A way to construct MzMLReader without buffering?

[TheLostLambda]

It seems like there is always a BufReader constructed, regardless of if the source is already in memory or not! In my case, I just have a byte vector, so the BufReader is unneeded! https://docs.rs/mzdata/0.60.3/src/mzdata/io/mzml/reader.rs.html#1228

Would you like me to try to implement some other way of constructing unbuffered readers?

[mobiusklein]

Double-buffering isn't ideal and while there might be a way to do it without breaking changes with generic aliasing, I'm not certain. Is this scenario something you're dealing with to load data faster, or are you generating mzML documents in-memory?

What happens if you use MzMLReader::with_buffer_capacity_and_detail_level(src, 0, DetailLevel::Full)?

[TheLostLambda]

For the moment it's about flexibility and reading performance! I'm not sure about what would happen with a buffer size of 0, I can try to check that out, but I suppose it's not the ideal API either way.

One thing that could help here is applying the builder pattern to these other readers as well? I know the generic MZReader has a builder version, but it might be nice to have that for the more specialised ones too! In fact, this was something I was thinking about when I first wanted to set the DetailLevel of the MzMLReader, but found that there was (seemingly?) no way to do that without also needing to specify the buffer size!

Let me know what you think, and if you think it would be worth having builders for these readers too!

There could be one grand builder (a version of the MZReader) that could just take the type of file as an argument given via a builder method? You could do .format(MassSpectrometryFormat::MzML) before the build, or leave it off if you wanted it to figure it out automatically?

[TheLostLambda]

As an update for @mobiusklein , I've not had the time to get the bottom of it yet, but using a buffer capacity of 0 like you suggested in your previous comment leads the reader to silently fail — it just returns an empty set of spectra. It seems like BufReader behaves mostly normally with a capacity of 0 (https://play.rust-lang.org/?version=stable&mode=debug&edition=2024&gist=4b6bbd9d5891c881cc2260ecd4ebdfd1), but something about how it's being used in mzdata is causing an issue...