//! DR playlists from the Wayback Machine: archived copies of dr.dk/playlister/{channel}/{date}/… //! (2012–2021, mostly P3 2013–2017). The old pages list each play with DR's track id, the exact //! UTC time, title, artists and DR artist id, the same ids as today, so archived plays join the //! artist pool directly. Kept apart from the live `plays` table (sampled days, not continuous). //! //! airplay collect --only archive use crate::{db::Db, http::Http}; use anyhow::Result; use chrono::Utc; use regex::Regex; use rusqlite::params; use std::collections::BTreeMap; const CHANNELS: &[&str] = &["p3", "p4kbh", "p4", "p5", "p6beat", "p7mix", "p8jazz", "p4fyn", "p4aarhus", "p4nord", "p4vest", "p4syd", "p4sjaelland", "p4trekanten", "p4bornholm", "p4esbjerg", "p4ostjylland", "p4midtvest"]; pub fn create_tables(db: &Db) -> Result<()> { db.conn.execute_batch( "CREATE TABLE IF NOT EXISTS archive_plays ( station TEXT NOT NULL, played_at TEXT NOT NULL, track_id TEXT NOT NULL, artist TEXT NOT NULL, title TEXT NOT NULL, artist_ids TEXT NOT NULL DEFAULT '', page TEXT NOT NULL, recorded TEXT, country TEXT, released TEXT, -- 2012-13 pages: 'Indspillet', country, 'UDGIVET: year, label' PRIMARY KEY (station, played_at, track_id) ); CREATE INDEX IF NOT EXISTS archive_plays_day ON archive_plays (station, substr(played_at, 1, 10)); CREATE TABLE IF NOT EXISTS archive_pages ( url TEXT PRIMARY KEY, snapshot TEXT NOT NULL, plays INTEGER, fetched_at TEXT NOT NULL );", )?; Ok(()) } fn unescape(s: &str) -> String { let s = s.replace("&", "&").replace(""", "\"").replace("'", "'").replace("<", "<").replace(">", ">"); // numeric entities (ø …) let re = Regex::new(r"&#(\d+);").unwrap(); re.replace_all(&s, |c: ®ex::Captures| char::from_u32(c[1].parse().unwrap_or(63)).map(String::from).unwrap_or_default()).into_owned() } pub struct Play { pub at: String, pub track_id: String, pub title: String, pub artists: String, pub artist_ids: String, pub recorded: Option, pub country: Option, pub released: Option } /// The plays on one archived page. Two layouts: from about 2014 each item carries its UTC time /// (data-fulltime) and track id; the 2012–13 pages show the local clock time (the date comes from /// the URL, rolling over at midnight) and add recording year, country and release. pub fn parse(html: &str, page_date: chrono::NaiveDate) -> Vec { let item = Regex::new(r#"(?s)]*itemtype="http://schema.org/MusicRecording"[^>]*>(.*?)"#).unwrap(); let head = Regex::new(r#"(?s)]*)itemtype"#).unwrap(); let full = Regex::new(r#"data-fulltime="(\d{4}-\d{2}-\d{2}) (\d{2}:\d{2}:\d{2})Z""#).unwrap(); let clock = Regex::new(r#"]*>(\d{2}):(\d{2})"#).unwrap(); let tid_attr = Regex::new(r#"data-trackId="([^"]+)""#).unwrap(); let tid_link = Regex::new(r#"/musik/titel/[^/"]+/(\d+-\d+-\d+)"#).unwrap(); let title = Regex::new(r#"(?s)\s*(?:]*>)?([^<]+)"#).unwrap(); let names = Regex::new(r#"([^<]+)"#).unwrap(); let ids = Regex::new(r"kunstner/async/(\d+)/").unwrap(); let details = Regex::new(r#"(?s)
(.*?)
"#).unwrap(); let span = Regex::new(r"([^<]*)").unwrap(); let tz = chrono_tz::Europe::Copenhagen; let (mut day, mut last_min) = (page_date, -1i32); let mut out = Vec::new(); for m in item.captures_iter(html) { let whole = m.get(0).unwrap().as_str(); let body = &m[1]; let attrs = head.captures(whole).map(|c| c[1].to_string()).unwrap_or_default(); let Some(ti) = title.captures(body) else { continue }; let Some(track_id) = tid_attr.captures(&attrs).or_else(|| tid_link.captures(body)).map(|c| c[1].to_string()) else { continue }; let at = if let Some(t) = full.captures(body) { format!("{}T{}Z", &t[1], &t[2]) } else if let Some(c) = clock.captures(body) { let (h, mi): (u32, u32) = (c[1].parse().unwrap_or(0), c[2].parse().unwrap_or(0)); let minute = (h * 60 + mi) as i32; if last_min >= 0 && minute + 6 * 60 < last_min { day = day.succ_opt().unwrap_or(day); } last_min = minute; let Some(local) = day.and_hms_opt(h, mi, 0) else { continue }; use chrono::TimeZone; match tz.from_local_datetime(&local).earliest() { Some(d) => d.with_timezone(&Utc).format("%Y-%m-%dT%H:%M:%SZ").to_string(), None => continue } } else { continue }; let info = body.find("class=\"trackInfo\"").map(|i| &body[i..]).unwrap_or(body); let info = info.find("trackDetails").map(|i| &info[..i]).unwrap_or(info); let artists: Vec = names.captures_iter(info).map(|c| unescape(c[1].trim())).collect(); let artist_ids: Vec = ids.captures_iter(whole).map(|c| c[1].to_string()).collect(); let d: Vec = details.captures(body).map(|c| span.captures_iter(&c[1]).map(|s| unescape(s[1].trim())).collect()).unwrap_or_default(); let get = |p: &str| d.iter().find(|x| x.starts_with(p)).map(|x| x[p.len()..].trim().to_string()); out.push(Play { at, track_id, title: unescape(ti[1].trim()), artists: artists.join(", "), artist_ids: artist_ids.join(","), recorded: get("Indspillet:"), released: get("UDGIVET:"), country: d.iter().find(|x| !x.contains(':') && !x.is_empty()).cloned(), }); } out } pub fn collect(db: &Db, http: &Http, log: &mut Vec) -> Result { create_tables(db)?; let path = Regex::new(r"/playlister/([a-z0-9]+)/(\d{4})-(\d{1,2})-(\d{1,2})")?; // Pages that gave no plays get another try each run (an archive outage looks like an empty page). db.conn.execute("DELETE FROM archive_pages WHERE plays = 0", [])?; let mut pages = 0; for ch in CHANNELS { // The CDX index is often overloaded (503): retry patiently. let url = format!("https://web.archive.org/cdx/search/cdx?url=www.dr.dk/playlister/{ch}/*&output=json&fl=timestamp,original&filter=statuscode:200&collapse=urlkey&limit=50000"); let mut cdx = None; for attempt in 0..8 { match http.get(&url) { Ok(Some(c)) if c.trim_start().starts_with('[') => { cdx = Some(c); break; } Ok(Some(_)) | Ok(None) if attempt >= 2 => break, // answered, but not an index: nothing archived Ok(_) => {} Err(e) => { if attempt == 7 { log.push(format!("archive cdx {ch}: {e}")); } } } std::thread::sleep(std::time::Duration::from_secs(30)); } let Some(cdx) = cdx else { continue }; eprintln!("archive {ch}: index read"); let rows: Vec> = serde_json::from_str(&cdx).unwrap_or_default(); for row in rows.iter().skip(1) { let (ts, url) = (&row[0], &row[1]); if !path.is_match(url) { continue; } let clean = url.replace(":80/", "/").replace("http://", "https://"); let done: i64 = db.conn.query_row("SELECT count(*) FROM archive_pages WHERE url = ?1", [&clean], |r| r.get(0))?; if done > 0 { continue; } let html = match http.get(&format!("https://web.archive.org/web/{ts}id_/{url}")) { Ok(h) => h.unwrap_or_default(), Err(e) => { log.push(format!("archive {url}: {e}")); continue; } }; if html.is_empty() || html.contains("Internet Archive: Temporarily Offline") { std::thread::sleep(std::time::Duration::from_secs(60)); continue; } let Some(c) = path.captures(url) else { continue }; let Some(page_date) = chrono::NaiveDate::from_ymd_opt(c[2].parse().unwrap_or(0), c[3].parse().unwrap_or(0), c[4].parse().unwrap_or(0)) else { continue }; let plays = parse(&html, page_date); let tx = db.conn.unchecked_transaction()?; for p in &plays { tx.execute("INSERT OR IGNORE INTO archive_plays VALUES (?1, ?2, ?3, ?4, ?5, ?6, ?7, ?8, ?9, ?10)", params![ch, p.at, p.track_id, p.artists, p.title, p.artist_ids, clean, p.recorded, p.country, p.released])?; } tx.execute("INSERT OR REPLACE INTO archive_pages VALUES (?1, ?2, ?3, ?4)", params![clean, ts, plays.len() as i64, Utc::now().to_rfc3339()])?; tx.commit()?; pages += 1; } } Ok(pages) } /// Archived days per channel and how much of each day is covered (hours with at least one play). #[allow(dead_code)] pub fn coverage(db: &Db) -> Result> { create_tables(db)?; let mut out = BTreeMap::new(); for r in db.conn.prepare("SELECT station, substr(played_at, 1, 10), count(DISTINCT substr(played_at, 12, 2)) FROM archive_plays GROUP BY 1, 2")? .query_map([], |r| Ok((r.get::<_, String>(0)?, r.get::<_, String>(1)?, r.get::<_, i64>(2)?)))? { let (s, d, h) = r?; out.insert((s, d), h as usize); } Ok(out) }