I am new to Span and Memory, and am trying to migrate my library to them. I discovered PooledList when looking into how to AddRange a Span to a List. My initial migration increased the warm of my program by 30% (parsing a large data file and loading it into different lists and dictionaries in memory). Thanks!
However I did find one point that I needed to work around:
It currently isn't possible to get a Memory from PooledList.
Being able to access a Memory from the PooledList will allow me to pass around heap storable references of sub-ranges which makes the PooledList even more flexible. I'd be able to use references to it with Linq IEnumerable and also be able to use it in local functions. Ie. it would free the references for the restrictions that are required by Span.
I understand that PooledList is built upon ArrayPool which means that returning a Memory will cause a memory leak when expanding the PooledList.
Is it possible to implement this functionality - or would it come as an unacceptable tradeoff to the performance/complexity of the library?
I am new to Span and Memory, and am trying to migrate my library to them. I discovered PooledList when looking into how to AddRange a Span to a List. My initial migration increased the warm of my program by 30% (parsing a large data file and loading it into different lists and dictionaries in memory). Thanks! However I did find one point that I needed to work around: It currently isn't possible to get a Memory from PooledList.
Being able to access a Memory from the PooledList will allow me to pass around heap storable references of sub-ranges which makes the PooledList even more flexible. I'd be able to use references to it with Linq IEnumerable and also be able to use it in local functions. Ie. it would free the references for the restrictions that are required by Span.
I understand that PooledList is built upon ArrayPool which means that returning a Memory will cause a memory leak when expanding the PooledList.
Is it possible to implement this functionality - or would it come as an unacceptable tradeoff to the performance/complexity of the library?
Thanks again