I'd like to get some guidance on correct, clear, and concise initialization of singleton, read-only access objects wrapping cached run-once computations, with either lazy or eager initialization.
I brought this up peripherally in http://groups.google.com/forum/?fromgroups#!topic/google-guice/u0V97-FZBTQ but I wanted to avoid taking my attempted threadjacking any further and am starting a new topic. Here's a sample use case: A Database provides a table of String keys and values. I'd like to provide a read-only get(String)->String access object, and I'd like to initialize it safely, so that the resulting object is thread-safe. I'd like to express this concisely, in a way that is clear to code readers so they will be incented to copy the pattern. Lazy init can be done in Java using the static hack, as descrbied by Bob here http://blog.crazybob.org/2007/01/lazy-loading-singletons.html and in Wikipedia here http://en.wikipedia.org/wiki/Initialization-on-demand_holder_idiom but I'd like to do it using Guice, for all the usual reasons, such as so I can integrate these providers in with normal injection. I hope my problem statement is clear. I'm looking for guidance on the best solution. Here are three unsatisfactory ideas and one maybe OK one. Proposal A: Proposal A does the work of reading the database in the constructor. This code is clear, and it's certainly concise, and as near as I can understand, it's thread-safe. Unfortunately, I belive it isn't correct in Guice because of the problems associated with doing work in constructors (proxy objects, for example). @LazySingleton class CachedThings { private final Map<String,String> cache; @Inject CachedThings(DB db) { this.cache = readCache(db); } public String get(String x) { return cache.get(x); } private Map<String,String> readCache(DB db) { Map<String,String> result = new HashMap<String,String>(); for (Row row : db) { map.add(row.a, row.b); } return result; } } Proposal B: Proposal B splits the data access object from the database reader operation, moving the database read into its own Provider, where it can operate safely. This appears to be just as thread-safe as the previous version, but is considerably less concise. It also exposes a @Named TypeLiteral that is not only ugly, but also opens the possibility of someone directly injecting that Map and causing unwanted expensive database reads. @LazySingleton class CachedThings { private final Map<String,String> cache; @Inject CachedThings(@Named("hack") Map<String,String> cache) { this.cache = cache; } public String get(String x) { return cache.get(x); } static class MyModule extends AbstractModule { protected void configure() { bind(new TypeLiteral<Map<String,String>>(){/**/}.annotatedWith(Names.named("hack")).toProvider(ThingProvider.class).in(LazySingleton.class); } } @LazySingleton static class ThingProvider implements Provider<Map<String,String>> { private final DB db; @Inject ThingProvider(DB db) { this.db = db; } private Map<String,String> get() { Map<String,String> result = new HashMap<String,String>(); for (Row row : db) { map.add(row.a, row.b); } return result; } } Proposal C: Proposal C uses @Provides methods so avoid to the problematic verbosity of TypeLiteral, and indeed the code is cleaner, but we still have the @Named hack and internal data exposure. (I've written the MyModule as a static class of CachedThings, which is questionable, so we might need to deduct a few points for the additional verbosity needed to move the module out.) @LazySingleton class CachedThings { private final Map<String,String> cache; @Inject CachedThings(@Named("hack") Provider<Map<String,String>> thingProvider) { this.cache = thingProvider.get(); } public String get(String x) { return cache.get(x); } static class MyModule extends AbstractModule { protected void configure() { } @LazySingleton @Provides @Named("hack") Provider<Map<String,String>> getThing(DB db) { Map<String,String> result = new HashMap<String,String>(); for (Row row : db) { map.add(row.a, row.b); } return result; } } } Proposal D: I don't know much about injected methods other than that they run after constructors and the results can't be final. Is this correct with regard to Guice initialization sequence? Is it safe for multi-threaded readonly access of the resulting HashMap? Maybe this is the right solution is to use an @Inject setCache(DB) method? @LazySingleton class CachedThings { private Map<String,String> cache; @Inject CachedThings(...) { ... other stuff here if necessary ... } @Inject void setCache(DB db) { Map<String,String> result = new HashMap<String,String>(); for (Row row : db) { map.add(row.a, row.b); } cache = result; } public String get(String x) { return cache.get(x); } } Thank you, Leigh. -- You received this message because you are subscribed to the Google Groups "google-guice" group. To view this discussion on the web visit https://groups.google.com/d/msg/google-guice/-/HUKKPENRVqcJ. To post to this group, send email to [email protected]. To unsubscribe from this group, send email to [email protected]. For more options, visit this group at http://groups.google.com/group/google-guice?hl=en.
