research

[lena-hinrichsen]

No description provided.

[joschrew]

I am using this ticket to write down what I have done so far. I have started to implement parts of the API and I have also thought about the API. This could also be called research.

In my opinion, we should implement the API if we want to use it (in whatever form). I think this is the only way to notice the weaknesses and flaws that the API will most likely have. A mock could possibly be enough, it depends on how detailed it is. I also don't know if you would call what I'm doing an implementation or if it's still a mock, since I'm not aiming for performance.

My approach is to do a minimal implementation first. I do this not only to find out how the API works and what is still missing. I think it's also a good way to get to know OCR-D better and thus it would be another step to somehow "step into OCR-D" better. I first want to provide OCR-D functionality with a Docker container, similar to what is shown in the user guide. Docker could also be used to run workflows. I realize that the performance of this will be very poor. But that's also the reason (i.e. that I want use Docker for the workflow) why I don't see a direct link to the workflow server (pull request from Robert) for now.

• The test-repo is here: https://github.com/joschrew/ocrd-webapi-test
• What I have done so far:
  • first I created the stubs of the functions from the WebAPI with a tool to have a starting point: https://github.com/koxudaxi/fastapi-code-generator. Very small changes were necessary so that it starts.
  • then I implemented individual functions:
    • get discovery: implemented it because it is an as a easy staring point
    • post workspace: upload a workspace
    • get workspaces: list uploaded workspaces
    • post processor: start a processor

[tdoan2010]

Hi, I've checked your work, thank you.

I will take over the code from here. In my opinion, there is a more urgent topic I hope you could investigate. I want to know how to create a Nextflow workflow via a REST endpoint. In short, please focus on this endpoint only.

You don't have to deliver a fully working endpoint at this time. Just come up with ideas how to implement it, which options are available, how the request body should look like, etc.

Some weeks ago, I thought about using DolphinNext to handle these workflow endpoints because the functions we need were already implemented there. So, that might be a starting point for you.

[tdoan2010]

@MehmedGIT if you have any idea how to create a Nextflow workflow via a REST endpoint, please let us know.

[MehmedGIT]

In the best case, I guess, we are aiming to have a good and extensive REST API? Not just a single workflow POST method which we then try to parse to a Nextflow script. For example, we should be able to define separate NextFlow processes (or rather OCR-D processors) with POST methods. Then with another POST method to combine the previously specified processors into a single workflow. In other words, building the Nextflow workflow script step by step and manipulating it with REST requests. If we do the parsing of the workflow on a process level, then it should be easier to parse. However, I am also not sure about this approach. It may end up making the workflow creation process a more challenging task for the users if there is no WebUI to ease the building of workflows.

DolphinNext does not support a good API for that purpose. Check here. From my current understanding, what they do is gather data via the WebUI and convert that data into a single POST request (Create a Run) which is then submitted to the server and parsed to a NextFlow script. I still have not found how exactly they do the parsing, I am checking that. Ideally, everything that the DolphinNext's WebUI does, we should also be able to do with a series of REST requests to the OCR-D server.

[tdoan2010]

This is the protocol of our meeting. In short, we agreed that:

• All implementation projects agree to use Nextflow.
• Therefore, Nextflow scripting language (Groovy) can be used as an exchange format.
• OCR-D/core will provide a tool to validate the workflow.
  • Validating means that the workflows must not contain script task, but only call to ocrd- processors.
• We need to check Robert's work and decide what to do with it.
• OCR-D/core needs to provide a mechanism so that all processors can expose their REST endpoints.