Is there a way to make lazy vals use custom synchronization scheme?

[Eugene Burmako]

synchronized(this) implicitly emitted by lazy val codegen provides an
excellent opportunity for deadlocks.

[Eugene Burmako]

The same questions goes to local objects.

[Eugene Burmako]

So far I have only one idea - explicitly force initialization of all
lazy stuff in JavaUniverse.init. Very tedious, but at least that will
make deadlocks impossible.

[Hubert Plociniczak]

... I realize I didn't reply to internals.

On 02/07/2013 11:06 AM, Hubert Plociniczak wrote:

On 02/07/2013 10:31 AM, Eugene Burmako wrote:

synchronized(this) implicitly emitted by lazy val codegen provides an
excellent opportunity for deadlocks.

not so easy.

https://issues.scala-lang.org/browse/SUGGEST-11

[Hubert Plociniczak]

I don't know much what's going on in JavaUniverse but won't you have the
same problems like in Definitions?
I think Paul or someone else wanted to do s/lazy val/def there and failed.

[Eugene Burmako]

Yes, definitions are the part of the problem. Do you remember why exactly he failed?

--
You received this message because you are subscribed to the Google Groups "scala-internals" group.
To unsubscribe from this group and stop receiving emails from it, send an email to scala-internals+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

[Hubert Plociniczak]

On 02/07/2013 11:34 AM, Eugene Burmako wrote:

Yes, definitions are the part of the problem. Do you remember why exactly he failed?

Sorry, I don't recall the details now. Must be somewhere on the mailing list (try old -devel one).

On 7 February 2013 11:30, Hubert Plociniczak <hubert.pl...@epfl.ch> wrote:

I don't know much what's going on in JavaUniverse but won't you have the same problems like in Definitions?
I think Paul or someone else wanted to do s/lazy val/def there and failed.

On 02/07/2013 10:42 AM, Eugene Burmako wrote:

So far I have only one idea - explicitly force initialization of all
lazy stuff in JavaUniverse.init. Very tedious, but at least that will
make deadlocks impossible.

On Feb 7, 10:32 am, Eugene Burmako <eugene.burm...@epfl.ch> wrote:

The same questions goes to local objects.

On Feb 7, 10:31 am, Eugene Burmako <eugene.burm...@epfl.ch> wrote:

synchronized(this) implicitly emitted by lazy val codegen provides an
excellent opportunity for deadlocks.

--
You received this message because you are subscribed to the Google Groups "scala-internals" group.

To unsubscribe from this group and stop receiving emails from it, send an email to scala-interna...@googlegroups.com.

For more options, visit https://groups.google.com/groups/opt_out.

--
You received this message because you are subscribed to the Google Groups "scala-internals" group.

To unsubscribe from this group and stop receiving emails from it, send an email to scala-interna...@googlegroups.com.

[Paul Phillips]

On Thu, Feb 7, 2013 at 2:30 AM, Hubert Plociniczak <hubert.pl...@epfl.ch> wrote:

I think Paul or someone else wanted to do s/lazy val/def there and failed.

It was s/lazy // which I tried, because all those lazies are only to get the initialization order right, not because you actually care about deferring the computation. But it was too difficult. What I left as a future ambition is to have a means to write "locally lazy" vals - lazy in terms of what order they unroll during the initialization method, but after init is over they are all regular eager vals, i.e. no synchronization.

[Paul Phillips]

Oh, I just noticed the subject line of this thread, "Is there a way to make lazy vals use custom synchronization scheme?" It's something I've wanted to do for a long time, but no. See

  https://issues.scala-lang.org/browse/SUGGEST-11

possibly among others. My idea was to do it something like the pattern matcher uses MatchStrategy: if a __lazy in scope defines all the particulars of what the "lazy" means in "lazy val". But that idea predates macros, and macro annotations are probably a better way.

[√iktor Ҡlang]

On Thu, Feb 7, 2013 at 10:42 AM, Eugene Burmako <eugene....@epfl.ch> wrote:

So far I have only one idea - explicitly force initialization of all
lazy stuff in JavaUniverse.init. Very tedious, but at least that will
make deadlocks impossible.

Maybe a dumb question, but what's the point of making them lazy vals then?

Cheers,

√

 

On Feb 7, 10:32 am, Eugene Burmako <eugene.burm...@epfl.ch> wrote:
> The same questions goes to local objects.
>
> On Feb 7, 10:31 am, Eugene Burmako <eugene.burm...@epfl.ch> wrote:
>
>
>
>
>
>
>
> > synchronized(this) implicitly emitted by lazy val codegen provides an
> > excellent opportunity for deadlocks.

--
You received this message because you are subscribed to the Google Groups "scala-internals" group.
To unsubscribe from this group and stop receiving emails from it, send an email to scala-interna...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--

Viktor Klang
Director of Engineering

Typesafe - The software stack for applications that scale
Twitter: @viktorklang

[Eugene Burmako]

Mostly initialization order issues, since they are a part of a big cake with non-linear initialization control flow. Definitions in Definitions.scala are lazy probably because of symbol initialization issues.

[√iktor Ҡlang]

On Thu, Feb 7, 2013 at 9:07 PM, Eugene Burmako <eugene....@epfl.ch> wrote:

Mostly initialization order issues, since they are a part of a big cake with non-linear initialization control flow. Definitions in Definitions.scala are lazy probably because of symbol initialization issues.

In my experience, interdependicies between lazy vals is a deadlock waiting to wreak havoc in production.

There has to be another way.