Azure Functions Python Blob image

Question

I am trying to load two images in, one uploaded image (trigger) and one from a hardcoded blob location.

def main(blobtrig: func.InputStream, medianimage: func.InputStream, blobout: func.Out[bytes]):
    logging.info(f"Python blob trigger function processed blob \n"
                 f"Stuff: {blobtrig.__dict__}\n")

    logging.info(f"Next Blob Name \n"
                 f"Stuff: {medianimage.__dict__}\n")

    input_image = blobtrig
    base_image = Image.open(input_image)

    med_image = medianimage
    logging.info(f"Med image read is {type(med_image)}")
    median_image = Image.open(med_image)

The first Image.open(input_image) works, but the second doesn't. I think the medianimage Inputstream is not pointing to the right place. My functions.json is like this...

{"scriptFile": "__init__.py","bindings": [
{
  "name": "blobtrig",
  "type": "blobTrigger",
  "direction": "in",
  "path": "facility-model-image-data/{name}",
  "connection": "AzureWebJobsStorage"
},
{
  "type": "blob",
  "direction": "in",
  "name": "medianimage",
  "path": "facility-model-image-data/median.jpg",
  "connection": "AzureWebJobsStorage"
},
{
  "type": "blob",
  "direction": "out",
  "name": "blobout",
  "path": "processed-images/{rand-guid}.jpg",
  "connection": "AzureWebJobsStorage"
}],"disabled": false, "scriptFile": "__init__.py"}

But the files printed by python logging are both the same even though medianimage should be pointing to a completely different file:

[02/02/2020 18:39:36] Stuff: {'_io': <_io.BytesIO object at 0x7f2cabdc0620>,
'_name': 'facility-model-image-data/test_gray.jpg', '_length': 152606,
'_uri': 'https://functionimageprocessor.blob.core.windows.net/facility-model-image-data/test_gray.jpg'}
[02/02/2020 18:39:36] 
[02/02/2020 18:39:36] Next Blob Name 
[02/02/2020 18:39:36] Stuff: {'_io': <_io.BytesIO object at 0x7f2cabdc0f10>,
'_name': 'facility-model-image-data/test_gray.jpg', '_length': 152606,
'_uri': 'https://functionimageprocessor.blob.core.windows.net/facility-model-image-data/test_gray.jpg'}

and the script fails here...

[02/02/2020 18:39:36] Stack:   File "/usr/lib/azure-functions-core-tools/workers/python/3.7/LINUX/X64/azure_functions_worker/dispatcher.py", line 312, in _handle__invocation_request
[02/02/2020 18:39:36]     self.__run_sync_func, invocation_id, fi.func, args)
[02/02/2020 18:39:36]   File "/usr/lib/python3.7/concurrent/futures/thread.py", line 57, in run
[02/02/2020 18:39:36]     result = self.fn(*self.args, **self.kwargs)
[02/02/2020 18:39:36]   File "/usr/lib/azure-functions-core-tools/workers/python/3.7/LINUX/X64/azure_functions_worker/dispatcher.py", line 431, in __run_sync_func
[02/02/2020 18:39:36]     return func(**params)
[02/02/2020 18:39:36]   File "/home/peter/Documents/Python/dataset-processing/functions/imageuploadtrigger/__init__.py", line 35, in main
[02/02/2020 18:39:36]     median_image = Image.open(med_image)#Image.frombytes('RGBA', (636,795), med_image, 'raw')

Been struggling with this for ages so any help would be great! Also sorry if the question formatting isn't the best (first time posting here).

Answer 1

This looks about right. I was unable to find any examples of using 2 different blobBindings with python, unfortunately. but both the blobTrigger and blobBinding are read via func.InputStream. My guess would be - this doesn't really work with python (perhaps a bug?). You can file an issue on Github.

Couple of workarounds:

1. if that median.jpg is relatively static - you can just copy it locally to the function storage and read it without having to have a blobBinding
2. if it's not - you can use Azure Storage SDK for python to read the file from the blob using the SDK, not the binding.