How to emit json output instead of html

[draganrakas]

Hi,

We are running Karumi Shot tests with ./gradlew executeScreenshotTests and would like a more parseable way to postprocess these results. Is there a way (maybe a CLI arg) to emit a json output instead of an html?

I scanned the readme and didn't notice anything related, but saw some Json parsing related code, so I thought I'd ask the team here before parsing HTML.

Thanks!

[pedrovgs]

Thanks for reporting @draganrakas 😃 If any external contributor would like to take this one, is more than welcome. Please, before start coding discuss the implementation details here.

[SamCosta1]

I also think this would be really useful, I resorted to just ignoring Shot's report and just generating my own json output based on the file structure. In hindsight it would have been better to actually update the plugin

[SamCosta1]

@pedrovgs I think I should have some time to pick things up. My proposed solution will involve passing a commandline argument to specify which output formats should be produced (defaulting to HTML only for backward compatibility).

Then I'm thinking it'll spit something out in the following format

{
  "deleted": [
    {
      "testClass": "...",
      "testName": "...",
      "originalScreenshot": "./images/recorded/...."
    }
  ],
  "new": [
    {
      "testClass": "...",
      "testName": "...",
      "newScreenshot": "./images/recorded/...."
    }
  ],
  "modified": [
    {
      "testClass": "...",
      "testName": "...",
      "originalScreenshot": "./images/recorded/....",
      "newScreenshot": "./images/....",
      "diff": "./...."
    }
  ],
  "resized": [
    {
      "testClass": "...",
      "testName": "...",
      "originalScreenshot": "./images/recorded/....",
      "newScreenshot": "./images/...."
    }
  ]
}

This will be generated in a JSONExecutionReporter class alongside the current reporters. (I think I'll also need to tweak the ScreenshotsComparator in order to spit out results for new screenshots too).

[pedrovgs]

Instead of adding different dictionaries for every screenshot, why just not add an array of screenshots with the execution result inside?