pyblish / pyblish-deadline Goto Github PK

Add name attribute to all plugins for nicer experience for users.

Goal

To make it easier to complete a full publish with Deadline, without having to install event plugins.

Motivation

Event plugins are good for fine-tune control of the Deadline events, but can be a lot of manual setup to complete before a full publish with Deadline can be done.
This extension could provide a post-job script by default which will continue the publishing process after Deadline has executed the job.

Implemention

This can currently be implemented manually by providing the submission with a path to a post-job script. What this extension could help with, would be to standardize this script and add it by default.

If "deadlineData" is a list of dictionaries, we could submit multiple jobs per instance.

Currently the default implementation of Draft only takes in one template per job. With some simple modifications, Draft could support multiple outputs, so an image sequence could output high and low resolutions on different jobs.

The thinking is to have a list of draft templates in the submission parameters.

Currently the extension is writing temporary submission files to disk. These are purely for submission and the data is visible in the Monitor afterwards, so we don't need to keep these.

Goal

To be able to run tests on Travis or other CI's.

Current querying host with pyblish.plugin, need to convert to pyblish.api

https://github.com/pyblish/pyblish-deadline/blob/master/pyblish_deadline/plugins/conform_deadline_data.py#L32-L36

Having multiple extractors operating on the same data, seems to overwrite each other.

Solution atm is offsetting the extractors, but this is very intuitive, so need a solution for simultaneous execution. Adding a "deadlineDraftTemplate" data member might be best.

process_context and process_instance needs to be changed in favour of dependency injection.

Write an introduction to this extension

I'd say that plugins like this should probably not be a part of extension. It's not really doing anything related to deadline and is only necessary in some hosts. For example, nuke needs folder created before rendering, but houdini creates them on the fly. This means that every single time we publish, it fails, we have to repair and run publish again.

How about we either remove it, or move it out of the plugins folder to 'optional' or the 'examples' folder. that way it would stay available if user wants to put it in their studio plugins, is not in the way for general deadline extension.

Discard built-in ftrack plugins and use official ftrack extension;
https://github.com/pyblish/pyblish-ftrack
https://github.com/mkolar/pyblish-ftrack

We should be storing and creating deadline job, per instance.

Currently job_data get's appended to context, which however means, that every instance to be rendered, gets the same job_data, even though they might be very different.

It's also causing troubles when extracting ftrack data for deadline as they only get extracted for the first processed instance.

Inform users about how to install and the workflow of this extension

Hi,

1. I want to submit multiple jobs per Instance. As the instance.data("deadlineData")["job"] suggests, it is not possible at the moment.
2. I want to submit jobs for the Context.

Both seem doable, I am committed to implementing it, and submit a PR when ready. Do you have any pre-implementation notes?

Goal

Get Deadline Submission to process instances, instead of context, so its easier to see how far the submission process is.

Implemetation

Change the order of the instances in the context in what is now Deadline Passthrough. This will ensure the processing of the instances are in the correct order, so setting up dependencies in Deadline Submission is possible without processing the context.

@mottosso guessing you can change the order of the instance in context, similar to how you might do it in a collector?

Goal

Reduce code of current plugin by getting rid of new_context.

Implementation

for instance in sorted(context, key=lambda instance: instance["deadlineData"]["order"]):
   # do things in your custom order.

Instead of this, and get rid of above code.

Output from submitting to Deadline only gets displayed to user.

Need to process string output and raise errors if any present.

If both remote and local is setup, two jobs will get submitted.

Need to continue to local submission if remote fails. Else return.

Include or document an example collector for Maya/Nuke, to show how all the functionality of the extension is working.

There are 3 extract_deadline_draft plugins (vanilla, maya, nuke).

They practically do the same thing apart from setting the width and height of the frame. I think thse can easily be merge to just one plugin. To deal with the width and height then we could either do it by querying hosts and running different snippets based on that, or even omit the attributes altogether. I'm quite certain that deadline doesn't need either of them.