# Inconsistency with multidimensional comprehensions and iterators

**URL:** <https://discourse.julialang.org/t/inconsistency-with-multidimensional-comprehensions-and-iterators/3188>\
**Category:** General Usage\
**Created:** [April 12, 2017, 10:39pm UTC](https://discourse.julialang.org/t/inconsistency-with-multidimensional-comprehensions-and-iterators/3188 "2017-04-12T22:39:14Z")\
**Posts on this page:** 6\
**Page:** 1

<div class="post-metadata">

**Author:** ![rdeits](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/rdeits/32/286_2.png) [@rdeits](https://discourse.julialang.org/u/rdeits)\
**Post date:** [April 12, 2017, 10:39pm UTC](https://discourse.julialang.org/t/inconsistency-with-multidimensional-comprehensions-and-iterators/3188/1 "2017-04-12T22:39:14Z")

</div>

I found this behavior surprising:

```julia
julia> using Iterators: product

julia> p = product(1:2, 1:2)
Iterators.Product{Tuple{UnitRange{Int64},UnitRange{Int64}}}([(1:2, 1:2))

julia> [1 for x in p, y in 1:2]
8-element Array{Int64,1}:
 1
 1
 1
 1
 1
 1
 1
 1

julia> [1 for x in collect(p), y in 1:2]
4×2 Array{Int64,2}:
 1 1
 1 1
 1 1
 1 1

```

I expected both versions to behave the way `collect(p)` does. I don’t think this has anything to do with `Iterators.product` itself, rather I guess comprehensions with AbstractArray ranges behave differently than comprehensions with arbitrary iterator ranges? Is this an intentional distinction?

---

<div class="post-metadata">

**Author:** ![fengyang.wang](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/fengyang.wang/32/104_2.png) [@fengyang.wang](https://discourse.julialang.org/u/fengyang.wang)\
**Post date:** [April 12, 2017, 10:47pm UTC](https://discourse.julialang.org/t/inconsistency-with-multidimensional-comprehensions-and-iterators/3188/2 "2017-04-12T22:47:22Z")

</div>

I suspect the former is a bug. Perhaps report it as an issue?

Edit:

For what it’s worth, I can’t reproduce this issue. I get the expected 3-dimensional array in both cases.

```julia
julia> p = Iterators.product(1:2, 1:2)
Base.Iterators.Prod2{UnitRange{Int64},UnitRange{Int64}}(1:2, 1:2)

julia> [1 for x in p, y in 1:2]
2×2×2 Array{Int64,3}:
[:, :, 1] =
 1 1
 1 1

[:, :, 2] =
 1 1
 1 1

julia> [1 for x in collect(p), y in 1:2]
2×2×2 Array{Int64,3}:
[:, :, 1] =
 1 1
 1 1

[:, :, 2] =
 1 1
 1 1

```

---

<div class="post-metadata">

**Author:** ![traktofon](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/traktofon/32/591_2.png) [@traktofon](https://discourse.julialang.org/u/traktofon)\
**Post date:** [April 13, 2017, 2:45am UTC](https://discourse.julialang.org/t/inconsistency-with-multidimensional-comprehensions-and-iterators/3188/3 "2017-04-13T02:45:21Z")

</div>

FWIW, I get the same behavior as @rdeits, both on julia-0.5.1 and on latest master (b91b411). Also, Pkg.checkout(“Iterators”) makes no difference for me. @fengyang.wang, which julia version are you using?

---

<div class="post-metadata">

**Author:** ![fengyang.wang](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/fengyang.wang/32/104_2.png) [@fengyang.wang](https://discourse.julialang.org/u/fengyang.wang)\
**Post date:** [April 13, 2017, 2:52am UTC](https://discourse.julialang.org/t/inconsistency-with-multidimensional-comprehensions-and-iterators/3188/4 "2017-04-13T02:52:55Z")

</div>

Ah, the discreptancy is that you are using the `Iterators` package. I recommend that the `Compat.Iterators` (`Base.Iterators` in 0.6 and newer) package be used instead of `Iterators`. `Iterators.product` lacks shape information, whereas `Compat.Iterators.product` retains it.

Edit: To clarify, the difference is not between arrays and other iterables; the difference is between objects with shape information and objects without it. Iterators.jl was written, in large part, before the mechanisms for preserving shape information existed. In general, `Base.Iterators` (or `Compat.Iterators` for 0.5 support) is preferred.

---

<div class="post-metadata">

**Author:** ![rdeits](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/rdeits/32/286_2.png) [@rdeits](https://discourse.julialang.org/u/rdeits)\
**Post date:** [April 13, 2017, 3:25am UTC](https://discourse.julialang.org/t/inconsistency-with-multidimensional-comprehensions-and-iterators/3188/5 "2017-04-13T03:25:58Z")

</div>

Ah, I see, thanks! I had no idea that `Base.Iterators` existed in 0.6.

It still seems odd that shape-less iterators behave differently than 1-dimensional arrays in this case. Do you think it’s worth opening an issue about that?

---

<div class="post-metadata">

**Author:** ![fengyang.wang](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/fengyang.wang/32/104_2.png) [@fengyang.wang](https://discourse.julialang.org/u/fengyang.wang)\
**Post date:** [April 13, 2017, 3:35am UTC](https://discourse.julialang.org/t/inconsistency-with-multidimensional-comprehensions-and-iterators/3188/6 "2017-04-13T03:35:32Z")

</div>

I think this is intentional. It’s not possible to compute the shape ahead of time for length-less iterators like the `Iterators.Product` from Iterators.jl.