Log manager for Python based nodes

[ChemicalXandco]

Description

There is currently no easy way for nodes that are executed in python to use the log file, this system makes adding a message as simple as chunk.logger.info('message'), currently the Publish node uses print() but this is not very helpful because it is not stored in a file that can be viewed afterwards.

chunk.logManager.waitUntilCleared()
chunk.logger.setLevel(chunk.logManager.textToLevel(chunk.node.verboseLevel.value))

chunk.logger.info('info message')
chunk.logger.warning('warning message')
chunk.logger.error('error message')

chunk.logManager.makeProgressBar(100)
for i in range(101):
    chunk.logManager.updateProgressBar(i)
chunk.logManager.completeProgressBar()

becomes

[18:06:47.016][info] info message
[18:06:47.022][warning] warning message
[18:06:47.028][error] error message
0%   10   20   30   40   50   60   70   80   90   100%
|----|----|----|----|----|----|----|----|----|----|
***************************************************

Features list

• LogManager class to manage the logs
• Simulation of boost::progress_display within LogManager
• Updated Publish node to use this system

Implementation remarks

• Progress bar can be used while adding messages
• Log clears automatically every time the chunk is ran

I do welcome feedback because I am not sure if this is the best way to implement this.

[fabiencastan]

Hi @ChemicalXandco ,
That's an interesting feature that will be useful to create more python-only nodes.

In term of implementation, it would be nice if we could rely on a standard logging object.
With logging, you can directly write to file, like
logging.basicConfig(filename='/path/to/node/cache/log', level=logging.INFO).
And create a logging object in your class with something like log = logging.getLogger(uniqueNodeId).

It would be good also to expose a verbose level parameter on the node, as on the C++ nodes.
Best,

[fabiencastan]

    def __init__(self, node, range, parent=None):
        ...
        self.logManager = LogManager(self)
        ...

    @property
    def logging(self):
        return self.logManager.logger

[ChemicalXandco]

I have called the property logger instead of logging because it makes more sense imo.

[fabiencastan]

yes, I agree.

[fabiencastan]

chunk.logging.info('Publish end')

[TigerVersusT]

@ChemicalXandco a stupid question, it seems that your logManager class is used for python based nodes only, how about CLI based node? Is it suitable to use your class for log implementtation in my own CLI nodes? And could you tell me somthing about how do nodes like "Texturing" in meshroom handle the log message.Thanks.

[ChemicalXandco]

@TigerVersusT The C++ side of the software (AliceVision framework) uses it's own logging system. By 'CLI based node', I'm assuming you mean an executable file not from the AliceVision framework. In that case, the best approach imo would be to forward the messages from stdout of the process to the logger implemented here.

[TigerVersusT]

@ChemicalXandco Thanks for your help!