Expand description
§Planned Breaking Changes
This page lists (incomplete) list of planned changes on the next major release (most likely version 0.4).
§Major design change (and prelude module)
The next major release will heavily utilize traits and for convenience, you
will want to import ssdeep::prelude::*.
This module is added in version 0.3.13.
§Removal of the feature opt-reduce-fnv-table
This feature is rather subtractive than recommended additive feature rules (although not completely against it) and the change is too subtle (memory footprint of 4KiB can be easily justified on the environment which ssdeep hashes are generated).
So, this feature is going to be removed on the next major release.
It may return in the future when fine-tuning is absolutely necessary in this crate (but unlikely; since considered optimal algorithms are far better than past alternatives).
§Removal of the feature: nightly (use unstable instead)
Currently, the unstable feature is preferred to use Nightly Rust features.
nightly is preserved as an alias of unstable (although being undocumented)
but will be removed on the next major release.
§Raising MSRV
Depending on the Rust edition target (we haven’t decided yet), MSRV on the next major release will be raised to the version specific to that edition.
- Edition 2021 case: at least 1.79 (most likely 1.81)
Rust 1.79’s associated type bounds feature makes our new trait-based design far more easier. - Edition 2024 case: 1.85 (not released yet)
§Removal of += operator overload
Since it’s clear that this operator overload is not helpful enough, they are now deprecated and will be removed on the next major release.
§Going to be Private: Internal Hashes / Comparison Structures
They were useful to experiment with ssdeep internals but we could not justify keeping them public. If you need to use those, just vendor the source code for your needs (they are relatively simple and will be easy to vendor them).
For specific needs, a comparison target storing just one block hash (instead of two block hashes) will be public as a new type.
§Stabilization of RLE-based compression in dual fuzzy hashes
We found no issues on dual fuzzy hashes and recognized that separately storing “reverse normalization” data is helpful on some cases.
So, the next major release will stabilize currently internal RLE-based compression used by dual fuzzy hashes.
Also, we will attempt to reduce memory footprint and if (and only if) we are sure that this is performant enough (considering the use cases), the compression ratio will be improved from ≒5/8 to ≒1/2(=4/8) (compared to both normalized and raw hashes).