How do I capture the script editor output of a mel command, using in python?

Coding
, ,
#1

We have some plugins that run as mel commands, and when things go wrong they spew useful info to the script editor, but return 0. How do I capture the script editor spew (not the return value) of a mel command, using python?

import pymel.core as pm
mel_cmd= 'print("hello world\\n");'
result=pm.mel.eval(mel_cmd) # no this is just the return value
print "the output of mel '{}' is: '{}'".format(mel_cmd, result)

#script editor output:
#hello world
#the output of mel 'print("hello world\n");' is: 'None'
#2

so I came up with this clunky decorator function, but it feels…clunky.

def capture(function):
    """
    a decorator fn that captures output sent to the maya script editor
    :param function: a function whose script editor output we want to capture
    :return:
    """

    TEMP_FILE = os.getenv("TEMP") + ("/maya_script_editor_capture.txt")


    # empty the temp file so we only capture current command
    with open(TEMP_FILE, "w") as f:
        f.write("")

    #wrap the function that outputs to the script editor
    def captured_output(*args, **kw):
        pm.scriptEditorInfo(writeHistory=True, historyFilename=TEMP_FILE)
        result= function(*args, **kw)
        pm.scriptEditorInfo(writeHistory=False)
        with open(TEMP_FILE, "r") as f:
            history=f.read()
            return history
    return captured_output

@capture
def test():
    mel_cmd= 'print("hello world\\n");'
    pm.mel.eval(mel_cmd)

captured_result=test()
print("captured result: {}".format(captured_result))

Seems to give me what I want but I wonder if there’s some other way to redirect the script editor, so that I’m not writing and reading files just to get the text.

#3

You might be able to use StringIO.StringIO() instance instead of the temp file.

1 Like
#4

I’ll give StringIO a shot.

the plugin programmer explained that the plugin returns MStatus::kSuccess or MStatus::kFailure
but I can’t seem to capture a value from this using MEL (or python wrapped mel). the return values appear to be null /None

#5

StringIO class instances can’t be passed as the historyFilename parameter.
odd that scriptEditorInfo() has no direct ‘give me your text’ function.

#6

Ah, didn’t realize it took a file path and not a file object.
Yeah StringIO won’t work in that case.

Another option would be OpenMaya.MGlobal.executeCommand which lets you run a mel command and potentially capture the success/failure.
Pymel uses this under the hood to capture errors / convert to Exceptions.

#7

What happens if you stick an imposter into sys.stdout in python… ? It will allow you to, for example, record the log there before passing it along. Does mel show up in that? never tried it , but perhaps someone has.

#8

I was able to use MCommandMessage to capture a filepath from arnold. This doesn’t seem to work with the new api.

# capture stdout in a global var instead of writing to a file
# om2 doesn't work
global cmd_output
cmd_output = []

def _callback(nativeMsg, messageType, data):
    global cmd_output
    cmd_output.append(nativeMsg)

id = om.MCommandMessage.addCommandOutputCallback(_callback, None)

# fire the render
mc.arnoldRenderToTexture(
    folder=outputFolder, 
    shader=shader, 
    resolution=resolution, 
    aa_samples=samples, 
    filter='gaussian', 
    normal_offset=.1, 
    enable_aovs=False, 
    extend_edges=False,
    sequence= False
    )

# !! deregister the callback !!
# copy command output to string
om.MMessage.removeCallback(id)
# print cmd_output

# parse and check output for path
for line in cmd_output:
    path = line.split('[mtoa] Render to Texture : Rendered to ')[-1]
#9

I happened across this problem again, cleaned up implementation a bit. Posting here in case anyone encounters.

import maya.api.OpenMaya as om
from contextlib import contextmanager

# Build a simple data class, initialize a simple list
class CustomData:
    def __init__(self):
        self.value = []

# Create a custom callback, this will be passed three arguments from MCommand message
def customCommandOutputCallback(message, filter, customData):
    if isinstance(customData, CustomData):
        customData.value.append(message)

# Setup context manager
@contextmanager
def capture_output():
    # setup the custom data
    data = CustomData()
    try:
        # >>>> IMPORTANT DO NOT PRINT<<<<
        # Any print statements in this block will create ain infinite loop and crash
        # you are trying to capture the the item you are printing...
        id = om.MCommandMessage.addCommandOutputCallback(customCommandOutputCallback, data)
        yield data
    finally:
        om.MCommandMessage.removeCallback(id)
        # print("Exiting the context manager")
        
    return data

# example
def get_queue(redo=False, strip=False):
    queue = []
    if redo:
         with capture_output() as output:
            mc.undoInfo(q=True, prq=True)
    else:
        with capture_output() as output:
            mc.undoInfo(q=True, pq=True)

    
    if output.value is not None:
        queue = output.value
        
        # optionally remove the queue id
        if strip:
            queue = [v.split(':')[-1].lstrip() for v in output.value]
    
    return queue

get_queue()
get_queue(redo=True)
2 Likes