build: bump version to 0.7.15

fixes: https://todo.sr.ht/~gheartsfield/nostr-rs-relay/33

.pre-commit-config.yaml

@@ -11,6 +11,6 @@ repos:
 - repo: https://github.com/doublify/pre-commit-rust
   rev: v1.0
   hooks:
-  - id: fmt
+#  - id: fmt
   - id: cargo-check
   - id: clippy

Cargo.lock

@@ -1096,7 +1096,7 @@ checksum = "38bf9645c8b145698bb0b18a4637dcacbc421ea49bef2317e4fd8065a387cf21"
 
 [[package]]
 name = "nostr-rs-relay"
-version = "0.7.13"
+version = "0.7.15"
 dependencies = [
  "anyhow",
  "bitcoin_hashes",

Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "nostr-rs-relay"
-version = "0.7.13"
+version = "0.7.15"
 edition = "2021"
 authors = ["Greg Heartsfield <scsibug@imap.cc>"]
 description = "A relay implementation for the Nostr protocol"
@@ -28,7 +28,7 @@ secp256k1 = {version = "0.21", features = ["rand", "rand-std", "serde", "bitcoin
 serde = { version = "1.0", features = ["derive"] }
 serde_json = {version = "1.0", features = ["preserve_order"]}
 hex = "0.4"
-rusqlite = { version = "0.26", features = ["limits","bundled"]}
+rusqlite = { version = "0.26", features = ["limits","bundled","modern_sqlite", "trace"]}
 r2d2 = "0.8"
 r2d2_sqlite = "0.19"
 lazy_static = "1.4"

Dockerfile

@@ -1,4 +1,4 @@
-FROM docker.io/library/rust:1.66.0@sha256:359949280cebefe93ccb33089fe25111a3aadfe99eac4b6cbe8ec3e1b571dacb as builder
+FROM docker.io/library/rust:1.66.0 as builder
 
 RUN USER=root cargo install cargo-auditable
 RUN USER=root cargo new --bin nostr-rs-relay
@@ -17,7 +17,7 @@ COPY ./src ./src
 RUN rm ./target/release/deps/nostr*relay*
 RUN cargo auditable build --release --locked
 
-FROM docker.io/library/debian:bullseye-20221205-slim@sha256:25f10b4f1ded5341a3ca0a30290ff3cd5639415f0c5a2222d5e7d5dd72952aa1
+FROM docker.io/library/debian:bullseye-slim
 
 ARG APP=/usr/src/app
 ARG APP_DATA=/usr/src/app/db

README.md

@@ -29,6 +29,7 @@ mirrored on [GitHub](https://github.com/scsibug/nostr-rs-relay).
 - [x] NIP-20: [Command Results](https://github.com/nostr-protocol/nips/blob/master/20.md)
 - [x] NIP-22: [Event `created_at` limits](https://github.com/nostr-protocol/nips/blob/master/22.md) (_future-dated events only_)
 - [x] NIP-26: [Event Delegation](https://github.com/nostr-protocol/nips/blob/master/26.md)
+- [x] NIP-28: [Public Chat](https://github.com/nostr-protocol/nips/blob/master/28.md)
 
 ## Quick Start
 
@@ -81,6 +82,38 @@ Text Note [81cf...2652] from 296a...9b92 5 seconds ago
 A pre-built container is also available on DockerHub:
 https://hub.docker.com/r/scsibug/nostr-rs-relay
 
+## Build and Run (without Docker)
+
+Building `nostr-rs-relay` requires an installation of Cargo & Rust: https://www.rust-lang.org/tools/install
+
+Clone this repository, and then build a release version of the relay:
+
+```console
+$ git clone -q https://git.sr.ht/\~gheartsfield/nostr-rs-relay
+$ cd nostr-rs-relay
+$ cargo build -q -r
+```
+
+The relay executable is now located in
+`target/release/nostr-rs-relay`.  In order to run it with logging
+enabled, execute it with the `RUST_LOG` variable set:
+
+```console
+$ RUST_LOG=warn,nostr_rs_relay=info ./target/release/nostr-rs-relay
+Dec 26 10:31:56.455  INFO nostr_rs_relay: Starting up from main
+Dec 26 10:31:56.464  INFO nostr_rs_relay::server: listening on: 0.0.0.0:8080
+Dec 26 10:31:56.466  INFO nostr_rs_relay::server: db writer created
+Dec 26 10:31:56.466  INFO nostr_rs_relay::db: Built a connection pool "event writer" (min=1, max=2)
+Dec 26 10:31:56.466  INFO nostr_rs_relay::db: opened database "./nostr.db" for writing
+Dec 26 10:31:56.466  INFO nostr_rs_relay::schema: DB version = 11
+Dec 26 10:31:56.467  INFO nostr_rs_relay::db: Built a connection pool "maintenance writer" (min=1, max=2)
+Dec 26 10:31:56.467  INFO nostr_rs_relay::server: control message listener started
+Dec 26 10:31:56.468  INFO nostr_rs_relay::db: Built a connection pool "client query" (min=4, max=8)
+```
+
+You now have a running relay, on port `8080`.  Use a `nostr` client or
+`websocat` to connect and send/query for events.
+
 ## Configuration
 
 The sample [`config.toml`](config.toml) file demonstrates the

config.toml

@@ -36,8 +36,9 @@ data_directory = "."
 # Minimum number of SQLite reader connections
 #min_conn = 4
 
-# Maximum number of SQLite reader connections
+# Maximum number of SQLite reader connections.  Recommend setting this
-#max_conn = 128
+# to approx the number of cores.
+#max_conn = 8
 
 [network]
 # Bind to this network address

docs/database-maintenance.md

@@ -0,0 +1,125 @@
+# Database Maintenance
+
+`nostr-rs-relay` uses the SQLite embedded database to minimize
+dependencies and overall footprint of running a relay.  If traffic is
+light, the relay should just run with very little need for
+intervention.  For heavily trafficked relays, there are a number of
+steps that the operator may need to take to maintain performance and
+limit disk usage.
+
+This maintenance guide is current as of version `0.7.14`.  Future
+versions may incorporate and automate some of these steps.
+
+## Backing Up the Database
+
+To prevent data loss, the database should be backed up regularly.  The
+recommended method is to use the `sqlite3` command to perform an
+"Online Backup".  This can be done while the relay is running, queries
+can still run and events will be persisted during the backup.
+
+The following commands will perform a backup of the database to a
+dated file, and then compress to minimize size:
+
+```console
+BACKUP_FILE=/var/backups/nostr/`date +%Y%m%d_%H%M`.db
+sqlite3 -readonly /apps/nostr-relay/nostr.db ".backup $BACKUP_FILE
+sqlite3 $BACKUP_FILE "vacuum;"
+bzip2 -9 $BACKUP_FILE
+```
+
+Nostr events are very compressible.  Expect a compression ratio on the
+order of 4:1, resulting in a 75% space saving.
+
+## Vacuuming the Database
+
+As the database is updated, it can become fragmented.  Performing a
+full `vacuum` will rebuild the entire database file, and can reduce
+space.  Running this may reduce the size of the database file,
+especially if a large amount of data was updated or deleted.
+
+```console
+vacuum;
+```
+
+## Clearing Hidden Events
+
+When events are deleted, either through deletion events, metadata or
+follower updates, or a replaceable event kind, the event is not
+actually removed from the database.  Instead, a flag `HIDDEN` is set
+to true for the event, which excludes it from search results.  The
+original intent was to ensure that subsequent rebroadcasts of the
+event would be easily detected as having been deleted, and would not
+need to be stored again.  In practice, this decision causes excessive
+growth of the `tags` table, since all the previous followers are
+retained for those `HIDDEN` events.
+
+The `event` and especially the `tag` table can be significantly
+reduced in size by running these commands:
+
+```console
+PRAGMA foreign_keys = ON;
+delete from event where HIDDEN=true;
+```
+
+## Manually Removing Events
+
+For a variety of reasons, an operator may wish to remove some events
+from the database.  The only way of achieving this today is with
+manually run SQL commands.
+
+It is recommended to have a good backup prior to manually running SQL
+commands!
+
+In all cases, it is mandatory to enable foreign keys, and this must be
+done for every connection.  Otherwise, you will likely orphan rows in
+the `tag` table.
+
+### Deleting Specific Event
+
+```console
+PRAGMA foreign_keys = ON;
+delete from event where event_hash=x'00000000000c1271675dc86e3e1dd1336827bccabb90dc4c9d3b4465efefe00e';
+```
+
+### Deleting All Events for Pubkey
+
+```console
+PRAGMA foreign_keys = ON;
+delete from event where author=x'000000000002c7831d9c5a99f183afc2813a6f69a16edda7f6fc0ed8110566e6';
+```
+
+### Deleting All Events of a Kind
+
+
+```console
+PRAGMA foreign_keys = ON;
+delete from event where kind=70202;
+```
+
+### Deleting Old Events
+
+In this scenario, we wish to delete any event that has been stored by
+our relay for more than 1 month.  Crucially, this is based on when the
+event was stored, not when the event says it was created.  If an event
+has a `created` field of 2 years ago, but was first sent to our relay
+yesterday, it would not be deleted in this scenario.  Keep in mind, we
+do not track anything for re-broadcast events that we already have, so
+this is not a very effective way of implementing a "least recently
+seen" policy.
+
+```console
+PRAGMA foreign_keys = ON;
+TODO!
+```
+
+### Delete Profile Events with No Recent Events
+
+Many users create profiles, post a "hello world" event, and then never
+appear again (likely using an ephemeral keypair that was lost in the
+browser cache).  We can find these accounts and remove them after some
+time.
+
+```console
+PRAGMA foreign_keys = ON;
+TODO!
+```

mk-platform-agnostic-dockerfile.sh

@@ -1,3 +0,0 @@
-#!/usr/bin/env bash
-sed -E 's/@sha256:[[:alnum:]]+//g' Dockerfile > Dockerfile.any-platform
-echo "Created platform-agnostic Dockerfile in 'Dockerfile.any-platform'"

src/config.rs

@@ -207,7 +207,7 @@ impl Default for Settings {
                 data_directory: ".".to_owned(),
                 in_memory: false,
                 min_conn: 4,
-                max_conn: 128,
+                max_conn: 8,
             },
             network: Network {
                 port: 8080,

src/db.rs

@@ -19,8 +19,10 @@ use r2d2_sqlite::SqliteConnectionManager;
 use rusqlite::params;
 use rusqlite::types::ToSql;
 use rusqlite::OpenFlags;
+use tokio::sync::{Mutex, MutexGuard};
 use std::fmt::Write as _;
 use std::path::Path;
+use std::sync::Arc;
 use std::thread;
 use std::time::Duration;
 use std::time::Instant;
@@ -39,9 +41,8 @@ pub struct SubmittedEvent {
 /// Database file
 pub const DB_FILE: &str = "nostr.db";
 
-/// How frequently to run maintenance
+/// How frequently to attempt checkpointing
-/// How many persisted events before DB maintenannce is triggered.
+pub const CHECKPOINT_FREQ_SEC: u64 = 60;
-pub const EVENT_MAINTENANCE_FREQ_SEC: u64 = 60;
 
 /// How many persisted events before we pause for backups.
 /// It isn't clear this is enough to make the online backup API work yet.
@@ -94,6 +95,16 @@ pub fn build_pool(
     pool
 }
 
+/// Display database pool stats every 1 minute
+pub async fn monitor_pool(name: &str, pool: SqlitePool) {
+    let sleep_dur = Duration::from_secs(60);
+    loop {
+	log_pool_stats(name, &pool);
+	tokio::time::sleep(sleep_dur).await;
+    }
+}
+
+
 /// Perform normal maintenance
 pub fn optimize_db(conn: &mut PooledConnection) -> Result<()> {
     let start = Instant::now();
@@ -102,15 +113,15 @@ pub fn optimize_db(conn: &mut PooledConnection) -> Result<()> {
     Ok(())
 }
 #[derive(Debug)]
-enum SqliteReturnStatus {
+enum SqliteStatus {
-    SqliteOk,
+    Ok,
-    SqliteBusy,
+    Busy,
-    SqliteError,
+    Error,
-    SqliteOther(u64),
+    Other(u64),
 }
 
-/// Checkpoint/Truncate WAL
+/// Checkpoint/Truncate WAL.  Returns the number of WAL pages remaining.
-pub fn checkpoint_db(conn: &mut PooledConnection) -> Result<()> {
+pub fn checkpoint_db(conn: &mut PooledConnection) -> Result<usize> {
     let query = "PRAGMA wal_checkpoint(TRUNCATE);";
     let start = Instant::now();
     let (cp_result, wal_size, _frames_checkpointed) = conn.query_row(query, [], |row| {
@@ -120,10 +131,10 @@ pub fn checkpoint_db(conn: &mut PooledConnection) -> Result<()> {
         Ok((checkpoint_result, wal_size, frames_checkpointed))
     })?;
     let result = match cp_result {
-        0 => SqliteReturnStatus::SqliteOk,
+        0 => SqliteStatus::Ok,
-        1 => SqliteReturnStatus::SqliteBusy,
+        1 => SqliteStatus::Busy,
-        2 => SqliteReturnStatus::SqliteError,
+        2 => SqliteStatus::Error,
-        x => SqliteReturnStatus::SqliteOther(x),
+        x => SqliteStatus::Other(x),
     };
     info!(
         "checkpoint ran in {:?} (result: {:?}, WAL size: {})",
@@ -131,7 +142,7 @@ pub fn checkpoint_db(conn: &mut PooledConnection) -> Result<()> {
         result,
         wal_size
     );
-    Ok(())
+    Ok(wal_size as usize)
 }
 
 /// Spawn a database writer that persists events to the SQLite store.
@@ -399,7 +410,7 @@ pub fn write_event(conn: &mut PooledConnection, e: &Event) -> Result<usize> {
     // if this event is replaceable update, hide every other replaceable
     // event with the same kind from the same author that was issued
     // earlier than this.
-    if e.kind == 0 || e.kind == 3 || (e.kind >= 10000 && e.kind < 20000) {
+    if e.kind == 0 || e.kind == 3 || e.kind == 41 || (e.kind >= 10000 && e.kind < 20000) {
         let update_count = tx.execute(
             "UPDATE event SET hidden=TRUE WHERE id!=? AND kind=? AND author=? AND created_at <= ? and hidden!=TRUE",
             params![ev_id, e.kind, hex::decode(&e.pubkey).ok(), e.created_at],
@@ -671,32 +682,67 @@ fn _pool_at_capacity(pool: &SqlitePool) -> bool {
 fn log_pool_stats(name: &str, pool: &SqlitePool) {
     let state: r2d2::State = pool.state();
     let in_use_cxns = state.connections - state.idle_connections;
-    trace!(
+    debug!(
-        "DB pool {:?} usage (in_use: {}, available: {})",
+        "DB pool {:?} usage (in_use: {}, available: {}, max: {})",
         name,
         in_use_cxns,
-        state.connections
+        state.connections,
+	pool.max_size()
     );
-    if state.connections == in_use_cxns {
-        debug!("DB pool {:?} is empty (in_use: {})", name, in_use_cxns);
-    }
 }
 
+
 /// Perform database maintenance on a regular basis
-pub async fn db_maintenance(pool: SqlitePool) {
+pub async fn db_optimize_task(pool: SqlitePool) {
     tokio::task::spawn(async move {
         loop {
             tokio::select! {
-                    _ = tokio::time::sleep(Duration::from_secs(EVENT_MAINTENANCE_FREQ_SEC)) => {
+                _ = tokio::time::sleep(Duration::from_secs(60*60)) => {
                     if let Ok(mut conn) = pool.get() {
 			// the busy timer will block writers, so don't set
 			// this any higher than you want max latency for event
 			// writes.
-                conn.busy_timeout(Duration::from_secs(1)).ok();
+                        info!("running database optimizer");
-                            debug!("running database optimizer");
                         optimize_db(&mut conn).ok();
+                    }
+		}
+            };
+        }
+    });
+}
+
+/// Perform database WAL checkpoint on a regular basis
+pub async fn db_checkpoint_task(pool: SqlitePool, safe_to_read: Arc<Mutex<u64>>) {
+    tokio::task::spawn(async move {
+	// WAL size in pages.
+	let mut current_wal_size = 0;
+	// WAL threshold for more aggressive checkpointing (10,000 pages, or about 40MB)
+	let wal_threshold = 1000*10;
+	// default threshold for the busy timer
+	let busy_wait_default = Duration::from_secs(1);
+	// if the WAL file is getting too big, switch to this
+	let busy_wait_default_long = Duration::from_secs(10);
+        loop {
+            tokio::select! {
+                _ = tokio::time::sleep(Duration::from_secs(CHECKPOINT_FREQ_SEC)) => {
+                    if let Ok(mut conn) = pool.get() {
+                        let mut _guard:Option<MutexGuard<u64>> = None;
+			// the busy timer will block writers, so don't set
+			// this any higher than you want max latency for event
+			// writes.
+			if current_wal_size <= wal_threshold {
+			    conn.busy_timeout(busy_wait_default).ok();
+			} else {
+			    // if the wal size has exceeded a threshold, increase the busy timeout.
+			    conn.busy_timeout(busy_wait_default_long).ok();
+			    // take a lock that will prevent new readers.
+			    info!("blocking new readers to perform wal_checkpoint");
+			    _guard = Some(safe_to_read.lock().await);
+			}
                         debug!("running wal_checkpoint(TRUNCATE)");
-                            checkpoint_db(&mut conn).ok();
+                        if let Ok(new_size) = checkpoint_db(&mut conn) {
+			    current_wal_size = new_size;
+			}
                     }
 		}
             };
@@ -716,9 +762,14 @@ pub async fn db_query(
     pool: SqlitePool,
     query_tx: tokio::sync::mpsc::Sender<QueryResult>,
     mut abandon_query_rx: tokio::sync::oneshot::Receiver<()>,
+    safe_to_read: Arc<Mutex<u64>>,
 ) {
     let pre_spawn_start = Instant::now();
     task::spawn_blocking(move || {
+	{
+	    // if we are waiting on a checkpoint, stop until it is complete
+	    let _ = safe_to_read.blocking_lock();
+	}
         let db_queue_time = pre_spawn_start.elapsed();
         // if the queue time was very long (>5 seconds), spare the DB and abort.
         if db_queue_time > Duration::from_secs(5) {
@@ -743,8 +794,6 @@ pub async fn db_query(
         if sql_gen_elapsed > Duration::from_millis(10) {
             debug!("SQL (slow) generated in {:?}", start.elapsed());
         }
-        // show pool stats
-        log_pool_stats("reader", &pool);
         // cutoff for displaying slow queries
         let slow_cutoff = Duration::from_millis(2000);
         // any client that doesn't cause us to generate new rows in 5
@@ -753,10 +802,13 @@ pub async fn db_query(
         let start = Instant::now();
         let mut slow_first_event;
         let mut last_successful_send = Instant::now();
-        if let Ok(conn) = pool.get() {
+        if let Ok(mut conn) = pool.get() {
-            // execute the query. Don't cache, since queries vary so much.
+            // execute the query.
-            let mut stmt = conn.prepare(&q)?;
+            // make the actual SQL query (with parameters inserted) available
+            conn.trace(Some(|x| {trace!("SQL trace: {:?}", x)}));
+            let mut stmt = conn.prepare_cached(&q)?;
             let mut event_rows = stmt.query(rusqlite::params_from_iter(p))?;
+
             let mut first_result = true;
             while let Some(row) = event_rows.next()? {
                 let first_event_elapsed = start.elapsed();
@@ -770,29 +822,23 @@ pub async fn db_query(
                 }
                 // logging for slow queries; show sub and SQL.
                 // to reduce logging; only show 1/16th of clients (leading 0)
-                if row_count == 0 && slow_first_event && client_id.starts_with("0") {
+                if row_count == 0 && slow_first_event && client_id.starts_with('0') {
                     debug!(
                         "query req (slow): {:?} (cid: {}, sub: {:?})",
                         sub, client_id, sub.id
                     );
-                    debug!(
-                        "query string (slow): {} (cid: {}, sub: {:?})",
-                        q, client_id, sub.id
-                    );
-                } else {
-                    trace!(
-                        "query req: {:?} (cid: {}, sub: {:?})",
-                        sub,
-                        client_id,
-                        sub.id
-                    );
-                    trace!(
-                        "query string: {} (cid: {}, sub: {:?})",
-                        q,
-                        client_id,
-                        sub.id
-                    );
 		}
+		// check if a checkpoint is trying to run, and abort
+		if row_count % 100 == 0 {
+		    {
+			if let Err(_) = safe_to_read.try_lock() {
+			    // lock was held, abort this query
+			    debug!("query aborted due to checkpoint (cid: {}, sub: {:?})", client_id, sub.id);
+			    return Ok(());
+			}
+		    }
+		}
+
                 // check if this is still active; every 100 rows
                 if row_count % 100 == 0 && abandon_query_rx.try_recv().is_ok() {
                     debug!("query aborted (cid: {}, sub: {:?})", client_id, sub.id);
@@ -809,10 +855,17 @@ pub async fn db_query(
                         trace!("db reader thread is stalled");
                         if last_successful_send + abort_cutoff < Instant::now() {
                             // the queue has been full for too long, abort
-                            info!("aborting database query due to slow client");
+                            info!("aborting database query due to slow client (cid: {}, sub: {:?})",
+				  client_id, sub.id);
                             let ok: Result<()> = Ok(());
                             return ok;
                         }
+			// check if a checkpoint is trying to run, and abort
+			if let Err(_) = safe_to_read.try_lock() {
+			    // lock was held, abort this query
+			    debug!("query aborted due to checkpoint (cid: {}, sub: {:?})", client_id, sub.id);
+			    return Ok(());
+			}
                         // give the queue a chance to clear before trying again
                         thread::sleep(Duration::from_millis(100));
                     }

src/delegation.rs

@@ -80,7 +80,7 @@ impl FromStr for Operator {
 
 #[derive(Serialize, Deserialize, PartialEq, Eq, Debug, Clone)]
 pub struct ConditionQuery {
-    pub(crate) conditions: Vec<Condition>,
+    pub conditions: Vec<Condition>,
 }
 
 impl ConditionQuery {
@@ -137,9 +137,9 @@ pub fn validate_delegation(
 /// An example complex condition would be:  kind=1,2,3&created_at<1665265999
 #[derive(Serialize, Deserialize, PartialEq, Eq, Debug, Clone)]
 pub struct Condition {
-    pub(crate) field: Field,
+    pub field: Field,
-    pub(crate) operator: Operator,
+    pub operator: Operator,
-    pub(crate) values: Vec<u64>,
+    pub values: Vec<u64>,
 }
 
 impl Condition {
@@ -332,19 +332,6 @@ mod tests {
         assert_eq!(parsed, cq);
         Ok(())
     }
-    fn simple_event() -> Event {
-        Event {
-            id: "0".to_owned(),
-            pubkey: "0".to_owned(),
-            delegated_by: None,
-            created_at: 0,
-            kind: 0,
-            tags: vec![],
-            content: "".to_owned(),
-            sig: "0".to_owned(),
-            tagidx: None,
-        }
-    }
     // Check for condition logic on event w/ empty values
     #[test]
     fn condition_with_empty_values() {
@@ -353,7 +340,7 @@ mod tests {
             operator: Operator::GreaterThan,
             values: vec![],
         };
-        let e = simple_event();
+        let e = Event::simple_event();
         assert!(!c.allows_event(&e));
         c.operator = Operator::LessThan;
         assert!(!c.allows_event(&e));
@@ -373,7 +360,7 @@ mod tests {
             operator: Operator::GreaterThan,
             values: vec![10],
         };
-        let mut e = simple_event();
+        let mut e = Event::simple_event();
         // kind is not greater than 10, not allowed
         e.kind = 1;
         assert!(!c.allows_event(&e));
@@ -392,7 +379,7 @@ mod tests {
             operator: Operator::Equals,
             values: vec![0, 10, 20],
         };
-        let mut e = simple_event();
+        let mut e = Event::simple_event();
         // Allow if event kind is in list for Equals
         e.kind = 10;
         assert!(c.allows_event(&e));

src/event.rs

@@ -37,19 +37,19 @@ impl EventCmd {
 #[derive(Serialize, Deserialize, PartialEq, Eq, Debug, Clone)]
 pub struct Event {
     pub id: String,
-    pub(crate) pubkey: String,
+    pub pubkey: String,
     #[serde(skip)]
-    pub(crate) delegated_by: Option<String>,
+    pub delegated_by: Option<String>,
-    pub(crate) created_at: u64,
+    pub created_at: u64,
-    pub(crate) kind: u64,
+    pub kind: u64,
     #[serde(deserialize_with = "tag_from_string")]
     // NOTE: array-of-arrays may need to be more general than a string container
-    pub(crate) tags: Vec<Vec<String>>,
+    pub tags: Vec<Vec<String>>,
-    pub(crate) content: String,
+    pub content: String,
-    pub(crate) sig: String,
+    pub sig: String,
     // Optimization for tag search, built on demand.
     #[serde(skip)]
-    pub(crate) tagidx: Option<HashMap<char, HashSet<String>>>,
+    pub tagidx: Option<HashMap<char, HashSet<String>>>,
 }
 
 /// Simple tag type for array of array of strings.
@@ -101,6 +101,21 @@ impl From<EventCmd> for Result<Event> {
 }
 
 impl Event {
+    #[cfg(test)]
+    pub fn simple_event() -> Event {
+        Event {
+            id: "0".to_owned(),
+            pubkey: "0".to_owned(),
+            delegated_by: None,
+            created_at: 0,
+            kind: 0,
+            tags: vec![],
+            content: "".to_owned(),
+            sig: "0".to_owned(),
+            tagidx: None,
+        }
+    }
+
     pub fn is_kind_metadata(&self) -> bool {
         self.kind == 0
     }
@@ -226,7 +241,7 @@ impl Event {
     }
 
     /// Check if this event has a valid signature.
-    fn validate(&self) -> Result<()> {
+    pub fn validate(&self) -> Result<()> {
         // TODO: return a Result with a reason for invalid events
         // validation is performed by:
         // * parsing JSON string into event fields
@@ -319,31 +334,18 @@ impl Event {
 #[cfg(test)]
 mod tests {
     use super::*;
-    fn simple_event() -> Event {
-        Event {
-            id: "0".to_owned(),
-            pubkey: "0".to_owned(),
-            delegated_by: None,
-            created_at: 0,
-            kind: 0,
-            tags: vec![],
-            content: "".to_owned(),
-            sig: "0".to_owned(),
-            tagidx: None,
-        }
-    }
 
     #[test]
     fn event_creation() {
         // create an event
-        let event = simple_event();
+        let event = Event::simple_event();
         assert_eq!(event.id, "0");
     }
 
     #[test]
     fn event_serialize() -> Result<()> {
         // serialize an event to JSON string
-        let event = simple_event();
+        let event = Event::simple_event();
         let j = serde_json::to_string(&event)?;
         assert_eq!(j, "{\"id\":\"0\",\"pubkey\":\"0\",\"created_at\":0,\"kind\":0,\"tags\":[],\"content\":\"\",\"sig\":\"0\"}");
         Ok(())
@@ -351,14 +353,14 @@ mod tests {
 
     #[test]
     fn empty_event_tag_match() {
-        let event = simple_event();
+        let event = Event::simple_event();
         assert!(!event
             .generic_tag_val_intersect('e', &HashSet::from(["foo".to_owned(), "bar".to_owned()])));
     }
 
     #[test]
     fn single_event_tag_match() {
-        let mut event = simple_event();
+        let mut event = Event::simple_event();
         event.tags = vec![vec!["e".to_owned(), "foo".to_owned()]];
         event.build_index();
         assert_eq!(
@@ -373,7 +375,7 @@ mod tests {
     #[test]
     fn event_tags_serialize() -> Result<()> {
         // serialize an event with tags to JSON string
-        let mut event = simple_event();
+        let mut event = Event::simple_event();
         event.tags = vec![
             vec![
                 "e".to_owned(),

src/server.rs

@@ -25,10 +25,13 @@ use hyper::{
 use rusqlite::OpenFlags;
 use serde::{Deserialize, Serialize};
 use serde_json::json;
+use tokio::sync::Mutex;
 use std::collections::HashMap;
 use std::convert::Infallible;
 use std::net::SocketAddr;
 use std::path::Path;
+use std::sync::Arc;
+use std::sync::atomic::Ordering;
 use std::sync::mpsc::Receiver as MpscReceiver;
 use std::time::Duration;
 use std::time::Instant;
@@ -53,6 +56,7 @@ async fn handle_web_request(
     broadcast: Sender<Event>,
     event_tx: tokio::sync::mpsc::Sender<SubmittedEvent>,
     shutdown: Receiver<()>,
+    safe_to_read: Arc<Mutex<u64>>,
 ) -> Result<Response<Body>, Infallible> {
     match (
         request.uri().path(),
@@ -88,6 +92,7 @@ async fn handle_web_request(
                                     Some(config),
                                 )
                                 .await;
+                                let origin = get_header_string("origin", request.headers());
                                 let user_agent = get_header_string("user-agent", request.headers());
                                 // determine the remote IP from headers if the exist
                                 let header_ip = settings
@@ -101,6 +106,7 @@ async fn handle_web_request(
                                 let client_info = ClientInfo {
                                     remote_ip,
                                     user_agent,
+                                    origin,
                                 };
                                 // spawn a nostr server with our websocket
                                 tokio::spawn(nostr_server(
@@ -111,6 +117,7 @@ async fn handle_web_request(
                                     broadcast,
                                     event_tx,
                                     shutdown,
+				    safe_to_read,
                                 ));
                             }
                             // todo: trace, don't print...
@@ -245,14 +252,19 @@ pub fn start_server(settings: Settings, shutdown_rx: MpscReceiver<()>) -> Result
     // configure tokio runtime
     let rt = Builder::new_multi_thread()
         .enable_all()
-        .thread_name("tokio-ws")
+        .thread_name_fn(|| {
+	    // give each thread a unique numeric name
+	    static ATOMIC_ID: std::sync::atomic::AtomicUsize = std::sync::atomic::AtomicUsize::new(0);
+	    let id = ATOMIC_ID.fetch_add(1,Ordering::SeqCst);
+	    format!("tokio-ws-{}", id)
+		})
         // limit concurrent SQLite blocking threads
         .max_blocking_threads(settings.limits.max_blocking_threads)
         .on_thread_start(|| {
-            trace!("started new thread");
+            trace!("started new thread: {:?}", std::thread::current().name());
         })
         .on_thread_stop(|| {
-            trace!("stopping thread");
+	    trace!("stopped thread: {:?}", std::thread::current().name());
         })
         .build()
         .unwrap();
@@ -317,10 +329,16 @@ pub fn start_server(settings: Settings, shutdown_rx: MpscReceiver<()>) -> Result
             &settings,
             OpenFlags::SQLITE_OPEN_READ_WRITE | OpenFlags::SQLITE_OPEN_CREATE,
             1,
-            1,
+            2,
             false,
         );
-        db::db_maintenance(maintenance_pool).await;
+
+	// Create a mutex that will block readers, so that a
+	// checkpoint can be performed quickly.
+	let safe_to_read = Arc::new(Mutex::new(0));
+
+        db::db_optimize_task(maintenance_pool.clone()).await;
+        db::db_checkpoint_task(maintenance_pool, safe_to_read.clone()).await;
 
         // listen for (external to tokio) shutdown request
         let controlled_shutdown = invoke_shutdown.clone();
@@ -356,6 +374,10 @@ pub fn start_server(settings: Settings, shutdown_rx: MpscReceiver<()>) -> Result
             db_max_conn,
             true,
         );
+	// spawn a task to check the pool size.
+	let pool_monitor = pool.clone();
+	tokio::spawn(async move {db::monitor_pool("reader", pool_monitor).await;});
+
         // A `Service` is needed for every connection, so this
         // creates one from our `handle_request` function.
         let make_svc = make_service_fn(|conn: &AddrStream| {
@@ -365,6 +387,7 @@ pub fn start_server(settings: Settings, shutdown_rx: MpscReceiver<()>) -> Result
             let event = event_tx.clone();
             let stop = invoke_shutdown.clone();
             let settings = settings.clone();
+            let safe_to_read = safe_to_read.clone();
             async move {
                 // service_fn converts our function into a `Service`
                 Ok::<_, Infallible>(service_fn(move |request: Request<Body>| {
@@ -376,6 +399,7 @@ pub fn start_server(settings: Settings, shutdown_rx: MpscReceiver<()>) -> Result
                         bcast.clone(),
                         event.clone(),
                         stop.subscribe(),
+			safe_to_read.clone(),
                     )
                 }))
             }
@@ -439,6 +463,7 @@ fn make_notice_message(notice: Notice) -> Message {
 struct ClientInfo {
     remote_ip: String,
     user_agent: Option<String>,
+    origin: Option<String>,
 }
 
 /// Handle new client connections.  This runs through an event loop
@@ -451,6 +476,7 @@ async fn nostr_server(
     broadcast: Sender<Event>,
     event_tx: mpsc::Sender<SubmittedEvent>,
     mut shutdown: Receiver<()>,
+    safe_to_read: Arc<Mutex<u64>>,
 ) {
     // the time this websocket nostr server started
     let orig_start = Instant::now();
@@ -502,9 +528,14 @@ async fn nostr_server(
     let mut client_published_event_count: usize = 0;
     let mut client_received_event_count: usize = 0;
     debug!("new client connection (cid: {}, ip: {:?})", cid, conn.ip());
-    if let Some(ua) = client_info.user_agent {
+    let origin = client_info.origin.unwrap_or_else(|| "<unspecified>".into());
-        debug!("cid: {}, user-agent: {:?}", cid, ua);
+    let user_agent = client_info
-    }
+        .user_agent
+        .unwrap_or_else(|| "<unspecified>".into());
+    debug!(
+        "cid: {}, origin: {:?}, user-agent: {:?}",
+        cid, origin, user_agent
+    );
     loop {
         tokio::select! {
             _ = shutdown.recv() => {
@@ -655,8 +686,10 @@ async fn nostr_server(
                                     if let Some(previous_query) = running_queries.insert(s.id.to_owned(), abandon_query_tx) {
                     previous_query.send(()).ok();
                                     }
+		    if s.needs_historical_events() {
                                     // start a database query.  this spawns a blocking database query on a worker thread.
-                                    db::db_query(s, cid.to_owned(), pool.clone(), query_tx.clone(), abandon_query_rx).await;
+					db::db_query(s, cid.to_owned(), pool.clone(), query_tx.clone(), abandon_query_rx,safe_to_read.clone()).await;
+				    }
                 },
                 Err(e) => {
                     info!("Subscription error: {} (cid: {}, sub: {:?})", e, cid, s.id);

src/subscription.rs

@@ -65,12 +65,21 @@ impl<'de> Deserialize<'de> for ReqFilter {
             tags: None,
             force_no_match: false,
         };
+	let empty_string = "".into();
         let mut ts = None;
         // iterate through each key, and assign values that exist
         for (key, val) in filter.into_iter() {
             // ids
             if key == "ids" {
-                rf.ids = Deserialize::deserialize(val).ok();
+		let raw_ids: Option<Vec<String>>= Deserialize::deserialize(val).ok();
+		if let Some(a) = raw_ids.as_ref() {
+		    if a.contains(&empty_string) {
+		    return Err(serde::de::Error::invalid_type(
+                Unexpected::Other("prefix matches must not be empty strings"),
+                &"a json object"));
+		    }
+		}
+		rf.ids =raw_ids;
             } else if key == "kinds" {
                 rf.kinds = Deserialize::deserialize(val).ok();
             } else if key == "since" {
@@ -80,7 +89,15 @@ impl<'de> Deserialize<'de> for ReqFilter {
             } else if key == "limit" {
                 rf.limit = Deserialize::deserialize(val).ok();
             } else if key == "authors" {
-                rf.authors = Deserialize::deserialize(val).ok();
+                let raw_authors: Option<Vec<String>>= Deserialize::deserialize(val).ok();
+		if let Some(a) = raw_authors.as_ref() {
+		    if a.contains(&empty_string) {
+			return Err(serde::de::Error::invalid_type(
+			    Unexpected::Other("prefix matches must not be empty strings"),
+			    &"a json object"));
+		    }
+		}
+		rf.authors = raw_authors;
             } else if key.starts_with('#') && key.len() > 1 && val.is_array() {
                 if let Some(tag_search) = tag_search_char_from_filter(key) {
                     if ts.is_none() {
@@ -183,6 +200,13 @@ impl Subscription {
     pub fn get_id(&self) -> String {
         self.id.clone()
     }
+
+    /// Determine if any filter is requesting historical (database)
+    /// queries.  If every filter has limit:0, we do not need to query the DB.
+    pub fn needs_historical_events(&self) -> bool {
+	self.filters.iter().any(|f| f.limit!=Some(0))
+    }
+
     /// Determine if this subscription matches a given [`Event`].  Any
     /// individual filter match is sufficient.
     pub fn interested_in_event(&self, event: &Event) -> bool {
@@ -294,6 +318,24 @@ mod tests {
         assert!(serde_json::from_str::<Subscription>(raw_json).is_err());
     }
 
+    #[test]
+    fn req_empty_authors_prefix() {
+	let raw_json = "[\"REQ\",\"some-id\",{\"authors\": [\"\"]}]";
+        assert!(serde_json::from_str::<Subscription>(raw_json).is_err());
+    }
+
+    #[test]
+    fn req_empty_ids_prefix() {
+	let raw_json = "[\"REQ\",\"some-id\",{\"ids\": [\"\"]}]";
+        assert!(serde_json::from_str::<Subscription>(raw_json).is_err());
+    }
+
+    #[test]
+    fn req_empty_ids_prefix_mixed() {
+	let raw_json = "[\"REQ\",\"some-id\",{\"ids\": [\"\",\"aaa\"]}]";
+        assert!(serde_json::from_str::<Subscription>(raw_json).is_err());
+    }
+
     #[test]
     fn legacy_filter() {
         // legacy field in filter