submit_publish_job.py

# -*- coding: utf-8 -*-
"""Submit publishing job to farm."""

import os
import json
import re
from copy import copy, deepcopy
import requests
import clique

import pyblish.api

from openpype.client import (
    get_last_version_by_subset_name,
    get_representations,
)
from openpype.pipeline import (
    get_representation_path,
    legacy_io,
)
from openpype.tests.lib import is_in_tests
from openpype.pipeline.farm.patterning import match_aov_pattern
from openpype.lib import is_running_from_build


def get_resources(project_name, version, extension=None):
    """Get the files from the specific version."""

    # TODO this functions seems to be weird
    #   - it's looking for representation with one extension or first (any)
    #       representation from a version?
    #  - not sure how this should work, maybe it does for specific use cases
    #       but probably can't be used for all resources from 2D workflows
    extensions = None
    if extension:
        extensions = [extension]
    repre_docs = list(get_representations(
        project_name, version_ids=[version["_id"]], extensions=extensions
    ))
    assert repre_docs, "This is a bug"

    representation = repre_docs[0]
    directory = get_representation_path(representation)
    print("Source: ", directory)
    resources = sorted(
        [
            os.path.normpath(os.path.join(directory, fname))
            for fname in os.listdir(directory)
        ]
    )

    return resources


def get_resource_files(resources, frame_range=None):
    """Get resource files at given path.

    If `frame_range` is specified those outside will be removed.

    Arguments:
        resources (list): List of resources
        frame_range (list): Frame range to apply override

    Returns:
        list of str: list of collected resources

    """
    res_collections, _ = clique.assemble(resources)
    assert len(res_collections) == 1, "Multiple collections found"
    res_collection = res_collections[0]

    # Remove any frames
    if frame_range is not None:
        for frame in frame_range:
            if frame not in res_collection.indexes:
                continue
            res_collection.indexes.remove(frame)

    return list(res_collection)


class ProcessSubmittedJobOnFarm(pyblish.api.InstancePlugin):
    """Process Job submitted on farm.

    These jobs are dependent on a deadline or muster job
    submission prior to this plug-in.

    - In case of Deadline, it creates dependend job on farm publishing
      rendered image sequence.

    - In case of Muster, there is no need for such thing as dependend job,
      post action will be executed and rendered sequence will be published.

    Options in instance.data:
        - deadlineSubmissionJob (dict, Required): The returned .json
          data from the job submission to deadline.

        - musterSubmissionJob (dict, Required): same as deadline.

        - outputDir (str, Required): The output directory where the metadata
            file should be generated. It's assumed that this will also be
            final folder containing the output files.

        - ext (str, Optional): The extension (including `.`) that is required
            in the output filename to be picked up for image sequence
            publishing.

        - publishJobState (str, Optional): "Active" or "Suspended"
            This defaults to "Suspended"

        - expectedFiles (list or dict): explained bellow

    """

    label = "Submit image sequence jobs to Deadline or Muster"
    order = pyblish.api.IntegratorOrder + 0.2
    icon = "tractor"
    deadline_plugin = "OpenPype"
    targets = ["local"]

    hosts = ["fusion", "max", "maya", "nuke",
             "celaction", "aftereffects", "harmony"]

    families = ["render.farm", "prerender.farm",
                "renderlayer", "imagesequence", "maxrender", "vrayscene"]

    aov_filter = {"maya": [r".*([Bb]eauty).*"],
                  "aftereffects": [r".*"],  # for everything from AE
                  "harmony": [r".*"],  # for everything from AE
                  "celaction": [r".*"],
                  "max": [r".*"]}

    environ_job_filter = [
        "OPENPYPE_METADATA_FILE"
    ]

    environ_keys = [
        "FTRACK_API_USER",
        "FTRACK_API_KEY",
        "FTRACK_SERVER",
        "AVALON_APP_NAME",
        "OPENPYPE_USERNAME",
        "OPENPYPE_SG_USER",
    ]

    # Add OpenPype version if we are running from build.
    if is_running_from_build():
        environ_keys.append("OPENPYPE_VERSION")

    # custom deadline attributes
    deadline_department = ""
    deadline_pool = ""
    deadline_pool_secondary = ""
    deadline_group = ""
    deadline_chunk_size = 1
    deadline_priority = None

    # regex for finding frame number in string
    R_FRAME_NUMBER = re.compile(r'.+\.(?P<frame>[0-9]+)\..+')

    # mapping of instance properties to be transfered to new instance for every
    # specified family
    instance_transfer = {
        "slate": ["slateFrames", "slate"],
        "review": ["lutPath"],
        "render2d": ["bakingNukeScripts", "version"],
        "renderlayer": ["convertToScanline"]
    }

    # list of family names to transfer to new family if present
    families_transfer = ["render3d", "render2d", "ftrack", "slate"]
    plugin_pype_version = "3.0"

    # script path for publish_filesequence.py
    publishing_script = None

    # poor man exclusion
    skip_integration_repre_list = []

    def _create_metadata_path(self, instance):
        ins_data = instance.data
        # Ensure output dir exists
        output_dir = ins_data.get(
            "publishRenderMetadataFolder", ins_data["outputDir"])

        try:
            if not os.path.isdir(output_dir):
                os.makedirs(output_dir)
        except OSError:
            # directory is not available
            self.log.warning("Path is unreachable: `{}`".format(output_dir))

        metadata_filename = "{}_metadata.json".format(ins_data["subset"])

        metadata_path = os.path.join(output_dir, metadata_filename)

        # Convert output dir to `{root}/rest/of/path/...` with Anatomy
        success, rootless_mtdt_p = self.anatomy.find_root_template_from_path(
            metadata_path)
        if not success:
            # `rootless_path` is not set to `output_dir` if none of roots match
            self.log.warning((
                "Could not find root path for remapping \"{}\"."
                " This may cause issues on farm."
            ).format(output_dir))
            rootless_mtdt_p = metadata_path

        return metadata_path, rootless_mtdt_p

    def _submit_deadline_post_job(self, instance, job, instances):
        """Submit publish job to Deadline.

        Deadline specific code separated from :meth:`process` for sake of
        more universal code. Muster post job is sent directly by Muster
        submitter, so this type of code isn't necessary for it.

        Returns:
            (str): deadline_publish_job_id
        """
        data = instance.data.copy()
        subset = data["subset"]
        job_name = "Publish - {subset}".format(subset=subset)

        # instance.data.get("subset") != instances[0]["subset"]
        # 'Main' vs 'renderMain'
        override_version = None
        instance_version = instance.data.get("version")  # take this if exists
        if instance_version != 1:
            override_version = instance_version
        output_dir = self._get_publish_folder(
            instance.context.data['anatomy'],
            deepcopy(instance.data["anatomyData"]),
            instance.data.get("asset"),
            instances[0]["subset"],
            'render',
            override_version
        )

        # Transfer the environment from the original job to this dependent
        # job so they use the same environment
        metadata_path, rootless_metadata_path = \
            self._create_metadata_path(instance)

        environment = {
            "AVALON_PROJECT": legacy_io.Session["AVALON_PROJECT"],
            "AVALON_ASSET": legacy_io.Session["AVALON_ASSET"],
            "AVALON_TASK": legacy_io.Session["AVALON_TASK"],
            "OPENPYPE_USERNAME": instance.context.data["user"],
            "OPENPYPE_PUBLISH_JOB": "1",
            "OPENPYPE_RENDER_JOB": "0",
            "OPENPYPE_REMOTE_JOB": "0",
            "OPENPYPE_LOG_NO_COLORS": "1",
            "IS_TEST": str(int(is_in_tests()))
        }

        # add environments from self.environ_keys
        for env_key in self.environ_keys:
            if os.getenv(env_key):
                environment[env_key] = os.environ[env_key]

        # pass environment keys from self.environ_job_filter
        job_environ = job["Props"].get("Env", {})
        for env_j_key in self.environ_job_filter:
            if job_environ.get(env_j_key):
                environment[env_j_key] = job_environ[env_j_key]

        # Add mongo url if it's enabled
        if instance.context.data.get("deadlinePassMongoUrl"):
            mongo_url = os.environ.get("OPENPYPE_MONGO")
            if mongo_url:
                environment["OPENPYPE_MONGO"] = mongo_url

        priority = self.deadline_priority or instance.data.get("priority", 50)

        args = [
            "--headless",
            'publish',
            rootless_metadata_path,
            "--targets", "deadline",
            "--targets", "farm"
        ]

        if is_in_tests():
            args.append("--automatic-tests")

        # Generate the payload for Deadline submission
        payload = {
            "JobInfo": {
                "Plugin": self.deadline_plugin,
                "BatchName": job["Props"]["Batch"],
                "Name": job_name,
                "UserName": job["Props"]["User"],
                "Comment": instance.context.data.get("comment", ""),

                "Department": self.deadline_department,
                "ChunkSize": self.deadline_chunk_size,
                "Priority": priority,

                "Group": self.deadline_group,
                "Pool": instance.data.get("primaryPool"),
                "SecondaryPool": instance.data.get("secondaryPool"),
                # ensure the outputdirectory with correct slashes
                "OutputDirectory0": output_dir.replace("\\", "/")
            },
            "PluginInfo": {
                "Version": self.plugin_pype_version,
                "Arguments": " ".join(args),
                "SingleFrameOnly": "True",
            },
            # Mandatory for Deadline, may be empty
            "AuxFiles": [],
        }

        # add assembly jobs as dependencies
        if instance.data.get("tileRendering"):
            self.log.info("Adding tile assembly jobs as dependencies...")
            job_index = 0
            for assembly_id in instance.data.get("assemblySubmissionJobs"):
                payload["JobInfo"]["JobDependency{}".format(job_index)] = assembly_id  # noqa: E501
                job_index += 1
        elif instance.data.get("bakingSubmissionJobs"):
            self.log.info("Adding baking submission jobs as dependencies...")
            job_index = 0
            for assembly_id in instance.data["bakingSubmissionJobs"]:
                payload["JobInfo"]["JobDependency{}".format(job_index)] = assembly_id  # noqa: E501
                job_index += 1
        else:
            payload["JobInfo"]["JobDependency0"] = job["_id"]

        if instance.data.get("suspend_publish"):
            payload["JobInfo"]["InitialStatus"] = "Suspended"

        for index, (key_, value_) in enumerate(environment.items()):
            payload["JobInfo"].update(
                {
                    "EnvironmentKeyValue%d"
                    % index: "{key}={value}".format(
                        key=key_, value=value_
                    )
                }
            )
        # remove secondary pool
        payload["JobInfo"].pop("SecondaryPool", None)

        self.log.info("Submitting Deadline job ...")

        url = "{}/api/jobs".format(self.deadline_url)
        response = requests.post(url, json=payload, timeout=10)
        if not response.ok:
            raise Exception(response.text)

        deadline_publish_job_id = response.json()["_id"]

        return deadline_publish_job_id

    def _copy_extend_frames(self, instance, representation):
        """Copy existing frames from latest version.

        This will copy all existing frames from subset's latest version back
        to render directory and rename them to what renderer is expecting.

        Arguments:
            instance (pyblish.plugin.Instance): instance to get required
                data from
            representation (dict): presentation to operate on

        """
        import speedcopy

        self.log.info("Preparing to copy ...")
        start = instance.data.get("frameStart")
        end = instance.data.get("frameEnd")
        project_name = legacy_io.active_project()

        # get latest version of subset
        # this will stop if subset wasn't published yet
        project_name = legacy_io.active_project()
        version = get_last_version_by_subset_name(
            project_name,
            instance.data.get("subset"),
            asset_name=instance.data.get("asset")
        )

        # get its files based on extension
        subset_resources = get_resources(
            project_name, version, representation.get("ext")
        )
        r_col, _ = clique.assemble(subset_resources)

        # if override remove all frames we are expecting to be rendered
        # so we'll copy only those missing from current render
        if instance.data.get("overrideExistingFrame"):
            for frame in range(start, end + 1):
                if frame not in r_col.indexes:
                    continue
                r_col.indexes.remove(frame)

        # now we need to translate published names from represenation
        # back. This is tricky, right now we'll just use same naming
        # and only switch frame numbers
        resource_files = []
        r_filename = os.path.basename(
            representation.get("files")[0])  # first file
        op = re.search(self.R_FRAME_NUMBER, r_filename)
        pre = r_filename[:op.start("frame")]
        post = r_filename[op.end("frame"):]
        assert op is not None, "padding string wasn't found"
        for frame in list(r_col):
            fn = re.search(self.R_FRAME_NUMBER, frame)
            # silencing linter as we need to compare to True, not to
            # type
            assert fn is not None, "padding string wasn't found"
            # list of tuples (source, destination)
            staging = representation.get("stagingDir")
            staging = self.anatomy.fill_root(staging)
            resource_files.append(
                (frame,
                 os.path.join(staging,
                              "{}{}{}".format(pre,
                                              fn.group("frame"),
                                              post)))
            )

        # test if destination dir exists and create it if not
        output_dir = os.path.dirname(representation.get("files")[0])
        if not os.path.isdir(output_dir):
            os.makedirs(output_dir)

        # copy files
        for source in resource_files:
            speedcopy.copy(source[0], source[1])
            self.log.info("  > {}".format(source[1]))

        self.log.info(
            "Finished copying %i files" % len(resource_files))

    def _create_instances_for_aov(
        self, instance_data, exp_files, additional_data
    ):
        """Create instance for each AOV found.

        This will create new instance for every aov it can detect in expected
        files list.

        Arguments:
            instance_data (pyblish.plugin.Instance): skeleton data for instance
                (those needed) later by collector
            exp_files (list): list of expected files divided by aovs

        Returns:
            list of instances

        """
        task = os.environ["AVALON_TASK"]
        subset = instance_data["subset"]
        cameras = instance_data.get("cameras", [])
        instances = []
        # go through aovs in expected files
        for aov, files in exp_files[0].items():
            cols, rem = clique.assemble(files)
            # we shouldn't have any reminders. And if we do, it should
            # be just one item for single frame renders.
            if not cols and rem:
                assert len(rem) == 1, ("Found multiple non related files "
                                       "to render, don't know what to do "
                                       "with them.")
                col = rem[0]
                ext = os.path.splitext(col)[1].lstrip(".")
            else:
                # but we really expect only one collection.
                # Nothing else make sense.
                assert len(cols) == 1, "only one image sequence type is expected"  # noqa: E501
                ext = cols[0].tail.lstrip(".")
                col = list(cols[0])

            self.log.debug(col)
            # create subset name `familyTaskSubset_AOV`
            group_name = 'render{}{}{}{}'.format(
                task[0].upper(), task[1:],
                subset[0].upper(), subset[1:])

            cam = [c for c in cameras if c in col.head]
            if cam:
                if aov:
                    subset_name = '{}_{}_{}'.format(group_name, cam, aov)
                else:
                    subset_name = '{}_{}'.format(group_name, cam)
            else:
                if aov:
                    subset_name = '{}_{}'.format(group_name, aov)
                else:
                    subset_name = '{}'.format(group_name)

            if isinstance(col, (list, tuple)):
                staging = os.path.dirname(col[0])
            else:
                staging = os.path.dirname(col)

            success, rootless_staging_dir = (
                self.anatomy.find_root_template_from_path(staging)
            )
            if success:
                staging = rootless_staging_dir
            else:
                self.log.warning((
                    "Could not find root path for remapping \"{}\"."
                    " This may cause issues on farm."
                ).format(staging))

            self.log.info("Creating data for: {}".format(subset_name))

            app = os.environ.get("AVALON_APP", "")

            preview = False

            if isinstance(col, list):
                render_file_name = os.path.basename(col[0])
            else:
                render_file_name = os.path.basename(col)
            aov_patterns = self.aov_filter

            preview = match_aov_pattern(app, aov_patterns, render_file_name)
            # toggle preview on if multipart is on

            if instance_data.get("multipartExr"):
                self.log.debug("Adding preview tag because its multipartExr")
                preview = True
            self.log.debug("preview:{}".format(preview))
            new_instance = deepcopy(instance_data)
            new_instance["subset"] = subset_name
            new_instance["subsetGroup"] = group_name
            if preview:
                new_instance["review"] = True

            # create represenation
            if isinstance(col, (list, tuple)):
                files = [os.path.basename(f) for f in col]
            else:
                files = os.path.basename(col)

            # Copy render product "colorspace" data to representation.
            colorspace = ""
            products = additional_data["renderProducts"].layer_data.products
            for product in products:
                if product.productName == aov:
                    colorspace = product.colorspace
                    break

            rep = {
                "name": ext,
                "ext": ext,
                "files": files,
                "frameStart": int(instance_data.get("frameStartHandle")),
                "frameEnd": int(instance_data.get("frameEndHandle")),
                # If expectedFile are absolute, we need only filenames
                "stagingDir": staging,
                "fps": new_instance.get("fps"),
                "tags": ["review"] if preview else [],
                "colorspaceData": {
                    "colorspace": colorspace,
                    "config": {
                        "path": additional_data["colorspaceConfig"],
                        "template": additional_data["colorspaceTemplate"]
                    },
                    "display": additional_data["display"],
                    "view": additional_data["view"]
                }
            }

            # support conversion from tiled to scanline
            if instance_data.get("convertToScanline"):
                self.log.info("Adding scanline conversion.")
                rep["tags"].append("toScanline")

            # poor man exclusion
            if ext in self.skip_integration_repre_list:
                rep["tags"].append("delete")

            self._solve_families(new_instance, preview)

            new_instance["representations"] = [rep]

            # if extending frames from existing version, copy files from there
            # into our destination directory
            if new_instance.get("extendFrames", False):
                self._copy_extend_frames(new_instance, rep)
            instances.append(new_instance)
            self.log.debug("instances:{}".format(instances))
        return instances

    def _get_representations(self, instance, exp_files, additional_data):
        """Create representations for file sequences.

        This will return representations of expected files if they are not
        in hierarchy of aovs. There should be only one sequence of files for
        most cases, but if not - we create representation from each of them.

        Arguments:
            instance (dict): instance data for which we are
                             setting representations
            exp_files (list): list of expected files

        Returns:
            list of representations

        """
        representations = []
        host_name = os.environ.get("AVALON_APP", "")
        collections, remainders = clique.assemble(exp_files)

        # create representation for every collected sequence
        for collection in collections:
            ext = collection.tail.lstrip(".")
            preview = False
            # TODO 'useSequenceForReview' is temporary solution which does
            #   not work for 100% of cases. We must be able to tell what
            #   expected files contains more explicitly and from what
            #   should be review made.
            # - "review" tag is never added when is set to 'False'
            if instance["useSequenceForReview"]:
                # toggle preview on if multipart is on
                if instance.get("multipartExr", False):
                    self.log.debug(
                        "Adding preview tag because its multipartExr"
                    )
                    preview = True
                else:
                    render_file_name = list(collection)[0]
                    # if filtered aov name is found in filename, toggle it for
                    # preview video rendering
                    preview = match_aov_pattern(
                        host_name, self.aov_filter, render_file_name
                    )

            staging = os.path.dirname(list(collection)[0])
            success, rootless_staging_dir = (
                self.anatomy.find_root_template_from_path(staging)
            )
            if success:
                staging = rootless_staging_dir
            else:
                self.log.warning((
                    "Could not find root path for remapping \"{}\"."
                    " This may cause issues on farm."
                ).format(staging))

            frame_start = int(instance.get("frameStartHandle"))
            if instance.get("slate"):
                frame_start -= 1

            rep = {
                "name": ext,
                "ext": ext,
                "files": [os.path.basename(f) for f in list(collection)],
                "frameStart": frame_start,
                "frameEnd": int(instance.get("frameEndHandle")),
                # If expectedFile are absolute, we need only filenames
                "stagingDir": staging,
                "fps": instance.get("fps"),
                "tags": ["review"] if preview else [],
            }

            # poor man exclusion
            if ext in self.skip_integration_repre_list:
                rep["tags"].append("delete")

            if instance.get("multipartExr", False):
                rep["tags"].append("multipartExr")

            # support conversion from tiled to scanline
            if instance.get("convertToScanline"):
                self.log.info("Adding scanline conversion.")
                rep["tags"].append("toScanline")

            representations.append(rep)

            self._solve_families(instance, preview)

        # add remainders as representations
        for remainder in remainders:
            ext = remainder.split(".")[-1]

            staging = os.path.dirname(remainder)
            success, rootless_staging_dir = (
                self.anatomy.find_root_template_from_path(staging)
            )
            if success:
                staging = rootless_staging_dir
            else:
                self.log.warning((
                    "Could not find root path for remapping \"{}\"."
                    " This may cause issues on farm."
                ).format(staging))

            rep = {
                "name": ext,
                "ext": ext,
                "files": os.path.basename(remainder),
                "stagingDir": staging,
            }

            preview = match_aov_pattern(
                host_name, self.aov_filter, remainder
            )
            if preview:
                rep.update({
                    "fps": instance.get("fps"),
                    "tags": ["review"]
                })
            self._solve_families(instance, preview)

            already_there = False
            for repre in instance.get("representations", []):
                # might be added explicitly before by publish_on_farm
                already_there = repre.get("files") == rep["files"]
                if already_there:
                    self.log.debug("repre {} already_there".format(repre))
                    break

            if not already_there:
                representations.append(rep)

        return representations

    def _solve_families(self, instance, preview=False):
        families = instance.get("families")

        # if we have one representation with preview tag
        # flag whole instance for review and for ftrack
        if preview:
            if "ftrack" not in families:
                if os.environ.get("FTRACK_SERVER"):
                    self.log.debug(
                        "Adding \"ftrack\" to families because of preview tag."
                    )
                    families.append("ftrack")
            if "review" not in families:
                self.log.debug(
                    "Adding \"review\" to families because of preview tag."
                )
                families.append("review")
            instance["families"] = families

    def process(self, instance):
        # type: (pyblish.api.Instance) -> None
        """Process plugin.

        Detect type of renderfarm submission and create and post dependend job
        in case of Deadline. It creates json file with metadata needed for
        publishing in directory of render.

        Args:
            instance (pyblish.api.Instance): Instance data.

        """
        data = instance.data.copy()
        context = instance.context
        self.context = context
        self.anatomy = instance.context.data["anatomy"]

        asset = data.get("asset") or legacy_io.Session["AVALON_ASSET"]
        subset = data.get("subset")

        start = instance.data.get("frameStart")
        if start is None:
            start = context.data["frameStart"]

        end = instance.data.get("frameEnd")
        if end is None:
            end = context.data["frameEnd"]

        handle_start = instance.data.get("handleStart")
        if handle_start is None:
            handle_start = context.data["handleStart"]

        handle_end = instance.data.get("handleEnd")
        if handle_end is None:
            handle_end = context.data["handleEnd"]

        fps = instance.data.get("fps")
        if fps is None:
            fps = context.data["fps"]

        if data.get("extendFrames", False):
            start, end = self._extend_frames(
                asset,
                subset,
                start,
                end,
                data["overrideExistingFrame"])

        try:
            source = data["source"]
        except KeyError:
            source = context.data["currentFile"]

        success, rootless_path = (
            self.anatomy.find_root_template_from_path(source)
        )
        if success:
            source = rootless_path

        else:
            # `rootless_path` is not set to `source` if none of roots match
            self.log.warning((
                "Could not find root path for remapping \"{}\"."
                " This may cause issues."
            ).format(source))

        family = "render"
        if "prerender" in instance.data["families"]:
            family = "prerender"
        families = [family]

        # pass review to families if marked as review
        if data.get("review"):
            families.append("review")

        instance_skeleton_data = {
            "family": family,
            "subset": subset,
            "families": families,
            "asset": asset,
            "frameStart": start,
            "frameEnd": end,
            "handleStart": handle_start,
            "handleEnd": handle_end,
            "frameStartHandle": start - handle_start,
            "frameEndHandle": end + handle_end,
            "comment": instance.data["comment"],
            "fps": fps,
            "source": source,
            "extendFrames": data.get("extendFrames"),
            "overrideExistingFrame": data.get("overrideExistingFrame"),
            "pixelAspect": data.get("pixelAspect", 1),
            "resolutionWidth": data.get("resolutionWidth", 1920),
            "resolutionHeight": data.get("resolutionHeight", 1080),
            "multipartExr": data.get("multipartExr", False),
            "jobBatchName": data.get("jobBatchName", ""),
            "useSequenceForReview": data.get("useSequenceForReview", True),
            # map inputVersions `ObjectId` -> `str` so json supports it
            "inputVersions": list(map(str, data.get("inputVersions", [])))
        }

        # skip locking version if we are creating v01
        instance_version = instance.data.get("version")  # take this if exists
        if instance_version != 1:
            instance_skeleton_data["version"] = instance_version

        # transfer specific families from original instance to new render
        for item in self.families_transfer:
            if item in instance.data.get("families", []):
                instance_skeleton_data["families"] += [item]

        # transfer specific properties from original instance based on
        # mapping dictionary `instance_transfer`
        for key, values in self.instance_transfer.items():
            if key in instance.data.get("families", []):
                for v in values:
                    instance_skeleton_data[v] = instance.data.get(v)

        # look into instance data if representations are not having any
        # which are having tag `publish_on_farm` and include them
        for repre in instance.data.get("representations", []):
            staging_dir = repre.get("stagingDir")
            if staging_dir:
                success, rootless_staging_dir = (
                    self.anatomy.find_root_template_from_path(
                        staging_dir
                    )
                )
                if success:
                    repre["stagingDir"] = rootless_staging_dir
                else:
                    self.log.warning((
                        "Could not find root path for remapping \"{}\"."
                        " This may cause issues on farm."
                    ).format(staging_dir))
                    repre["stagingDir"] = staging_dir

            if "publish_on_farm" in repre.get("tags"):
                # create representations attribute of not there
                if "representations" not in instance_skeleton_data.keys():
                    instance_skeleton_data["representations"] = []

                instance_skeleton_data["representations"].append(repre)

        instances = None
        assert data.get("expectedFiles"), ("Submission from old Pype version"
                                           " - missing expectedFiles")

        """
        if content of `expectedFiles` are dictionaries, we will handle
        it as list of AOVs, creating instance from every one of them.

        Example:
        --------

        expectedFiles = [
            {
                "beauty": [
                    "foo_v01.0001.exr",
                    "foo_v01.0002.exr"
                ],

                "Z": [
                    "boo_v01.0001.exr",
                    "boo_v01.0002.exr"
                ]
            }
        ]

        This will create instances for `beauty` and `Z` subset
        adding those files to their respective representations.

        If we've got only list of files, we collect all filesequences.
        More then one doesn't probably make sense, but we'll handle it
        like creating one instance with multiple representations.

        Example:
        --------

        expectedFiles = [
            "foo_v01.0001.exr",
            "foo_v01.0002.exr",
            "xxx_v01.0001.exr",
            "xxx_v01.0002.exr"
        ]

        This will result in one instance with two representations:
        `foo` and `xxx`
        """

        self.log.info(data.get("expectedFiles"))

        additional_data = {
            "renderProducts": instance.data["renderProducts"],
            "colorspaceConfig": instance.data["colorspaceConfig"],
            "display": instance.data["colorspaceDisplay"],
            "view": instance.data["colorspaceView"],
            "colorspaceTemplate": instance.data["colorspaceConfig"].replace(
                str(context.data["anatomy"].roots["work"]), "{root[work]}"
            )
        }

        if isinstance(data.get("expectedFiles")[0], dict):
            # we cannot attach AOVs to other subsets as we consider every
            # AOV subset of its own.

            if len(data.get("attachTo")) > 0:
                assert len(data.get("expectedFiles")[0].keys()) == 1, (
                    "attaching multiple AOVs or renderable cameras to "
                    "subset is not supported")

            # create instances for every AOV we found in expected files.
            # note: this is done for every AOV and every render camere (if
            #       there are multiple renderable cameras in scene)
            instances = self._create_instances_for_aov(
                instance_skeleton_data,
                data.get("expectedFiles"),
                additional_data
            )
            self.log.info("got {} instance{}".format(
                len(instances),
                "s" if len(instances) > 1 else ""))

        else:
            representations = self._get_representations(
                instance_skeleton_data,
                data.get("expectedFiles")
            )

            if "representations" not in instance_skeleton_data.keys():
                instance_skeleton_data["representations"] = []

            # add representation
            instance_skeleton_data["representations"] += representations
            instances = [instance_skeleton_data]

        # if we are attaching to other subsets, create copy of existing
        # instances, change data to match thats subset and replace
        # existing instances with modified data
        if instance.data.get("attachTo"):
            self.log.info("Attaching render to subset:")
            new_instances = []
            for at in instance.data.get("attachTo"):
                for i in instances:
                    new_i = copy(i)
                    new_i["version"] = at.get("version")
                    new_i["subset"] = at.get("subset")
                    new_i["family"] = at.get("family")
                    new_i["append"] = True
                    # don't set subsetGroup if we are attaching
                    new_i.pop("subsetGroup")
                    new_instances.append(new_i)
                    self.log.info("  - {} / v{}".format(
                        at.get("subset"), at.get("version")))
            instances = new_instances

        r''' SUBMiT PUBLiSH JOB 2 D34DLiN3
          ____
        '     '            .---.  .---. .--. .---. .--..--..--..--. .---.
        |     |   --= \   |  .  \/   _|/    \|  .  \  ||  ||   \  |/   _|
        | JOB |   --= /   |  |  ||  __|  ..  |  |  |  |;_ ||  \   ||  __|
        |     |           |____./ \.__|._||_.|___./|_____|||__|\__|\.___|
        ._____.

        '''

        render_job = None
        submission_type = ""
        if instance.data.get("toBeRenderedOn") == "deadline":
            render_job = data.pop("deadlineSubmissionJob", None)
            submission_type = "deadline"

        if instance.data.get("toBeRenderedOn") == "muster":
            render_job = data.pop("musterSubmissionJob", None)
            submission_type = "muster"

        if not render_job and instance.data.get("tileRendering") is False:
            raise AssertionError(("Cannot continue without valid Deadline "
                                  "or Muster submission."))

        if not render_job:
            import getpass

            render_job = {}
            self.log.info("Faking job data ...")
            render_job["Props"] = {}
            # Render job doesn't exist because we do not have prior submission.
            # We still use data from it so lets fake it.
            #
            # Batch name reflect original scene name

            if instance.data.get("assemblySubmissionJobs"):
                render_job["Props"]["Batch"] = instance.data.get(
                    "jobBatchName")
            else:
                render_job["Props"]["Batch"] = os.path.splitext(
                    os.path.basename(context.data.get("currentFile")))[0]
            # User is deadline user
            render_job["Props"]["User"] = context.data.get(
                "deadlineUser", getpass.getuser())

            render_job["Props"]["Env"] = {
                "FTRACK_API_USER": os.environ.get("FTRACK_API_USER"),
                "FTRACK_API_KEY": os.environ.get("FTRACK_API_KEY"),
                "FTRACK_SERVER": os.environ.get("FTRACK_SERVER"),
            }

        if submission_type == "deadline":
            # get default deadline webservice url from deadline module
            self.deadline_url = instance.context.data["defaultDeadline"]
            # if custom one is set in instance, use that
            if instance.data.get("deadlineUrl"):
                self.deadline_url = instance.data.get("deadlineUrl")
            assert self.deadline_url, "Requires Deadline Webservice URL"

            deadline_publish_job_id = \
                self._submit_deadline_post_job(instance, render_job, instances)

        # publish job file
        publish_job = {
            "asset": asset,
            "frameStart": start,
            "frameEnd": end,
            "fps": context.data.get("fps", None),
            "source": source,
            "user": context.data["user"],
            "version": context.data["version"],  # this is workfile version
            "intent": context.data.get("intent"),
            "comment": context.data.get("comment"),
            "job": render_job or None,
            "session": legacy_io.Session.copy(),
            "instances": instances
        }

        if deadline_publish_job_id:
            publish_job["deadline_publish_job_id"] = deadline_publish_job_id

        # add audio to metadata file if available
        audio_file = context.data.get("audioFile")
        if audio_file and os.path.isfile(audio_file):
            publish_job.update({"audio": audio_file})

        # pass Ftrack credentials in case of Muster
        if submission_type == "muster":
            ftrack = {
                "FTRACK_API_USER": os.environ.get("FTRACK_API_USER"),
                "FTRACK_API_KEY": os.environ.get("FTRACK_API_KEY"),
                "FTRACK_SERVER": os.environ.get("FTRACK_SERVER"),
            }
            publish_job.update({"ftrack": ftrack})

        metadata_path, rootless_metadata_path = self._create_metadata_path(
            instance)

        self.log.info("Writing json file: {}".format(metadata_path))
        with open(metadata_path, "w") as f:
            json.dump(publish_job, f, indent=4, sort_keys=True)

    def _extend_frames(self, asset, subset, start, end):
        """Get latest version of asset nad update frame range.

        Based on minimum and maximuma values.

        Arguments:
            asset (str): asset name
            subset (str): subset name
            start (int): start frame
            end (int): end frame

        Returns:
            (int, int): upddate frame start/end

        """
        # Frame comparison
        prev_start = None
        prev_end = None

        project_name = legacy_io.active_project()
        version = get_last_version_by_subset_name(
            project_name,
            subset,
            asset_name=asset
        )

        # Set prev start / end frames for comparison
        if not prev_start and not prev_end:
            prev_start = version["data"]["frameStart"]
            prev_end = version["data"]["frameEnd"]

        updated_start = min(start, prev_start)
        updated_end = max(end, prev_end)

        self.log.info(
            "Updating start / end frame : "
            "{} - {}".format(updated_start, updated_end)
        )

        return updated_start, updated_end

    def _get_publish_folder(self, anatomy, template_data,
                            asset, subset,
                            family='render', version=None):
        """
            Extracted logic to pre-calculate real publish folder, which is
            calculated in IntegrateNew inside of Deadline process.
            This should match logic in:
                'collect_anatomy_instance_data' - to
                    get correct anatomy, family, version for subset and
                'collect_resources_path'
                    get publish_path

        Args:
            anatomy (openpype.pipeline.anatomy.Anatomy):
            template_data (dict): pre-calculated collected data for process
            asset (string): asset name
            subset (string): subset name (actually group name of subset)
            family (string): for current deadline process it's always 'render'
                TODO - for generic use family needs to be dynamically
                    calculated like IntegrateNew does
            version (int): override version from instance if exists

        Returns:
            (string): publish folder where rendered and published files will
                be stored
                based on 'publish' template
        """
        if not version:
            project_name = legacy_io.active_project()
            version = get_last_version_by_subset_name(
                project_name,
                subset,
                asset_name=asset
            )
            if version:
                version = int(version["name"]) + 1
            else:
                version = 1

        template_data["subset"] = subset
        template_data["family"] = "render"
        template_data["version"] = version

        anatomy_filled = anatomy.format(template_data)

        if "folder" in anatomy.templates["render"]:
            publish_folder = anatomy_filled["render"]["folder"]
        else:
            # solve deprecated situation when `folder` key is not underneath
            # `publish` anatomy
            project_name = legacy_io.Session["AVALON_PROJECT"]
            self.log.warning((
                "Deprecation warning: Anatomy does not have set `folder`"
                " key underneath `publish` (in global of for project `{}`)."
            ).format(project_name))

            file_path = anatomy_filled["render"]["path"]
            # Directory
            publish_folder = os.path.dirname(file_path)

        return publish_folder