Add Giese comparison script

examples/giese_comparison.py

@@ -0,0 +1,93 @@
+r"""
+Comparison of Giese and PTtools solvers
+=======================================
+"""
+
+
+import logging
+
+
+import matplotlib.pyplot as plt
+import numpy as np
+
+from examples.utils import save
+from pttools.analysis.parallel import create_bubbles
+from pttools.bubble import Bubble
+from pttools.models import ConstCSModel
+from pttools.speedup import GITHUB_ACTIONS
+
+logger = logging.getLogger(__name__)
+
+
+def get_kappa(bubble: Bubble) -> float:
+    if (not bubble.solved) or bubble.no_solution_found or bubble.solver_failed or bubble.numerical_error:
+        return np.nan
+    return bubble.kappa
+
+get_kappa.return_type = float
+get_kappa.fail_value = np.nan
+
+
+def main():
+    alpha_ns = np.array([0.01, 0.03, 0.1, 0.3, 1, 3])
+    colors = ["b", "y", "r", "g", "purple", "grey"]
+    n_v_walls = 20 if GITHUB_ACTIONS else 50
+    v_walls = np.linspace(0.2, 0.95, n_v_walls)
+    a_s = 5
+    a_b = 1
+    V_s = 1
+    models = [
+        ConstCSModel(css2=1 / 3, csb2=1 / 3, a_s=a_s, a_b=a_b, V_s=V_s, alpha_n_min=alpha_ns[0]),
+        ConstCSModel(css2=1 / 3, csb2=1 / 4, a_s=a_s, a_b=a_b, V_s=V_s, alpha_n_min=alpha_ns[0]),
+        ConstCSModel(css2=1 / 4, csb2=1 / 3, a_s=a_s, a_b=a_b, V_s=V_s, alpha_n_min=alpha_ns[0]),
+        ConstCSModel(css2=1 / 4, csb2=1 / 4, a_s=a_s, a_b=a_b, V_s=V_s, alpha_n_min=alpha_ns[0])
+    ]
+    logger.info(f"Minimum alpha_ns: %s", [model.alpha_n_min for model in models])
+    for model in models:
+        logger.info("Model parameters: %s", model.params_str())
+
+    figsize_x = 10
+    fig1: plt.Figure = plt.figure(figsize=(figsize_x, 6))
+    fig2: plt.Figure = plt.figure(figsize=(figsize_x, 6))
+    axs1 = fig1.subplots(1, 2)
+    ax1 = axs1[0]
+    ax2 = axs1[1]
+    ax3 = fig2.add_subplot()
+
+    kappas_pttools = np.empty((len(models), alpha_ns.size, v_walls.size))
+    kappas_giese = np.empty((len(models), alpha_ns.size, v_walls.size))
+    for i_model, model in enumerate(models):
+        ls = "--" if i_model in [2, 3] else "-"
+        bubbles_pttools, kappas_pttools[i_model, :, :] = create_bubbles(
+            model=model, v_walls=v_walls, alpha_ns=alpha_ns, func=get_kappa,
+            bubble_kwargs={"allow_invalid": False}, allow_bubble_failure=True
+        )
+        bubbles_giese, kappas_giese[i_model, :, :] = create_bubbles(
+            model=model, v_walls=v_walls, alpha_ns=alpha_ns, func=get_kappa,
+            bubble_kwargs={"allow_invalid": False, "use_giese_solver": True}, allow_bubble_failure=True
+        )
+        for i_alpha_n, (alpha_n, color) in enumerate(zip(alpha_ns, colors)):
+            kpt = kappas_pttools[i_model, i_alpha_n, :]
+            kg = kappas_giese[i_model, i_alpha_n, :]
+            ax1.plot(v_walls, kpt, ls=ls, color=color, alpha=0.5)
+            ax2.plot(v_walls, kg, ls=ls, color=color, alpha=0.5)
+            ax3.plot(v_walls, (kpt - kg)/kg, color=color, alpha=0.5)
+
+    for ax in axs1.flat:
+        ax.set_xlabel(r"$v_\text{wall}$")
+        ax.set_ylabel(r"$\kappa$")
+        ax.set_xlim(v_walls.min(), v_walls.max())
+        ax.set_yscale("log")
+    ax1.set_title("PTtools")
+    ax2.set_title("Giese et al.")
+    ax3.set_xlim(v_walls.min(), v_walls.max())
+    ax3.set_yscale("log")
+
+    return fig1, fig2
+
+
+if __name__ == "__main__":
+    fig, fig2 = main()
+    save(fig, "giese_comparison")
+    save(fig, "giese_comparison_diff")
+    plt.show()