tree 26b8d974311190bcc0e64669bb8ef5d48ef4b259
parent a94241f64edde1729fc152cc9bb38c1123e01edd
author Emmanuel T Odeke <emmanuel@orijtech.com> 1568255068 -0700
committer Emmanuel Odeke <emm.odeke@gmail.com> 1568675099 +0000

storage: remove expensive and slow reslicing iteration logic

Previously, on invoking .Next for both:
* ObjectIterator
* BucketIterator

to grab the single item at the head for the case of already
available items after an RPC e.g. after 1000 items, we
expensively iterated by reslicing i.e.

    head := items[0]
    rest = items[1:]

but really we can just keep an index that is reset on fetching
new items.

The results of the change are quite dramatic, if we use the benchmark at
https://github.com/googleapis/google-cloud-go/issues/1506#issue-467672838

$ benchstat before.txt after.txt

name        old time/op    new time/op     delta
Iterator-8    3.68µs ± 4%     0.06µs ± 2%    -98.47%  (p=0.000 n=10+10)

name        old speed      new speed       delta
Iterator-8   272MB/s ± 4%  17716MB/s ± 1%  +6422.92%  (p=0.000 n=10+10)

name        old alloc/op   new alloc/op    delta
Iterator-8     0.00B           0.00B            ~     (all equal)

name        old allocs/op  new allocs/op   delta
Iterator-8      0.00            0.00            ~     (all equal)

and for the TL;DR, this cuts down the iteration latency/speed to
    60 nanoseconds
instead of
    3.68µs aka 3,680 nanoseconds

Fixes #1506.

Change-Id: I68b3f6bcb4523da15374cee09166b6b158b12ed0
Reviewed-on: https://code-review.googlesource.com/c/gocloud/+/45313
Reviewed-by: Jean de Klerk <deklerk@google.com>
Reviewed-by: kokoro <noreply+kokoro@google.com>