On Saturday, May 28, 2016 at 2:38:38 PM UTC-5, Camilo Roca wrote:
>
> @alex-miller
> thanks, that explained a lot. I did check the UseStringDeduplication
> feature and that's exactly what I was looking for.
>
> Funny thing is that it can only be implemented because strings are *immutable
> *in Java :P thus it is cheaper to store a reference to it since it will
> never change.
>
> @Tassilo
> that's actually a rather interesting example. I always wondered about it
> but didn't bother to ask it until now. If you see your example, you can see
> that the JVM is actually recognizing the characters inside the vectors are
> the same and only allocating one for each occurrence. On the other hand,
> the vectors although being exactly the same are not being recognized by the
> JVM as so. Thus, two different allocations are done even though only one
> would suffice (given their immutability).
>
> That was precisely my point: If you could identify that something is an
> exact copy of something else, why not just point to the first thing? From
> Alex miller, answer I know now that this is true for strings and for
> primitive values present in the source code.
>
> But what about keywords? vectors, hash-maps, sets?
>
Keywords are interned in Clojure - when you "create" one it will return you
the identical instance from the cache if it already exists. There are a few
other special cases like this too - things like literal {}, [], (), #{}.
But in general this has all the same problems with any cache -
synchronization/concurrency, knowing when to clear, updating use
stats/refcounting, etc.
> Of course, complaining about allocating those in the source code is
> non-sense as it would need a huge amount of them to even start to matter.
> But if you create lots of them at runtime then it is probably worth to look
> at it.
>
> Say for example a web-server that receives json objects with 70% of its
> content being (predefined) configuration values. If you allocate new
> strings/vector/hash-map for every request that you get, then your memory
> rises quite quickly. On the other hand if you keep a sort of "pool of
> values" and replace the new values with references to the ones in the pool
> then the total memory consumed is less than before.
>
I believe there are some json parsers that do exactly this optimization,
not too sure about the Clojure ones though.
>From my understanding of the description of "UseStringDeduplication" this
> is what's been done for Strings, but I guess NOT for any other object.
>
--
You received this message because you are subscribed to the Google
Groups "Clojure" group.
To post to this group, send email to clojure@googlegroups.com
Note that posts from new members are moderated - please be patient with your
first post.
To unsubscribe from this group, send email to
clojure+unsubscr...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/clojure?hl=en
---
You received this message because you are subscribed to the Google Groups
"Clojure" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to clojure+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
