Support len() of multi-dim arrays in array analysis

[ehsantn]

As title. The check for 1D array in len handling of array analysis is unnecessary.

[esc]

This looks good, thank you for the patch. One tiny nitpick: you have used the 'blessed' repo numba/numba to store the branch for this pull-request. In general, we aim to use our own forks to store branches for pull-requests because this means the blessed repo stays clean and any fresh forks don't inherit development/feature branches. Here is a listing of the current branches for numba/numba:

💥 zsh» git branch -a | grep remotes/numba                                                                                                                          
  remotes/numba/0.11_maintenance
  remotes/numba/0.38-patches
  remotes/numba/HEAD -> numba/master
  remotes/numba/gh-pages
  remotes/numba/master
  remotes/numba/master_30_april_2019
  remotes/numba/multidim_len_analysis
  remotes/numba/newtargets
  remotes/numba/ocl-compiler
  remotes/numba/ocl-driver
  remotes/numba/release0.19.2
  remotes/numba/release0.26.0
  remotes/numba/release0.30.1
  remotes/numba/release0.40
  remotes/numba/release0.41
  remotes/numba/release0.42
  remotes/numba/release0.43
  remotes/numba/release0.43.1
  remotes/numba/release0.44

[ehsantn]

@esc looks like there was a problem with my remotes in this case. I'll delete the branch as soon as it's merged.

[stuartarchibald]

Thanks for the patch, change seems good, wondering if the test can be extended opposed to altered (see comment). Thanks again!

[stuartarchibald]

Perhaps this could be g = len(d) and then use it in the return, and then assert equivalence over 'e', 'f', 'g' so as to still retain the original check based on shapes and slices in the 2D array context?

[ehsantn]

done.

[seibert]

Also, merging master into this PR will fix the flake8 failure.

[ehsantn]

Thank you @stuartarchibald and @seibert . Applied the changes.

[stuartarchibald]

Thanks for the fixes.

[stuartarchibald]

I've marked this for merge, the only fails on CI were due to 504's on platforms that don't support ParallelAccelerator. CI has been restarted.