Increase Test Code Coverage

[arav-agarwal2]

• [x] 100% coverage in R/
• [ ] 100% coverage in src/
• [ ] ~100% coverage in inst/~

Original goals below:

• [x] Graph-Search
  • [x] Bfs
  • [x] Dfs
  • [x] MaxCardinalitySearch
  • [x] Check if Bfs/DfsVisit can be implemented
• [x] Shortest Path
  • [x] BellmanFord
  • [x] Dijkstra
  • [x] Suurballe
• [x] MinimumSpanningTree
  • [x] Kruskal
  • [x] minCostArborescence
• [x] MaxFlow
  • [x] Circulation
  • [x] EdmondsKarp
  • [x] Preflow
• [x] MinCostFlow
  • [x] CapacityScaling
  • [x] CostScaling
  • [x] CycleCanceling
  • [x] NetworkSimplex
• [x] MinCut
  • [x] HaoOrlin
  • [x] GomoryHu
  • [x] Nagamochilbaraki
• [x] MinMeanCycle
  • [x] HartmannOrlinMmc
  • [x] HowardMmc
  • [x] KarpMmc
• [x] Matching
  • [x] MaxMatching
  • [x] MaxWeightedMatching
  • [x] MaxWeightedPerfectMatching
  • [x] MaxFractionalMatching
  • [x] MaxWeightedFractionalMatching
  • [x] MaxWeightedPerfectFractionalMatching
• [x] Connectivity Algorithms
  • [x] TopologicalSort
  • [x] CheckedTopologicalSort
  • [x] Bipartite Partitions
  • [x] loopfree
  • [x] parallelfree
  • [x] simpleGraph
  • [x] eulerian
• [x] PlanarEmbedding and Drawing
  • [x] PlanarEmbedding
  • [x] PlanarDrawing
  • [x] PlanarColoring
• [x] TSP
  • [x] Christosfides
  • [x] Greedy
  • [x] Insertion
  • [x] NearestNeighbor
  • [x] Opt2
• [x] Approximation Algorithms
  • [x] GrossoLocatelliPullan
• [x] Check graph

[josherrickson]

Looking at some files, it seems like most of this is checked off?

• https://github.com/atewari7/rlemon/blob/5f968dcf102a86340b58ec3777dda80994d0ca4c/tests/testthat/test-shortestPath.R#L27
• https://github.com/atewari7/rlemon/blob/5f968dcf102a86340b58ec3777dda80994d0ca4c/tests/testthat/test-planarityembedding.R#L54
• https://github.com/atewari7/rlemon/blob/master/tests/testthat/test.check_graph.R

[josherrickson]

I've code-coverage against the package; please see the coverage.html file. Any insights into whether we can up that coverage?

[arav-agarwal2]

Regarding the former - that's likely the case. I'm unsure where I/Aditya left off on testing, but something tells me we tested the R/C++ interface much more than what the total coverage report indicates, as the source folder has coverage at 97.51%. ( And that's including the conversion function at src/conversion.h, which I do not think is being used and we should probably remove ).

Regarding the latter - it'll depend heavily on what is left in LEMON that we can test, as I'd wager some of it might be class-related stuff that we're not explicitly using in our interface. For sections like Matching, we should probably include tests on dense graphs, as the coverage report seems to indicate we're not hitting that section hard enough. Conversely, for sections like lemon/adaptors.h, I'm not convinced improving test coverage there will help substantially, as these seem to be mainly convenience functions that we're not handling due to how we process graphs.

[josherrickson]

Thanks. I just pushed up a commit expanding to cover 100% of the R/ directory. I think down the road if we could get 100% of src/ files covered, that'd be great, and whatever of inst/ we get, we get.

No rush on this.

[josherrickson]

Gonna close this; I'm happy with 100% on R/. Users will never enter src/ without first entering R/, so we can mostly test in there. We may see occasional errors in src/ that we haven't tested for, but can address those as they arrive.