spring_boot_starter_cache 0.5.0

Cache starter for spring_boot_dart: @Cacheable, @CachePut, @CacheEvict and @Caching on bean methods, SpEL keys, a CacheManager with TTL and size limits.

spring_boot_starter_cache

No backward compatibility before 1.0. Any release may rename or change APIs, properties and behaviour; the breaking changes are listed in each package's CHANGELOG.

Caching for spring_boot_dart applications: put @Cacheable on a bean method, and repeated calls with the same arguments return the stored result instead of running the method again.

Some methods are slow or costly and give the same answer for the same arguments: a network call, a database query, a heavy computation. Keeping their results by hand means a map per method, keys, invalidation and expiry, mixed into the business code. With this starter you annotate the method; the cache sits in front of it. The results live in memory by default, with an optional time to live and size limit, and you can put your own storage behind the same annotations.

It is the Dart counterpart of spring-boot-starter-cache, for servers and Flutter apps alike.

@Service()
class Books {
  Books(this.api);
  final CatalogApi api;

  // Slow: the API is called once per isbn, later calls read the cache.
  @Cacheable('books')
  Future<Book> find(String isbn) => api.load(isbn);

  // The book is gone: so is its cache entry.
  @CacheEvict('books', key: '#isbn')
  Future<void> delete(String isbn) => api.delete(isbn);
}

In the guides: server, 8. Cache and scheduling, app, 8. Cache & scheduling.

Setup

dependencies:
  spring_boot_starter_cache: ^0.5.0

dev_dependencies:
  build_runner: ^2.17.0
  spring_core_dart_generator: ^0.5.0   # required: generates the proxies caching runs on

dart run build_runner build

Add the dependency and run the generator again: the starter is found and switched on automatically, there is no @EnableCaching.

One import is enough: package:spring_boot_starter_cache/spring_boot_starter_cache.dart re-exports spring_boot_dart and spring_boot_starter_logging. The starter is the only dependency an application needs: it brings the others, logging included, as every Spring Boot starter does.

Calls through the proxy only

The generator writes an AOP proxy for each bean (generator option aop_proxies, all by default), and the starter registers an advisor that wraps every method with a cache annotation. So, exactly as in Spring:

• the cache applies to calls from other beans, which go through the proxy;
• a call through this inside the same class goes straight to the method and skips the cache. To call a cached method of your own class, inject the bean into itself with @Lazy():

@Service()
class Books implements BookService {
  Books(@Lazy() this._self); // the proxy of this same bean
  final BookService _self;

  @Cacheable('books')
  @override
  Book find(String isbn) => ...;

  List<Book> findAll(List<String> isbns) =>
      [for (final isbn in isbns) _self.find(isbn)]; // cached; find(isbn) would not be
}

Annotations

Annotation	What it does
@Cacheable('books')	Returns the cached result when there is one; otherwise runs the method and stores the result.
@CachePut('books', key: '#isbn')	Always runs the method and stores the result, to refresh the entry. Next to a @Cacheable hit, the method runs anyway only when the put's condition passes (a condition reading #result counts as passing), as in Spring.
@CacheEvict('books', key: '#isbn')	Removes the entry after the method returned. allEntries: true clears the whole cache; beforeInvocation: true evicts before the call, so also when it throws.
@Caching(cacheable: [...], put: [...], evict: [...])	Several operations on one method. Repeating an annotation works too.
@CacheConfig(cacheNames: ['books'])	On the class: default cache names, keyGenerator, cacheManager and cacheResolver for its methods.

@Cacheable, @CachePut, @CacheEvict and @Caching on a class apply to every public method that has no cache annotation of its own (not to getters, setters or toString), as in Spring.

Class-level annotations and @CacheConfig are read from the class that declares the method's body and, when it has none, from its superclasses (the nearest wins), as in Spring. So a base class with @CacheConfig or @Cacheable serves the beans that extend it, while a subclass's class-level annotations do not apply to the methods it inherits.

@Cacheable('books') names one cache. For several caches, or to take them from @CacheConfig, use the .of constructor: @Cacheable.of(cacheNames: ['books', 'isbn']) or @Cacheable.of(key: '#isbn'); the same for CachePut.of and CacheEvict.of.

Options of @Cacheable (and of the others, where they make sense):

Option	Meaning
key	SpEL computing the key. Default: the arguments (see Keys).
condition	SpEL; the cache is used only when it is true.
unless	SpEL with #result; the result is not stored when it is true. It does not prevent a hit.
sync	For Future methods: concurrent calls with the same key share one call (see Async methods).
keyGenerator	The name of a KeyGenerator bean computing the key (not with key).
cacheManager	The name of the CacheManager bean to use.
cacheResolver	The name of a CacheResolver bean that picks the caches (not with cacheManager).

@CacheConfig fills in only what an operation leaves open, as in Spring: its cache names when the operation names none, its keyGenerator when the operation sets neither key nor keyGenerator, and its cacheResolver (else its cacheManager) when the operation sets neither of those.

The annotations of every proxied bean are checked when the bean is created, so a wrong combination fails the startup, naming the member, rather than the first call: key with keyGenerator, cacheManager with cacheResolver, sync with other operations, several caches or unless, an operation without cache names (unless a cache resolver picks them), @Cacheable / @CachePut on a method returning a Stream.

Keys

Without key, the key is made of the arguments, as Spring's SimpleKeyGenerator does: no arguments → SimpleKey.empty, one argument → the argument itself, several → SimpleKey([...]), which compares by content (lists, sets and maps inside too).

key, condition and unless are SpEL expressions, written plain, without #{}:

Expression	Value
#isbn	the argument named isbn
#p0, #a0	the first argument
#result	the result (in unless; in key and condition of @CachePut; in @CacheEvict after the call)
#root.methodName	find
#root.method.name	find (also returnType, parameterTypes, parameterNames, parameterCount)
#root.args	all arguments, #root.args[1]
#root.target	the bean
#root.targetClass	the bean's class name, Books
#root.caches	the cache names of the operation
@region	the bean named region

A key that comes out null (from key or a key generator) fails the call with an ArgumentError, as in Spring.

@Cacheable('books', key: '#isbn')                         // ignore the other arguments
Book find(String isbn, bool includeReviews) => ...;

@Cacheable('books', key: "#isbn + '-' + #lang")            // a combined key
Book translated(String isbn, String lang) => ...;

@Cacheable('books', condition: '#isbn.length() == 13')     // cache 13-digit ISBNs only
Book findIsbn(String isbn) => ...;

@Cacheable('books', unless: '#result == null')            // do not remember misses
Book? maybe(String isbn) => ...;

@CachePut('books', key: '#result.isbn')                   // Book is @Introspected
Book save(Book book) => ...;

Reading properties or calling methods of your own objects (#book.isbn, #result.isbn, @clock.zone()) needs @Introspected() on their class: Dart has no reflection, so the generator writes the accessors SpEL uses. Strings, numbers, lists and maps work as they are (#isbn.length(), #tags.size()).

Async methods

Methods returning a Future cache the value, not the future: the value is stored when the future completes, a hit returns an already completed Future, and errors are never cached, so the next call tries again.

// Callers asking for the same price at the same time share one request.
@Cacheable('prices', sync: true)
Future<double> price(String isbn) => api.price(isbn);

With sync: true, calls that arrive while the first one is still running wait for it instead of calling api.price again. As in Spring, a sync operation must be the only cache annotation of the method, use one cache and have no unless. The waiting is shared per cache and key, not per method: two sync methods that use the same cache and compute the same key share one pending result, even though they are different methods.

A Stream cannot be cached: it can be listened to once, so the cached stream would fail every later caller. @Cacheable and @CachePut on a method returning a Stream fail the build (and the startup, for hand-written proxies). Collect the values instead:

@Cacheable('events')
Future<List<Event>> recent() => api.events().toList();

The default: caches in memory

Without any setup the starter creates the cacheManager bean, an InMemoryCacheManager (Spring's ConcurrentMapCacheManager):

• each cache is an InMemoryCache, a map in the memory of the running app, created the first time an annotation names it ('books', ...);
• entries are lost when the app stops, and when the context restarts (also on a hot reload that changes beans): the cache starts empty;
• nothing expires and there is no size limit unless you set them (below); the time to live is measured on a monotonic clock, as Caffeine's ticker, so setting the device's or server's clock neither keeps entries longer nor drops them early; expired entries are dropped when they are read, and writes sweep out the others now and then (after a burst of writes, the writes a time to live or two later do; a cache no longer written keeps them until they are read);
• null results are stored too: a stored null is a hit, as in Spring;
• one app, one isolate: the cache is not shared between processes, isolates or devices.

That is right for most caches: results of slow calls that are cheap to recompute after a restart.

Properties

The default manager is built from spring.cache.*, only when the application declares no CacheManager of its own:

Property	Default	Meaning
spring.cache.type	generic with Cache beans, else simple	generic: a SimpleCacheManager of the Cache beans (at least one needed); simple: caches in memory; none: cache nothing (the annotations stay, the methods always run). Another value fails the start
spring.cache.cache-names	—	a fixed set of caches, such as books,authors or a YAML list; an annotation naming another cache then fails. Without it (or empty), caches are created on first use
spring.cache.simple.time-to-live	none	how long entries live: 10m, 30s, PT1H, or milliseconds. 0 or more: 0 keeps nothing (as Caffeine's expireAfterWrite(Duration.ZERO)); a negative value fails the start with an error naming the property, as Caffeine refuses one
spring.cache.simple.max-size	none	the most entries per cache; the least recently used go first. 0 or more: 0 keeps nothing (as Caffeine's maximumSize(0)); a negative value fails the start with an error naming the property
spring.cache.simple.caches.<name>.time-to-live	—	the same for one cache (also checked at the start)
spring.cache.simple.caches.<name>.max-size	—	the same for one cache (also checked at the start)

A cache name with dots goes in brackets, as in any Spring Boot map (spring.cache.simple.caches[books.byIsbn].max-size=2; in YAML the key "[books.byIsbn]":); written plainly (spring.cache.simple.caches.books.byIsbn.max-size) it works too: the property is the end of the key (max-size, time-to-live, or a relaxed form: maxSize, max_size, MAXSIZE), the rest the cache name. Every per-cache value is checked at the start, whether the cache exists yet or not.

In environment variables, as in Spring Boot: SPRING_CACHE_SIMPLE_CACHES_BOOKS_MAX_SIZE=5 or ..._BOOKS_MAXSIZE=5, SPRING_CACHE_SIMPLE_CACHES_PRICES_TIME_TO_LIVE=30s or ..._PRICES_TIMETOLIVE=30s. The cache name comes out in lower case, so a name with capitals or dashes cannot be set this way (in Boot neither); a _ inside it is a dot here (..._BOOKS_BYISBN_MAX_SIZE sets the cache books.byisbn; Boot binds nothing for it).

# lib/resources/application.properties
spring.cache.simple.time-to-live=10m
spring.cache.simple.max-size=1000
spring.cache.simple.caches.prices.time-to-live=30s     # prices go stale faster

Inject the manager to look at or clear caches by hand:

// An admin action: drop every cached book.
@Component()
class CacheAdmin {
  CacheAdmin(this.caches);
  final CacheManager caches;

  void clearBooks() => caches.getCache('books')?.clear();
}

To adjust the starter's manager without replacing it, for example to create some caches at startup, declare a CacheManagerCustomizer bean: each one is called once with the new manager. The function cacheManagerFromProperties(environment) builds the same manager for a CacheManager bean of your own, such as one that wraps it.

Your own cache or manager

The annotations only talk to two interfaces, so anything can stand behind them:

Type	What it does	Built in
CacheManager	finds a cache by name (getCache('books'))	InMemoryCacheManager, SimpleCacheManager, NoOpCacheManager
Cache	one cache: get, put, evict, clear, ...	InMemoryCache, NoOpCache

Declare a CacheManager bean and the starter's one steps aside (it is created only when there is none). The spring.cache.* properties then no longer apply: your bean decides.

1. The default, configured in code

@Configuration()
class CacheSetup {
  @Bean()
  CacheManager cacheManager() => InMemoryCacheManager(
    cacheNames: ['books', 'prices'],                     // only these two
    defaultSpec: const CacheSpec(timeToLive: Duration(minutes: 10)),
    specs: {'prices': const CacheSpec(timeToLive: Duration(seconds: 30))},
  );
}

2. Your own storage

To keep entries somewhere else, such as a file or a key-value database so they survive a restart, implement Cache. With the CacheDefaults mixin four methods are enough; putIfAbsent, evictIfPresent and invalidate are derived from them. Then hand your caches to a SimpleCacheManager (Spring's SimpleCacheManager), alone or mixed with in-memory ones:

import 'dart:convert';

/// Any synchronous key-value storage: a file, a key-value database, ...
abstract interface class KeyValueStore {
  String? read(String key);
  void write(String key, String value);
  void delete(String key);
  void deleteAll({required String prefix});
}

/// Entries kept in the store as JSON, so they survive a restart.
final class StoreCache with CacheDefaults implements Cache {
  StoreCache(this.name, this.store);

  @override
  final String name;
  final KeyValueStore store;

  String _key(Object? key) => '$name:$key';

  @override
  ValueWrapper? get(Object? key) => switch (store.read(_key(key))) {
    null => null,                               // a miss: the method runs
    final json => ValueWrapper(jsonDecode(json)), // a hit (also a stored null)
  };

  @override
  void put(Object? key, Object? value) =>
      store.write(_key(key), jsonEncode(value));

  @override
  void evict(Object? key) => store.delete(_key(key));

  @override
  void clear() => store.deleteAll(prefix: '$name:');
}

@Configuration()
class CacheSetup {
  @Bean()
  CacheManager cacheManager(KeyValueStore store) => SimpleCacheManager([
    StoreCache('books', store),   // kept in the store
    InMemoryCache('prices'),      // in memory
  ]);
}

What to keep in mind:

• Return ValueWrapper for a hit and null for a miss. ValueWrapper(null) is a cached null; returning plain null makes the method run again.

• Keys are the method arguments (one argument: the argument itself; several: a SimpleKey), or what key: computes. A store that needs strings must turn them into strings: '$key' works for strings, numbers and SimpleKey; for your own classes override toString, or set key: '#isbn' so the key is a string anyway.

• Values are whatever the methods return. A store outside memory must serialize them: JSON for maps, lists, strings and numbers; your classes need toJson / fromJson (decode in get into the right type).

• invalidate() of CacheDefaults always returns true, also for an empty cache. The annotations never call it; implement it yourself if your code needs to know whether there were entries.

• The Cache interface is synchronous, as in Spring. For an asynchronous store (SQLite, shared_preferences in Flutter, a server), keep the entries in memory, load the stored ones in the background (until then calls are misses: the methods simply run) and write changes back in the background:

  import 'dart:async'; // unawaited
  
  final class WriteBehindCache with CacheDefaults implements Cache {
    WriteBehindCache(this.name, this.store) {
      unawaited(
        store.readAll(prefix: '$name:').then((entries) {
          entries.forEach((key, json) => _memory.putIfAbsent(
            key.substring(name.length + 1),
            () => jsonDecode(json),
          ));
        }),
      );
    }
  
    @override
    final String name;
    final AsyncStore store;               // Future-based API
    final _memory = <String, Object?>{};
  
    @override
    ValueWrapper? get(Object? key) =>
        _memory.containsKey('$key') ? ValueWrapper(_memory['$key']) : null;
  
    @override
    void put(Object? key, Object? value) {
      _memory['$key'] = value;
      unawaited(store.write('$name:$key', jsonEncode(value)));
    }
  
    @override
    void evict(Object? key) {
      _memory.remove('$key');
      unawaited(store.delete('$name:$key'));
    }
  
    @override
    void clear() {
      _memory.clear();
      unawaited(store.deleteAll(prefix: '$name:'));
    }
  }
  

• Errors thrown by your cache reach the caller of the cached method (Spring's default too); catch them inside the cache if a broken store should only mean "no cache", or see CacheErrorHandler in section 7.

3. Caches created on demand

SimpleCacheManager knows only the caches it was given: @Cacheable('other') then fails with "Cannot find cache named 'other'". To create caches when they are first named, implement CacheManager:

final class StoreCacheManager implements CacheManager {
  StoreCacheManager(this.store);
  final KeyValueStore store;
  final _caches = <String, Cache>{};

  @override
  Cache? getCache(String name) =>
      _caches.putIfAbsent(name, () => StoreCache(name, store));

  @override
  Iterable<String> get cacheNames => _caches.keys;
}

4. Several managers

To keep the starter's manager (or your main one) for most methods and use another one for some, give the other one a name, hide it from injection by type (autowireCandidate: false), and name it in the annotation:

@Configuration()
class CacheSetup {
  // Sessions are kept in the store; everything else stays in memory.
  @Bean(autowireCandidate: false)
  CacheManager sessionCaches(KeyValueStore store) =>
      SimpleCacheManager([StoreCache('sessions', store)]);
}

@Cacheable('sessions', cacheManager: 'sessionCaches')
Session load(String token) => ...;

5. Your own keys

When the arguments do not make a good key as they are, implement KeyGenerator, declare it as a bean and name it in the annotation:

// "978-0-00" and "978000" are the same book.
final class IsbnKeys implements KeyGenerator {
  const IsbnKeys();

  @override
  Object? generate(Object target, MethodSignature method, List<Object?> args) =>
      (args.first! as String).replaceAll('-', '');
}

@Configuration()
class Keys {
  @Bean()
  KeyGenerator isbnKeys() => const IsbnKeys();
}

@Cacheable('books', keyGenerator: 'isbnKeys')
Book find(String isbn) => ...;

6. Choosing caches per call: CacheResolver

A CacheResolver decides at each call which caches an operation uses, for example one cache per tenant. It sees the operation (its cache names, key, ...), the bean, the method and the arguments. Declare it as a bean and name it with cacheResolver:, in the annotation or in @CacheConfig:

// Each tenant gets its own caches: "acme.books", "globex.books", ...
final class TenantCacheResolver implements CacheResolver {
  TenantCacheResolver(this.manager, this.tenants);
  final CacheManager manager;
  final TenantContext tenants;

  @override
  Iterable<Cache> resolveCaches(CacheOperationInvocationContext context) => [
    for (final name in context.operation.cacheNames)
      manager.getCache('${tenants.current}.$name')!,
  ];
}

@Configuration()
class CacheSetup {
  @Bean()
  CacheResolver tenantCacheResolver(CacheManager manager, TenantContext tenants) =>
      TenantCacheResolver(manager, tenants);
}

@Cacheable('books', cacheResolver: 'tenantCacheResolver')
Book find(String isbn) => ...;

Spring (Java):

@Cacheable(cacheNames = "books", cacheResolver = "tenantCacheResolver")
Book find(String isbn) { ... }

Built in: SimpleCacheResolver(manager) (the default: the operation's names in the cache manager) and NamedCacheResolver(manager, ['audit']) (always the same caches; the operations then need no names). As in Spring Boot, the starter's cacheManager backs off when a bean named cacheResolver exists.

7. Defaults for everything: CachingConfigurer

One bean implementing CachingConfigurer sets the defaults of all cache operations; a method returning null (the default) keeps the starter's:

Method	Default
cacheManager()	the CacheManager bean
cacheResolver()	SimpleCacheResolver of the cache manager
keyGenerator()	SimpleKeyGenerator
errorHandler()	SimpleCacheErrorHandler: cache errors reach the caller

@Configuration()
class CacheSetup with CachingConfigurer {
  @override
  KeyGenerator keyGenerator() => const IsbnKeys();

  // A cache that is down: log, run the method, carry on.
  @override
  CacheErrorHandler errorHandler() => const LoggingCacheErrorHandler();
}

Spring (Java):

@Configuration
@EnableCaching
class CacheSetup implements CachingConfigurer {
  @Override
  public KeyGenerator keyGenerator() { return new IsbnKeys(); }

  @Override
  public CacheErrorHandler errorHandler() { return new LoggingCacheErrorHandler(); }
}

The methods are called once, when caching is first used. Return objects (or beans injected into the configuration) rather than marking the methods @Bean: Dart does not proxy @Configuration classes, so calling a @Bean method creates a second instance.

With a CacheErrorHandler that does not rethrow, a failed get counts as a miss (the method runs) and a failed put, evict or clear is skipped. LoggingCacheErrorHandler logs them at warning level to spring.cache (logStackTraces: true adds the stack trace).

Testing

Cached results carry over between calls, which can hide behaviour in tests. Switch caching off with a property:

final app = springBootTest(
  springApplication,
  properties: {'spring.cache.type': 'none'},
);

Or keep it on and check the cache through the CacheManager bean: app.getBean<CacheManager>().getCache('books')!.get('978-0').

Time to live in tests

The time to live runs on a monotonic stopwatch, in real time. A clock given to InMemoryCache or InMemoryCacheManager moves expiry forward but no longer holds entries alive (since 0.5.0): an entry expires once its time to live has passed on the stopwatch or on the clock. With a fixed clock and a time to live of 1 second, 0.4.3 still returned an entry 1.2 seconds later; now it is gone. So, to test expiry:

• Move a clock, don't wait. A clock you move by hand needs nothing else:

  var now = DateTime.utc(2026);
  final cache = InMemoryCache('books',
      spec: const CacheSpec(timeToLive: Duration(minutes: 10)),
      clock: () => now);
  cache.put('978-0', book);
  now = now.add(const Duration(minutes: 9));
  expect(cache.get('978-0')?.value, book);
  now = now.add(const Duration(minutes: 1)); // the whole time to live
  expect(cache.get('978-0'), isNull);
  

• fake_async alone expires nothing. async.elapse moves fake time, not the stopwatch: a cache without a clock keeps its entries however far you elapse. Give it a clock that follows fake time, () => clock.now() (package clock, which fakeAsync fakes) or async.getClock(start).now, and write and read inside the same fakeAsync. Write the closure: the tear-off clock.now keeps the clock of the zone it was taken in, the real one outside fakeAsync.

  fakeAsync((async) {
    final cache = InMemoryCache('books',
        spec: const CacheSpec(timeToLive: Duration(minutes: 10)),
        clock: () => clock.now());
    cache.put('978-0', book);
    async.elapse(const Duration(minutes: 9));
    expect(cache.get('978-0')?.value, book);
    async.elapse(const Duration(minutes: 1));
    expect(cache.get('978-0'), isNull);
  });
  

• Choose a long time to live. Real time expires entries too, whatever the clock, so a time to live of a second can expire an entry the test still expects on a slow machine. Give minutes and move the clock past them. A clock set back before the write also counts as expired (it tells nothing then), so don't start a fake clock earlier than the writes it reads.

• In an application test, the cacheManager the starter makes from the properties has no clock: declare your own InMemoryCacheManager bean with the specs and a clock (it replaces the starter's), then move that clock.

Differences from Spring

• No @EnableCaching: adding the starter turns caching on; spring.cache.type=none turns it off.
• @Cacheable('books') takes one name; several names go to @Cacheable.of(cacheNames: [...]), because a Dart constructor cannot have both an optional positional parameter and named ones.
• #root.caches holds cache names, #root.targetClass the class name, #root.method the MethodSignature (name, returnType, ... as text).
• CacheManagerCustomizer is not generic: every customizer gets the manager and checks its type. The container finds beans by their exact type, so CacheManagerCustomizer<InMemoryCacheManager> could not be matched against a manager.
• Future results are cached by value (Spring 6.1 does the same for CompletableFuture).
• The sync wait is shared per cache and key, so two different sync methods with the same cache and key share one pending result.
• Built-in store: memory, with time to live and size limit. Spring's ConcurrentMapCacheManager has neither; it needs Caffeine for that.
• spring.cache.simple.time-to-live, spring.cache.simple.max-size and spring.cache.simple.caches.<name>.* are this starter's: Spring Boot's simple type has no settings (time to live and size come from Caffeine or Redis there, under spring.cache.caffeine.* / spring.cache.redis.*). spring.cache.type knows generic, simple and none only.
• SimpleKeyGenerator wraps a single collection argument in a SimpleKey, because Dart lists, sets and maps compare by identity and the key compares by content; Spring wraps only arrays.
• Stream results cannot be cached: @Cacheable / @CachePut on a method returning a Stream fail the build. Spring 6.1 collects a Flux into a list and replays it; a generic interceptor cannot create a Stream<T> of the right type without reflection. Return a Future<List<T>>.
• The Cache API is a subset of Spring's: name, get, put, putIfAbsent, evict, evictIfPresent, clear, invalidate. There is no get(key, Class), get(key, Callable) (value loader), retrieve (asynchronous) or getNativeCache.
• invalidate() returns true for a cache that had nothing left to remove in two cases: CacheDefaults always returns true, and InMemoryCache counts entries that have expired but were neither read nor swept out yet.
• CachingConfigurer methods are called once; they must not be @Bean methods (no @Configuration proxies in Dart). CacheErrorHandler methods receive the stack trace next to the error.
• Annotations are checked when the bean is created, for every member; the sync rules too (Spring checks those at the first call).