Add .value() method for expression evaluation and refactor variable retrieval

.github/workflows/ci.yml

@@ -115,7 +115,7 @@ jobs:
       - name: Install Rust toolchain
         uses: dtolnay/rust-toolchain@master
         with:
-          toolchain: "1.70"
+          toolchain: "1.71"
 
       - name: Check build with MSRV
         run: cargo check --all-features

Cargo.toml

@@ -9,7 +9,7 @@ repository = "https://github.com/cvxpy/cvxrust"
 documentation = "https://docs.rs/cvxrust"
 keywords = ["optimization", "convex", "dcp", "solver", "math"]
 categories = ["mathematics", "algorithms", "science"]
-rust-version = "1.70"
+rust-version = "1.71"
 
 [dependencies]
 # Clarabel solver

examples/basic_lp.rs

@@ -45,7 +45,7 @@ fn main() {
     println!("  Status: {:?}", solution.status);
     println!("  Optimal profit: {:.4}", -solution.value.unwrap());
 
-    let x_vals = &solution[&x];
+    let x_vals = x.value(&solution);
     println!("  x1 = {:.4}", x_vals[(0, 0)]);
     println!("  x2 = {:.4}", x_vals[(1, 0)]);
     println!("  x3 = {:.4}", x_vals[(2, 0)]);

examples/lasso.rs

@@ -38,7 +38,7 @@ fn main() {
         .solve()
         .expect("Failed to solve");
 
-    let x_ls_vals = &solution_ls[&x_ls];
+    let x_ls_vals = x_ls.value(&solution_ls);
     println!("Coefficients:");
     for i in 0..10 {
         println!("  x{}: {:.6}", i + 1, x_ls_vals[(i, 0)]);
@@ -57,7 +57,7 @@ fn main() {
         .solve()
         .expect("Failed to solve");
 
-    let x_vals = &solution[&x];
+    let x_vals = x.value(&solution);
     println!("Coefficients:");
     for i in 0..10 {
         let val = x_vals[(i, 0)];
@@ -66,6 +66,12 @@ fn main() {
     }
     println!("  Objective: {:.6}", solution.value.unwrap());
 
+    // Expression values: inspect individual terms
+    let fit_err = sum_squares(&residual).value(&solution).as_scalar().unwrap();
+    let l1_pen = norm1(&x).value(&solution).as_scalar().unwrap();
+    println!("  Fit error (expression eval):   {:.6}", fit_err);
+    println!("  L1 penalty (expression eval):  {:.6}", l1_pen);
+
     // LASSO with strong regularization
     println!("\n--- LASSO (lambda = 1.0) ---\n");
 
@@ -78,7 +84,7 @@ fn main() {
         .solve()
         .expect("Failed to solve");
 
-    let x2_vals = &solution2[&x2];
+    let x2_vals = x2.value(&solution2);
     println!("Coefficients:");
     for i in 0..10 {
         let val = x2_vals[(i, 0)];

examples/least_squares.rs

@@ -28,9 +28,30 @@ fn main() {
     println!("  Status: {:?}", solution.status);
     println!("  Optimal value: {:.6}", solution.value.unwrap());
 
-    let w_vals = &solution[&w];
-    println!("  w0 (intercept) = {:.6}", w_vals[(0, 0)]);
-    println!("  w1 (slope) = {:.6}", w_vals[(1, 0)]);
+    // Two equivalent ways to retrieve variable values:
+    let w_vals = &solution[&w]; // index into solution directly
+    let w_vals2 = w.value(&solution); // call .value() on the variable
+    println!(
+        "  w0 (intercept) = {:.6} / {:.6}",
+        w_vals[(0, 0)],
+        w_vals2[(0, 0)]
+    );
+    println!(
+        "  w1 (slope)     = {:.6} / {:.6}",
+        w_vals[(1, 0)],
+        w_vals2[(1, 0)]
+    );
+
+    // Expression values: evaluate any sub-expression at the solution
+    let fitted = matmul(&a, &w).value(&solution);
+    println!(
+        "  Fitted values: [{:.2}, {:.2}, {:.2}, {:.2}, {:.2}]",
+        fitted[(0, 0)],
+        fitted[(1, 0)],
+        fitted[(2, 0)],
+        fitted[(3, 0)],
+        fitted[(4, 0)]
+    );
 
     // Constrained least squares (w >= 0)
     println!("\n--- Constrained Least Squares (w >= 0) ---\n");
@@ -47,7 +68,7 @@ fn main() {
     println!("  Status: {:?}", solution2.status);
     println!("  Optimal value: {:.6}", solution2.value.unwrap());
 
-    let w2_vals = &solution2[&w2];
+    let w2_vals = w2.value(&solution2);
     println!("  w0 (intercept) = {:.6}", w2_vals[(0, 0)]);
-    println!("  w1 (slope) = {:.6}", w2_vals[(1, 0)]);
+    println!("  w1 (slope)     = {:.6}", w2_vals[(1, 0)]);
 }

examples/portfolio.rs

@@ -47,13 +47,20 @@ fn main() {
 
     // Results
     println!("Optimal Portfolio:");
-    let portfolio = &solution[&x];
+    let portfolio = x.value(&solution);
     let assets = ["A", "B", "C", "D"];
 
     for i in 0..4 {
         println!("  Asset {}: {:.2}%", assets[i], portfolio[(i, 0)] * 100.0);
     }
 
+    // Expression values: evaluate return and risk directly
+    let actual_return = dot(&mu, &x).value(&solution).as_scalar().unwrap();
+    println!(
+        "  Actual return (expression eval): {:.2}%",
+        actual_return * 100.0
+    );
+
     let variance = solution.value.unwrap();
     let std_dev = variance.sqrt();
     println!("\nPortfolio Statistics:");

examples/quadratic_program.rs

@@ -33,7 +33,7 @@ fn main() {
     println!("  Status: {:?}", solution.status);
     println!("  Optimal value: {:.6}", solution.value.unwrap());
 
-    let x_vals = &solution[&x];
+    let x_vals = x.value(&solution);
     println!("  x1 = {:.6}", x_vals[(0, 0)]);
     println!("  x2 = {:.6}", x_vals[(1, 0)]);
     println!("  Sum: {:.6}", x_vals[(0, 0)] + x_vals[(1, 0)]);