fix: support create FBA with no data

Author: pablormier

corneto/_data.py

@@ -141,6 +141,10 @@ def from_cdict(cls, d: dict) -> "Data":
             samples[sample_key] = Sample(features)
         return cls(samples)
 
+    @classmethod
+    def empty(cls) -> "Data":
+        return cls.from_cdict({"": {}})
+
     def to_json(self, **kwargs) -> str:
         return json.dumps(self.to_dict(), **kwargs)
 

corneto/methods/future/method.py

@@ -142,7 +142,7 @@ def preprocess(self, graph: BaseGraph, data: Data) -> Tuple[BaseGraph, Data]:
         pass
 
     @abstractmethod
-    def create_problem(self, graph: BaseGraph, data: Data):
+    def create_problem(self, graph: BaseGraph, data: Data) -> ProblemDef:
         """Create the optimization problem.
 
         This method should define variables, constraints, and objectives according to the method's
@@ -157,7 +157,7 @@ def create_problem(self, graph: BaseGraph, data: Data):
         """
         pass
 
-    def build(self, graph: BaseGraph, data: Data) -> ProblemDef:
+    def build(self, graph: BaseGraph, data: Optional[Data] = None) -> ProblemDef:
         """Build the complete optimization problem.
 
         The process involves:
@@ -175,6 +175,8 @@ def build(self, graph: BaseGraph, data: Data) -> ProblemDef:
         Raises:
             ValueError: If lambda_reg > 0 but no regularization variable name is provided.
         """
+        if data is None:
+            data = Data.empty()
         self.processed_graph, self.processed_data = self.preprocess(graph, data)
         self.problem = self.create_problem(self.processed_graph, self.processed_data)
 
@@ -502,7 +504,7 @@ def create_problem(self, graph: BaseGraph, data: Data):
                 # is selected or not, to the edges in the original graph, using the selected_edges list that
                 # contains the indexes of the edges in the original graph. The variable selected_edges contains
                 # None for edges that are not selected in the original graph, those we ignore.
-                # --- Step 1: Filter out None entries ---
+
                 # Create a boolean mask that is True for valid indices (not None)
                 valid_mask = np.array([edge is not None for edge in selected_edges])
 
@@ -512,7 +514,7 @@ def create_problem(self, graph: BaseGraph, data: Data):
                 # Extract the valid edge indices (they are already 0-indexed)
                 valid_edges = np.array([edge for edge in selected_edges if edge is not None])
 
-                # --- Step 2: Build the indicator matrix E ---
+                # Build the indicator matrix E
                 # m is the number of edges in the original graph
                 m = graph.num_edges  # Total number of rows in the target vector
 
@@ -524,7 +526,7 @@ def create_problem(self, graph: BaseGraph, data: Data):
                 # resulting in an indicator matrix E of shape (m, n_valid).
                 E = (rows == valid_edges.reshape(1, -1)).astype(int)
 
-                # --- Step 3: Matrix multiplication to scatter the values ---
+                # Matrix multiplication to scatter the values
                 # Here the "@" operator multiplies the indicator matrix E with valid_vec.
                 # Y is then a vector of shape (m, 1) with the values from vec mapped
                 # to their corresponding positions in the original graph.

corneto/methods/future/pcst.py

@@ -10,7 +10,7 @@
 
 
 class PrizeCollectingSteinerTree(FlowMethod):
-    """Prize-Collecting Steiner Tree optimization method as a flow-based problem.
+    """Prize-Collecting Steiner Tree optimization method built on top of FlowMethod.
 
     In a prize-collecting Steiner tree problem, terminals can have prizes (values > 0),
     making them optional terminals that provide a benefit if included in the solution.
@@ -298,7 +298,7 @@ def create_flow_based_problem(self, flow_problem: ProblemDef, graph: BaseGraph,
             if other_flow_edges:
                 flow_problem += F[other_flow_edges] == 0
 
-            # --- Root Flow Constraints ---
+            # Root Flow Constraints
             if sample_selected_root is not None:
                 if self.force_flow_through_root:
                     flow_problem += F[self.flow_edges_in[sample_selected_root]] == sample_max_flow
@@ -331,7 +331,7 @@ def create_flow_based_problem(self, flow_problem: ProblemDef, graph: BaseGraph,
                                 terminal_neg
                             ][t_idx].sum() == len(t_idx)
 
-            # --- Costs Objective ---
+            # Costs Objective
             edge_costs = np.ones((graph.num_edges)) * self.default_edge_cost
             selected = with_flow if len(with_flow.shape) == 1 else with_flow[:, i]
 
@@ -345,7 +345,7 @@ def create_flow_based_problem(self, flow_problem: ProblemDef, graph: BaseGraph,
 
             flow_problem.add_objective(edge_costs[edge_ids] @ selected[edge_ids], name="edge_cost")
 
-            # --- Prizes Objective (PCST) ---
+            # Prizes Objective (PCST)
             if prized_terminals:
                 prized_idx = []
                 prized_vertices = []

corneto/methods/metabolism/fba.py

@@ -81,10 +81,11 @@ def fba_problem(G, create_reaction_indicators=False, num_fluxes=1, eps=1e-4, bac
 
     Args:
         G (BaseGraph): The metabolic network graph.
-        create_reaction_indicators (bool, optional): Whether to create binary indicators
-            for reactions. Defaults to False.
+        create_reaction_indicators (bool, optional): Whether to create binary positive
+            and negative indicators for reaction fluxes. Defaults to False.
         num_fluxes (int, optional): Number of flux variables per reaction. Defaults to 1.
-        eps (float, optional): Tolerance for considering a flux as non-zero. Defaults to 1e-4.
+        eps (float, optional): Tolerance for considering a flux as non-zero. Only used
+            with indicators. Defaults to 1e-4.
         backend (Backend, optional): The optimization backend to use. Defaults to K.
 
     Returns:

tests/methods/future/test_fba_future.py

@@ -17,6 +17,14 @@ def metabolic_network():
     return import_miom_model(file_path)
 
 
+def test_fba_without_data(metabolic_network, backend):
+    """Test a standard FBA problem with no objective."""
+    fba = MultiSampleFBA(backend=backend)
+    P = fba.build(metabolic_network)
+    result = P.solve()
+    assert result.status == "optimal"
+
+
 def test_single_sample_standard_fba(metabolic_network, backend):
     """Test the standard FBA method with a single sample."""
     fba = MultiSampleFBA(backend=backend)