api: Use subs for origin in case index is a function

[mloubout]

Fixes #1950

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[codecov]

Codecov Report

Merging #2120 (ad29693) into master (c80b86c) will increase coverage by 0.00%.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master    #2120   +/-   ##
=======================================
  Coverage   87.17%   87.17%           
=======================================
  Files         222      222           
  Lines       39368    39378   +10     
  Branches     5117     5118    +1     
=======================================
+ Hits        34319    34328    +9     
- Misses       4478     4480    +2     
+ Partials      571      570    -1     

Impacted Files	Coverage Δ
devito/operations/interpolators.py	98.79% <100.00%> (+<0.01%)	⬆️
devito/tools/data_structures.py	72.66% <100.00%> (+0.17%)	⬆️
devito/types/basic.py	93.77% <100.00%> (ø)
devito/types/dense.py	93.53% <100.00%> (ø)
tests/test_symbolics.py	100.00% <100.00%> (ø)

... and 1 file with indirect coverage changes

[FabioLuporini]

types isn't a valid tag anymore; you probably should use compiler or dsl

[FabioLuporini]

Can we add a small MFE somewhere?

[EdCaunt]

Would it be worth having a test to double-check that indices are not affected by staggering? Possibly create two simple 1D kernels containing the absolute of an index, one where the function is staggered and one where it is not and check the generated C string is the same.

[mloubout]

Added test

[FabioLuporini]

equivalent to:

subs = {v: v.subs({k: c - v.origin.get(k, 0) for k, c in mapper.items()})}

?

[FabioLuporini]

why does this issue emerge only now?

[mloubout]

Before we were just doing index - origin during indexification which was making the assumption "indices are of the form i + o" so it didn't matter, but now without the assumption it needs the dimension which is removed and replaced by the ii_rec_... indices so we need to take care of it here when the dimension is removed

[FabioLuporini]

OK, thanks