Discrepancy in numbers between cobertura (input) and generated report

[bvgheluwe]

When we process a cobertura xml file with ReportGenerator, we notice a discrepancy between the general numbers about covered lines, branches, etc.
Example:

The numbers 11494 (covered) and 35927 (coverable) are not found in the coverage element of the corresponding cobertura file (lines-covered and lines-valid attributes):

How does ReportGenerator calculate these numbers and why do they differ from Cobertura?

[danielpalme]

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

I don't know how the coverage tool that you have used to generate the cobertura xml, counts the lines-covered number.
ReportGenerator counts all lines with a hits attribute greater than 0 as covered.

E.g.

<line number="20" hits="2" branch="false" />

Maybe you can count the line elements with hits greater 0 in your coverage file. What number do you get?

[bvgheluwe]

For our frontend codebase we generate a cobertura file using Karma and Istanbul.
I counted the lines with the following regexes in Notepad++:

• hit count > 0 => \<line\snumber="\d*"\shits="[1-9][0-9]*"\sbranch => 9652
• any hit count => \<line\snumber="\d*"\shits="[0-9]*"\sbranch => 29934

The numbers I get are exactly the same as those within the xml file itself:

<coverage lines-valid="29934" lines-covered="9652" line-rate="0.3224" branches-valid="12516" branches-covered="1833" branch-rate="0.1464" timestamp="1734539251822" complexity="0" version="0.1">

When I run ReportGenerator (v5.4.1) on this file, then it reports the number of coverable lines as 35742 and the number of covered as 11476

Each class has its own file.

[danielpalme]

Can you share the Cobertura file? Then I can have a closer look.

You can send it privately via email: reportgenerator@palmmedia.de

[bvgheluwe]

Done.
It's a 3844 kb file. I hope the email isn't blocked.

[danielpalme]

Thank you. Will have a look as soon as possible.

[danielpalme]

I looked at your coverage file.
Your regex only counts the <line> elements that exist in in the <class> -> <lines> elements.

But additional <line> elements exist in the <class> -> <methods> -> <method> -> <lines> elements.

In total 39807 <line> elements exist in the file. 12201 of them with a hit count > 0.

Not all of the <line> elements within the <methods> elements are unique. Most of them also appear in the <class> -> <lines> elements with the same line number.
ReportGenerator primarily uses the <class> -> <lines> -> <line> elements.
Elements from <class> -> <methods> -> <method> -> <lines> -> <line> are only considered if not present in the first list.
In your file 35742 unique <line>elements exist. 11476 of them are covered.
Those numbers are also reported by ReportGenerator.