store/postgres: warn if locale is not en_US.utf8

docs/environment-variables.md

@@ -187,3 +187,4 @@ happens, subgraphs might process inconsistent data. Defaults to 250.
   identified as unused, `graph-node` will wait at least this long before
   actually deleting the data (value is in minutes, defaults to 360, i.e. 6
   hours)
+- `GRAPH_STORE_LOCAL`: default is `en_US.utf8`. This is used to mandate the locale for database. This is useful for comparing strings fields.

graph/src/env/store.rs

@@ -91,6 +91,10 @@ pub struct EnvVarsStore {
     /// once the new behavior has run in the hosted service for a few days
     /// without issues.
     pub disable_error_for_toplevel_parents: bool,
+
+    /// This is used to mandate what locale the Database run.
+    /// Set by the environment variable `GRAPH_STORE_LOCALE`.
+    pub store_locale: String,
 }
 
 // This does not print any values avoid accidentally leaking any sensitive env vars
@@ -128,6 +132,7 @@ impl From<InnerStore> for EnvVarsStore {
             connection_idle_timeout: Duration::from_secs(x.connection_idle_timeout_in_secs),
             write_queue_size: x.write_queue_size,
             disable_error_for_toplevel_parents: x.disable_error_for_toplevel_parents.0,
+            store_locale: x.store_locale,
         }
     }
 }
@@ -173,4 +178,6 @@ pub struct InnerStore {
     write_queue_size: usize,
     #[envconfig(from = "GRAPH_DISABLE_ERROR_FOR_TOPLEVEL_PARENTS", default = "false")]
     disable_error_for_toplevel_parents: EnvVarBoolean,
+    #[envconfig(from = "GRAPH_STORE_LOCALE", default = "en_US.utf8")]
+    store_locale: String,
 }

store/postgres/src/catalog.rs

@@ -37,6 +37,13 @@ table! {
         nspname -> Text,
     }
 }
+// Readonly; only mapping the columns we want
+table! {
+    pg_database(datname) {
+        datname -> Text,
+        datcollate -> Text,
+    }
+}
 
 table! {
     subgraphs.table_stats {
@@ -47,6 +54,17 @@ table! {
     }
 }
 
+/// Get collate for current database
+pub fn get_collation(conn: &PgConnection) -> Result<String, StoreError> {
+    use pg_database as db;
+
+    return db::table
+        .filter(db::datname.eq(diesel::dsl::sql("current_database()")))
+        .select(db::datcollate)
+        .get_result(conn)
+        .map_err(StoreError::from);
+}
+
 // In debug builds (for testing etc.) create exclusion constraints, in
 // release builds for production, skip them
 #[cfg(debug_assertions)]

store/postgres/src/connection_pool.rs

@@ -977,6 +977,15 @@ impl PoolInner {
         let pool = self.clone();
         let conn = self.get().map_err(|_| StoreError::DatabaseUnavailable)?;
 
+        let db_locale = catalog::get_collation(&conn)?;
+
+        if !db_locale.eq(&ENV_VARS.store.store_locale) {
+            panic!(
+                "Database locale is not '{}', but {}",
+                ENV_VARS.store.store_locale, db_locale
+            );
+        }
+
         advisory_lock::lock_migration(&conn)
             .unwrap_or_else(|err| die(&pool.logger, "failed to get migration lock", &err));
         let result = pool