Bugfix vanishing instances after elimination

[kidiki]

Description

Summary: Solving and RQAOA bug when vanishing problems.
Context: unweighted graphs have a lot of edge cases when reduced during the RQAOA algorithm.
Bug description:
Approach to fixing: check if the new problem is an empty dictionary, which would mean that a total elimination happened. If so, set the new problem to the old one and break the optimization cycle. Hence, solving for the smallest non-vanishing instance or equivalently, increasing the cutoff.

Note: this might cause a potential problem if the smallest non-vanishing instance is too big to be solved classically. TODO

Checklist

• I have performed a self-review of my code.
• I have commented my code and used numpy-style docstrings
• I have made corresponding updates to the documentation.
• My changes generate no new warnings
• I have added/updated tests to make sure bugfix/feature works.
• New and existing unit tests pass locally with my changes.

Type of change

Please delete options that are not relevant.

• Bug fix (non-breaking change which fixes an issue)

How Has This Been Tested?

• Write unit tests for the specific instances which were failing previously.

[codecov]

Codecov Report

Merging #158 (a513760) into dev (13f8f4d) will increase coverage by 0.28%.
The diff coverage is 97.76%.

@@            Coverage Diff             @@
##              dev     #158      +/-   ##
==========================================
+ Coverage   81.70%   81.98%   +0.28%     
==========================================
  Files          95       95              
  Lines       11794    11994     +200     
==========================================
+ Hits         9636     9833     +197     
- Misses       2158     2161       +3     

Impacted Files	Coverage Δ
openqaoa/rqaoa/rqaoa.py	96.95% <92.15%> (+0.42%)	⬆️
openqaoa/workflows/optimizer.py	93.81% <96.20%> (-0.26%)	⬇️
openqaoa/problems/problem.py	98.16% <96.29%> (+0.28%)	⬆️
openqaoa/optimizers/result.py	92.85% <100.00%> (-0.78%)	⬇️
openqaoa/rqaoa/rqaoa_results.py	100.00% <100.00%> (ø)
tests/test_problems.py	99.16% <100.00%> (+0.04%)	⬆️
tests/test_rqaoa.py	100.00% <100.00%> (ø)
tests/test_workflows.py	99.84% <100.00%> (+0.01%)	⬆️

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[kidiki]

"Spiders are the only web developers that enjoy finding bugs." + Kristina ;) :D

[kidiki]

Note that this bug occurs for 27 out of 1100 random instances for unweighted graphs with p=0.7 probability for an edge.