ipodderx-rs/src/db.rs

//! SQLite state. Replaces the per-feed .ipxd plists, history.dat and qmcache.dat.

use anyhow::{Context, Result};
use rusqlite::{Connection, OptionalExtension};
use std::path::Path;
use std::sync::Mutex;

/// ponytail: one global connection mutex. Writes here are tiny and rare; move to a
/// spawn_blocking pool if a large feed count ever makes it contend.
pub struct Db {
    conn: Mutex<Connection>,
}

const SCHEMA: &str = "
CREATE TABLE IF NOT EXISTS feeds (
    id            TEXT PRIMARY KEY,
    url           TEXT NOT NULL,
    title         TEXT,
    etag          TEXT,
    last_modified TEXT,
    last_checked  INTEGER,
    ttl_mins      INTEGER,
    last_error    TEXT
);

CREATE TABLE IF NOT EXISTS entries (
    feed_id     TEXT    NOT NULL,
    guid        TEXT    NOT NULL,
    title       TEXT,
    link        TEXT,
    published   INTEGER,
    description TEXT,
    first_seen  INTEGER NOT NULL,
    read        INTEGER NOT NULL DEFAULT 0,
    flagged     INTEGER NOT NULL DEFAULT 0,
    PRIMARY KEY (feed_id, guid)
);

-- url is UNIQUE: this is the dedupe key, and it subsumes the old history.dat pickle.
-- A reaped file keeps its row with path = NULL and state = 'reaped', so a purged
-- episode is never fetched a second time.
CREATE TABLE IF NOT EXISTS enclosures (
    id            INTEGER PRIMARY KEY,
    feed_id       TEXT    NOT NULL,
    guid          TEXT    NOT NULL,
    url           TEXT    NOT NULL UNIQUE,
    mime          TEXT,
    length        INTEGER,
    path          TEXT,
    state         TEXT    NOT NULL,
    bytes_done    INTEGER NOT NULL DEFAULT 0,
    downloaded_at INTEGER,
    last_error    TEXT
);

CREATE INDEX IF NOT EXISTS enclosures_entry ON enclosures (feed_id, guid);
";

/// What `ipx list` shows next to each configured feed.
#[derive(Debug, Default)]
pub struct FeedSummary {
    pub title: Option<String>,
    pub last_checked: Option<i64>,
    pub last_error: Option<String>,
    pub entries: i64,
    pub downloaded: i64,
}

impl Db {
    pub fn open(path: &Path) -> Result<Self> {
        if let Some(dir) = path.parent() {
            std::fs::create_dir_all(dir)
                .with_context(|| format!("creating {}", dir.display()))?;
        }
        let conn = Connection::open(path)
            .with_context(|| format!("opening {}", path.display()))?;
        conn.pragma_update(None, "journal_mode", "WAL")?;
        conn.pragma_update(None, "foreign_keys", "ON")?;
        conn.pragma_update(None, "busy_timeout", 5000)?;
        conn.execute_batch(SCHEMA).context("creating schema")?;
        Ok(Self { conn: Mutex::new(conn) })
    }

    /// In-memory database, for tests.
    #[cfg(test)]
    pub fn memory() -> Result<Self> {
        let conn = Connection::open_in_memory()?;
        conn.execute_batch(SCHEMA)?;
        Ok(Self { conn: Mutex::new(conn) })
    }

    #[cfg(test)]
    pub fn exec_for_test(&self, sql: &str) -> Result<()> {
        self.conn.lock().unwrap().execute_batch(sql)?;
        Ok(())
    }

    pub fn feed_summary(&self, feed_id: &str) -> Result<FeedSummary> {
        let conn = self.conn.lock().unwrap();
        let mut sum: FeedSummary = conn
            .query_row(
                "SELECT title, last_checked, last_error FROM feeds WHERE id = ?1",
                [feed_id],
                |r| {
                    Ok(FeedSummary {
                        title: r.get(0)?,
                        last_checked: r.get(1)?,
                        last_error: r.get(2)?,
                        ..Default::default()
                    })
                },
            )
            .optional()?
            .unwrap_or_default();

        sum.entries = conn.query_row(
            "SELECT count(*) FROM entries WHERE feed_id = ?1",
            [feed_id],
            |r| r.get(0),
        )?;
        sum.downloaded = conn.query_row(
            "SELECT count(*) FROM enclosures WHERE feed_id = ?1 AND path IS NOT NULL",
            [feed_id],
            |r| r.get(0),
        )?;
        Ok(sum)
    }
}


/// Everything `fetch` needs to decide whether to poll a feed, and how.
#[derive(Debug, Default)]
pub struct HttpState {
    pub etag: Option<String>,
    pub last_modified: Option<String>,
    pub last_checked: Option<i64>,
    pub ttl_mins: Option<u64>,
}

impl Db {
    pub fn http_state(&self, feed_id: &str) -> Result<HttpState> {
        let conn = self.conn.lock().unwrap();
        Ok(conn
            .query_row(
                "SELECT etag, last_modified, last_checked, ttl_mins FROM feeds WHERE id = ?1",
                [feed_id],
                |r| {
                    Ok(HttpState {
                        etag: r.get(0)?,
                        last_modified: r.get(1)?,
                        last_checked: r.get(2)?,
                        ttl_mins: r.get::<_, Option<i64>>(3)?.map(|t| t.max(0) as u64),
                    })
                },
            )
            .optional()?
            .unwrap_or_default())
    }

    /// Upsert after a successful poll. Clears any previous error.
    pub fn record_feed(
        &self,
        feed_id: &str,
        url: &str,
        title: Option<&str>,
        etag: Option<&str>,
        last_modified: Option<&str>,
        ttl_mins: Option<u64>,
    ) -> Result<()> {
        let conn = self.conn.lock().unwrap();
        conn.execute(
            "INSERT INTO feeds (id, url, title, etag, last_modified, last_checked, ttl_mins, last_error)
             VALUES (?1, ?2, ?3, ?4, ?5, ?6, ?7, NULL)
             ON CONFLICT(id) DO UPDATE SET
                url = excluded.url,
                title = coalesce(excluded.title, feeds.title),
                etag = excluded.etag,
                last_modified = excluded.last_modified,
                last_checked = excluded.last_checked,
                ttl_mins = excluded.ttl_mins,
                last_error = NULL",
            rusqlite::params![feed_id, url, title, etag, last_modified, now(), ttl_mins.map(|t| t as i64)],
        )?;
        Ok(())
    }

    /// 304, or any other poll that produced no new data: only the clock moves.
    pub fn touch_feed(&self, feed_id: &str, url: &str) -> Result<()> {
        let conn = self.conn.lock().unwrap();
        conn.execute(
            "INSERT INTO feeds (id, url, last_checked) VALUES (?1, ?2, ?3)
             ON CONFLICT(id) DO UPDATE SET last_checked = excluded.last_checked, last_error = NULL",
            rusqlite::params![feed_id, url, now()],
        )?;
        Ok(())
    }

    pub fn set_feed_error(&self, feed_id: &str, url: &str, msg: &str) -> Result<()> {
        let conn = self.conn.lock().unwrap();
        conn.execute(
            "INSERT INTO feeds (id, url, last_checked, last_error) VALUES (?1, ?2, ?3, ?4)
             ON CONFLICT(id) DO UPDATE SET last_checked = excluded.last_checked, last_error = excluded.last_error",
            rusqlite::params![feed_id, url, now(), msg],
        )?;
        Ok(())
    }

    /// Returns true when this entry had not been seen before.
    ///
    /// A changed description or title flips `read` back to 0, which is what the original's
    /// textDiff dance was ultimately for -- minus the diff markup, which the UI can do.
    pub fn record_entry(&self, feed_id: &str, e: &crate::feed::Entry) -> Result<bool> {
        let conn = self.conn.lock().unwrap();
        let inserted = conn.execute(
            "INSERT OR IGNORE INTO entries
                (feed_id, guid, title, link, published, description, first_seen, read, flagged)
             VALUES (?1, ?2, ?3, ?4, ?5, ?6, ?7, 0, 0)",
            rusqlite::params![feed_id, e.guid, e.title, e.link, e.published, e.description, now()],
        )?;
        if inserted == 0 {
            // The SET expressions see the pre-update row, so this compares old vs new.
            conn.execute(
                "UPDATE entries SET
                    title = coalesce(?3, title),
                    description = coalesce(?4, description),
                    read = CASE WHEN description IS NOT ?4 OR title IS NOT ?3 THEN 0 ELSE read END
                 WHERE feed_id = ?1 AND guid = ?2",
                rusqlite::params![feed_id, e.guid, e.title, e.description],
            )?;
        }
        Ok(inserted == 1)
    }

    /// Returns true when this enclosure URL is new. False means we have downloaded it
    /// before, or deliberately reaped it -- either way it is not fetched again.

    pub fn mark_downloaded(&self, url: &str, path: &std::path::Path, bytes: u64) -> Result<()> {
        let conn = self.conn.lock().unwrap();
        conn.execute(
            "UPDATE enclosures SET state = 'done', path = ?2, bytes_done = ?3,
                downloaded_at = ?4, last_error = NULL WHERE url = ?1",
            rusqlite::params![url, path.to_string_lossy(), bytes as i64, now()],
        )?;
        Ok(())
    }

    /// The row stays -- a failed URL is still a URL we have seen. `state` says why it has
    /// no file, and a retry is an explicit act rather than something a rescan does silently.
    pub fn mark_enclosure(&self, url: &str, state: &str, error: Option<&str>) -> Result<()> {
        let conn = self.conn.lock().unwrap();
        conn.execute(
            "UPDATE enclosures SET state = ?2, last_error = ?3 WHERE url = ?1",
            rusqlite::params![url, state, error],
        )?;
        Ok(())
    }

    pub fn record_enclosure(
        &self,
        feed_id: &str,
        guid: &str,
        enc: &crate::feed::Enclosure,
    ) -> Result<bool> {
        let conn = self.conn.lock().unwrap();
        let inserted = conn.execute(
            "INSERT OR IGNORE INTO enclosures (feed_id, guid, url, mime, length, state)
             VALUES (?1, ?2, ?3, ?4, ?5, 'pending')",
            rusqlite::params![feed_id, guid, enc.url, enc.mime, enc.length],
        )?;
        Ok(inserted == 1)
    }
}


/// An enclosure waiting to be downloaded.
#[derive(Debug)]
pub struct Pending {
    pub url: String,
    pub mime: Option<String>,
}

impl Db {
    /// The download queue is the table, not the parse result: an enclosure held back by
    /// `max_new_per_check` is simply picked up by the next scan, in feed order.
    pub fn pending(&self, feed_id: &str, limit: usize) -> Result<Vec<Pending>> {
        let conn = self.conn.lock().unwrap();
        let mut stmt = conn.prepare(
            "SELECT url, mime FROM enclosures
             WHERE feed_id = ?1 AND state = 'pending' ORDER BY id LIMIT ?2",
        )?;
        let rows = stmt
            .query_map(rusqlite::params![feed_id, limit as i64], |r| {
                Ok(Pending { url: r.get(0)?, mime: r.get(1)? })
            })?
            .collect::<rusqlite::Result<Vec<_>>>()?;
        Ok(rows)
    }
}


/// A downloaded file the reaper may consider.
#[derive(Debug, Clone, PartialEq)]
pub struct Candidate {
    pub id: i64,
    pub url: String,
    pub path: String,
    pub bytes: i64,
    /// When it landed; the reaper works oldest-first.
    pub age_key: i64,
    pub read: bool,
}

impl Db {
    /// Files on disk, flagged ones excluded, read before unread and oldest first within
    /// each group.
    ///
    /// The Python intended `read = 1 AND flagged = 0` but never achieved it (a missing
    /// plistlib import and an `EntreiesData` typo meant the filter always threw). Requiring
    /// `read = 1` outright would be just as dead here, since nothing marks episodes read
    /// until a UI exists -- so `flagged` is the keep-forever marker, and `read` only decides
    /// what goes first.
    pub fn reap_candidates(&self) -> Result<Vec<Candidate>> {
        let conn = self.conn.lock().unwrap();
        let mut stmt = conn.prepare(
            "SELECT e.id, e.url, e.path, e.bytes_done,
                    coalesce(e.downloaded_at, 0), coalesce(n.read, 0), coalesce(n.flagged, 0)
             FROM enclosures e
             LEFT JOIN entries n ON n.feed_id = e.feed_id AND n.guid = e.guid
             WHERE e.path IS NOT NULL AND coalesce(n.flagged, 0) = 0
             ORDER BY coalesce(n.read, 0) DESC, coalesce(e.downloaded_at, 0) ASC, e.id ASC",
        )?;
        let rows = stmt
            .query_map([], |r| {
                Ok(Candidate {
                    id: r.get(0)?,
                    url: r.get(1)?,
                    path: r.get(2)?,
                    bytes: r.get(3)?,
                    age_key: r.get(4)?,
                    read: r.get::<_, i64>(5)? != 0,
                })
            })?
            .collect::<rusqlite::Result<Vec<_>>>()?;
        Ok(rows)
    }

    /// The row survives the file: that is what stops a reaped episode being re-downloaded.
    pub fn mark_reaped(&self, id: i64) -> Result<()> {
        let conn = self.conn.lock().unwrap();
        conn.execute(
            "UPDATE enclosures SET state = 'reaped', path = NULL WHERE id = ?1",
            [id],
        )?;
        Ok(())
    }

    /// Rows claiming a file that is no longer there (someone deleted it by hand).
    pub fn missing_files(&self) -> Result<Vec<(i64, String)>> {
        let conn = self.conn.lock().unwrap();
        let mut stmt =
            conn.prepare("SELECT id, path FROM enclosures WHERE path IS NOT NULL")?;
        let rows = stmt
            .query_map([], |r| Ok((r.get(0)?, r.get::<_, String>(1)?)))?
            .collect::<rusqlite::Result<Vec<_>>>()?;
        Ok(rows
            .into_iter()
            .filter(|(_, p)| !std::path::Path::new(p).exists())
            .collect())
    }

    /// Old entries that never had a file, or no longer have one. Enclosure rows stay --
    /// they are the dedupe history.
    pub fn prune_entries(&self, older_than: i64) -> Result<usize> {
        let conn = self.conn.lock().unwrap();
        let n = conn.execute(
            "DELETE FROM entries WHERE flagged = 0
               AND coalesce(published, first_seen) < ?1
               AND NOT EXISTS (
                 SELECT 1 FROM enclosures e
                 WHERE e.feed_id = entries.feed_id AND e.guid = entries.guid
                   AND e.path IS NOT NULL)",
            [older_than],
        )?;
        Ok(n)
    }
}

/// Unix seconds. Everything time-shaped in the DB is stored this way.
pub fn now() -> i64 {
    std::time::SystemTime::now()
        .duration_since(std::time::UNIX_EPOCH)
        .map(|d| d.as_secs() as i64)
        .unwrap_or(0)
}

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

    #[test]
    fn schema_is_idempotent_and_summary_handles_unknown_feeds() {
        let db = Db::memory().unwrap();
        // Re-running the schema must not fail: open() does this on every start.
        db.conn.lock().unwrap().execute_batch(SCHEMA).unwrap();

        let sum = db.feed_summary("never-seen").unwrap();
        assert_eq!(sum.last_checked, None);
        assert_eq!(sum.entries, 0);
        assert_eq!(sum.downloaded, 0);
    }

    #[test]
    fn enclosure_url_is_the_dedupe_key() {
        let db = Db::memory().unwrap();
        let conn = db.conn.lock().unwrap();
        let insert = "INSERT INTO enclosures (feed_id, guid, url, state) VALUES ('f', 'g', 'http://x/a.mp3', 'pending')";
        conn.execute(insert, []).unwrap();
        assert!(conn.execute(insert, []).is_err(), "duplicate url must be rejected");
    }
}