WIP Add support for recursive aggregate

[remysucre]

No description provided.

[codecov]

Codecov Report

Merging #2263 (917a36f) into master (bec3083) will decrease coverage by 0.22%.
The diff coverage is 32.67%.

❗ Current head 917a36f differs from pull request most recent head dd9c087. Consider uploading reports for the commit dd9c087 to get more accurate results

@@            Coverage Diff             @@
##           master    #2263      +/-   ##
==========================================
- Coverage   76.92%   76.69%   -0.23%     
==========================================
  Files         455      456       +1     
  Lines       28658    28638      -20     
==========================================
- Hits        22045    21964      -81     
- Misses       6613     6674      +61     

Impacted Files	Coverage Δ
src/RelationTag.h	46.05% <0.00%> (-6.19%)	⬇️
src/parser/parser.yy	94.23% <0.00%> (-0.92%)	⬇️
src/ram/AggregateExistenceCheck.h	0.00% <0.00%> (ø)
src/ram/analysis/Complexity.cpp	90.90% <0.00%> (-9.10%)	⬇️
src/ram/analysis/Index.h	98.66% <ø> (ø)
src/ram/analysis/Level.cpp	59.16% <0.00%> (-2.58%)	⬇️
src/synthesiser/Synthesiser.cpp	81.78% <6.25%> (-1.35%)	⬇️
src/ram/analysis/Index.cpp	87.95% <7.69%> (-3.65%)	⬇️
src/ram/utility/Visitor.h	54.47% <50.00%> (-0.07%)	⬇️
src/synthesiser/Relation.cpp	95.17% <50.00%> (-3.04%)	⬇️
... and 15 more

[b-scholz]

I looked at the WIP again. As discussed, we need a semantic check for recursive-aggregate relations. I would check the following:

• The arity of recursive-aggregate relations must be greater than or equal to two.
• The type of the last attribute cannot be of type symbol if the aggregation is a sum
• If the type of the last attribute is symbol, and the aggregation is min/max perhaps a warning should be issued that this is not the usual lexorder.

[b-scholz]

Please bring the source files into the right format using clang-format -i <file>. Following source files are not formatted:

src/RelationTag.h
src/synthesiser/Relation.h

[b-scholz]

Please also check whether I/O is working. The auxiliary arity of one may hide the last column.

[remysucre]

Please also check whether I/O is working. The auxiliary arity of one may hide the last column.

I/O is working fine - all attributes are printed.

[quentin]

The technical choice of introducing new btree representations does not ease adding other aggregation operators in the future.