[FIX] Revert index (entity format) definition to be a non-negative number, permitting zero

[TheChymera]

In continuation of previous discussions, including #535

A bit of theoretical background:

• An index simply labels elements of another set. There's no particular reason for zero to not be an index, in fact it might be the most used one :)
• The confusion zero sometimes causes is when it is used as an ordinal, since it leads to an asymmetric break in the number line by virtue of being its own inverse additive. Notably there is no year 0AD (or BC). Mathematically this can be considered an atavism, 0th is a valid ordinal, the problem with having a year 0 is that it causes asymmetry since distances between years would need to be calculated based on their start/end regardless of their position, and not based on their start if BC and end if AD.

More practically, there are a few datasets which include 0 values of an index:

• https://openneuro.org/datasets/ds000017
• https://openneuro.org/datasets/ds003722
• https://openneuro.org/datasets/ds003810
• https://openneuro.org/datasets/ds003020
• https://openneuro.org/datasets/ds000244
• https://zenodo.org/record/3575149
• https://zenodo.org/record/3601531
• https://zenodo.org/record/3598424

and supporting this doesn't break or encumber anything (if anything it makes the regex matching marginally shorter).

If we were to explicitly exclude zero, I think ordinal would be a better notion, since zero is clearly what people think of as a valid index when reading the word index — though as explained above, even that is a bit arbitrary. The zeroth (nought) week is a thing at some colleges.

@effigies @yarikoptic

[codecov]

Codecov Report

Patch and project coverage have no change.

Comparison is base (3d00184) 87.88% compared to head (86e06a5) 87.88%.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #1482   +/-   ##
=======================================
  Coverage   87.88%   87.88%           
=======================================
  Files          14       14           
  Lines        1279     1279           
=======================================
  Hits         1124     1124           
  Misses        155      155           

☔ View full report in Codecov by Sentry.
📢 Do you have feedback about the report comment? Let us know in this issue.

[effigies]

I approve in principle. A wording suggestion and a fix for the regular expression.

[effigies]

Some additional context: #590 further used the non-negative definition. #885 converted to positive integer without any discussion that I noticed. Possibly I missed something though.

[TheChymera]

@effigies non-negative I think is more accurate. Zero is commonly unsigned, so we can have very esoteric discussions on the issues page about whether or not it's positive :3

[sappelhoff]

LGTM, non-negative is what we need, I think.

[effigies]

Yes, it should be non-negative. I was pointing to the PR where it was changed to positive, so people could evaluate this fix.

[Remi-Gau]

Good with me.

[bendhouseart]

👍

[TheChymera]

can we merge this to make sure it's in 1.9.0 ? :)

[effigies]

Yes. We just have rules for changes to the spec to give people adequate time to comment.