#1720 Carry over hash from literal maps in cache key computation Pull request opened by nicholasjng TL;DR Until now, the hash of a literal map is not carried over into the representative literal map entering the cache key computation, despite the code clearly stating so. This commit enables carrying hashes from `LiteralMap`s over by explicitly setting the hashes on the representer literal map. Also fixes a shadowing bug in the key-value loop over the literal map, which happened since the value argument and the input name were both

literal

. Type ☑︎ Bug Fix ☐ Feature ☐ Plugin Are all requirements met? ☑︎ Code completed ☐ Smoke tested ☑︎ Unit tests added ☐ Code documentation added ☑︎ Any pending items have an associated Issue (there aren't any I guess) Complete description This came up during internal tests, where it broke existing tests when I changed from a single

Literal

to a

LiteralMap

in a type transformer with an explicitly set hash value. Minimal reproducer:

# test.py
from flytekit import Literal
from flytekit.core.local_cache import _recursive_hash_placement
from flytekit.models.literals import LiteralMap, Scalar, Primitive

lit = Literal(
    map=LiteralMap(
        literals={
            "hello": Literal(scalar=Scalar(primitive=Primitive(string_value="hello"))),
        }
    ),
    hash="0xffff",
)

if __name__ == "__main__":
    print(lit)
# --> prints <FlyteLiteral map { literals { key: "hello" value { scalar { primitive { string_value: "hello" } } } } } hash: "0xffff">
    print(_recursive_hash_placement(lit))
# --> prints <FlyteLiteral map { literals { key: "hello" value { scalar { primitive { string_value: "hello" } } } } }>

This means that the

_recursive_hash_placement

function, which facilitates part of the hash key computation, silently scrubs hash values from input literal maps, which can cause involuntary cache hits if the values inside the map stay the same. I would be happy for a discussion - maybe the repro can go in as a unit test in a suitable location. Tracking Issue N/A Follow-up issue N/A flyteorg/flytekit ✅ All checks have passed 30/30 successful checks