Question: Is it safe to use PrefixingBufferWriter without calling Commit()?

[rmja]

Hi,

I am using the PrefixingBufferWriter in a context that may throw after bytes have been written to it.
It does not implement IDisposable, yet it seems that it may have a sequence memory pool allocation if writing has started. I can see that the sequence is reset/memory returned when Commit() is called. But, how do I ensure that the allocated sequence is correctly returned to the pool in the case where Commit() is not called on a writer?

[AArnott]

I'm not sure. If an exception were thrown, I would tend to distrust anything in a writing chain. If the exception were thrown by Commit() itself for example, there's no telling how much data was written to the underlying writer. What kind of recovery did you have in mind after a writing exception is thrown?

[rmja]

I have something like this:

using var sequnce = SequencePool.Get();
var writer =  new PrefixingBufferWriter(sequence, 4);

// Do something that writes - it may throw

writer.Commit();
sequence.Flush();
...

In this case I do not plan to rely on any of the bytes written to the writer in case that the intermediate writing logic throws, but I would like to have guarantees that the writer does not leak. Does this make sense?

[AArnott]

Well, it certainly won't leak but it may not recycle arrays without adding a member that can recycle without trying to write again.