Understanding lazy-seq
Edmund Jackson
I have been trying to get a proper grip on the operation of lazy-seq and hope somebody will have the time to clarify a point for me. The references indicate that you should not hold onto the head of a lazy sequence as it blocks the GC. This has lead to me to believe that a lazy sequence, even while being active 'downstream' can be GCd 'upstream'. Is this so ?
An example: in this post
in the first version of fibo, the call (nth (fibo) 1000000) will cause a seq to proceed to the millionth element. If memory were tight could earlier elements be GC'd before nth had reached the end ? My understand is that it can. I ask because on my machine on Clojure 1.2, (nth (fibo) 1000000) causes a heap overflow which I don't understand.
Thanks in advance,
Edmund
Stuart Halloway
This is a regression since last Tuesday's commit
f81e612cc9ff91ddefc1d86e270cd7f018701802. Thanks for catching it!
Stu
> --
> You received this message because you are subscribed to the Google
> Groups "Clojure" group.
> To post to this group, send email to clo...@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+u...@googlegroups.com
> For more options, visit this group at
> http://groups.google.com/group/clojure?hl=en
>
> To unsubscribe, reply using "remove me" as the subject.
Sean Devlin
On Apr 12, 7:53 pm, Stuart Halloway <stuart.hallo...@gmail.com> wrote:
> Hi Edmund,
>
> This is a regression since last Tuesday's commit
> f81e612cc9ff91ddefc1d86e270cd7f018701802. Thanks for catching it!
>
> Stu
>
> > Dear Clojurians,
>
> > I have been trying to get a proper grip on the operation of lazy-
> > seq and hope somebody will have the time to clarify a point for me.
> > The references indicate that you should not hold onto the head of a
> > lazy sequence as it blocks the GC. This has lead to me to believe
> > that a lazy sequence, even while being active 'downstream' can be
> > GCd 'upstream'. Is this so ?
>
> > An example: in this post
>
> >http://groups.google.com/group/clojure/browse_thread/thread/57a12f1a0...
Rich Hickey
On Apr 12, 7:53 pm, Stuart Halloway <stuart.hallo...@gmail.com> wrote:
> Hi Edmund,
>
> This is a regression since last Tuesday's commit
> f81e612cc9ff91ddefc1d86e270cd7f018701802. Thanks for catching it!
>
> Stu
>
>
>
> > Dear Clojurians,
>
> > I have been trying to get a proper grip on the operation of lazy-
> > seq and hope somebody will have the time to clarify a point for me.
> > The references indicate that you should not hold onto the head of a
> > lazy sequence as it blocks the GC. This has lead to me to believe
> > that a lazy sequence, even while being active 'downstream' can be
> > GCd 'upstream'. Is this so ?
>
> > An example: in this post
>
> >http://groups.google.com/group/clojure/browse_thread/thread/57a12f1a0...
>
> > in the first version of fibo, the call (nth (fibo) 1000000) will
> > cause a seq to proceed to the millionth element. If memory were
> > tight could earlier elements be GC'd before nth had reached the
> > end ? My understand is that it can. I ask because on my machine on
> > Clojure 1.2, (nth (fibo) 1000000) causes a heap overflow which I
> > don't understand.
>
> > Thanks in advance,
>
> > Edmund
>
Now fixed - thanks for the report!
Rich
Edmund Jackson
> --
> You received this message because you are subscribed to the Google
> Groups "Clojure" group.
> To post to this group, send email to clo...@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+u...@googlegroups.com
> For more options, visit this group at
> http://groups.google.com/group/clojure?hl=en
>
> To unsubscribe, reply using "remove me" as the subject.
Edmund