Tag values always decoded as immutable

[oyvindronningstad]

Things to check first

• I have searched the existing issues and didn't find my bug already reported there

• I have checked that my bug is still present in the latest release

cbor2 version

6.0.1

Python version

3.14.4

What happened?

When decoding tagged maps, they are always returned as frozendict, regardless of the value of the immutable flag.

I assume this is a bug? Maybe related to #295?

How can we reproduce the bug?

>>> from cbor2 import loads
>>> print(loads(bytes.fromhex("CAA10000")))
CBORTag(10, frozendict({0: 0}))
>>> print(loads(bytes.fromhex("A10000")))
{0: 0}

[agronholm]

No, this one was a conscious decision. What problems is this causing for you?

[oyvindronningstad]

@agronholm Personally, I'm having 3 different issues:

1. I'm traversing the decoded structure and detecting maps via their type (dict). This is quite trivial to fix.
2. I'm decoding data, changing it, and reencoding, and frozendicts cannot be modified.
3. I'm encoding the decoding result (or a subset of it) to JSON and the json library doesn't accept frozendict (at least not cbor2.frozendict)

The last two are fixable by taking all results of cbor2 decoding and passing it through a recursive function that rebuilds it with dicts and lists instead of frozendicts and tuples.

Can I ask why this was done? I cannot really see any reason why it shouldn't be controllable. My thinking is that, in principle, there's no difference between cbor data by itself, and cbor data wrapped in a bstr. A protocol could easily include a structure that is sent bare in certain cases, and wrapped in a bstr in others (as part of a bigger structure), and handling these will need to be different with the current change. (unless you resolve to use only frozendicts).

[agronholm]

I truly wish you had tried the RC version. It was out for a month but hardly anyone seemed to try it, and the day I shipped 6.0.0 final I get a bunch of issues reported. Now I can't switch the default without breaking backwards compatibility before 7.0.

The reasoning was that immutable structures can be more efficient and tags tend to just contain static data not intended to be mutable anyways.

[oyvindronningstad]

Can you add a separate immutable flag for tags and bstrs that is default True? It's quite restrictive that it's not controllable in any way, and if there are significant performance reasons, maybe the flag should be separate anyways? Just an idea

[agronholm]

Just clarifying, what does "bstrs" mean?

[oyvindronningstad]

@agronholm Sorry, my mistake, my head was thinking about CBOR-encoded byte strings in another context, so I brought that in here for a moment. My points should still stand though.