search

tuffy / bitstream-io

Library for reading/writing un-aligned values from/to streams in big-endian and little-endian formats
Apache License 2.0
52 stars 20 forks source link

`FromByteStream` depends on caller-selected endianness #18

Closed sdroege closed 10 months ago

sdroege commented 10 months ago

My situation here is that I have a FromByteStream implementation where the endianness of the integers that are going to be read is fixed and not something the caller should be able to decide. However when creating a ByteReader, exactly that is happening.

Currently I work around this by using u32::from_be_bytes(r.read::<[u8; 4]>()?) but that is of course not as nice as r.read::<u32>()?.

Do you have any suggestions how to handle this? Would it maybe make sense to add something like an explicit r.read_le::<u32>() / r.read_be::<u32>() / r.read_ne::<u32>() etc for this situation?

tuffy commented 10 months ago

I've pushed out a fresh version with a handy new read_as method in the ByteRead trait which takes a second endianness generic argument, overriding whatever the reader was using previously for that specific read.

Hopefully that'll work a bit better.

sdroege commented 10 months ago

Thanks, that solves it and is a bit more consistent with the existing APIs than my suggestion.

sdroege commented 10 months ago

@tuffy I saw you only added that for the byte read-side of things. Can you add the same to the write trait, and the two bit-based traits? Thanks again :)

tuffy commented 10 months ago

This is now fixed in version 2.2.0, least for reading/writing whole values. Reading/writing values with a partial number of bytes in a different endianness could also be added, but its output is nonintuitive and I'm unaware of any actual file formats that work that way.

sdroege commented 10 months ago

I agree. Thanks again :)