Question about sequenceCoverage, Line coverage and branchCoverage, branch coverage differences about reportgenerator

[ddur]

AFAIK and remember, when branch coverage is improved, original branch coverage calculated for method is left unchanged for legacy reasons. I believe correct values comes from ReportGenerator.
Take look at your <BranchPoints> in report

[danielpalme]

This wiki page explains some metrics of ReportGenerator: https://github.com/danielpalme/ReportGenerator/wiki/Calculation-details

Here you'll find the relevant code:
https://github.com/danielpalme/ReportGenerator/blob/master/ReportGenerator/Parser/Analysis/CodeFile.cs
https://github.com/danielpalme/ReportGenerator/blob/master/ReportGenerator/Parser/OpenCoverParser.cs

I don't know the details of how OpenCover calculates its metrics, so I can't explain any differences.
But: Line coverage is not the same as Sequence coverage. One sequence point can span several lines of code. So there might be a (little) difference.

[draftsteve]

@danielpalme Thank you. I'll go over these.

@ddur I'm having trouble parsing what you're saying. What do I look for with the BranchPoints in the OpenCover output?

[ddur]

@draftsteve
I mean, % of branch coverage should be number of visited <branch point>'s divided by number of all <branch point>'s, multiplied by 100

edit:
Coverage percentage totals are probably averaged and rounded for each level up from method (class, namespace, assembly, total). You should not care much for exact number, unless is 100% :)

[danielpalme]

@ddur:
Coverage percentage totals are not averaged. For each level (assembly, class, file), the exact numbers are calculated.

[ddur]

@danielpalme Sorry, I said "probably" cause I'm not sure.
So you accumulate visited/unvisited numbers of <SequencePoint>'s & <BranchPoint>'s for each level up and then compute?

What I meant under averaged is: sum number of methods, sum percentages and divide both sums.

[danielpalme]

So you accumulate visited/unvisited numbers of 's & 's for each level up and then compute?

Yes.