edsl 0.1.14__py3-none-any.whl → 0.1.40__py3-none-any.whl

edsl/questions/QuestionDict.py

@@ -0,0 +1,343 @@
+from __future__ import annotations
+from typing import Union, Optional, Dict, List, Any, Type
+from pydantic import BaseModel, Field, field_validator
+from jinja2 import Environment, FileSystemLoader, TemplateNotFound
+from pathlib import Path
+
+from edsl.questions.QuestionBase import QuestionBase
+from edsl.questions.descriptors import (
+    AnswerKeysDescriptor,
+    ValueTypesDescriptor,
+    ValueDescriptionsDescriptor,
+    QuestionTextDescriptor,
+)
+from edsl.questions.response_validator_abc import ResponseValidatorABC
+from edsl.exceptions.questions import QuestionCreationValidationError
+from edsl.questions.decorators import inject_exception
+
+
+class DictResponseValidator(ResponseValidatorABC):
+    required_params = ["answer_keys", "permissive"]
+
+    valid_examples = [
+        (
+            {
+                "answer": {
+                    "name": "Hot Chocolate",
+                    "num_ingredients": 5,
+                    "ingredients": ["milk", "cocoa", "sugar"]
+                }
+            },
+            {
+                "answer_keys": ["name", "num_ingredients", "ingredients"],
+                "value_types": ["str", "int", "list[str]"]
+            },
+        )
+    ]
+    invalid_examples = [
+        (
+            {"answer": {"name": 123}},  # Name should be a string
+            {"answer_keys": ["name"], "value_types": ["str"]},
+            "Key 'name' has value of type int, expected str",
+        ),
+        (
+            {"answer": {"ingredients": "milk"}},  # Should be a list
+            {"answer_keys": ["ingredients"], "value_types": ["list"]},
+            "Key 'ingredients' should be a list, got str",
+        )
+    ]
+
+
+class QuestionDict(QuestionBase):
+    question_type = "dict"
+    question_text: str = QuestionTextDescriptor()
+    answer_keys: List[str] = AnswerKeysDescriptor()
+    value_types: Optional[List[str]] = ValueTypesDescriptor()
+    value_descriptions: Optional[List[str]] = ValueDescriptionsDescriptor()
+
+    _response_model = None
+    response_validator_class = DictResponseValidator
+
+    def _get_default_answer(self) -> Dict[str, Any]:
+        """Get default answer based on types."""
+        answer = {}
+        if not self.value_types:
+            return {
+                "title": "Sample Recipe",
+                "ingredients": ["ingredient1", "ingredient2"],
+                "num_ingredients": 2,
+                "instructions": "Sample instructions"
+            }
+
+        for key, type_str in zip(self.answer_keys, self.value_types):
+            if type_str.startswith(('list[', 'list')):
+                if '[' in type_str:
+                    element_type = type_str[type_str.index('[') + 1:type_str.rindex(']')].lower()
+                    if element_type == 'str':
+                        answer[key] = ["sample_string"]
+                    elif element_type == 'int':
+                        answer[key] = [1]
+                    elif element_type == 'float':
+                        answer[key] = [1.0]
+                    else:
+                        answer[key] = []
+                else:
+                    answer[key] = []
+            else:
+                if type_str == 'str':
+                    answer[key] = "sample_string"
+                elif type_str == 'int':
+                    answer[key] = 1
+                elif type_str == 'float':
+                    answer[key] = 1.0
+                else:
+                    answer[key] = None
+                    
+        return answer
+
+    def create_response_model(
+        self,
+    ) -> Type[BaseModel]:
+        """Create a response model for dict questions."""
+        default_answer = self._get_default_answer()
+        
+        class DictResponse(BaseModel):
+            answer: Dict[str, Any] = Field(
+                default_factory=lambda: default_answer.copy()
+            )
+            comment: Optional[str] = None
+            
+            @field_validator("answer")
+            def validate_answer(cls, v, values, **kwargs):
+                # Ensure all keys exist
+                missing_keys = set(self.answer_keys) - set(v.keys())
+                if missing_keys:
+                    raise ValueError(f"Missing required keys: {missing_keys}")
+                    
+                # Validate value types if not permissive
+                if not self.permissive and self.value_types:
+                    for key, type_str in zip(self.answer_keys, self.value_types):
+                        if key not in v:
+                            continue
+                            
+                        value = v[key]
+                        type_str = type_str.lower()  # Normalize to lowercase
+                        
+                        # Handle list types
+                        if type_str.startswith(('list[', 'list')):
+                            if not isinstance(value, list):
+                                raise ValueError(f"Key '{key}' should be a list, got {type(value).__name__}")
+                                
+                            # If it's a parameterized list, check element types
+                            if '[' in type_str:
+                                element_type = type_str[type_str.index('[') + 1:type_str.rindex(']')]
+                                element_type = element_type.lower().strip()
+                                
+                                for i, elem in enumerate(value):
+                                    expected_type = {
+                                        'str': str,
+                                        'int': int,
+                                        'float': float,
+                                        'list': list
+                                    }.get(element_type)
+                                    
+                                    if expected_type and not isinstance(elem, expected_type):
+                                        raise ValueError(
+                                            f"List element at index {i} for key '{key}' "
+                                            f"has type {type(elem).__name__}, expected {element_type}"
+                                        )
+                        else:
+                            # Handle basic types
+                            expected_type = {
+                                'str': str,
+                                'int': int,
+                                'float': float,
+                                'list': list,
+                            }.get(type_str)
+                            
+                            if expected_type and not isinstance(value, expected_type):
+                                raise ValueError(
+                                    f"Key '{key}' has value of type {type(value).__name__}, expected {type_str}"
+                                )
+                return v
+
+            model_config = {
+                "json_schema_extra": {
+                    "examples": [{
+                        "answer": default_answer,
+                        "comment": None
+                    }]
+                }
+            }
+
+        DictResponse.__name__ = "DictResponse"
+        return DictResponse
+
+    def __init__(
+        self,
+        question_name: str,
+        question_text: str,
+        answer_keys: List[str],
+        value_types: Optional[List[Union[str, type]]] = None,
+        value_descriptions: Optional[List[str]] = None,
+        include_comment: bool = True,
+        question_presentation: Optional[str] = None,
+        answering_instructions: Optional[str] = None,
+        permissive: bool = False,
+    ):
+        self.question_name = question_name
+        self.question_text = question_text
+        self.answer_keys = answer_keys
+        self.value_types = self._normalize_value_types(value_types)
+        self.value_descriptions = value_descriptions
+        self.include_comment = include_comment
+        self.question_presentation = question_presentation or self._render_template(
+            "question_presentation.jinja"
+        )
+        self.answering_instructions = answering_instructions or self._render_template(
+            "answering_instructions.jinja"
+        )
+        self.permissive = permissive
+
+        # Validation
+        if self.value_types and len(self.value_types) != len(self.answer_keys):
+            raise QuestionCreationValidationError(
+                "Length of value_types must match length of answer_keys."
+            )
+        if self.value_descriptions and len(self.value_descriptions) != len(self.answer_keys):
+            raise QuestionCreationValidationError(
+                "Length of value_descriptions must match length of answer_keys."
+            )
+
+    @staticmethod
+    def _normalize_value_types(value_types: Optional[List[Union[str, type]]]) -> Optional[List[str]]:
+        """Convert all value_types to string representations, including type hints."""
+        if not value_types:
+            return None
+
+        def normalize_type(t) -> str:
+            # Handle string representations of List
+            t_str = str(t)
+            if t_str == 'List':
+                return 'list'
+                
+            # Handle string inputs
+            if isinstance(t, str):
+                t = t.lower()
+                # Handle list types
+                if t.startswith(('list[', 'list')):
+                    if '[' in t:
+                        # Normalize the inner type
+                        inner_type = t[t.index('[') + 1:t.rindex(']')].strip().lower()
+                        return f"list[{inner_type}]"
+                    return "list"
+                return t
+
+            # Handle List the same as list
+            if t_str == "<class 'List'>":
+                return "list"
+
+            # If it's list type
+            if t is list:
+                return "list"
+
+            # If it's a basic type
+            if hasattr(t, "__name__"):
+                return t.__name__.lower()
+            
+            # If it's a typing.List
+            if t_str.startswith(('list[', 'list')):
+                return t_str.replace('typing.', '').lower()
+
+            # Handle generic types
+            if hasattr(t, "__origin__"):
+                origin = t.__origin__.__name__.lower()
+                args = [
+                    arg.__name__.lower() if hasattr(arg, "__name__") else str(arg).lower()
+                    for arg in t.__args__
+                ]
+                return f"{origin}[{', '.join(args)}]"
+
+            raise QuestionCreationValidationError(
+                f"Invalid type in value_types: {t}. Must be a type or string."
+            )
+
+        normalized = []
+        for t in value_types:
+            try:
+                normalized.append(normalize_type(t))
+            except Exception as e:
+                raise QuestionCreationValidationError(f"Error normalizing type {t}: {str(e)}")
+        
+        return normalized
+
+    def _render_template(self, template_name: str) -> str:
+        """Render a template using Jinja."""
+        try:
+            template_dir = Path(__file__).parent / "templates" / "dict"
+            env = Environment(loader=FileSystemLoader(template_dir))
+            template = env.get_template(template_name)
+            return template.render(
+                question_name=self.question_name,
+                question_text=self.question_text,
+                answer_keys=self.answer_keys,
+                value_types=self.value_types,
+                value_descriptions=self.value_descriptions,
+                include_comment=self.include_comment,
+            )
+        except TemplateNotFound:
+            return f"Template {template_name} not found in {template_dir}."
+
+    def to_dict(self, add_edsl_version: bool = True) -> dict:
+        """Serialize to JSON-compatible dictionary."""
+        return {
+            "question_type": self.question_type,
+            "question_name": self.question_name,
+            "question_text": self.question_text,
+            "answer_keys": self.answer_keys,
+            "value_types": self.value_types or [],
+            "value_descriptions": self.value_descriptions or [],
+            "include_comment": self.include_comment,
+            "permissive": self.permissive,
+        }
+
+    @classmethod
+    def from_dict(cls, data: dict) -> 'QuestionDict':
+        """Recreate from a dictionary."""
+        return cls(
+            question_name=data["question_name"],
+            question_text=data["question_text"],
+            answer_keys=data["answer_keys"],
+            value_types=data.get("value_types"),
+            value_descriptions=data.get("value_descriptions"),
+            include_comment=data.get("include_comment", True),
+            permissive=data.get("permissive", False),
+        )
+
+    @classmethod
+    @inject_exception
+    def example(cls) -> 'QuestionDict':
+        """Return an example question."""
+        return cls(
+            question_name="example",
+            question_text="Please provide a simple recipe for hot chocolate.",
+            answer_keys=["title", "ingredients", "num_ingredients", "instructions"],
+            value_types=["str", "list[str]", "int", "str"],
+            value_descriptions=[
+                "The title of the recipe.",
+                "A list of ingredients.",
+                "The number of ingredients.",
+                "The instructions for making the recipe."
+            ],
+        )
+
+    def _simulate_answer(self) -> dict:
+            """Simulate an answer for the question."""
+            return {
+                "answer": self._get_default_answer(),
+                "comment": None
+            }
+
+if __name__ == "__main__":
+    q = QuestionDict.example()
+    print(q.to_dict())

edsl/questions/QuestionExtract.py

@@ -1,71 +1,152 @@
 from __future__ import annotations
-import re
 import json
-from typing import Any
-from edsl.questions import Question
+import re
+
+from typing import Any, Optional, Dict
+from edsl.questions.QuestionBase import QuestionBase
 from edsl.questions.descriptors import AnswerTemplateDescriptor
-from edsl.scenarios import Scenario
-from edsl.utilities import random_string
 
+from edsl.questions.response_validator_abc import ResponseValidatorABC
+from edsl.questions.data_structures import BaseResponse
+from edsl.questions.decorators import inject_exception
+
+from typing import Dict, Any
+from pydantic import create_model, Field
+
+
+def extract_json(text, expected_keys, verbose=False):
+    # Escape special regex characters in keys
+    escaped_keys = [re.escape(key) for key in expected_keys]
+
+    # Create a pattern that looks for all expected keys
+    pattern = r"\{[^}]*" + r"[^}]*".join(escaped_keys) + r"[^}]*\}"
+
+    json_match = re.search(pattern, text)
+
+    if json_match:
+        json_str = json_match.group(0)
+        try:
+            # Parse the extracted string as JSON
+            json_data = json.loads(json_str)
+
+            # Verify that all expected keys are present
+            if all(key in json_data for key in expected_keys):
+                return json_data
+            else:
+                if verbose:
+                    print(
+                        "Error: Not all expected keys were found in the extracted JSON."
+                    )
+                return None
+        except json.JSONDecodeError:
+            if verbose:
+                print("Error: The extracted content is not valid JSON.")
+            return None
+    else:
+        if verbose:
+            print("Error: No JSON-like structure found with all expected keys.")
+        return None
+
+
+def dict_to_pydantic_model(input_dict: Dict[str, Any]) -> Any:
+    field_definitions = {
+        key: (type(value), Field(default=value)) for key, value in input_dict.items()
+    }
+
+    DynamicModel = create_model("DynamicModel", **field_definitions)
+
+    class AnswerModel(BaseResponse):
+        answer: "DynamicModel"
+        generated_tokens: Optional[str] = None
+        comment: Optional[str] = None
 
-class QuestionExtract(Question):
-    """
-    This question asks the user to extract values from a string, and return them in a given template.
+    return AnswerModel
 
-    Arguments:
-    - `question_name` is the name of the question (string)
-    - `question_text` is the text of the question (string)
-    - `answer_template` is the template for the answer (dictionary mapping strings to strings)
 
-    Optional arguments:
-    - `instructions` are the instructions for the question (string). If not provided, the default instructions are used. To view them, run `QuestionExtract.default_instructions`
+class ExtractResponseValidator(ResponseValidatorABC):
+    required_params = ["answer_template"]
+    valid_examples = [({"answer": "This is great"}, {})]
+    invalid_examples = [
+        (
+            {"answer": None},
+            {"answer_template": {"name": "John Doe", "profession": "Carpenter"}},
+            "Result cannot be empty",
+        ),
+    ]
 
-    For an example, run `QuestionExtract.example()`
-    """
+    def custom_validate(self, response) -> BaseResponse:
+        return response.dict()
+
+    def fix(self, response, verbose=False):
+        raw_tokens = response["generated_tokens"]
+        if verbose:
+            print(f"Invalid response of QuestionExtract was: {raw_tokens}")
+        extracted_json = extract_json(raw_tokens, self.answer_template.keys(), verbose)
+        if verbose:
+            print("Proposed solution is: ", extracted_json)
+        return {
+            "answer": extracted_json,
+            "comment": response.get("comment", None),
+            "generated_tokens": raw_tokens,
+        }
+
+
+class QuestionExtract(QuestionBase):
+    """This question prompts the agent to extract information from a string and return it in a given template."""
 
     question_type = "extract"
     answer_template: dict[str, Any] = AnswerTemplateDescriptor()
+    _response_model = None
+    response_validator_class = ExtractResponseValidator
 
     def __init__(
         self,
         question_text: str,
         answer_template: dict[str, Any],
         question_name: str,
+        answering_instructions: str = None,
+        question_presentation: str = None,
     ):
+        """Initialize the question.
+
+        :param question_name: The name of the question.
+        :param question_text: The text of the question.
+        :param answer_template: The template for the answer.
+        :param answering_instructions: Instructions for answering the question.
+        :param question_presentation: The presentation of the question.
+        """
         self.question_name = question_name
         self.question_text = question_text
         self.answer_template = answer_template
+        self.answering_instructions = answering_instructions
+        self.question_presentation = question_presentation
 
-    ################
-    # Answer methods
-    ################
-    def validate_answer(self, answer: Any) -> dict[str, Any]:
-        """Validates the answer"""
-        # raw_json = answer["answer"]
-        # fixed_json_data = re.sub(r"\'", '"', raw_json)
-        # answer["answer"] = json.loads(fixed_json_data)
-        self.validate_answer_template_basic(answer)
-        # self.validate_answer_key_value(answer, "answer", dict)
-
-        self.validate_answer_extract(answer)
-        return answer
-
-    def translate_answer_code_to_answer(self, answer, scenario: Scenario = None):
-        """Returns the answer in a human-readable format"""
-        return answer
-
-    def simulate_answer(self, human_readable: bool = True) -> dict[str, str]:
-        """Simulates a valid answer for debugging purposes"""
-        return {
-            "answer": {key: random_string() for key in self.answer_template.keys()},
-            "comment": random_string(),
-        }
+    def create_response_model(self):
+        return dict_to_pydantic_model(self.answer_template)
+
+    @property
+    def question_html_content(self) -> str:
+        from jinja2 import Template
+
+        question_html_content = Template(
+            """
+        {% for field, placeholder in answer_template.items() %}
+        <div>
+        <label for="{{ field }}">{{ field }}</label>
+        <input type="text" id="{{ field }}" name="{{ question_name }}[{{ field }}]" placeholder="{{ placeholder }}">
+        </div>
+        {% endfor %}
+        """
+        ).render(
+            question_name=self.question_name,
+            answer_template=self.answer_template,
+        )
+        return question_html_content
 
-    ################
-    # Helpful methods
-    ################
     @classmethod
+    @inject_exception
     def example(cls) -> QuestionExtract:
+        """Return an example question."""
         return cls(
             question_name="extract_name",
             question_text="My name is Moby Dick. I have a PhD in astrology, but I'm actually a truck driver",
@@ -73,22 +154,27 @@ class QuestionExtract(Question):
         )
 
 
-# main
 def main():
+    """Administer a question and validate the answer."""
     from edsl.questions.QuestionExtract import QuestionExtract
 
     q = QuestionExtract.example()
     q.question_text
     q.question_name
     q.answer_template
-    q.validate_answer({"answer": {"name": "Moby", "profession": "truck driver"}})
-    q.translate_answer_code_to_answer(
+    q._validate_answer({"answer": {"name": "Moby", "profession": "truck driver"}})
+    q._translate_answer_code_to_answer(
         {"answer": {"name": "Moby", "profession": "truck driver"}}
     )
-    # simulate answer
-    q.simulate_answer()
-    q.simulate_answer(human_readable=False)
-    q.validate_answer(q.simulate_answer(human_readable=False))
+    q._simulate_answer()
+    q._simulate_answer(human_readable=False)
+    q._validate_answer(q._simulate_answer(human_readable=False))
     # serialization (inherits from Question)
     q.to_dict()
     assert q.from_dict(q.to_dict()) == q
+
+
+if __name__ == "__main__":
+    import doctest
+
+    doctest.testmod(optionflags=doctest.ELLIPSIS)