istanbul-rewire-test

A test repository for debugging why istanbul and rewire are fighting with each other

Introduction

I noticed in one of my projects that I was getting some weird behavior while using the nyc test coverage in the same project that included a dependency for rewire@3. In this repository, I am trying to provide an example demo for this problem with the minimum setup I can create.

The following tests were performed on mac OS using node version 8.9.1 and npm version 5.5.1

With rewire@3

Failing nyc tests (foo -> baz)

Run the following test: npm run nyc-test-fail. This should have the following coverage output

----------|----------|----------|----------|----------|----------------|
File      |  % Stmts | % Branch |  % Funcs |  % Lines |Uncovered Lines |
----------|----------|----------|----------|----------|----------------|
All files |    57.14 |      100 |        0 |    57.14 |                |
 bar.js   |       50 |      100 |        0 |       50 |              4 |
 baz.js   |      100 |      100 |      100 |      100 |                |
 foo.js   |       50 |      100 |        0 |       50 |            6,7 |
----------|----------|----------|----------|----------|----------------|

Notice that only 50% of the lines in foo.js are marked as covered.

Passing nyc tests (baz -> foo)

Run the following test: npm run nyc-test-pass. This should have the following coverage output

----------|----------|----------|----------|----------|----------------|
File      |  % Stmts | % Branch |  % Funcs |  % Lines |Uncovered Lines |
----------|----------|----------|----------|----------|----------------|
All files |    85.71 |      100 |    66.67 |    85.71 |                |
 bar.js   |       50 |      100 |        0 |       50 |              4 |
 baz.js   |      100 |      100 |      100 |      100 |                |
 foo.js   |      100 |      100 |      100 |      100 |                |
----------|----------|----------|----------|----------|----------------|

Notice that only 100% of the lines in foo.js are marked as covered.

Passing istanbul test (foo -> baz)

Run the following test: npm run istanbul-test-pass1

----------|----------|----------|----------|----------|----------------|
File      |  % Stmts | % Branch |  % Funcs |  % Lines |Uncovered Lines |
----------|----------|----------|----------|----------|----------------|
 app/     |     87.5 |      100 |       50 |     87.5 |                |
  bar.js  |       50 |      100 |        0 |       50 |              4 |
  baz.js  |      100 |      100 |      100 |      100 |                |
  foo.js  |      100 |      100 |      100 |      100 |                |
----------|----------|----------|----------|----------|----------------|
All files |     87.5 |      100 |       50 |     87.5 |                |
----------|----------|----------|----------|----------|----------------|

Notice that unlike with nyc, istanbul is able to cover all lines in foo.js

Passing istanbul test (baz -> foo)

Run the following test: npm run istanbul-test-pass2

----------|----------|----------|----------|----------|----------------|
File      |  % Stmts | % Branch |  % Funcs |  % Lines |Uncovered Lines |
----------|----------|----------|----------|----------|----------------|
 app/     |     87.5 |      100 |       50 |     87.5 |                |
  bar.js  |       50 |      100 |        0 |       50 |              4 |
  baz.js  |      100 |      100 |      100 |      100 |                |
  foo.js  |      100 |      100 |      100 |      100 |                |
----------|----------|----------|----------|----------|----------------|
All files |     87.5 |      100 |       50 |     87.5 |                |
----------|----------|----------|----------|----------|----------------|

This test passes for both nyc and istanbul

With rewire@2

Next, run npm install rewire@2 and re-run the above tests. With rewire@2, both nyc and istanbul code cover tools work as expected and cover 100% of the lines in foo.js