Conversation
granted, one more step, but will help in disentangling json from parser
only used in tests
only used in tests
only used in tests. remove some of the tests, adjust others
|
jenkins build this please |
|
if you prefer to see the whole thing, i will publish the rest as follow-up PR's. |
🎉 So the goal is, to some extent, to remove every reference to
Yes please. That'd be great. |
Will this impact users' ability to define custom keywords through JSON? If so, we should at least think about the implications. This doesn't matter at all for the primary use case of deploying the parser as part of the simulator, but could be important for those who wish to use the parser in a more stand-alone context. |
|
i'm not going to break it, but you will need to link an additional opmjson library which contains the json to keyword stuff. |
That's a perfectly fine solution. Thanks! |
This is a first step in making cjson a private dependency of opm-common, that is, to remove the dependency on cjson from the exported opm-common library. Eventually only the python bindings (to not break the api) and the keyword generator need cjson.
Less dependencies are good, and we avoid publishing vendored code when cjson is not on the system.