Bug #7877: E::Lazy#with_index should be lazy

• Status: Closed
• Priority: Normal
• Assignee: jeremyevans0 (Jeremy Evans)
• Target version: -
• ruby -v: 2.1.0-dev
• Backport: 2.0.0: UNKNOWN

[shyouhei (Shyouhei Urabe)]

So I wanted some real benefit of being lazy. I wrote a Leibniz formula:

def leibniz(n)
  (0..Float::INFINITY).lazy.with_index {|i, j| (-1 ** j) / (2*i+1).to_f }.take(n).reduce(:+)
end

But it doesn't work (well, it does, indeed. It just doesn't stop working). I got frustrated.
How about it? Don't you feel it nifty?

Of course I can wait for the release next to 2.0.0.

[nobu (Nobuyoshi Nakada)]

• File 0001-enumerator.c-Enumerator-Lazy-with_index.patch 0001-enumerator.c-Enumerator-Lazy-with_index.patch added
• Description updated (diff)

[marcandre (Marc-Andre Lafortune)]

See #7696 on how to handle state...

[shyouhei (Shyouhei Urabe)]

• Description updated (diff)

OK, so @marcandre (Marc-Andre Lafortune) is interested in. I re-wrote the description in English.

[marcandre (Marc-Andre Lafortune)]

Note that (thanks to #7715), you can use with_index without a block and follow it with map:

def leibniz(n)
  (0..Float::INFINITY).lazy.with_index.map {|i, j| (-1 ** j) / (2*i+1).to_f }.take(n).reduce(:+)
end

I'm neutral about this feature request. The problem I see is that it's too late for 2.0.0 and it would introduce potential incompatibility in next minor.

[shyouhei (Shyouhei Urabe)]

@marcandre (Marc-Andre Lafortune) oh, thank you! You saved my day.

Still I want this though.

[zzak (Zak Scott)]

Propose to move this to next major?

[duerst (Martin Dürst)]

zzak (Zachary Scott) wrote:

Propose to move this to next major?

Do you mean because of "potential incompatibility" (https://bugs.ruby-lang.org/issues/7877#note-4)? What exactly would this incompatibility be? To me, it seems that lazy.with_index would just work, so we should just fix it. Next major seems way too long to wait.

[zzak (Zak Scott)]

@duerst (Martin Dürst) You're probably right, since this feature was introduced in 2.0.0

If yhara-san wants to implement #with_index with a block then I see no problem with introducing this in 2.1.0

[zzak (Zak Scott)]

• Tracker changed from Feature to Bug
• Subject changed from E::Lazy#with_index needed to E::Lazy#with_index should be lazy
• Target version changed from 2.6 to 2.1.0
• ruby -v set to 2.1.0-dev
• Backport set to 2.0.0: UNKNOWN

[hsbt (Hiroshi SHIBATA)]

• Target version changed from 2.1.0 to 2.2.0

[Nakilon (Victor Maslov)]

Is it somehow related? http://stackoverflow.com/q/24782712/322020
... .lazy ... .take_while.with_index{ ...
ArgumentError - tried to call lazy take_while without a block:`

Nevermind, swaping the chain in this way .with_index.take_while saved me.

[shyouhei (Shyouhei Urabe)]

Is there reason this is not loved? I'd like to have this.

[shyouhei (Shyouhei Urabe)]

• Assignee changed from yhara (Yutaka HARA) to nobu (Nobuyoshi Nakada)

[jeremyevans0 (Jeremy Evans)]

• File lazy-with-index-block-7877.patch lazy-with-index-block-7877.patch added

This bug is still present in the master branch. I've updated nobu's patch to apply to the master branch, which required a significant rewrite. The updated patch is attached.

[nobu (Nobuyoshi Nakada)]

• Description updated (diff)
• Assignee changed from nobu (Nobuyoshi Nakada) to jeremyevans0 (Jeremy Evans)

Thank you, I've missed it.

[nobu (Nobuyoshi Nakada)]

• Status changed from Assigned to Closed