STORE_RECV_CHUNKED_DATASET

[stefanoostwegel]

Hello,

We have a tool that transfers many DICOM series concurrent.
This could beup to 20 concurrent treads.
In update 2.0 i see a new configuration item:
https://pydicom.github.io/pynetdicom/dev/reference/generated/pynetdicom._config.STORE_RECV_CHUNKED_DATASET.html

`pynetdicom._config.STORE_RECV_CHUNKED_DATASET¶
pynetdicom._config.STORE_RECV_CHUNKED_DATASET: bool= False¶
Chunk a dataset file when receiving it to minimise memory usage.

New in version 2.0.

If True, then when receiving C-STORE requests as an SCP, don’t decode the dataset and instead write the raw data to a temporary file in the DICOM File Format. The path to the dataset is available to the evt.EVT_C_STORE handler using the Event.dataset_path attribute. This should minimise the amount of memory required when:

Receiving large datasets

Receiving many datasets concurrently

Default: False

Examples

from pynetdicom import _config
_config.STORE_RECV_CHUNKED_DATASET = True`

This sounds like a very nice feature for us to use since we have quite a lot of 'stress' on our SCP.
Is it just as simple as setting this flag to true? Or does it require more postprocessing or anything else after that?
Im not quite sure if i can implement it this easy, or do i need to do anything extra with the data, after recieving it?

[scaramallion]

Yeah, its just as simple as setting the configuration option to True. Obviously you'll need to do something with the received dataset like read/move/copy it from the temporary location.

You may want to run some sort of cleanup action on the temporary directory used by Python as well, just in case.

[stefanoostwegel]

So, i use pydicom to write the files to disk, and then use it to distrubute it to other locations.
Is that the temp folder you talk about, or are you talking about an extra tempfolder?

[scaramallion]

The files will be saved using NamedTemporaryFile to the path where Python typically writes them (varies by OS). It's probably good practice to clean up that directory every now and then.

The full path to the NamedTemporaryFiles is given by Event.dataset_path but the dataset file should only exist while performing the C-STORE operation, so you need to copy/read (or whatever) it or lose it.