Calendar units (`m`, `y`) are imprecise; for windows we use these
/// conventions:
///
/// - `s` — seconds
/// - `min` — minutes (use this rather than `m` to avoid the months-vs-minutes
///   ambiguity)
/// - `h` — hours
/// - `d` — days
/// - `w` — weeks (= 7 days)
/// - `m` — **months** (= 30 days; matches what most users mean by "1m" in
///   analytics windows)
/// - `y` — years (= 365 days)
///
/// For exact calendar math, pass an ISO-8601 timestamp instead.
pub fn parse_relative_duration(input: &str) -> Option<chrono::Duration> {
    let s = input.trim();
    if s.len() < 2 {
        return None;
    }
    // Order matters: `min` must be tried before `m` so we don't read
    // "10min" as "10mi" + "n".
    let multi_char_units = [("min", 60i64)];
    for (suffix, secs_per_unit) in multi_char_units {
        if let Some(digits) = s.strip_suffix(suffix) {
            let n: i64 = digits.parse().ok()?;
            return Some(chrono::Duration::seconds(n * secs_per_unit));
        }
    }
    let (digits, unit) = s.split_at(s.len() - 1);
    let n: i64 = digits.parse().ok()?;
    match unit {
        "s" => Some(chrono::Duration::seconds(n)),
        "h" => Some(chrono::Duration::hours(n)),
        "d" => Some(chrono::Duration::days(n)),
        "w" => Some(chrono::Duration::weeks(n)),
        "m" => Some(chrono::Duration::days(n * 30)),
        "y" => Some(chrono::Duration::days(n * 365)),
        _ => None,
    }
}

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

    #[test]
    fn slugify_simple_ascii() {
        assert_eq!(slugify("Hello World"), "hello-world");
    }

    #[test]
    fn slugify_collapses_runs_of_non_alnum() {
        assert_eq!(slugify("a   b___c!!!d"), "a-b-c-d");
    }

    #[test]
    fn slugify_trims_leading_and_trailing_dashes() {
        assert_eq!(slugify("   hello   "), "hello");
        assert_eq!(slugify("!!!foo!!!"), "foo");
    }

    #[test]
    fn slugify_empty_input_returns_untitled() {
        assert_eq!(slugify(""), "untitled");
        assert_eq!(slugify("   "), "untitled");
        assert_eq!(slugify("!!!"), "untitled");
    }

    #[test]
    fn slugify_preserves_numbers() {
        assert_eq!(slugify("Topic 42 - intro"), "topic-42-intro");
    }

    #[test]
    fn slugify_lowercases() {
        assert_eq!(slugify("ABCxyz"), "abcxyz");
    }

    #[test]
    fn slugify_transliterates_unicode() {
        // The whole reason for adopting the `slug` crate: pre-existing
        // ASCII behaviour preserved, plus accented Latin, Cyrillic, and
        // CJK now produce meaningful slugs instead of "untitled".
        assert_eq!(slugify("Café Tonight"), "cafe-tonight");
        assert_eq!(slugify("Привет мир"), "privet-mir");
        assert_eq!(slugify("日本語"), "ri-ben-yu");
    }

    #[test]
    fn slugify_trims_both_ends_of_dashes() {
        // Regression guard: catches a latent bug from a contributor PR
        // that only trimmed trailing dashes. The `slug` crate handles
        // both ends correctly.
        assert_eq!(slugify("-foo-"), "foo");
        assert_eq!(slugify("---foo---bar---"), "foo-bar");
    }

    #[test]
    fn normalize_baseurl_strips_trailing_slashes() {
        assert_eq!(normalize_baseurl("https://example.com/"), "https://example.com");
        assert_eq!(normalize_baseurl("https://example.com///"), "https://example.com");
        assert_eq!(normalize_baseurl("https://example.com"), "https://example.com");
    }

    #[test]
    fn normalize_baseurl_preserves_no_trailing() {
        assert_eq!(normalize_baseurl(""), "");
    }

    #[test]
    fn resolve_topic_path_uses_title_when_no_path_given() {
        let default_dir = Path::new("/tmp/dsc-test");
        let out = resolve_topic_path(None, "Hello World", default_dir).unwrap();
        assert_eq!(out, default_dir.join("hello-world.md"));
    }

    #[test]
    fn resolve_topic_path_uses_given_path_with_extension() {
        let default_dir = Path::new("/tmp/dsc-test");
        let explicit = Path::new("/tmp/custom.md");
        let out = resolve_topic_path(Some(explicit), "Ignored", default_dir).unwrap();
        assert_eq!(out, explicit);
    }

    #[test]
    fn parse_relative_duration_common_units() {
        assert_eq!(
            parse_relative_duration("7d"),
            Some(chrono::Duration::days(7))
        );
        assert_eq!(
            parse_relative_duration("24h"),
            Some(chrono::Duration::hours(24))
        );
        assert_eq!(
            parse_relative_duration("30min"),
            Some(chrono::Duration::minutes(30))
        );
        assert_eq!(
            parse_relative_duration("1w"),
            Some(chrono::Duration::weeks(1))
        );
        assert_eq!(
            parse_relative_duration("90s"),
            Some(chrono::Duration::seconds(90))
        );
    }

    #[test]
    fn parse_relative_duration_rejects_nonsense() {
        assert!(parse_relative_duration("").is_none());
        assert!(parse_relative_duration("d").is_none());
        assert!(parse_relative_duration("7x").is_none());
        assert!(parse_relative_duration("abc").is_none());
        assert!(parse_relative_duration("3M").is_none()); // case-sensitive
    }

    #[test]
    fn parse_relative_duration_treats_m_as_months() {
        // `m` = months (= 30 days). Users naturally write `1m` for "one
        // month" in analytics windows; we match that. Use `min` for the
        // rare minutes case.
        assert_eq!(
            parse_relative_duration("1m"),
            Some(chrono::Duration::days(30))
        );
        assert_eq!(
            parse_relative_duration("3m"),
            Some(chrono::Duration::days(90))
        );
    }

    #[test]
    fn parse_relative_duration_minutes_via_min_suffix() {
        assert_eq!(
            parse_relative_duration("5min"),
            Some(chrono::Duration::minutes(5))
        );
        assert_eq!(
            parse_relative_duration("90min"),
            Some(chrono::Duration::minutes(90))
        );
    }

    #[test]
    fn parse_relative_duration_accepts_years_as_365d() {
        assert_eq!(
            parse_relative_duration("1y"),
            Some(chrono::Duration::days(365))
        );
        assert_eq!(
            parse_relative_duration("2y"),
            Some(chrono::Duration::days(730))
        );
    }

    #[test]
    fn parse_since_cutoff_iso_date() {
        let cutoff = parse_since_cutoff("2026-01-01").unwrap();
        assert_eq!(cutoff.to_rfc3339(), "2026-01-01T00:00:00+00:00");
    }

    #[test]
    fn parse_since_cutoff_iso_timestamp() {
        let cutoff = parse_since_cutoff("2026-04-15T12:30:00Z").unwrap();
        assert_eq!(cutoff.to_rfc3339(), "2026-04-15T12:30:00+00:00");
    }

    #[test]
    fn parse_since_cutoff_relative_is_in_the_past() {
        let now = chrono::Utc::now();
        let cutoff = parse_since_cutoff("7d").unwrap();
        let diff = now - cutoff;
        // Should be very close to 7 days (within a second).
        assert!(
            (diff - chrono::Duration::days(7)).num_seconds().abs() < 2,
            "expected ~7 day delta, got {}",
            diff
        );
    }

    #[test]
    fn parse_since_cutoff_rejects_garbage() {
        assert!(parse_since_cutoff("not a date").is_err());
        assert!(parse_since_cutoff("").is_err());
    }

    #[test]
    fn yaml_scalar_leaves_simple_values_bare() {
        assert_eq!(yaml_scalar("Dependency management"), "Dependency management");
        assert_eq!(yaml_scalar("Topic 42"), "Topic 42");
    }

    #[test]
    fn yaml_scalar_quotes_when_needed() {
        assert_eq!(yaml_scalar("a: b"), "\"a: b\"");
        assert_eq!(yaml_scalar("# hash"), "\"# hash\"");
        assert_eq!(yaml_scalar("- leading dash"), "\"- leading dash\"");
        // Interior quotes alone do not trigger quoting; a quote that coincides
        // with another trigger (here the colon) is escaped inside the wrap.
        assert_eq!(yaml_scalar("she said \"hi\""), "she said \"hi\"");
        assert_eq!(yaml_scalar("a: \"b\""), "\"a: \\\"b\\\"\"");
    }

    #[test]
    fn strip_frontmatter_parses_block_and_body() {
        let raw = "---\ntitle: Dependency management\ntopic_id: 412\nurl: https://forum.rcpch.tech/t/dependency-management/412\npulled_at: 2026-06-22T09:19:00Z\n---\n\nBody line one.\nBody line two.\n";
        let (front, body) = strip_frontmatter(raw);
        assert_eq!(front.get("topic_id").map(String::as_str), Some("412"));
        assert_eq!(
            front.get("title").map(String::as_str),
            Some("Dependency management")
        );
        assert_eq!(
            front.get("url").map(String::as_str),
            Some("https://forum.rcpch.tech/t/dependency-management/412")
        );
        assert_eq!(body, "Body line one.\nBody line two.\n");
    }

    #[test]
    fn strip_frontmatter_absent_returns_empty_map_and_full_body() {
        let raw = "# Heading\n\nNo front matter here.\n";
        let (front, body) = strip_frontmatter(raw);
        assert!(front.is_empty());
        assert_eq!(body, raw);
    }

    #[test]
    fn strip_frontmatter_unclosed_fence_is_not_front_matter() {
        // Opening `---` but never closed: treat the whole thing as body.
        let raw = "---\ntitle: oops\nstill body, no closing fence\n";
        let (front, body) = strip_frontmatter(raw);
        assert!(front.is_empty());
        assert_eq!(body, raw);
    }

    #[test]
    fn strip_frontmatter_preserves_horizontal_rules_in_body() {
        // A `---` inside the body (after the real close) must survive intact.
        let raw = "---\ntopic_id: 7\n---\n\nIntro.\n\n---\n\nAfter the rule.\n";
        let (front, body) = strip_frontmatter(raw);
        assert_eq!(front.get("topic_id").map(String::as_str), Some("7"));
        assert_eq!(body, "Intro.\n\n---\n\nAfter the rule.\n");
    }

    #[test]
    fn strip_frontmatter_unquotes_yaml_scalar_values() {
        // yaml_scalar quotes a title containing a colon; strip must invert it.
        let title = "Intro: getting started";
        let raw = format!("---\ntitle: {}\ntopic_id: 3\n---\n\nbody\n", yaml_scalar(title));
        let (front, body) = strip_frontmatter(&raw);
        assert_eq!(front.get("title").map(String::as_str), Some(title));
        assert_eq!(front.get("topic_id").map(String::as_str), Some("3"));
        assert_eq!(body, "body\n");
    }

    #[test]
    fn strip_frontmatter_leaves_url_with_colons_intact() {
        // URLs are written bare (not via yaml_scalar) and only the first colon
        // separates key from value, so the scheme colon must survive.
        let raw = "---\nurl: https://forum.rcpch.tech/t/x/9\n---\n\nbody\n";
        let (front, _) = strip_frontmatter(raw);
        assert_eq!(
            front.get("url").map(String::as_str),
            Some("https://forum.rcpch.tech/t/x/9")
        );
    }

    #[test]
    fn strip_frontmatter_tolerates_leading_bom() {
        let raw = "\u{feff}---\ntopic_id: 99\n---\n\nbody\n";
        let (front, body) = strip_frontmatter(raw);
        assert_eq!(front.get("topic_id").map(String::as_str), Some("99"));
        assert_eq!(body, "body\n");
    }

    #[test]
    fn current_utc_iso8601_has_expected_shape() {
        let s = current_utc_iso8601();
        assert_eq!(s.len(), 20, "got {s:?}");
        assert!(s.ends_with('Z'));
        assert_eq!(&s[4..5], "-");
        assert_eq!(&s[10..11], "T");
    }

    #[test]
    fn civil_from_days_matches_known_dates() {
        // 1970-01-01 is day 0.
        assert_eq!(civil_from_days(0), (1970, 1, 1));
        // 2026-06-10 = 20614 days from epoch (well-known via cal / date).
        assert_eq!(civil_from_days(20614), (2026, 6, 10));
        // Leap-day check: 2024-02-29.
        assert_eq!(civil_from_days(19782), (2024, 2, 29));
    }
}