Compress public functions in json artifact, but still deploy uncompressed

[fcarreiro]

See https://aztecprotocol.slack.com/archives/C0183F0V42V/p1736508632678059?thread_ts=1735850380.203639&cid=C0183F0V42V .

Compressing is easy, uncompressing on every usage of this data is more complicated. In particular, don't forget to update the public bytecode diff util.

Reference for self: #8623

[fcarreiro]

Not as easy as I thought. There's no single place that loads an artifact as a TS structure and there are several other complications. How I think things work

• The artifact is loaded via a schema which does decode the base64 for the functions (both noir and public, there's no distinction)
• The functions' bytecode is kept as is and eventually passed to noir or the avm. Therefore, Noir's bytecode stays compressed, noir/bb decompress it. For the AVM, we don't want to add decompression logic to it, so that will not work.
• Even if it did, there's another complication, there are free functions to compute things like a contract class id, etc, that will use whatever bytecode is in the class. If we kept public bytecode compressed, the hash will be wrong (you actually want the hash of the uncompressed bytecode since that's what gets deployed)

@spalladino, do you think it would be possible to add decompression via the schema, only for public functions? Would this make sense to achieve what I want? Any pointers?

[spalladino]

Every JSON artifact goes through the loadContractArtifact function before being used in ts code. In there, you got a generateFunctionArtifact that is used to tweak each function in the artifact. You could check there if the fn is public, and deal with the bytecode as you see fit.