google / guice Goto Github PK

From kevinb9n on March 02, 2007 01:03:34

Support a development mode that reports to you how much time was spent
initializing instances that were never used.  I think this would help
people find refactorings that will speed up their runtime performance and
test runs.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=50

From crazyboblee on August 25, 2006 00:46:05

To support web applications.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=5

From kevinb9n on February 27, 2007 03:03:06

Right now if a class is, say, @ContainerScoped you can override this in the
binding to give a different scope like @SessionScoped but you can't say
that you want it to have no scope.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=35

From kevinb9n on February 27, 2007 03:21:14

It would be good if certain keys could be multivalued; where multiple
modules could append one or more bindings each, and these could all be
injected as a Collection<Foo> or something.  This would be useful for some
things Laura is doing, and for replacing excalibur, and for
AdWordsPaymentOptions, and . . .

I'm going ahead and calling this priority "high" for now.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=37

From crazyboblee on February 15, 2007 21:35:22

For example:

@ContainerScoped(preload=true)
@SessionScoped(transient=true)

Original issue: http://code.google.com/p/google-guice/issues/detail?id=20

From kevinb9n on February 22, 2007 14:19:44

When a chain of dependencies produces a LocationException you get a whole
chain of LocationExceptions; instead we could keep pushing additional
context onto the initial LEx, then display a more unified and readable
report to the user.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=23

From kevinb9n on February 27, 2007 02:59:57

if anything is thrown out of Module.configure(), the binder might want to
catch that and record it with addError() instead of bombing out completely.
 However, this may not be so great either as there could be a ripple effect
of other errors that happen only because of the missing bindings from this
module.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=33

From crazyboblee on February 14, 2007 14:10:28

Proxy objects in a given scope during development and throw an exception if
the object is used out of scope.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=11

From kevinb9n on February 20, 2007 01:04:08

Write unit tests for error conditions / messages.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=21

From crazyboblee on February 15, 2007 13:47:46

Write an APT processor which generates an error if an annotation is
annotated with @Binder but not @Retention(RUNTIME).

Original issue: http://code.google.com/p/google-guice/issues/detail?id=19

From kevinb9n on February 27, 2007 03:28:48

Support eager loading generically for any scope.  Rename
eagerlyInContainer() to just eagerly().

Original issue: http://code.google.com/p/google-guice/issues/detail?id=38

From kevinb9n on February 26, 2007 22:17:14

if module A installs B and C and B and C both install D, you'll get errors
and you won't be able to create the container.  We should be able to detect
this situation somehow and handle it correctly.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=31

From crazyboblee on February 27, 2007 20:19:26

We can leave behind a half constructed implicit binding.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=42

From crazyboblee on August 25, 2006 00:48:18

For example a request scope object into a session scoped object. Should we
use proxies, throw exceptions?

Original issue: http://code.google.com/p/google-guice/issues/detail?id=6

From kevinb9n on February 26, 2007 22:14:47

It's possible someone did something in their Module that caught an
exception, and they should be able to include this in addError(), shouldn't
they?

Original issue: http://code.google.com/p/google-guice/issues/detail?id=30

From crazyboblee on February 14, 2007 18:36:32

So IDEs don't expose the classes to users.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=16

From kevinb9n on February 27, 2007 01:58:35

Someone making such a binding seems like it could only lead to pain ("so
much pain").  We should probably forbid it?  Without breaking the unit tests.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=32

From kevinb9n on February 27, 2007 03:35:57

This may be a long shot.  Just floating it out there.

Theoretically it's possible no one would ever need
Container.getProvider() because they can always write a nested class with
the @Injections they need and use Container.injectMembers().  The appeal of
this is that they're then specifying the binding they want in the "usual"
way, not with some crappy "Key.get(Foo.class, new
AnnotationImpl(god_knows_what?))" syntax.  There's less for people to learn
and stuff.

...

Original issue: http://code.google.com/p/google-guice/issues/detail?id=39

From crazyboblee on February 25, 2007 13:43:43

Like in Eclipse's debugger so you can keep iterating during development
without restarting. We can disable this in production mode and run full speed.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=26

From crazyboblee on February 25, 2007 17:38:40

So they know who they're creating an instance for.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=27

From crazyboblee on February 22, 2007 14:13:17

Maybe support conversion for normal bindings, too.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=22

From limpbizkit on February 14, 2007 19:06:44

Guice makes classes like util/Objects available in its public API, even
though they are implementation details.

If you think it's worthwhile, the .jar for Guice could hide these classes
from OSGi-compliant tools like Eclipse. Here's a sample Ant task that
creates the necessary meta information for the .jar file:

  <taskdef name="osgiBundleInfo"
classname="org.knopflerfish.ant.taskdefs.bundle.BundleInfoTask"
classpath="tools/OSGiBundleInfo.jar"/>
  <osgiBundleInfo exports="packagelist">
     <fileset dir="${basedir}">
        <include name="/source//.java"/>
        <exclude name="/util//.java"/>
     </fileset>
  </osgiBundleInfo>

Original issue: http://code.google.com/p/google-guice/issues/detail?id=17

From crazyboblee on August 25, 2006 00:52:07

So you can inject code for which you don't have the source.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=8

From crazyboblee on February 28, 2007 14:09:33

For example, I could pass a Container to a Binder and the resulting
Container would include all the bindings from the original Container. This
can be used the implement layering (i.e. view layer container includes the
application layer container) as well as initialization stages.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=45

From crazyboblee on February 25, 2007 17:59:35

From Ant. This will help with our mult-module build, and we'll be able to
generate our IntelliJ configuration instead of maintaining it separately.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=28

From kevinb9n on February 27, 2007 03:08:09

We should tighten up the API so that you can only call things that are
valid; this will make auto-completion a breeze.  A downside is that this
usually forces us to mandate a particular order that the calls happen in
as well, but I don't think this is a problem.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=36

From kevinb9n on February 28, 2007 15:40:58

I'm a little concerned about Binder, in that 90% of what newbies want to
use it for is simply binding types to classes, instances and providers.
But when they go to look at it (Binder) they're confronted with all this
stuff about interceptors and binding their own custom scopes and installing
other modules and more.  Is there anything we can do about this?

Bob and I already thought about this and came up with nothing good but I
thought it wouldn't hurt to put it here in case someone from the group at
large will give us some ideas.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=46

From crazyboblee on February 14, 2007 17:25:17

Right now we can proxy the root object but not the nested object. This is
low priority. I don't know how I feel about using proxies to enable
circular dependencies anyway.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=15

From kevinb9n on February 28, 2007 17:09:35

It would be helpful if we could configure an injector in a special mode
where whenever it encounters a request for a type it doesn't have bound, it
will just create a dummy for that type.  It could do this with Proxy or
cglib or easy mock or whatever.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=49

From kevinb9n on February 27, 2007 03:01:30

Arguable. Right now if you bind a null you get a NullPointerException and
you won't get to see the rest of the errors that were/would have been
collected.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=34

From crazyboblee on August 25, 2006 00:50:54

Allow isolation and overrides on a per-dependency basis.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=7

From crazyboblee on February 22, 2007 16:37:00

For unit testing.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=25

From kevinb9n on February 27, 2007 14:58:17

Right now you can have a custom provider return null and I think guice will
inject the null value.  We should add sanity checking to make sure guice
never injects null.  This way people do not need to check for it in their
constructors etc., which in turn allows them to use null when they want
to in their unit tests (which is quite a helpful ability to have).

Original issue: http://code.google.com/p/google-guice/issues/detail?id=41

From crazyboblee on February 28, 2007 13:52:31

Something like this would be nice (we may have to wait for closures):

inScope(Scopes.SINGLETON) {
  bind(...).to(...);
  ...
}

Original issue: http://code.google.com/p/google-guice/issues/detail?id=43

From crazyboblee on August 25, 2006 00:42:26

To enable standalone use of the framework.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=3

From kevinb9n on February 28, 2007 15:51:25

This would be a scope that anything that reads from modifiable
configuration sources (database, xml/props file, etc.) could live at.  Then
we'd provide a means by which an administrator can send a reload message to
a running server (JMX, perhaps?).  When the message is received, the
current configuration scope would become a "lame duck" while the objects
for the next configuration scope are all reconstructred; when
reconstruction is finished we'd invalidate the old configuration scope and
start serving requests out of the new.

I wonder if we'd have to be careful that each single request scope uses
only one configuration instance and never a mixture of old and new.

This is the rough idea though; comments?

Original issue: http://code.google.com/p/google-guice/issues/detail?id=48

From crazyboblee on February 28, 2007 15:51:22

Let's make Spring integration trival, so it's easy to share or migrate.
Here's how it will work:

The user must provide a binding to Spring's BeanFactory.

Then, we'll have the following class:

class SpringIntegration {

  static <T> Provider<T> fromSpring(Class<T> type, String name) {
    ...
  }
}

The returned provider will look up the Bean with the given name in the
bound BeanFactory.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=47

From crazyboblee on August 25, 2006 00:57:01

Implement an annotation processor which validates injections at compile
time. Accompany with IDE plugins.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=10

From crazyboblee on February 14, 2007 14:41:00

Sometimes it makes more sense to point to a member than a binding. Chris
Nokleberg implemented LineNumbers to support this.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=14

From crazyboblee on February 14, 2007 14:37:05

Where stages could be like development, production, etc.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=13

From crazyboblee on February 14, 2007 14:13:01

So you can store proxies on the HTTP session, etc.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=12

From crazyboblee on August 25, 2006 00:41:55

Because AOP and DI are a good match.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=2

From crazyboblee on February 15, 2007 00:15:30

And annotation type.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=18

From crazyboblee on February 22, 2007 15:47:21

This will help us catch problems where the user forgets to bind a scope.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=24

From crazyboblee on February 28, 2007 13:56:15

For example:

@DefaultImplementation(ServiceImpl.class)
interface Service {
  ...
}

This can be used when a compile time dependency on the implementation is
acceptable, but you still want the flexibility of proxying, mocking, etc.,
that comes with interfaces.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=44

From kevinb9n on February 26, 2007 22:12:58

To support tools, e.g. graph-drawing.  And what about the inverse,
Set<Binding> getDependents(Key) (in Container?)

Original issue: http://code.google.com/p/google-guice/issues/detail?id=29

From kevinb9n on February 27, 2007 03:45:37

We ought to compile warning-free and only @SuppressWarnings where necessary
and where we can explain how we know the suppression is safe.  I think the
refactoring and commenting necessary to handle this issue will make the
code easiser to understand.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=40

From crazyboblee on August 25, 2006 00:54:20

- Create nested spi package

• Use "dependency" or a shorter term in ContainerBuilder
• Use nested builders in ContainerBuilder instead of overloaded methods
• Break down Scope.Strategy

Original issue: http://code.google.com/p/google-guice/issues/detail?id=9

From crazyboblee on August 25, 2006 00:44:32

So you can dump a bunch of string values and have them converted to enums,
ints, booleans, etc.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=4

From crazyboblee on August 25, 2006 00:41:20

To enable extensions, debugging, tools, etc.

Original issue: http://code.google.com/p/google-guice/issues/detail?id=1