Also, I don't see much value in definitions like this:

emptySTRLESeq :: ST s (STRef s (Seq (Maybe b)))
emptySTRLESeq = newSTRef DS.empty

I would just replace uses of emptySTRLESeq with newSTRef DS.empty everywhere, and similarly for other functions. Furthermore, I don't think you need to encode the type in symbol names like brleseqstackempty - just call it stackRef, and rename brleseqstackemptyr just stack.

@ddssff do you think we should move away from implementing via the ST monad right (replace with Writer monad)? If so a lot of these functions probably wont be necessary anymore. I still wonder about performance though, doesn't the ST monad allow for scoped mutation (should allow for a Haskellish way to maintain near C performance)? Maybe the current ST monad implementation isn't any faster than say implementing via the Writer monad (I like the Writer monad way much better).

Honestly, I don't understand how the ST monad affects performance. I did the writer monad implementation by incrementally transforming the ST monad implementation, so first I had to do the work described in this ticket - parameterize the ByteString / Text type, inline trival functions and types, etc. Once that is done we can convert to Writer monad but keep the ST monad version. Then perhaps some benchmarking.

My guess is that high performance is something you will have to do afterwards. But regular pure haskell does a pretty good job on its own.

One more suggestion - use pattern matching and the maybe function instead of fromJust and isNothing.

Sorry for the delay, will be looking into this soon.

No problem, I've been very busy.