The code in https://github.com/qri-io/dataset/tree/master/detect is used to detect schemas for datasets that don't have them, including types and column names, amongst some other things. We've recently starting doing these sorts of "intelligent inference" tasks more over in the mainline qri repo (for example: dsref/generate), and personally I feel like that's a good thing: perhaps the dataset repo should be less fancy, more pure data representation, and leave opinionated stances to other packages.
The code in https://github.com/qri-io/dataset/tree/master/detect is used to detect schemas for datasets that don't have them, including types and column names, amongst some other things. We've recently starting doing these sorts of "intelligent inference" tasks more over in the mainline qri repo (for example: dsref/generate), and personally I feel like that's a good thing: perhaps the dataset repo should be less fancy, more pure data representation, and leave opinionated stances to other packages.
Moving detect into qri would allow us to use dsref/generated instead of varName, which would enable us to fix https://github.com/qri-io/desktop/issues/508. It would also make it easier to fix https://github.com/qri-io/qri/issues/1322 and other such problems.