feat: Add to_map constructor function #6885

Is your feature request related to a problem?

Daft supports the Map datatype and provides map_get for accessing map values, but there is no way to manually construct a map column from expressions. We already have to_list and to_struct constructor functions that allow users to build list and struct columns from expressions, but the equivalent to_map is missing.

This means users who want to create map columns must either:

• Read them from an external source (e.g. Parquet, Iceberg)
• Use workarounds with structs/lists and casts

Describe the solution you'd like

Add a to_map function (exposed via daft.functions.to_map) that constructs a map column from key-value expression pairs.

Proposed API:

from daft.functions import to_map

df = daft.from_pydict({"k1": ["a", "b"], "v1": [1, 2], "k2": ["c", "d"], "v2": [3, 4]})

# Option A: alternating key/value expressions
df.select(to_map(df["k1"], df["v1"], df["k2"], df["v2"]).alias("my_map"))

# Option B: keyword-style (keys as strings, values as expressions)  
df.select(to_map(a=df["v1"], b=df["v2"]).alias("my_map"))

This should follow the same patterns established by to_list (daft/functions/list.py) and to_struct (daft/functions/struct.py).

Describe alternatives you've considered

• Constructing a list of structs and casting to map — verbose and error-prone
• Using from_pydict with pre-built Python dicts — doesn't work for expression-level construction within a query plan

Component(s)

Built-in Functions

Additional Context

• Existing map support: MapType dtype, map_get accessor
• Reference implementations: to_list in daft/functions/list.py, to_struct in daft/functions/struct.py

Would you like to implement a fix?

Yes