ignore trailing .0 in version comparison

[wolfv]

Not sure if this is the cleanest way of implementing this behavior, but in order to fix #194 we need to ignore trailing 0 when comparing versions for equality.

[baszalmstra]

Can this code also be found somewhere in the conda codebase? I followed that pretty closely.

[baszalmstra]

Can this code also be found somewhere in the conda codebase? I followed that pretty closely.

[wolfv]

I can try to find the place in the conda code but this was an issue i observed while debugging resolvelib stuff

[wolfv]

We actually did copy the docs for it! :) https://github.com/mamba-org/rattler/blob/001f326c7c3160637c8778c5a1c33f715266434e/crates/rattler_conda_types/src/version/mod.rs#L71-L72

[wolfv]

And this is the code in conda:

    def _eq(self, t1, t2):
        for v1, v2 in zip_longest(t1, t2, fillvalue=[]):
            for c1, c2 in zip_longest(v1, v2, fillvalue=self.fillvalue):
                if c1 != c2:
                    return False
        return True

https://github.com/conda/conda/blob/61edfac743a2461a24fafe48b821a99af61b6bd4/conda/models/version.py#L244-L249

[wolfv]

Btw. I think we're also missing an compare of the epoch of the two versions.

[baszalmstra]

Nice!

• I thinkthe epoch is always included in the components even if not specified. So it is also taken into account.

• if you can implement the comparison as part of the PartialEq of VersionComponent this should be good. This comparison holds for both the version amd the local.

[wolfv]

I am confused by the follownig:

    /// The epoch of this version. This is an optional number that can be used to indicate a change
    /// in the versioning scheme.
    norm: String,

The norm doesn't seem to be the epoch but rather the string representation of the version?

[wolfv]

E.g. this is what I see in debug print:

Version { norm: "1.2.0", version: [[0], [1], [2], [0]], local: [] }
Version { norm: "1!1.2.3", version: [[1], [1], [2], [3]], local: [] }

[wolfv]

I attempted what you asked but there I am a bit unsure about a couple things.

I had to add #[derive(Eq)] to the Version for some reason.

Also I had to implement a custom hash function. I am now removing all default values (0) from the hash. Clippy was warning that the hash function needs to match the PartialEq function.

[baszalmstra]

This looks good to me! The warning about the hash is also very correct! Very nice!

The norm is a bit of an odd duck. It represents "sorta" the string it was initially created from. This can be important because (as you've seen with this issue) multiple version representations can represent the exact same version. The norm sorta enables keeping track of that. However, modifications to the Version are also not expressed in the norm which is actually just plain wrong. Maybe we should remove it and implement a proper Display for Version.