Ignoring Statements, Branches, etc.?

[jednano]

@nathancahill and I are seeing statements and branches ignored, even though there are no ignore comments in our source code. It looks like this:

=============================== Coverage summary ===============================
Statements   : 100% ( 138/138 ), 10 ignored
Branches     : 98.61% ( 71/72 ), 9 ignored
Functions    : 100% ( 12/12 )
Lines        : 100% ( 69/69 )
================================================================================

Any idea why code is being ignored for seemingly no reason? Or maybe the report is off somehow?

[jamesplease]

@jedmao, in my tests it appeared to me that isparta was reporting inconsistent stats. Let me explain.

When you compile your source code, you end up with two versions of the same file: the original, and the compiled version.

The compiled version is usually more verbose and messy than the original version.

After using isparta, you will get data on Lines for the original file. Because of this, the "visualizer" that displays the lines covered can show the original source.

However, every other data point will reference the compiled code. What you're seeing here is likely due to Babel adding an internal function with a few branches that your tests aren't fully covering.

You can see this by running babel my-file.js --out-file build.js and comparing what the stats are reporting to your original and compiled code. You might even be able to step through and figure out which branch is being missed in the compiled version.

So is it fixable? I'm not sure, honestly. I know very little about the internals of isparta and Istanbul.

As for me, I don't worry too much about the mismatch. I use code coverage to get an approximation of how complete my tests are. Even a value of 100% on everything doesn't mean that the code has no bugs.

Anyway, I don't think I'll convince anyone to change their opinion on whether or not 100% code coverage matters, but it's worth keeping in mind that the coverage data will partially reflect the transpiled code.

Does that make sense?

[jednano]

@jmeas yeah that makes sense, but it's still not ideal. I wonder if isparta could more gracefully report what's going on instead of saying "skipped" for babel-generated code. Until then, I'm forced to "ignore" the number of ignored statements and branches, trusting that it's indeed accurate. Again, not ideal. Still waiting to hear back from @douglasduteil.

[HipsterZipster]

This still seems to be happening in the latest version of isparta ^4.0.0
@jedmao Were you ever able to get around it?