Should .nyc_output be deleted?

[sarbbottam]

Since .nyc_output is a temp directory, I am wondering if nyc should delete it, once the process ends.

[jamestalmage]

We allow nyc report commands to be run at a later time, and .nyc_output still needs to be around for that.

[sarbbottam]

Wondering if it would be possible to run nyc report on coverage dir. Please excuse my ignorance.
What could the list of potential usage of nyc report dir. Thanks!

[jamestalmage]

nyc report operates basically the same as istanbul report, except it first combines all the contents of .nyc_output before running the report. Running .nyc_output on the coverage directory wouldn't make a lot of sense (you already have istanbul report for that)

[bcoe]

Hey @sarbbottam, as @jamestalmage says the reason for not deleting .nyc_output on each run is that it can be handy to output reports multiple times:

nyc npm test.
nyc report lcov-text
nyc report lcov

If you want to cleanup after nyc, I recommend adding a posttest command that uses rimraf to nuke the folder 👍

[yf-hk]

Shall we provide an option to not generate .nyc_output at all?

[jamestalmage]

Not possible. We need it to store coverage info

[mmc41]

Would it be possible to configure where to store the coverage info ? F.x. in the same folder as the reports ?

[vamonke]

You can specify where .nyc_output is generated under temp-directory in the config. e.g.

"nyc": {
  "temp-directory": "./coverage/.nyc_output"
}

source: https://github.com/istanbuljs/nyc/blob/master/index.js#L43

[ozzyogkush]

The "temp-directory" option is not documented in the repository README.

[aalexgabi]

I'm not a fan of creating a directory under my project in general and especially when I specify explicitly --report-dir. This should be an option and not the default behavior. For me nyc should use a tmp directory by default generated by something like mktemp -d and the equivalent on Windows.

The problem here is the assumption that there will be only one instance of the covered process running at a given path. I wonder how nyc handles this case but I suppose that there are concurrency issues with this approach.

In my case I launch two instances of the same service under nyc and then I launch the integration tests as a separate process. At the end I use istanbul-combine to combine the two coverage reports into one and I use nyc check-coverage for the combined coverage.

In general I feel that nyc is the only option for the NodeJS community and it's quality is questionable compared to istanbul and in general. I started using it for a week and I already stumbled on a couple of bugs, undocumented features in the man page, strange imposed defaults like the number of lines at 90% when running nyc check-coverage, uncovered js files without extension etc.

I know this is an open source and anybody is welcome to contribute but I'm still disappointed that such a large community does not have what I would call "proper tools".

[IsFlwrs]

I have resolved it moving the .nyc_output folder under the node_modules.
All the node_modules is ignored by the .gitignore file.

"nyc": {
  "temp-directory": "./node_modules/coverage/.nyc_output"
}

[jml6m]

      The "temp-directory" option is not documented in the repository README.

It looks like the property has changed to temp-dir

[timbeadle]

I have resolved it moving the .nyc_output folder under the node_modules.

@IsFlwrs This isn't a very good idea. node_modules contains your dependencies, which you may wish to remove and reinstall at any time.

It's better to add .nyc_output to your .gitignore file.