is POI HMEF memory constant

[marijusGood]

Hello POI team,

I Would like to use POI HMEF, but before that, I would need to know if it is memory constant. As in, if I have a SharedInputStream(or any other InputStream) will it be able to use it without converting it to a byte array. As an example, if we have a SharedInputStream that is memory constant as it reads from a file that is of size 10MB, will POI HMEF not read all of it to memory for processing?

Thanks!

[pjfanning]

Hi Marijus,
POI is a 20 year old project, well past peak development and community participation. Maybe someone will do this work for you but from experience, it will go unanswered. I will only advise that you are better off testing this for yourself instead of expecting unpaid volunteers to test it for you. Few if any of the POI APIs are written to fully stream the input data - ie most APIs that take InputStream inputs will still read all the data into memory before processing it. The old formats like HMEF tend to be structured so that streaming implementations will gain little benefit anyway. The X**F formats are more streamable and we have some stream friendly support for XSSF (and SXXSF).

[marijusGood]

Hi @pjfanning

Thanks for the honest response — I appreciate the context and completely understand the limits of time and community bandwidth.

Given that, do you have any recommendations on how I might enhance POI HMEF to support memory-constant streaming? I realize the project is quite large, so any guidance on which parts of the codebase would be the most relevant to focus on would be extremely helpful. If there's a particular abstraction layer or component responsible for buffering or parsing the input stream, I’d be happy to start there and see what’s feasible.

Thanks again