issue #324 - fix how composer settings are created in the GeneratorOptions class

[ajoh504]

The GeneratorOptions::setOptionValue method contains a new condtion to check if the current input contains composer settings, and internally calls setComposerSettings if so. The unit tests attempt to test the following possibilites:

1. Composer settings are sent via command line
2. Composer settings are sent via a PHP project script

This fixes issue #324 where, when adding composer settings via the command line, the composer.json file was written as this:

"0": "config.disable-tls:true",
"1": "autoload.psr-4.Custom\\SoapClient\\:./Custom/SoapClient/"

instead of this:

"autoload": {
    "psr-4": {
        "Custom\\SoapClient\\": "./Custom/SoapClient/"
    }
},
"config": {
  "disable-tls": true
}

[mikaelcom]

Thanks for you work @ajoh504 , could you add at least one test in the WsdlToPhp\PackageGenerator\Tests\File\ComposerTest to see an actual composer.json result?

[ajoh504]

Thanks for you work @ajoh504 , could you add at least one test in the WsdlToPhp\PackageGenerator\Tests\File\ComposerTest to see an actual composer.json result?

Hi @mikaelcom, my pleasure. I've added an additional test to assertJson, however I don't believe this properly covers the code changes since the primary change was in GeneratorOptions::setOptionValue.

I considered creating an instance of GeneratorOptionsTest inside the ComposerTest, but I did not see anywhere in the Test suite where test classes referenced other test classes. Please let me know if you have additional recommendations for code test coverage.

[mikaelcom]

To be adapted 😉

[ajoh504]

I'm finding some failures with the PHP Unit tests for the ComposerTest class. One example is that the expected value does not have an "author" entry but the actual value does. Another example has to do with Windows vs. *nix path separators. Should I create a separate issue for this? I could work on this issue next.

[mikaelcom]

I'm finding some failures with the PHP Unit tests for the ComposerTest class. One example is that the expected value does not have an "author" entry but the actual value does. Another example has to do with Windows vs. *nix path separators. Should I create a separate issue for this? I could work on this issue next.

Feel free to create a new issue then 😉

[mikaelcom]

let me know if you can't merge by yourself

[ajoh504]

It looks like I only have the option to "Close with comment."