Hi David, If it is straightforward to use Nuxeo's REST API, I actually prefer that approach over the nuxeo java client. MCF already has a huge number of dependencies, and managing them is a challenge. Also, client code often introduces issues with thread blocking/interruptibility, since it's not usually designed with that use case in mind.
Thanks!! Karl On Tue, Jun 21, 2016 at 7:12 AM, David Arroyo <[email protected]> wrote: > Hi, > > > Last week I started to develop the repository connector. I have been taking > a look to existent connectors for checking how they work, code structure > and so on. > > > For developing the repository connector, I have decided to implement the > seeding incrementally, since although the API not support it directly, > Nuxeo allows to do queries in a languague called NXQL. > > > NXQL will be used a query to get the ids of the documents to be processed. > We can differentiate two cases: > > * The first job execution, in this case all documents are processed. > > * The following executions will use the last execution date, that is the > last modified date of the last processed document, which will be stored in > timestamp format, to get the documents modified from that date. > > > On the other hand, Nuxeo doesn´t delete the documents of the system, but > the state property of the document is set as deleted. So we know what > documents should be removed. > > > The next week I will try to finish a first version of repository connector > using httpclient as client. This first version will be then tested and > improved. A possible improvement could be to use the official Nuxeo Java > Client[1] for implementing the connector. > > > Regards. > > > [1] https://github.com/nuxeo/nuxeo-java-client/ >
