CollectNukeWrites plugin fails for collection render targets if no frames/directory exist

[sebastianbrandhuber]

Is there an existing issue for this?

• I have searched the existing issues and added correct labels.

Description

Current Behavior

The CollectNukeWrites plugin fails, if render target is either "frames" or "frames_farm" and either the directory or the frames inside an existing directory do not exist.

Expected Behavior

Should not fail at this stage of collection OR raise a descriptive PublishError.

Steps To Reproduce:

1. Open Nuke, create a write instance but do not render.
2. Set its render target to "Use existing frames" or "Use existing frames - farm".
3. Publish fails with a FileNotFoundError because the directory does not exist -
   or might fail with an IndexError if the directory exists, but no files.

Additional context:

Version

ayon-nuke 0.4.10

What platform were you running when you found the bug?

Windows 11

I am not sure where this issue should be fixed first,
since a validator exists:
https://github.com/ynput/ayon-nuke/blob/develop/client/ayon_nuke/plugins/publish/validate_rendered_frames.py

This validator only validates existing representations, that might not exist yet for validation because the
Use existing frames - farm render target does not create one during local publish.

But even if using Use existing frames, the collect write plugin still fails, so the collector plugin needs to be fixed anyways to do a better error handling.

As an idea, we could either raise a PublishError directly in the collector,
or store the error information on the instance and raise the Error during validation in the
validate_rendered_frames plugin.

Relevant log output:

Traceback (most recent call last):
  File "C:\Users\brandhuber\AppData\Local\Ynput\AYON\dependency_packages\ayon_2602231726_windows.zip\dependencies\pyblish\plugin.py", line 528, in __explicit_process
    runner(*args)
  File "[dev]\ayon-nuke\client\ayon_nuke\plugins\publish\collect_writes.py", line 51, in process
    self._set_existing_files_data(instance, colorspace)
  File "[dev]\ayon-nuke\client\ayon_nuke\plugins\publish\collect_writes.py", line 76, in _set_existing_files_data
    collected_frames = self._get_collected_frames(instance)
  File "[dev]\ayon_nuke\plugins\publish\collect_writes.py", line 421, in _get_collected_frames
    for filename in os.listdir(output_dir)
FileNotFoundError: [WinError 3] The system cannot find the path specified: 'xxx/0001_010_comp_sbr'


Traceback (most recent call last):
  File "C:\Users\brandhuber\AppData\Local\Ynput\AYON\dependency_packages\ayon_2602231726_windows.zip\dependencies\pyblish\plugin.py", line 528, in __explicit_process
    runner(*args)
  File "[dev]\ayon-nuke\client\ayon_nuke\plugins\publish\collect_writes.py", line 51, in process
    self._set_existing_files_data(instance, colorspace)
  File "[dev]\ayon-nuke\client\ayon_nuke\plugins\publish\collect_writes.py", line 83, in _set_existing_files_data
    self.set_representation_colorspace(
  File "C:\Users\brandhuber\AppData\Local\Ynput\AYON\addons\core_1.9.7\ayon_core\pipeline\publish\publish_plugins.py", line 494, in set_representation_colorspace
    set_colorspace_data_to_representation(
  File "C:\Users\brandhuber\AppData\Local\Ynput\AYON\addons\core_1.9.7\ayon_core\pipeline\colorspace.py", line 1314, in set_colorspace_data_to_representation
    filename = filename[0]
IndexError: list index out of range

[sebastianbrandhuber]

As a quick sketch:

Option 1 – Validate during collection - raise PublishError directly

# Raise PublishError if output dir does not exist
# or no rendered frames are collected
error_msg = (
    f"No frames were collected for instance '{instance.name}'.\n"
    "You need to render them first."
)
if not os.path.exists(output_dir):
    raise PublishError(error_msg)

# make sure files are existing at folder
collected_frames = [
    filename
    for filename in os.listdir(output_dir)
    if filename in expected_filenames
]

if not collected_frames:
    raise PublishError(error_msg)

This approach is more straight forward - but would most likely make the validate_rendered_frames plugin redundant and does not quite follow the intended pyblish logic.

Option 2 – Validate during collection - update instance - raise in validate plugin

# Validate if output dir exists or rendered frames are collected
# and store result on instance data
# to ensure validation happens in 'validate_rendered_frames' plugin.
if not os.path.exists(output_dir):
    collected_frames = []
else:
    # make sure files are existing at folder
    collected_frames = [
        filename
        for filename in os.listdir(output_dir)
        if filename in expected_filenames
    ]

instance.data["validate_rendered_frames"] = bool(collected_frames)

This approach somehow feels like a hack to enforce validation later on in the publish process and handle validation information without an actual existing representation.
Also it would require a bit more work in the two mentioned plugins to make sure set_representation_colorspace
handles empty collection lists correctly (or is not called at all) and validate_rendered_frames knows how to handle the newly created instance data like so:

if not instance.data.get("validate_rendered_frames", True):
    raise PublishXmlValidationError(...)