langchain-ai/langgraph
1
0
Fork 0
mirror of https://github.com/langchain-ai/langgraph.git synced 2026-07-15 12:35:43 -04:00
Code Issues 314 Packages Projects Releases 17 Wiki Activity

[PR #6886] fix: support Pydantic default values with reducer functions #5379

New Issue
Open
opened 2026-02-20 17:51:37 -05:00 by yindo · 0 comments
yindo commented 2026-02-20 17:51:37 -05:00
Owner
Copy Link
Copy Source

Original Pull Request: https://github.com/langchain-ai/langgraph/pull/6886

State: open
Merged: No

Summary

Fixes #5225

When a Pydantic BaseModel state field has both a default value (Field(default=...) or Field(default_factory=...)) and a reducer function (Annotated[type, reducer]), the default value was being ignored. The channel was always initialized with the type's zero-argument constructor (e.g., [] for list, 0 for int).

Root Cause

BinaryOperatorAggregate.__init__ always initialized its value by calling typ() (the bare type constructor). It had no mechanism to receive or use schema-defined default values. While _get_channels and _get_channel had partial plumbing for passing the schema parameter, _is_field_binop didn't use it and BinaryOperatorAggregate didn't accept a default parameter.

Changes

  • _fields.py: Added Pydantic BaseModel support to get_field_default() to extract Field(default=...) and Field(default_factory=...) values via model_fields
  • binop.py: Added default parameter to BinaryOperatorAggregate.__init__(). When provided, uses it instead of typ(). Uses copy.deepcopy to prevent mutable default sharing. Preserves defaults through copy() and from_checkpoint()
  • state.py: Connected the schema context through _get_channels → _get_channel → _is_field_binop so field defaults are resolved and passed to BinaryOperatorAggregate

Example

from pydantic import BaseModel, Field
from typing import Annotated
import operator
from langgraph.graph import StateGraph, START, END

class State(BaseModel):
    messages: Annotated[list[str], operator.add] = Field(
        default_factory=lambda: ["welcome"]
    )
    count: Annotated[int, operator.add] = Field(default=100)

def node(state):
    return {"messages": ["hello"], "count": 5}

graph = StateGraph(State)
graph.add_node("process", node)
graph.add_edge(START, "process")
graph.add_edge("process", END)

result = graph.compile().invoke({})
# result["messages"] == ["welcome", "hello"]  ✅
# result["count"] == 105                       ✅

Test plan

  • Unit test for BinaryOperatorAggregate with default values (test_binop_with_default)
  • Integration test: default_factory with custom reducer (exact issue scenario)
  • Integration test: static Field(default=...) with operator.add
  • Integration test: default + user-provided initial input
  • Integration test: mutable default safety across multiple invocations
  • Integration test: backward compat - no default still works
  • Integration test: mixed fields with/without defaults
  • All existing tests pass (33/33 in test_channels, test_pydantic, test_state)
**Original Pull Request:** https://github.com/langchain-ai/langgraph/pull/6886 **State:** open **Merged:** No --- ## Summary Fixes #5225 When a Pydantic `BaseModel` state field has both a default value (`Field(default=...)` or `Field(default_factory=...)`) and a reducer function (`Annotated[type, reducer]`), the default value was being ignored. The channel was always initialized with the type's zero-argument constructor (e.g., `[]` for `list`, `0` for `int`). ### Root Cause `BinaryOperatorAggregate.__init__` always initialized its value by calling `typ()` (the bare type constructor). It had no mechanism to receive or use schema-defined default values. While `_get_channels` and `_get_channel` had partial plumbing for passing the `schema` parameter, `_is_field_binop` didn't use it and `BinaryOperatorAggregate` didn't accept a `default` parameter. ### Changes - **`_fields.py`**: Added Pydantic `BaseModel` support to `get_field_default()` to extract `Field(default=...)` and `Field(default_factory=...)` values via `model_fields` - **`binop.py`**: Added `default` parameter to `BinaryOperatorAggregate.__init__()`. When provided, uses it instead of `typ()`. Uses `copy.deepcopy` to prevent mutable default sharing. Preserves defaults through `copy()` and `from_checkpoint()` - **`state.py`**: Connected the schema context through `_get_channels` → `_get_channel` → `_is_field_binop` so field defaults are resolved and passed to `BinaryOperatorAggregate` ### Example ```python from pydantic import BaseModel, Field from typing import Annotated import operator from langgraph.graph import StateGraph, START, END class State(BaseModel): messages: Annotated[list[str], operator.add] = Field( default_factory=lambda: ["welcome"] ) count: Annotated[int, operator.add] = Field(default=100) def node(state): return {"messages": ["hello"], "count": 5} graph = StateGraph(State) graph.add_node("process", node) graph.add_edge(START, "process") graph.add_edge("process", END) result = graph.compile().invoke({}) # result["messages"] == ["welcome", "hello"] ✅ # result["count"] == 105 ✅ ``` ## Test plan - [x] Unit test for `BinaryOperatorAggregate` with default values (`test_binop_with_default`) - [x] Integration test: `default_factory` with custom reducer (exact issue scenario) - [x] Integration test: static `Field(default=...)` with `operator.add` - [x] Integration test: default + user-provided initial input - [x] Integration test: mutable default safety across multiple invocations - [x] Integration test: backward compat - no default still works - [x] Integration test: mixed fields with/without defaults - [x] All existing tests pass (33/33 in test_channels, test_pydantic, test_state)
yindo added the pull-request label 2026-02-20 17:51:37 -05:00
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
agents
awaiting author revision
bug
change
chat models
checkpointer
ci
code quality
code quality
dependencies
documentation
drawing
duplicate
enhancement
good first issue
help wanted
invalid
maintainer
meta
missing information
not maintained by langchain
open-swe-auto
open-swe-max
open-swe-max-auto
pending
performance
pull-request
Mirrored from GitHub Pull Request
 question
requires design
sdk
server
stale
studio
tests
type checking
usage question
v0.5
v1.1
No Label pull-request
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
yindo
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: langchain-ai/langgraph#5379
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?