Like for the plugin, we wanted to have the same hot reload system for the migration schemas.
We are doing this :

• for consistency regarding the way we are loading the plugin
• because reading file from the FS can take some time, it's interesting to have them already loaded in the memory. Like that when the endpoint /migrate is called, all script and schema are already loaded and so the user don't have to wait for this loading step.

The idea here is to reuse the reloading system from the schemas package.
The only problem with this idea, is that the loading of the migration schema is done in two steps:

1. Load every file mig.cuepart
2. Once everything is loaded, use these values to complete the final migration schema.

So everytime a file mig.cuepart is changing, the step 2 must be triggered. That's why we are introducing a callback function in the watcher. Like that we are able to trigger a last step once every Loader have been called.

Finally we are introducing a new package migrate so we can reuse the migration logic in the CLI (for later). Like that we will have the same behaviour between the command lint and migrate. Meaning that you will have an online mode but it can also totally work offline.

Signed-off-by: Augustin Husson husson.augustin@gmail.com