[ENH] Convert InterpolationOptions to Pydantic model (#12)

# Getting the first example of InterpolationOptions serializable

This PR converts the `InterpolationOptions` class from a dataclass to a Pydantic model to enable serialization. The changes include:

- Replaced dataclass with Pydantic's `BaseModel` for `InterpolationOptions`
- Converted the constructor to a class method `from_args()` to maintain backward compatibility
- Added proper type annotations and Pydantic Field definitions
- Fixed a minor type issue in `octree_curvature_threshold` by adding a decimal point
- Added proper model configuration for enum handling
- Made `TempInterpolationValues` a proper dataclass

These changes are the first step toward making the model serializable while maintaining the existing functionality.

Author: Leguark

gempy_engine/API/server/main_server_pro.py

@@ -27,7 +27,7 @@
 
 # Default interpolation options
 range_ = 1
-default_interpolation_options: InterpolationOptions = InterpolationOptions(
+default_interpolation_options: InterpolationOptions = InterpolationOptions.from_args(
     range=range_,
     c_o=(range_ ** 2) / 14 / 3,
     number_octree_levels=4,

gempy_engine/core/data/options/evaluation_options.py

@@ -14,7 +14,7 @@ class MeshExtractionMaskingOptions(enum.Enum):
 class EvaluationOptions:
     _number_octree_levels: int = 1
     _number_octree_levels_surface: int = 4
-    octree_curvature_threshold: float = -1  #: Threshold to do octree refinement due to curvature to deal with angular geometries. This curvature assumes that 1 is the maximum curvature of any voxel
+    octree_curvature_threshold: float = -1.  #: Threshold to do octree refinement due to curvature to deal with angular geometries. This curvature assumes that 1 is the maximum curvature of any voxel
     octree_error_threshold: float = 1.  #: Number of standard deviations to consider a voxel as candidate to refine
     octree_min_level: int = 2
 

gempy_engine/core/data/options/interpolation_options.py

@@ -1,6 +1,7 @@
 import enum
 import warnings
-from dataclasses import dataclass, asdict, field
+
+from pydantic import BaseModel, ConfigDict, Field, model_validator, PrivateAttr
 
 import gempy_engine.config
 from .evaluation_options import MeshExtractionMaskingOptions, EvaluationOptions
@@ -10,40 +11,50 @@
 from ..raw_arrays_solution import RawArraysSolution
 
 
-@dataclass
-class InterpolationOptions:
-    __slots__ = ['kernel_options', 'evaluation_options', 'temp_interpolation_values', 'debug',
-                 'cache_mode', 'cache_model_name', 'block_solutions_type', 'sigmoid_slope']
-
+class InterpolationOptions(BaseModel):
     class CacheMode(enum.Enum):
         """ Cache mode for the interpolation"""
         NO_CACHE: int = enum.auto()  #: No cache at all even during the interpolation computation. This is quite expensive for no good reason.
         CACHE = enum.auto()
         IN_MEMORY_CACHE = enum.auto()
         CLEAR_CACHE = enum.auto()
 
+    model_config = ConfigDict(
+        arbitrary_types_allowed=False,
+        use_enum_values=False,
+        json_encoders={
+                CacheMode: lambda e: e.value,
+                AvailableKernelFunctions: lambda e: e.name
+        }
+    )
+
     # @off
-    kernel_options: KernelOptions  # * This is the compression of the fields above and the way to go in the future
-    evaluation_options: EvaluationOptions
-    temp_interpolation_values: TempInterpolationValues
+    kernel_options: KernelOptions = Field(init=True, exclude=False)  # * This is the compression of the fields above and the way to go in the future
+    evaluation_options: EvaluationOptions = Field(init=True, exclude= False)
 
     debug: bool
     cache_mode: CacheMode
     cache_model_name: str  # : Model name for the cache
-
     block_solutions_type: RawArraysSolution.BlockSolutionType
-
     sigmoid_slope: int
-
     debug_water_tight: bool = False
 
-    def __init__(
-            self,
+    # region Volatile
+    temp_interpolation_values: TempInterpolationValues = Field(
+        default_factory=TempInterpolationValues,
+        exclude=True
+    )
+   
+    # endregion
+
+    @classmethod
+    def from_args(
+            cls,
             range: int | float,
             c_o: float,
             uni_degree: int = 1,
-            i_res: float = 4,
-            gi_res: float = 2,  # ! This should be DEP
+            i_res: float = 4.,
+            gi_res: float = 2.,  # ! This should be DEP
             number_dimensions: int = 3,  # ? This probably too
             number_octree_levels: int = 1,
             kernel_function: AvailableKernelFunctions = AvailableKernelFunctions.cubic,
@@ -52,7 +63,7 @@ def __init__(
             compute_condition_number: bool = False,
     ):
 
-        self.kernel_options = KernelOptions(
+        kernel_options = KernelOptions(
             range=range,
             c_o=c_o,
             uni_degree=uni_degree,
@@ -63,7 +74,7 @@ def __init__(
             compute_condition_number=compute_condition_number
         )
 
-        self.evaluation_options = EvaluationOptions(
+        evaluation_options = EvaluationOptions(
             _number_octree_levels=number_octree_levels,
             _number_octree_levels_surface=4,
             mesh_extraction=mesh_extraction,
@@ -73,18 +84,30 @@ def __init__(
 
         )
 
-        self.temp_interpolation_values = TempInterpolationValues()
-        self.debug = gempy_engine.config.DEBUG_MODE
-        self.cache_mode = InterpolationOptions.CacheMode.IN_MEMORY_CACHE
-        self.cache_model_name = ""
-        self.block_solutions_type = RawArraysSolution.BlockSolutionType.OCTREE
-        self.sigmoid_slope = 5_000_000
+        temp_interpolation_values = TempInterpolationValues()
+        debug = gempy_engine.config.DEBUG_MODE
+        cache_mode = InterpolationOptions.CacheMode.IN_MEMORY_CACHE
+        cache_model_name = ""
+        block_solutions_type = RawArraysSolution.BlockSolutionType.OCTREE
+        sigmoid_slope = 5_000_000
+
+        return InterpolationOptions(
+            kernel_options=kernel_options,
+            evaluation_options=evaluation_options,
+            # temp_interpolation_values=temp_interpolation_values,
+            debug=debug,
+            cache_mode=cache_mode,
+            cache_model_name=cache_model_name,
+            block_solutions_type=block_solutions_type,
+            sigmoid_slope=sigmoid_slope,
+            debug_water_tight=False,
+        )
 
     # @on
 
     @classmethod
     def init_octree_options(cls, range=1.7, c_o=10, refinement: int = 1):
-        return InterpolationOptions(
+        return InterpolationOptions.from_args(
             range=range,
             c_o=c_o,
             mesh_extraction=True,
@@ -93,7 +116,7 @@ def init_octree_options(cls, range=1.7, c_o=10, refinement: int = 1):
 
     @classmethod
     def init_dense_grid_options(cls):
-        options = InterpolationOptions(
+        options = InterpolationOptions.from_args(
             range=1.7,
             c_o=10,
             mesh_extraction=False,
@@ -107,17 +130,17 @@ def probabilistic_options(cls):
         # TODO: This should have the sigmoid slope different
         raise NotImplementedError("Probabilistic interpolation is not yet implemented.")
 
-    def __repr__(self):
-        return f"InterpolationOptions({', '.join(f'{k}={v}' for k, v in asdict(self).items())})"
-
-    def _repr_html_(self):
-        html = f"""
-                <table>
-                    <tr><td colspan='2' style='text-align:center'><b>InterpolationOptions</b></td></tr>
-                    {''.join(f'<tr><td>{k}</td><td>{v._repr_html_() if isinstance(v, KernelOptions) else v}</td></tr>' for k, v in asdict(self).items())}
-                </table>
-                """
-        return html
+    # def __repr__(self):
+    #     return f"InterpolationOptions.from_args({', '.join(f'{k}={v}' for k, v in asdict(self).items())})"
+
+    # def _repr_html_(self):
+    #     html = f"""
+    #             <table>
+    #                 <tr><td colspan='2' style='text-align:center'><b>InterpolationOptions</b></td></tr>
+    #                 {''.join(f'<tr><td>{k}</td><td>{v._repr_html_() if isinstance(v, KernelOptions) else v}</td></tr>' for k, v in asdict(self).items())}
+    #             </table>
+    #             """
+    #     return html
 
     def update_options(self, **kwargs):
         """

gempy_engine/core/data/options/kernel_options.py

@@ -1,14 +1,17 @@
 import warnings
 
 from dataclasses import dataclass, asdict
+from typing import Optional
+
+from pydantic import field_validator
 
 from gempy_engine.core.data.kernel_classes.kernel_functions import AvailableKernelFunctions
 from gempy_engine.core.data.kernel_classes.solvers import Solvers
 
 
 @dataclass(frozen=False)
 class KernelOptions:
-    range: int  # TODO: have constructor from RegularGrid
+    range: int | float  # TODO: have constructor from RegularGrid
     c_o: float  # TODO: This should be a property
     uni_degree: int = 1
     i_res: float = 4.
@@ -20,7 +23,25 @@ class KernelOptions:
 
     compute_condition_number: bool = False
     optimizing_condition_number: bool = False
-    condition_number: float = None
+    condition_number: Optional[float] = None
+
+    @field_validator('kernel_function', mode='before')
+    @classmethod
+    def _deserialize_kernel_function_from_name(cls, value):
+        """
+        Ensures that a string input (e.g., "cubic" from JSON)
+        is correctly converted to an AvailableKernelFunctions enum member.
+        """
+        if isinstance(value, str):
+            try:
+                return AvailableKernelFunctions[value]  # Lookup enum member by name
+            except KeyError:
+                # This provides a more specific error if the name doesn't exist
+                valid_names = [member.name for member in AvailableKernelFunctions]
+                raise ValueError(f"Invalid kernel function name '{value}'. Must be one of: {valid_names}")
+        # If it's already an AvailableKernelFunctions member (e.g., during direct model instantiation),
+        # or if it's another type that Pydantic's later validation will catch as an error.
+        return value
 
     @property
     def n_uni_eq(self):
@@ -65,16 +86,16 @@ def update_options(self, **kwargs):
     def __hash__(self):
         # Using a tuple to hash all the values together
         return hash((
-            self.range,
-            self.c_o,
-            self.uni_degree,
-            self.i_res,
-            self.gi_res,
-            self.number_dimensions,
-            self.kernel_function,
-            self.compute_condition_number,
+                self.range,
+                self.c_o,
+                self.uni_degree,
+                self.i_res,
+                self.gi_res,
+                self.number_dimensions,
+                self.kernel_function,
+                self.compute_condition_number,
         ))
-    
+
     def __repr__(self):
         return f"KernelOptions({', '.join(f'{k}={v}' for k, v in asdict(self).items())})"
 

gempy_engine/core/data/options/temp_interpolation_values.py

@@ -1,2 +1,6 @@
+from dataclasses import dataclass
+
+
+@dataclass
 class TempInterpolationValues:
     current_octree_level: int = 0  # * Make this a read only property 

gempy_engine/modules/kernel_constructor/_pykeops_cov_compiler.py

@@ -40,7 +40,7 @@ def simple_model_2():
     ori_i = Orientations(dip_positions, nugget_effect_grad)
 
     range = 5 ** 2
-    kri = InterpolationOptions(range, 1, 0, i_res=1, gi_res=1,
+    kri = InterpolationOptions.from_args(range, 1, 0, i_res=1, gi_res=1,
                                number_dimensions=2)
 
     _ = np.ones(3)

requirements/requirements.txt

@@ -1,2 +1,3 @@
 numpy
+pydantic
 python-dotenv

tests/benchmark/one_fault_model.py

@@ -93,7 +93,7 @@ def one_fault_model():
     range_ = 7 ** 2  # ? Since we are not getting the square root should we also square this? 
     c_o = 1
 
-    options = InterpolationOptions(
+    options = InterpolationOptions.from_args(
         range_, c_o,
         uni_degree=1,
         number_dimensions=3,

tests/fixtures/complex_geometries.py

@@ -77,7 +77,7 @@ def one_fault_model():
     range_ = 7 ** 2  # ? Since we are not getting the square root should we also square this? 
     c_o = 1
 
-    options = InterpolationOptions(
+    options = InterpolationOptions.from_args(
         range_, c_o,
         uni_degree=1,
         number_dimensions=3,
@@ -144,7 +144,7 @@ def one_finite_fault_model():
     range_ = 7 ** 2  # ? Since we are not getting the square root should we also square this? 
     c_o = 1
 
-    options = InterpolationOptions(
+    options = InterpolationOptions.from_args(
         range_, c_o,
         uni_degree=1,
         number_dimensions=3,
@@ -211,7 +211,7 @@ def graben_fault_model():
     range_ = 7 ** 2  # ? Since we are not getting the square root should we also square this? 
     c_o = 1
 
-    options = InterpolationOptions(
+    options = InterpolationOptions.from_args(
         range_, c_o,
         uni_degree=1,
         number_dimensions=3,

tests/fixtures/heavy_models.py

@@ -102,7 +102,7 @@ def moureze_model_factory(path_to_root: str, pick_every=8, octree_lvls=3, solver
     # endregion
 
     # region InterpolationOptions
-    interpolation_options: InterpolationOptions = InterpolationOptions(
+    interpolation_options: InterpolationOptions = InterpolationOptions.from_args(
         range=100.,
         c_o=10.,
         number_octree_levels=octree_lvls,