QuizGenerator 0.1.0__py3-none-any.whl

QuizGenerator/question.py

@@ -0,0 +1,657 @@
+#!env python
+from __future__ import annotations
+
+import abc
+import io
+import dataclasses
+import datetime
+import enum
+import importlib
+import itertools
+import os
+import pathlib
+import pkgutil
+import random
+import re
+import uuid
+
+import pypandoc
+import yaml
+from typing import List, Dict, Any, Tuple, Optional
+import canvasapi.course, canvasapi.quiz
+
+from QuizGenerator.misc import OutputFormat, Answer
+from QuizGenerator.contentast import ContentAST
+from QuizGenerator.performance import timer, PerformanceTracker
+
+import logging
+log = logging.getLogger(__name__)
+
+
+# Spacing presets for questions
+SPACING_PRESETS = {
+    "NONE": 0,
+    "SHORT": 4,
+    "MEDIUM": 6,
+    "LONG": 9,
+    "PAGE": 99,  # Special value that will be handled during bin-packing
+    "EXTRA_PAGE": 199,  # Special value that adds a full blank page after the question
+}
+
+
+def parse_spacing(spacing_value) -> float:
+    """
+    Parse spacing value from YAML config.
+
+    Args:
+        spacing_value: Either a preset name ("NONE", "SHORT", "LONG", "PAGE")
+                      or a numeric value in cm
+
+    Returns:
+        Spacing in cm as a float
+
+    Examples:
+        parse_spacing("SHORT") -> 5.0
+        parse_spacing("NONE") -> 1.0
+        parse_spacing(3.5) -> 3.5
+        parse_spacing("3.5") -> 3.5
+    """
+    if isinstance(spacing_value, str):
+        # Check if it's a preset
+        if spacing_value.upper() in SPACING_PRESETS:
+            return float(SPACING_PRESETS[spacing_value.upper()])
+        # Try to parse as a number
+        try:
+            return float(spacing_value)
+        except ValueError:
+            log.warning(f"Invalid spacing value '{spacing_value}', defaulting to 0")
+            return 0.0
+    elif isinstance(spacing_value, (int, float)):
+        return float(spacing_value)
+    else:
+        log.warning(f"Invalid spacing type {type(spacing_value)}, defaulting to 0")
+        return 0.0
+
+
+class QuestionRegistry:
+  _registry = {}
+  _scanned = False
+  
+  @classmethod
+  def register(cls, question_type=None):
+    def decorator(subclass):
+      # Use the provided name or fall back to the class name
+      name = question_type.lower() if question_type else subclass.__name__.lower()
+      cls._registry[name] = subclass
+      return subclass
+    return decorator
+    
+  @classmethod
+  def create(cls, question_type, **kwargs) -> Question:
+    """Instantiate a registered subclass."""
+    # If we haven't already loaded our premades, do so now
+    if not cls._scanned:
+      cls.load_premade_questions()
+
+    # Check to see if it's in the registry
+    question_key = question_type.lower()
+    if question_key not in cls._registry:
+      # Try stripping common course prefixes and module paths for backward compatibility
+      for prefix in ['cst334.', 'cst463.']:
+        if question_key.startswith(prefix):
+          stripped_name = question_key[len(prefix):]
+          if stripped_name in cls._registry:
+            question_key = stripped_name
+            break
+          # Also try extracting just the final class name after dots
+          if '.' in stripped_name:
+            final_name = stripped_name.split('.')[-1]
+            if final_name in cls._registry:
+              question_key = final_name
+              break
+      else:
+        # As a final fallback, try just the last part after dots
+        if '.' in question_key:
+          final_name = question_key.split('.')[-1]
+          if final_name in cls._registry:
+            question_key = final_name
+          else:
+            raise ValueError(f"Unknown question type: {question_type}")
+        else:
+          raise ValueError(f"Unknown question type: {question_type}")
+
+    new_question : Question = cls._registry[question_key](**kwargs)
+    # Note: Don't call refresh() here - it will be called by get_question()
+    # Calling it here would consume RNG calls and break QR code regeneration
+    return new_question
+    
+    
+  @classmethod
+  def load_premade_questions(cls):
+    package_name = "QuizGenerator.premade_questions"  # Fully qualified package name
+    package_path = pathlib.Path(__file__).parent / "premade_questions"
+
+    def load_modules_recursively(path, package_prefix):
+      # Load modules from the current directory
+      for _, module_name, _ in pkgutil.iter_modules([str(path)]):
+        # Import the module
+        module = importlib.import_module(f"{package_prefix}.{module_name}")
+
+      # Recursively load modules from subdirectories
+      for subdir in path.iterdir():
+        if subdir.is_dir() and not subdir.name.startswith('_'):
+          subpackage_name = f"{package_prefix}.{subdir.name}"
+          load_modules_recursively(subdir, subpackage_name)
+
+    load_modules_recursively(package_path, package_name)
+
+    # Load user-registered questions via entry points (Option 1: Robust PyPI approach)
+    # Users can register custom questions in their package's pyproject.toml:
+    # [project.entry-points."quizgenerator.questions"]
+    # my_custom_question = "my_package.questions:CustomQuestion"
+    try:
+      # Python 3.10+ approach
+      from importlib.metadata import entry_points
+      eps = entry_points()
+      # Handle both Python 3.10+ (dict-like) and 3.12+ (select method)
+      if hasattr(eps, 'select'):
+        question_eps = eps.select(group='quizgenerator.questions')
+      else:
+        question_eps = eps.get('quizgenerator.questions', [])
+
+      for ep in question_eps:
+        try:
+          # Loading the entry point will trigger @QuestionRegistry.register() decorator
+          ep.load()
+          log.debug(f"Loaded custom question type from entry point: {ep.name}")
+        except Exception as e:
+          log.warning(f"Failed to load entry point '{ep.name}': {e}")
+    except ImportError:
+      # Python < 3.10 fallback using pkg_resources
+      try:
+        import pkg_resources
+        for ep in pkg_resources.iter_entry_points('quizgenerator.questions'):
+          try:
+            ep.load()
+            log.debug(f"Loaded custom question type from entry point: {ep.name}")
+          except Exception as e:
+            log.warning(f"Failed to load entry point '{ep.name}': {e}")
+      except ImportError:
+        # If pkg_resources isn't available either, just skip entry points
+        log.debug("Entry points not supported (importlib.metadata and pkg_resources unavailable)")
+
+    cls._scanned = True
+
+
+class RegenerableChoiceMixin:
+  """
+  Mixin for questions that need to make random choices from enums/lists that are:
+  1. Different across multiple refreshes (when the same Question instance is reused for multiple PDFs)
+  2. Reproducible from QR code config_params
+
+  The Problem:
+  ------------
+  When generating multiple PDFs, Quiz.from_yaml() creates Question instances ONCE.
+  These instances are then refresh()ed multiple times with different RNG seeds.
+  If a question randomly selects an algorithm/policy in __init__(), all PDFs get the same choice
+  because __init__() only runs once with an unseeded RNG.
+
+  The Solution:
+  -------------
+  1. In __init__(): Register choices with fixed values (if provided) or None (for random)
+  2. In refresh(): Make random selections using the seeded RNG, store in config_params
+  3. Result: Each refresh gets a different random choice, and it's captured for QR codes
+
+  Usage Example:
+  --------------
+  class SchedulingQuestion(Question, RegenerableChoiceMixin):
+      class Kind(enum.Enum):
+          FIFO = enum.auto()
+          SJF = enum.auto()
+
+      def __init__(self, scheduler_kind=None, **kwargs):
+          # Register the choice BEFORE calling super().__init__()
+          self.register_choice('scheduler_kind', self.Kind, scheduler_kind, kwargs)
+          super().__init__(**kwargs)
+
+      def refresh(self, **kwargs):
+          super().refresh(**kwargs)
+          # Get the choice (randomly selected or from config_params)
+          self.scheduler_algorithm = self.get_choice('scheduler_kind', self.Kind)
+          # ... rest of refresh logic
+  """
+
+  def __init__(self, *args, **kwargs):
+    # Initialize the choices registry if it doesn't exist
+    if not hasattr(self, '_regenerable_choices'):
+      self._regenerable_choices = {}
+    super().__init__(*args, **kwargs)
+
+  def register_choice(self, param_name: str, enum_class: type[enum.Enum], fixed_value: str | None, kwargs_dict: dict):
+    """
+    Register a choice parameter that needs to be regenerable.
+
+    Args:
+        param_name: The parameter name (e.g., 'scheduler_kind', 'policy')
+        enum_class: The enum class to choose from (e.g., SchedulingQuestion.Kind)
+        fixed_value: The fixed value if provided, or None for random selection
+        kwargs_dict: The kwargs dictionary to update (for config_params capture)
+
+    This should be called in __init__() BEFORE super().__init__().
+    """
+    # Store the enum class for later use
+    if not hasattr(self, '_regenerable_choices'):
+      self._regenerable_choices = {}
+
+    self._regenerable_choices[param_name] = {
+      'enum_class': enum_class,
+      'fixed_value': fixed_value
+    }
+
+    # Add to kwargs so config_params captures it
+    if fixed_value is not None:
+      kwargs_dict[param_name] = fixed_value
+
+  def get_choice(self, param_name: str, enum_class: type[enum.Enum]) -> enum.Enum:
+    """
+    Get the choice for a registered parameter.
+    Should be called in refresh() AFTER super().refresh().
+
+    Args:
+        param_name: The parameter name registered earlier
+        enum_class: The enum class to choose from
+
+    Returns:
+        The selected enum value (either fixed or randomly chosen)
+    """
+    choice_info = self._regenerable_choices.get(param_name)
+    if choice_info is None:
+      raise ValueError(f"Choice '{param_name}' not registered. Call register_choice() in __init__() first.")
+
+    # Check for temporary fixed value (set during backoff loop in get_question())
+    fixed_value = choice_info.get('_temp_fixed_value', choice_info['fixed_value'])
+
+    # CRITICAL: Always consume an RNG call to keep RNG state synchronized between
+    # original generation and QR code regeneration. During original generation,
+    # we pick randomly. During regeneration, we already know the answer from
+    # config_params, but we still need to consume the RNG call.
+    enum_list = list(enum_class)
+    random_choice = self.rng.choice(enum_list)
+
+    if fixed_value is None:
+      # No fixed value - use the random choice we just picked
+      self.config_params[param_name] = random_choice.name
+      return random_choice
+    else:
+      # Fixed value provided - ignore the random choice, use the fixed value
+      # (but we still consumed the RNG call above to keep state synchronized)
+
+      # If already an enum instance, return it directly
+      if isinstance(fixed_value, enum_class):
+        return fixed_value
+
+      # If it's a string, look up the enum member by name
+      if isinstance(fixed_value, str):
+        try:
+          # Try exact match first (handles "RoundRobin", "FIFO", etc.)
+          return enum_class[fixed_value]
+        except KeyError:
+          # Try uppercase as fallback (handles "roundrobin" -> "ROUNDROBIN")
+          try:
+            return enum_class[fixed_value.upper()]
+          except KeyError:
+            log.warning(
+              f"Invalid {param_name} '{fixed_value}'. Valid options are: {[k.name for k in enum_class]}. Defaulting to random"
+            )
+            self.config_params[param_name] = random_choice.name
+            return random_choice
+
+      # Unexpected type
+      log.warning(
+        f"Invalid {param_name} type {type(fixed_value)}. Expected enum or string. Defaulting to random"
+      )
+      self.config_params[param_name] = random_choice.name
+      return random_choice
+
+
+class Question(abc.ABC):
+  """
+  Base class for all quiz questions with cross-format rendering support.
+
+  CRITICAL: When implementing Question subclasses, ALWAYS use ContentAST elements
+  for all content in get_body() and get_explanation() methods.
+
+  NEVER create manual LaTeX, HTML, or Markdown strings. The ContentAST system
+  ensures consistent rendering across PDF/LaTeX and Canvas/HTML formats.
+
+  Required Methods:
+    - get_body(): Return ContentAST.Section with question content
+    - get_explanation(): Return ContentAST.Section with solution steps
+
+  Required Class Attributes:
+    - VERSION (str): Question version number (e.g., "1.0")
+      Increment when RNG logic changes to ensure reproducibility
+
+  ContentAST Usage Examples:
+    def get_body(self):
+        body = ContentAST.Section()
+        body.add_element(ContentAST.Paragraph(["Calculate the matrix:"]))
+
+        # Use ContentAST.Matrix for math, NOT manual LaTeX
+        matrix_data = [[1, 2], [3, 4]]
+        body.add_element(ContentAST.Matrix(data=matrix_data, bracket_type="b"))
+
+        # Use ContentAST.Answer for input fields
+        body.add_element(ContentAST.Answer(answer=self.answers["result"]))
+        return body
+
+  Common ContentAST Elements:
+    - ContentAST.Paragraph: Text blocks
+    - ContentAST.Equation: Mathematical expressions
+    - ContentAST.Matrix: Matrices and vectors (use instead of manual LaTeX!)
+    - ContentAST.Table: Data tables
+    - ContentAST.OnlyHtml/OnlyLatex: Platform-specific content
+
+  Versioning Guidelines:
+    - Increment VERSION when changing:
+      * Order of random number generation calls
+      * Question generation logic
+      * Answer calculation methods
+    - Do NOT increment for:
+      * Cosmetic changes (formatting, wording)
+      * Bug fixes that don't affect answer generation
+      * Changes to get_explanation() only
+
+  See existing questions in premade_questions/ for patterns and examples.
+  """
+
+  # Default version - subclasses should override this
+  VERSION = "1.0"
+  
+  class Topic(enum.Enum):
+    # CST334 (Operating Systems) Topics
+    SYSTEM_MEMORY = enum.auto()      # Virtual memory, paging, segmentation, caching
+    SYSTEM_PROCESSES = enum.auto()   # Process management, scheduling
+    SYSTEM_CONCURRENCY = enum.auto() # Threads, synchronization, locks
+    SYSTEM_IO = enum.auto()          # File systems, persistence, I/O operations
+    SYSTEM_SECURITY = enum.auto()    # Access control, protection mechanisms
+
+    # CST463 (Machine Learning/Data Science) Topics
+    ML_OPTIMIZATION = enum.auto()    # Gradient descent, optimization algorithms
+    ML_LINEAR_ALGEBRA = enum.auto()  # Matrix operations, vector mathematics
+    ML_STATISTICS = enum.auto()      # Probability, distributions, statistical inference
+    ML_ALGORITHMS = enum.auto()      # Classification, regression, clustering
+    DATA_PREPROCESSING = enum.auto() # Data cleaning, transformation, feature engineering
+
+    # General/Shared Topics
+    MATH_GENERAL = enum.auto()       # Basic mathematics, calculus, algebra
+    PROGRAMMING = enum.auto()        # General programming concepts
+    LANGUAGES = enum.auto()          # Programming languages specifics
+    MISC = enum.auto()              # Uncategorized questions
+
+    # Legacy aliases for backward compatibility
+    PROCESS = SYSTEM_PROCESSES
+    MEMORY = SYSTEM_MEMORY
+    CONCURRENCY = SYSTEM_CONCURRENCY
+    IO = SYSTEM_IO
+    SECURITY = SYSTEM_SECURITY
+    MATH = MATH_GENERAL
+
+    @classmethod
+    def from_string(cls, string) -> Question.Topic:
+      mappings = {
+        member.name.lower() : member for member in cls
+      }
+      mappings.update({
+        # Legacy mappings
+        "processes": cls.SYSTEM_PROCESSES,
+        "process": cls.SYSTEM_PROCESSES,
+        "threads": cls.SYSTEM_CONCURRENCY,
+        "concurrency": cls.SYSTEM_CONCURRENCY,
+        "persistance": cls.SYSTEM_IO,
+        "persistence": cls.SYSTEM_IO,
+        "io": cls.SYSTEM_IO,
+        "memory": cls.SYSTEM_MEMORY,
+        "security": cls.SYSTEM_SECURITY,
+        "math": cls.MATH_GENERAL,
+        "mathematics": cls.MATH_GENERAL,
+
+        # New mappings
+        "optimization": cls.ML_OPTIMIZATION,
+        "gradient_descent": cls.ML_OPTIMIZATION,
+        "machine_learning": cls.ML_ALGORITHMS,
+        "ml": cls.ML_ALGORITHMS,
+        "linear_algebra": cls.ML_LINEAR_ALGEBRA,
+        "matrix": cls.ML_LINEAR_ALGEBRA,
+        "statistics": cls.ML_STATISTICS,
+        "stats": cls.ML_STATISTICS,
+        "data": cls.DATA_PREPROCESSING,
+        "programming" : cls.PROGRAMMING,
+        "misc": cls.MISC,
+      })
+      
+      if string.lower() in mappings:
+        return mappings.get(string.lower())
+      return cls.MISC
+  
+  def __init__(self, name: str, points_value: float, topic: Question.Topic = Topic.MISC, *args, **kwargs):
+    if name is None:
+      name = self.__class__.__name__
+    self.name = name
+    self.points_value = points_value
+    self.topic = topic
+    self.spacing = parse_spacing(kwargs.get("spacing", 0))
+    self.answer_kind = Answer.AnswerKind.BLANK
+
+    # Support for multi-part questions (defaults to 1 for normal questions)
+    self.num_subquestions = kwargs.get("num_subquestions", 1)
+
+    self.extra_attrs = kwargs # clear page, etc.
+
+    self.answers = {}
+    self.possible_variations = float('inf')
+
+    self.rng_seed_offset = kwargs.get("rng_seed_offset", 0)
+
+    # To be used throughout when generating random things
+    self.rng = random.Random()
+
+    # Track question-specific configuration parameters (excluding framework parameters)
+    # These will be included in QR codes for exam regeneration
+    framework_params = {
+      'name', 'points_value', 'topic', 'spacing', 'num_subquestions',
+      'rng_seed_offset', 'rng_seed', 'class', 'kwargs', 'kind'
+    }
+    self.config_params = {k: v for k, v in kwargs.items() if k not in framework_params}
+  
+  @classmethod
+  def from_yaml(cls, path_to_yaml):
+    with open(path_to_yaml) as fid:
+      question_dicts = yaml.safe_load_all(fid)
+  
+  def get_question(self, **kwargs) -> ContentAST.Question:
+    """
+    Gets the question in AST format
+    :param kwargs:
+    :return: (ContentAST.Question) Containing question.
+    """
+    # Generate the question, retrying with incremented seeds until we get an interesting one
+    with timer("question_refresh", question_name=self.name, question_type=self.__class__.__name__):
+      base_seed = kwargs.get("rng_seed", None)
+
+      # Pre-select any regenerable choices using the base seed
+      # This ensures the policy/algorithm stays constant across backoff attempts
+      if hasattr(self, '_regenerable_choices') and self._regenerable_choices:
+        # Seed a temporary RNG with the base seed to make the choices
+        choice_rng = random.Random(base_seed)
+        for param_name, choice_info in self._regenerable_choices.items():
+          if choice_info['fixed_value'] is None:
+            # No fixed value - pick randomly and store it as fixed for this get_question() call
+            enum_class = choice_info['enum_class']
+            random_choice = choice_rng.choice(list(enum_class))
+            # Temporarily set this as the fixed value so all refresh() calls use it
+            choice_info['_temp_fixed_value'] = random_choice.name
+            # Store in config_params
+            self.config_params[param_name] = random_choice.name
+
+      backoff_counter = 0
+      is_interesting = False
+      while not is_interesting:
+        # Increment seed for each backoff attempt to maintain deterministic behavior
+        current_seed = None if base_seed is None else base_seed + backoff_counter
+        self.refresh(rng_seed=current_seed, hard_refresh=(backoff_counter > 0))
+        is_interesting = self.is_interesting()
+        backoff_counter += 1
+
+      # Clear temporary fixed values
+      if hasattr(self, '_regenerable_choices') and self._regenerable_choices:
+        for param_name, choice_info in self._regenerable_choices.items():
+          if '_temp_fixed_value' in choice_info:
+            del choice_info['_temp_fixed_value']
+
+    with timer("question_body", question_name=self.name, question_type=self.__class__.__name__):
+      body = self.get_body()
+
+    with timer("question_explanation", question_name=self.name, question_type=self.__class__.__name__):
+      explanation = self.get_explanation()
+
+    # Store the actual seed used and question metadata for QR code generation
+    actual_seed = None if base_seed is None else base_seed + backoff_counter - 1
+    question_ast = ContentAST.Question(
+      body=body,
+      explanation=explanation,
+      value=self.points_value,
+      spacing=self.spacing,
+      topic=self.topic
+    )
+
+    # Attach regeneration metadata to the question AST
+    question_ast.question_class_name = self.__class__.__name__
+    question_ast.generation_seed = actual_seed
+    question_ast.question_version = self.VERSION
+    # Make a copy of config_params so each question AST has its own
+    # (important when the same Question instance is reused for multiple PDFs)
+    question_ast.config_params = dict(self.config_params)
+
+    return question_ast
+  
+  @abc.abstractmethod
+  def get_body(self, **kwargs) -> ContentAST.Section:
+    """
+    Gets the body of the question during generation
+    :param kwargs:
+    :return: (ContentAST.Section) Containing question body
+    """
+    pass
+  
+  def get_explanation(self, **kwargs) -> ContentAST.Section:
+    """
+    Gets the body of the question during generation
+    :param kwargs:
+    :return: (ContentAST.Section) Containing question explanation or None
+    """
+    return ContentAST.Section(
+      [ContentAST.Text("[Please reach out to your professor for clarification]")]
+    )
+  
+  def get_answers(self, *args, **kwargs) -> Tuple[Answer.AnswerKind, List[Dict[str,Any]]]:
+    return (
+      self.answer_kind,
+      list(itertools.chain(*[a.get_for_canvas() for a in self.answers.values()]))
+    )
+
+  def refresh(self, rng_seed=None, *args, **kwargs):
+    """If it is necessary to regenerate aspects between usages, this is the time to do it.
+    This base implementation simply resets everything.
+    :param rng_seed: random number generator seed to use when regenerating question
+    :param *args:
+    :param **kwargs:
+    :return: bool - True if the generated question is interesting, False otherwise
+    """
+    self.answers = {}
+    # Seed the RNG directly with the provided seed (no offset)
+    self.rng.seed(rng_seed)
+    # Note: We don't call is_interesting() here because child classes need to
+    # generate their workloads first. Child classes should call it at the end
+    # of their refresh() and return the result.
+    return self.is_interesting()  # Default: assume interesting if no override
+    
+  def is_interesting(self) -> bool:
+    return True
+  
+  def get__canvas(self, course: canvasapi.course.Course, quiz : canvasapi.quiz.Quiz, interest_threshold=1.0, *args, **kwargs):
+    # Get the AST for the question
+    with timer("question_get_ast", question_name=self.name, question_type=self.__class__.__name__):
+      questionAST = self.get_question(**kwargs)
+    log.debug("got question ast")
+    # Get the answers and type of question
+    question_type, answers = self.get_answers(*args, **kwargs)
+
+    # Define a helper function for uploading images to canvas
+    def image_upload(img_data) -> str:
+
+      course.create_folder(f"{quiz.id}", parent_folder_path="Quiz Files")
+      file_name = f"{uuid.uuid4()}.png"
+
+      with io.FileIO(file_name, 'w+') as ffid:
+        ffid.write(img_data.getbuffer())
+        ffid.flush()
+        ffid.seek(0)
+        upload_success, f = course.upload(ffid, parent_folder_path=f"Quiz Files/{quiz.id}")
+      os.remove(file_name)
+
+      img_data.name = "img.png"
+      # upload_success, f = course.upload(img_data, parent_folder_path=f"Quiz Files/{quiz.id}")
+      log.debug("path: " + f"/courses/{course.id}/files/{f['id']}/preview")
+      return f"/courses/{course.id}/files/{f['id']}/preview"
+
+    # Render AST to HTML for Canvas
+    with timer("ast_render_body", question_name=self.name, question_type=self.__class__.__name__):
+      question_html = questionAST.render("html", upload_func=image_upload)
+
+    with timer("ast_render_explanation", question_name=self.name, question_type=self.__class__.__name__):
+      explanation_html = questionAST.explanation.render("html", upload_func=image_upload)
+
+    # Build appropriate dictionary to send to canvas
+    return {
+      "question_name": f"{self.name} ({datetime.datetime.now().strftime('%m/%d/%y %H:%M:%S.%f')})",
+      "question_text": question_html,
+      "question_type": question_type.value,
+      "points_possible": self.points_value,
+      "answers": answers,
+      "neutral_comments_html": explanation_html
+    }
+
+
+class QuestionGroup():
+  
+  def __init__(self, questions_in_group: List[Question], pick_once : bool):
+    self.questions = questions_in_group
+    self.pick_once = pick_once
+  
+    self._current_question : Optional[Question] = None
+    
+  def instantiate(self, *args, **kwargs):
+    
+    # todo: Make work with rng_seed (or at least verify)
+    random.seed(kwargs.get("rng_seed", None))
+    
+    if not self.pick_once or self._current_question is None:
+      self._current_question = random.choice(self.questions)
+    
+  def __getattr__(self, name):
+    if self._current_question is None or name == "generate":
+      self.instantiate()
+    try:
+      attr = getattr(self._current_question, name)
+    except AttributeError:
+      raise AttributeError(
+        f"Neither QuestionGroup nor Question has attribute '{name}'"
+      )
+    
+    if callable(attr):
+      def wrapped_method(*args, **kwargs):
+        return attr(*args, **kwargs)
+      return wrapped_method
+    
+    return attr