python 3.8 compatibility issue

[idc9]

Thanks for making tooldantic!

The current version of the package does not work for python 3.8 because of an import from typing import Annotated. See below for some examples.

One potential solution (chatgpt suggested) is

from typing_extensions import Annotated

Example code

from tooldantic import ToolBaseModel as BaseModel

---------------------------------------------------------------------------
ImportError                               Traceback (most recent call last)
Cell In[4], line 2
      1 # drop in replacement for pydantic.BaseModel
----> 2 from tooldantic import ToolBaseModel as BaseModel
      5 class InnerModel(BaseModel):
      6     """nested inner model"""

File ~/anaconda3/envs/cpath/lib/python3.8/site-packages/tooldantic/__init__.py:3
      1 from pydantic import Field, ValidationError
----> 3 from .builder import ModelBuilder
      4 from .decorators import AsyncToolWrapper, ToolWrapper, ToolWrapperBase
      5 from .dispatch import ToolDispatch

File ~/anaconda3/envs/cpath/lib/python3.8/site-packages/tooldantic/builder.py:4
      2 import inspect
      3 import logging
----> 4 from typing import (
      5     Annotated,
      6     Any,
      7     Callable,
      8     Dict,
      9     List,
     10     Literal,
     11     Optional,
     12     Tuple,
     13     Type,
     14     Union,
     15     get_args,
     16     get_origin,
     17 )
     19 import docstring_parser
     20 import pydantic

ImportError: cannot import name 'Annotated' from 'typing' (/Users/iain/anaconda3/envs/cpath/lib/python3.8/typing.py)

or for a more simple demonstration

from typing import Annotated

ImportError                               Traceback (most recent call last)
Cell In[5], line 1
----> 1 from typing import Annotated

ImportError: cannot import name 'Annotated' from 'typing' (/Users/iain/anaconda3/envs/cpath/lib/python3.8/typing.py)

[idc9]

here are some other python 3.8 compatibility issues. Looks like you can fix these by just changing the type hints e.g. tuple -> Tuple from the typing package etc.

File ~/Library/CloudStorage/Dropbox/Research/local_packages/third_party/tooldantic-main/tooldantic/schema_generators.py:123, in CompatibilitySchemaGenerator()
    117         return [self._inline_all_of(item) for item in schema]
    118     return schema
    120 def _ensure_strict_json_schema(
    121     self,
    122     json_schema: object,
--> 123     path: tuple[str, ...] = (),
    124 ) -> dict[str, Any]:
    125     """Mutates the given JSON schema to ensure it conforms to the `strict` standard
    126     that the API expects.
    127     """
    128     if not isinstance(json_schema, dict):

TypeError: 'type' object is not subscriptable

File ~/Library/CloudStorage/Dropbox/Research/local_packages/third_party/tooldantic-main/tooldantic/schema_generators.py:123, in CompatibilitySchemaGenerator()
    117         return [self._inline_all_of(item) for item in schema]
    118     return schema
    120 def _ensure_strict_json_schema(
    121     self,
    122     json_schema: object,
--> 123     path: tuple[str, ...] = (),
    124 ) -> Dict[str, Any]:
    125     """Mutates the given JSON schema to ensure it conforms to the `strict` standard
    126     that the API expects.
    127     """
    128     if not isinstance(json_schema, dict):

TypeError: 'type' object is not subscriptable

File ~/Library/CloudStorage/Dropbox/Research/local_packages/third_party/tooldantic-main/tooldantic/dispatch.py:40, in ToolDispatch()
      9 class ToolDispatch:
     10     """
     11     A class representing a tool dispatcher.
     12     The ToolDispatch class is responsible for managing a collection of tool functions
   (...)
     34         values(): Returns an iterator over the values of the tool functions.
     35     """
     37     def __init__(
     38         self,
     39         *funcs: Union[ToolWrapperBase, Callable],
---> 40         base_model: type[ToolBaseModel] | None = None,
     41         **pydantic_kwargs
     42     ):
     43         if funcs and isinstance(funcs[0], Iterable):
     44             funcs = list(funcs[0])

TypeError: 'type' object is not subscriptable

[nicholishen]

Thanks for bringing this up and for providing the detailed examples!

The current setup was initially intended to support Python 3.9 and above, which led to these compatibility issues with 3.8. I appreciate your suggestion about typing_extensions for Annotated, as well as the other adjustments for type hints.

Would a 3.8-compatible version be helpful for your work? If there's enough demand, I'm open to working on compatibility changes or collaborating on a 3.8-compatible branch if you're interested in contributing.

Thanks again for highlighting this!