These are chat archives for ipython/ipython
My question really boils down to: Is there a natural way within the protocol to split/partition large execute_results and/or display_data messages into smaller messages that could be streamed/reassembled later w/o having to allocate a heap equal to the data size. You might imagine I have a multi-gigabyte image, or perhaps more likely converting a data frame to CSV/JSON.
I realize that one option would be to write that data to disk rather than trying to pass it through the IOPUB. That might even be the best option but at the moment my security/resource model assumes that python is running in a container w/ a read only FS.