NOT MAINTAINED I'm sorry, this project is not maintained since 2014. The code is not complicated though. If you need to make changes, please fork it.
The project adds transaction support to node-sqlite3.
Node-sqlite3 is a great way how to access a SQLite database but id doesn't provide support for transactions yet. The underlying SQLite database can handle transactions easily so you can still do:
db.serialize(function() {
db.exec("BEGIN");
...
db.exec("COMMIT");
});
This works fine until you add an async operation between BEGIN
and COMMIT
. The problem is that node-sqlite3 uses single connection to the database and thus queries from other places can end up intermixing with each other messing up the transactions.
Sqlite3-transactions solves this problem by transparently locking the database while in a transaction. If database is locked all queries which don't belong to the transaction are queued and executed after the transaction is finished. As a bonus you got a clean nice API for transactions.
npm install sqlite3
npm install sqlite3-transactions
var sqlite3 = require("sqlite3"),
TransactionDatabase = require("sqlite3-transactions").TransactionDatabase;
// Wrap sqlite3 database
var db = new TransactionDatabase(
new sqlite3.Database("test.sqlite", sqlite3.OPEN_READWRITE | sqlite3.OPEN_CREATE)
);
// Use db as normal sqlite3.Database object.
db.exec("CREATE TABLE ...", function(err) {
// table created
});
// Begin a transaction.
db.beginTransaction(function(err, transaction) {
// Now we are inside a transaction.
// Use transaction as normal sqlite3.Database object.
transaction.run("INSERT ...");
// All calls db.exec(), db.run(), db.beginTransaction(), ... are
// queued and executed after you do transaction.commit() or transaction.rollback()
// This will be executed after the transaction is finished.
database.run("INSERT ...");
// Feel free to do any async operations.
someAsync(function() {
// Remember to .commit() or .rollback()
transaction.commit(function(err) {
if (err) return console.log("Sad panda :-( commit() failed.", err);
console.log("Happy panda :-) commit() was successful.");
});
// or transaction.rollback()
});
});
I haven't mentioned one feature in the example - on error
event automatic rollback()
is performed on the current transaction.
Call this method to start a transaction. The database is locked and all queries are queued until the transaction is over. callback
receives two parameters error
and transaction
. Use transaction
to perform operations directly with database. You must call transaction.commit()
or transaction.rollback()
to finish the transaction and unlock the database. You don't have to rollback()
in the case:
- there is and
error
passed to the callback. - there was an
error
event during the transaction.
Commits the transaction. The callback(error)
is called after the transaction is committed. If error
is set then the commit failed for some reason and rollback was performed instead.
Queued operations are executed after the actual commit but before the callback
. This helps to prevent starvation.
Rolls back the transaction. The callback(error)
is called after the operation is completed.
Remember to call commit()
or rollback()
on each transaction otherwise you lock your DB forever.
sqlite3-transactions is in very early stage of development. Please help me test it.