Transactions and batched writes
Firestore supports atomic operations for reading and writing data. In a set of atomic operations, either all of the operations succeed, or none of them are applied. There are two types of atomic operations in Firestore:
- Transactions: a transaction is a set of read and write operations on one or more documents.
- Batched Writes: a batched write is a set of write operations on one or more documents.
Updating data with transactions
Using the Firestore client libraries, you can group multiple operations into a single transaction. Transactions are useful when you want to update a field's value based on its current value, or the value of some other field.
A transaction consists of any number of
get()
operations followed by any number of write operations such as set()
,
update()
, or delete()
. In the case of a concurrent edit,
Firestore runs the entire transaction again. For example, if a
transaction reads documents and another client modifies any of those documents,
Firestore retries the transaction. This feature ensures that the
transaction runs on up-to-date and consistent data.
Transactions never partially apply writes. All writes execute at the end of a successful transaction.
When using transactions, note that:
- Read operations must come before write operations.
- A function calling a transaction (transaction function) might run more than once if a concurrent edit affects a document that the transaction reads.
- Transaction functions should not directly modify application state.
- Transactions will fail when the client is offline.
The following example shows how to create and run a transaction:
Web version 9
import { runTransaction } from "firebase/firestore"; try { await runTransaction(db, async (transaction) => { const sfDoc = await transaction.get(sfDocRef); if (!sfDoc.exists()) { throw "Document does not exist!"; } const newPopulation = sfDoc.data().population + 1; transaction.update(sfDocRef, { population: newPopulation }); }); console.log("Transaction successfully committed!"); } catch (e) { console.log("Transaction failed: ", e); }
Web version 8
// Create a reference to the SF doc. var sfDocRef = db.collection("cities").doc("SF"); // Uncomment to initialize the doc. // sfDocRef.set({ population: 0 }); return db.runTransaction((transaction) => { // This code may get re-run multiple times if there are conflicts. return transaction.get(sfDocRef).then((sfDoc) => { if (!sfDoc.exists) { throw "Document does not exist!"; } // Add one person to the city population. // Note: this could be done without a transaction // by updating the population using FieldValue.increment() var newPopulation = sfDoc.data().population + 1; transaction.update(sfDocRef, { population: newPopulation }); }); }).then(() => { console.log("Transaction successfully committed!"); }).catch((error) => { console.log("Transaction failed: ", error); });
Swift
let sfReference = db.collection("cities").document("SF") do { let _ = try await db.runTransaction({ (transaction, errorPointer) -> Any? in let sfDocument: DocumentSnapshot do { try sfDocument = transaction.getDocument(sfReference) } catch let fetchError as NSError { errorPointer?.pointee = fetchError return nil } guard let oldPopulation = sfDocument.data()?["population"] as? Int else { let error = NSError( domain: "AppErrorDomain", code: -1, userInfo: [ NSLocalizedDescriptionKey: "Unable to retrieve population from snapshot \(sfDocument)" ] ) errorPointer?.pointee = error return nil } // Note: this could be done without a transaction // by updating the population using FieldValue.increment() transaction.updateData(["population": oldPopulation + 1], forDocument: sfReference) return nil }) print("Transaction successfully committed!") } catch { print("Transaction failed: \(error)") }
Objective-C
FIRDocumentReference *sfReference = [[self.db collectionWithPath:@"cities"] documentWithPath:@"SF"]; [self.db runTransactionWithBlock:^id (FIRTransaction *transaction, NSError **errorPointer) { FIRDocumentSnapshot *sfDocument = [transaction getDocument:sfReference error:errorPointer]; if (*errorPointer != nil) { return nil; } if (![sfDocument.data[@"population"] isKindOfClass:[NSNumber class]]) { *errorPointer = [NSError errorWithDomain:@"AppErrorDomain" code:-1 userInfo:@{ NSLocalizedDescriptionKey: @"Unable to retreive population from snapshot" }]; return nil; } NSInteger oldPopulation = [sfDocument.data[@"population"] integerValue]; // Note: this could be done without a transaction // by updating the population using FieldValue.increment() [transaction updateData:@{ @"population": @(oldPopulation + 1) } forDocument:sfReference]; return nil; } completion:^(id result, NSError *error) { if (error != nil) { NSLog(@"Transaction failed: %@", error); } else { NSLog(@"Transaction successfully committed!"); } }];
Kotlin
Android
val sfDocRef = db.collection("cities").document("SF") db.runTransaction { transaction -> val snapshot = transaction.get(sfDocRef) // Note: this could be done without a transaction // by updating the population using FieldValue.increment() val newPopulation = snapshot.getDouble("population")!! + 1 transaction.update(sfDocRef, "population", newPopulation) // Success null }.addOnSuccessListener { Log.d(TAG, "Transaction success!") } .addOnFailureListener { e -> Log.w(TAG, "Transaction failure.", e) }
Java
Android
final DocumentReference sfDocRef = db.collection("cities").document("SF"); db.runTransaction(new Transaction.Function<Void>() { @Override public Void apply(@NonNull Transaction transaction) throws FirebaseFirestoreException { DocumentSnapshot snapshot = transaction.get(sfDocRef); // Note: this could be done without a transaction // by updating the population using FieldValue.increment() double newPopulation = snapshot.getDouble("population") + 1; transaction.update(sfDocRef, "population", newPopulation); // Success return null; } }).addOnSuccessListener(new OnSuccessListener<Void>() { @Override public void onSuccess(Void aVoid) { Log.d(TAG, "Transaction success!"); } }) .addOnFailureListener(new OnFailureListener() { @Override public void onFailure(@NonNull Exception e) { Log.w(TAG, "Transaction failure.", e); } });
Dart
final sfDocRef = db.collection("cities").doc("SF"); db.runTransaction((transaction) async { final snapshot = await transaction.get(sfDocRef); // Note: this could be done without a transaction // by updating the population using FieldValue.increment() final newPopulation = snapshot.get("population") + 1; transaction.update(sfDocRef, {"population": newPopulation}); }).then( (value) => print("DocumentSnapshot successfully updated!"), onError: (e) => print("Error updating document $e"), );
Java
Python
Python
(Async)
C++
DocumentReference sf_doc_ref = db->Collection("cities").Document("SF"); db->RunTransaction([sf_doc_ref](Transaction& transaction, std::string& out_error_message) -> Error { Error error = Error::kErrorOk; DocumentSnapshot snapshot = transaction.Get(sf_doc_ref, &error, &out_error_message); // Note: this could be done without a transaction by updating the // population using FieldValue::Increment(). std::int64_t new_population = snapshot.Get("population").integer_value() + 1; transaction.Update( sf_doc_ref, {{"population", FieldValue::Integer(new_population)}}); return Error::kErrorOk; }).OnCompletion([](const Future<void>& future) { if (future.error() == Error::kErrorOk) { std::cout << "Transaction success!" << std::endl; } else { std::cout << "Transaction failure: " << future.error_message() << std::endl; } });
Node.js
Go
PHP
Unity
DocumentReference cityRef = db.Collection("cities").Document("SF"); db.RunTransactionAsync(transaction => { return transaction.GetSnapshotAsync(cityRef).ContinueWith((snapshotTask) => { DocumentSnapshot snapshot = snapshotTask.Result; long newPopulation = snapshot.GetValue<long>("Population") + 1; Dictionary<string, object> updates = new Dictionary<string, object> { { "Population", newPopulation} }; transaction.Update(cityRef, updates); }); });
C#
Ruby
Passing information out of transactions
Do not modify application state inside of your transaction functions. Doing so will introduce concurrency issues, because transaction functions can run multiple times and are not guaranteed to run on the UI thread. Instead, pass information you need out of your transaction functions. The following example builds on the previous example to show how to pass information out of a transaction:
Web version 9
import { doc, runTransaction } from "firebase/firestore"; // Create a reference to the SF doc. const sfDocRef = doc(db, "cities", "SF"); try { const newPopulation = await runTransaction(db, async (transaction) => { const sfDoc = await transaction.get(sfDocRef); if (!sfDoc.exists()) { throw "Document does not exist!"; } const newPop = sfDoc.data().population + 1; if (newPop <= 1000000) { transaction.update(sfDocRef, { population: newPop }); return newPop; } else { return Promise.reject("Sorry! Population is too big"); } }); console.log("Population increased to ", newPopulation); } catch (e) { // This will be a "population is too big" error. console.error(e); }
Web version 8
// Create a reference to the SF doc. var sfDocRef = db.collection("cities").doc("SF"); db.runTransaction((transaction) => { return transaction.get(sfDocRef).then((sfDoc) => { if (!sfDoc.exists) { throw "Document does not exist!"; } var newPopulation = sfDoc.data().population + 1; if (newPopulation <= 1000000) { transaction.update(sfDocRef, { population: newPopulation }); return newPopulation; } else { return Promise.reject("Sorry! Population is too big."); } }); }).then((newPopulation) => { console.log("Population increased to ", newPopulation); }).catch((err) => { // This will be an "population is too big" error. console.error(err); });
Swift
let sfReference = db.collection("cities").document("SF") do { let object = try await db.runTransaction({ (transaction, errorPointer) -> Any? in let sfDocument: DocumentSnapshot do { try sfDocument = transaction.getDocument(sfReference) } catch let fetchError as NSError { errorPointer?.pointee = fetchError return nil } guard let oldPopulation = sfDocument.data()?["population"] as? Int else { let error = NSError( domain: "AppErrorDomain", code: -1, userInfo: [ NSLocalizedDescriptionKey: "Unable to retrieve population from snapshot \(sfDocument)" ] ) errorPointer?.pointee = error return nil } // Note: this could be done without a transaction // by updating the population using FieldValue.increment() let newPopulation = oldPopulation + 1 guard newPopulation <= 1000000 else { let error = NSError( domain: "AppErrorDomain", code: -2, userInfo: [NSLocalizedDescriptionKey: "Population \(newPopulation) too big"] ) errorPointer?.pointee = error return nil } transaction.updateData(["population": newPopulation], forDocument: sfReference) return newPopulation }) print("Population increased to \(object!)") } catch { print("Error updating population: \(error)") }
Objective-C
FIRDocumentReference *sfReference = [[self.db collectionWithPath:@"cities"] documentWithPath:@"SF"]; [self.db runTransactionWithBlock:^id (FIRTransaction *transaction, NSError **errorPointer) { FIRDocumentSnapshot *sfDocument = [transaction getDocument:sfReference error:errorPointer]; if (*errorPointer != nil) { return nil; } if (![sfDocument.data[@"population"] isKindOfClass:[NSNumber class]]) { *errorPointer = [NSError errorWithDomain:@"AppErrorDomain" code:-1 userInfo:@{ NSLocalizedDescriptionKey: @"Unable to retreive population from snapshot" }]; return nil; } NSInteger population = [sfDocument.data[@"population"] integerValue]; population++; if (population >= 1000000) { *errorPointer = [NSError errorWithDomain:@"AppErrorDomain" code:-2 userInfo:@{ NSLocalizedDescriptionKey: @"Population too big" }]; return @(population); } [transaction updateData:@{ @"population": @(population) } forDocument:sfReference]; return nil; } completion:^(id result, NSError *error) { if (error != nil) { NSLog(@"Transaction failed: %@", error); } else { NSLog(@"Population increased to %@", result); } }];
Kotlin
Android
val sfDocRef = db.collection("cities").document("SF") db.runTransaction { transaction -> val snapshot = transaction.get(sfDocRef) val newPopulation = snapshot.getDouble("population")!! + 1 if (newPopulation <= 1000000) { transaction.update(sfDocRef, "population", newPopulation) newPopulation } else { throw FirebaseFirestoreException( "Population too high", FirebaseFirestoreException.Code.ABORTED, ) } }.addOnSuccessListener { result -> Log.d(TAG, "Transaction success: $result") }.addOnFailureListener { e -> Log.w(TAG, "Transaction failure.", e) }
Java
Android
final DocumentReference sfDocRef = db.collection("cities").document("SF"); db.runTransaction(new Transaction.Function<Double>() { @Override public Double apply(@NonNull Transaction transaction) throws FirebaseFirestoreException { DocumentSnapshot snapshot = transaction.get(sfDocRef); double newPopulation = snapshot.getDouble("population") + 1; if (newPopulation <= 1000000) { transaction.update(sfDocRef, "population", newPopulation); return newPopulation; } else { throw new FirebaseFirestoreException("Population too high", FirebaseFirestoreException.Code.ABORTED); } } }).addOnSuccessListener(new OnSuccessListener<Double>() { @Override public void onSuccess(Double result) { Log.d(TAG, "Transaction success: " + result); } }) .addOnFailureListener(new OnFailureListener() { @Override public void onFailure(@NonNull Exception e) { Log.w(TAG, "Transaction failure.", e); } });
Dart
final sfDocRef = db.collection("cities").doc("SF"); db.runTransaction((transaction) { return transaction.get(sfDocRef).then((sfDoc) { final newPopulation = sfDoc.get("population") + 1; transaction.update(sfDocRef, {"population": newPopulation}); return newPopulation; }); }).then( (newPopulation) => print("Population increased to $newPopulation"), onError: (e) => print("Error updating document $e"), );
Java
Python
Python
(Async)
C++
// This is not yet supported.
Node.js
Go
PHP
Unity
DocumentReference cityRef = db.Collection("cities").Document("SF"); db.RunTransactionAsync(transaction => { return transaction.GetSnapshotAsync(cityRef).ContinueWith((task) => { long newPopulation = task.Result.GetValue<long>("Population") + 1; if (newPopulation <= 1000000) { Dictionary<string, object> updates = new Dictionary<string, object> { { "Population", newPopulation} }; transaction.Update(cityRef, updates); return true; } else { return false; } }); }).ContinueWith((transactionResultTask) => { if (transactionResultTask.Result) { Console.WriteLine("Population updated successfully."); } else { Console.WriteLine("Sorry! Population is too big."); } });
C#
Ruby
Transaction failure
A transaction can fail for the following reasons:
- The transaction contains read operations after write operations. Read operations must always come before any write operations.
- The transaction read a document that was modified outside of the transaction. In this case, the transaction automatically runs again. The transaction is retried a finite number of times.
The transaction exceeded the maximum request size of 10 MiB.
Transaction size depends on the sizes of documents and index entries modified by the transaction. For a delete operation, this includes the size of the target document and the sizes of the index entries deleted in response to the operation.
A failed transaction returns an error and does not write anything to the database. You do not need to roll back the transaction; Firestore does this automatically.
Batched writes
If you do not need to read any documents in your operation set, you can execute
multiple write operations as a single batch that contains any combination of
set()
, update()
, or delete()
operations.
Each operation in the batch counts separately towards your
Firestore usage. A batch of writes completes
atomically and can write to multiple documents. The following
example shows how to build and commit a write batch:
Web version 9
import { writeBatch, doc } from "firebase/firestore"; // Get a new write batch const batch = writeBatch(db); // Set the value of 'NYC' const nycRef = doc(db, "cities", "NYC"); batch.set(nycRef, {name: "New York City"}); // Update the population of 'SF' const sfRef = doc(db, "cities", "SF"); batch.update(sfRef, {"population": 1000000}); // Delete the city 'LA' const laRef = doc(db, "cities", "LA"); batch.delete(laRef); // Commit the batch await batch.commit();
Web version 8
// Get a new write batch var batch = db.batch(); // Set the value of 'NYC' var nycRef = db.collection("cities").doc("NYC"); batch.set(nycRef, {name: "New York City"}); // Update the population of 'SF' var sfRef = db.collection("cities").doc("SF"); batch.update(sfRef, {"population": 1000000}); // Delete the city 'LA' var laRef = db.collection("cities").doc("LA"); batch.delete(laRef); // Commit the batch batch.commit().then(() => { // ... });
Swift
// Get new write batch let batch = db.batch() // Set the value of 'NYC' let nycRef = db.collection("cities").document("NYC") batch.setData([:], forDocument: nycRef) // Update the population of 'SF' let sfRef = db.collection("cities").document("SF") batch.updateData(["population": 1000000 ], forDocument: sfRef) // Delete the city 'LA' let laRef = db.collection("cities").document("LA") batch.deleteDocument(laRef) // Commit the batch do { try await batch.commit() print("Batch write succeeded.") } catch { print("Error writing batch: \(error)") }
Objective-C
// Get new write batch FIRWriteBatch *batch = [self.db batch]; // Set the value of 'NYC' FIRDocumentReference *nycRef = [[self.db collectionWithPath:@"cities"] documentWithPath:@"NYC"]; [batch setData:@{} forDocument:nycRef]; // Update the population of 'SF' FIRDocumentReference *sfRef = [[self.db collectionWithPath:@"cities"] documentWithPath:@"SF"]; [batch updateData:@{ @"population": @1000000 } forDocument:sfRef]; // Delete the city 'LA' FIRDocumentReference *laRef = [[self.db collectionWithPath:@"cities"] documentWithPath:@"LA"]; [batch deleteDocument:laRef]; // Commit the batch [batch commitWithCompletion:^(NSError * _Nullable error) { if (error != nil) { NSLog(@"Error writing batch %@", error); } else { NSLog(@"Batch write succeeded."); } }];
Kotlin
Android
val nycRef = db.collection("cities").document("NYC") val sfRef = db.collection("cities").document("SF") val laRef = db.collection("cities").document("LA") // Get a new write batch and commit all write operations db.runBatch { batch -> // Set the value of 'NYC' batch.set(nycRef, City()) // Update the population of 'SF' batch.update(sfRef, "population", 1000000L) // Delete the city 'LA' batch.delete(laRef) }.addOnCompleteListener { // ... }
Java
Android
// Get a new write batch WriteBatch batch = db.batch(); // Set the value of 'NYC' DocumentReference nycRef = db.collection("cities").document("NYC"); batch.set(nycRef, new City()); // Update the population of 'SF' DocumentReference sfRef = db.collection("cities").document("SF"); batch.update(sfRef, "population", 1000000L); // Delete the city 'LA' DocumentReference laRef = db.collection("cities").document("LA"); batch.delete(laRef); // Commit the batch batch.commit().addOnCompleteListener(new OnCompleteListener<Void>() { @Override public void onComplete(@NonNull Task<Void> task) { // ... } });
Dart
// Get a new write batch final batch = db.batch(); // Set the value of 'NYC' var nycRef = db.collection("cities").doc("NYC"); batch.set(nycRef, {"name": "New York City"}); // Update the population of 'SF' var sfRef = db.collection("cities").doc("SF"); batch.update(sfRef, {"population": 1000000}); // Delete the city 'LA' var laRef = db.collection("cities").doc("LA"); batch.delete(laRef); // Commit the batch batch.commit().then((_) { // ... });
Java
Python
Python
(Async)
C++
// Get a new write batch WriteBatch batch = db->batch(); // Set the value of 'NYC' DocumentReference nyc_ref = db->Collection("cities").Document("NYC"); batch.Set(nyc_ref, {}); // Update the population of 'SF' DocumentReference sf_ref = db->Collection("cities").Document("SF"); batch.Update(sf_ref, {{"population", FieldValue::Integer(1000000)}}); // Delete the city 'LA' DocumentReference la_ref = db->Collection("cities").Document("LA"); batch.Delete(la_ref); // Commit the batch batch.Commit().OnCompletion([](const Future<void>& future) { if (future.error() == Error::kErrorOk) { std::cout << "Write batch success!" << std::endl; } else { std::cout << "Write batch failure: " << future.error_message() << std::endl; } });
Node.js
Go
PHP
Unity
WriteBatch batch = db.StartBatch(); // Set the data for NYC DocumentReference nycRef = db.Collection("cities").Document("NYC"); Dictionary<string, object> nycData = new Dictionary<string, object> { { "name", "New York City" } }; batch.Set(nycRef, nycData); // Update the population for SF DocumentReference sfRef = db.Collection("cities").Document("SF"); Dictionary<string, object> updates = new Dictionary<string, object> { { "Population", 1000000} }; batch.Update(sfRef, updates); // Delete LA DocumentReference laRef = db.Collection("cities").Document("LA"); batch.Delete(laRef); // Commit the batch batch.CommitAsync();
C#
Ruby
Like transactions, batched writes are atomic. Unlike transactions, batched writes do not need to ensure that read documents remain un-modified which leads to fewer failure cases. They are not subject to retries or to failures from too many retries. Batched writes execute even when the user's device is offline.
A batched write with hundreds of documents might require many index updates and might exceed the limit on transaction size. In this case, reduce the number of documents per batch. To write a large number of documents, consider using a bulk writer or parallelized individual writes instead.
Data validation for atomic operations
For mobile/web client libraries, you can validate data using
Firestore Security Rules. You can ensure that related documents are
always updated atomically and always as part of a transaction or batched write.
Use the getAfter()
security rule function to access and validate
the state of a document after a set of operations completes but before
Firestore commits the operations.
For example, imagine that the database for the cities
example also contains a
countries
collection. Each country
document uses a last_updated
field to
keep track of the last time any city related to that country was updated. The
following security rules require that an update to a city
document must also
atomically update the related country's last_updated
field:
service cloud.firestore { match /databases/{database}/documents { // If you update a city doc, you must also // update the related country's last_updated field. match /cities/{city} { allow write: if request.auth != null && getAfter( /databases/$(database)/documents/countries/$(request.resource.data.country) ).data.last_updated == request.time; } match /countries/{country} { allow write: if request.auth != null; } } }
Security rules limits
In security rules for transactions or batched writes, there is a limit of 20 document access calls for the entire atomic operation in addition to the normal 10 call limit for each single document operation in the batch.
For example, consider the following rules for a chat application:
service cloud.firestore { match /databases/{db}/documents { function prefix() { return /databases/{db}/documents; } match /chatroom/{roomId} { allow read, write: if request.auth != null && roomId in get(/$(prefix())/users/$(request.auth.uid)).data.chats || exists(/$(prefix())/admins/$(request.auth.uid)); } match /users/{userId} { allow read, write: if request.auth != null && request.auth.uid == userId || exists(/$(prefix())/admins/$(request.auth.uid)); } match /admins/{userId} { allow read, write: if request.auth != null && exists(/$(prefix())/admins/$(request.auth.uid)); } } }
The snippets below illustrate the number of document access calls used for a few data access patterns:
// 0 document access calls used, because the rules evaluation short-circuits // before the exists() call is invoked. db.collection('user').doc('myuid').get(...); // 1 document access call used. The maximum total allowed for this call // is 10, because it is a single document request. db.collection('chatroom').doc('mygroup').get(...); // Initializing a write batch... var batch = db.batch(); // 2 document access calls used, 10 allowed. var group1Ref = db.collection("chatroom").doc("group1"); batch.set(group1Ref, {msg: "Hello, from Admin!"}); // 1 document access call used, 10 allowed. var newUserRef = db.collection("users").doc("newuser"); batch.update(newUserRef, {"lastSignedIn": new Date()}); // 1 document access call used, 10 allowed. var removedAdminRef = db.collection("admin").doc("otheruser"); batch.delete(removedAdminRef); // The batch used a total of 2 + 1 + 1 = 4 document access calls, out of a total // 20 allowed. batch.commit();
For more information on how to resolve latency issues caused by large writes and batched writes, errors due to contention from overlapping transactions, and other issues consider checking out the troubleshooting page.