Implementation of matplotlib backend for criterion_plot()

.tools/envs/testenv-linux.yml

@@ -19,6 +19,7 @@ dependencies:
   - numpy >= 2  # run, tests
   - pandas  # run, tests
   - plotly>=6.2  # run, tests
+  - matplotlib  # tests
   - pybaum>=0.1.2  # run, tests
   - scipy>=1.2.1  # run, tests
   - sqlalchemy  # run, tests

.tools/envs/testenv-nevergrad.yml

@@ -17,6 +17,7 @@ dependencies:
   - numpy >= 2  # run, tests
   - pandas  # run, tests
   - plotly>=6.2  # run, tests
+  - matplotlib  # tests
   - pybaum>=0.1.2  # run, tests
   - scipy>=1.2.1  # run, tests
   - sqlalchemy  # run, tests

.tools/envs/testenv-numpy.yml

@@ -17,6 +17,7 @@ dependencies:
   - cloudpickle  # run, tests
   - joblib  # run, tests
   - plotly>=6.2  # run, tests
+  - matplotlib  # tests
   - pybaum>=0.1.2  # run, tests
   - scipy>=1.2.1  # run, tests
   - sqlalchemy  # run, tests

.tools/envs/testenv-others.yml

@@ -17,6 +17,7 @@ dependencies:
   - numpy >= 2  # run, tests
   - pandas  # run, tests
   - plotly>=6.2  # run, tests
+  - matplotlib  # tests
   - pybaum>=0.1.2  # run, tests
   - scipy>=1.2.1  # run, tests
   - sqlalchemy  # run, tests

.tools/envs/testenv-pandas.yml

@@ -17,6 +17,7 @@ dependencies:
   - cloudpickle  # run, tests
   - joblib  # run, tests
   - plotly>=6.2  # run, tests
+  - matplotlib  # tests
   - pybaum>=0.1.2  # run, tests
   - scipy>=1.2.1  # run, tests
   - sqlalchemy  # run, tests

.tools/envs/testenv-plotly.yml

@@ -17,6 +17,7 @@ dependencies:
   - joblib  # run, tests
   - numpy >= 2  # run, tests
   - pandas  # run, tests
+  - matplotlib  # tests
   - pybaum>=0.1.2  # run, tests
   - scipy>=1.2.1  # run, tests
   - sqlalchemy  # run, tests

environment.yml

@@ -21,6 +21,7 @@ dependencies:
   - numpy >= 2  # run, tests
   - pandas  # run, tests
   - plotly>=6.2  # run, tests
+  - matplotlib  # tests
   - pybaum>=0.1.2  # run, tests
   - scipy>=1.2.1  # run, tests
   - sqlalchemy  # run, tests

pyproject.toml

@@ -290,6 +290,7 @@ module = [
 
     "optimagic.visualization",
     "optimagic.visualization.convergence_plot",
+    "optimagic.visualization.backends",
     "optimagic.visualization.deviation_plot",
     "optimagic.visualization.history_plots",
     "optimagic.visualization.plotting_utilities",
@@ -346,6 +347,8 @@ module = [
     "plotly.graph_objects",
     "plotly.express",
     "plotly.subplots",
+    "matplotlib",
+    "matplotlib.pyplot",
     "cyipopt",
     "nlopt",
     "bokeh",

src/optimagic/config.py

@@ -23,7 +23,7 @@ def _is_installed(module_name: str) -> bool:
 
 
 # ======================================================================================
-# Check Available Packages
+# Check Available Optimization Packages
 # ======================================================================================
 
 IS_PETSC4PY_INSTALLED = _is_installed("petsc4py")
@@ -40,6 +40,12 @@ def _is_installed(module_name: str) -> bool:
 IS_NEVERGRAD_INSTALLED = _is_installed("nevergrad")
 IS_BAYESOPT_INSTALLED = _is_installed("bayes_opt")
 
+# ======================================================================================
+# Check Available Visualization Packages
+# ======================================================================================
+
+IS_MATPLOTLIB_INSTALLED = _is_installed("matplotlib")
+
 
 # ======================================================================================
 # Check if pandas version is newer or equal to version 2.1.0

src/optimagic/exceptions.py

@@ -79,6 +79,10 @@ class InvalidAlgoInfoError(OptimagicError):
     """Exception for invalid user provided algorithm information."""
 
 
+class InvalidPlottingBackendError(OptimagicError):
+    """Exception for invalid user provided plotting backend."""
+
+
 class StopOptimizationError(OptimagicError):
     def __init__(self, message, current_status):
         super().__init__(message)

src/optimagic/visualization/backends.py

@@ -0,0 +1,145 @@
+import abc
+from typing import Any
+
+import plotly.express as px
+import plotly.graph_objects as go
+
+from optimagic.config import IS_MATPLOTLIB_INSTALLED
+from optimagic.exceptions import InvalidPlottingBackendError, NotInstalledError
+from optimagic.visualization.plotting_utilities import LineData
+
+if IS_MATPLOTLIB_INSTALLED:
+    import matplotlib as mpl
+    import matplotlib.pyplot as plt
+
+    # Handle the case where matplotlib is used in notebooks (inline backend)
+    # to ensure that interactive mode is disabled to avoid double plotting.
+    # (See: https://github.com/matplotlib/matplotlib/issues/26221)
+    if mpl.get_backend() == "module://matplotlib_inline.backend_inline":
+        plt.install_repl_displayhook()
+        plt.ioff()
+
+
+class PlotBackend(abc.ABC):
+    is_available: bool
+    default_template: str
+
+    @classmethod
+    @abc.abstractmethod
+    def get_default_palette(cls) -> list:
+        pass
+
+    @abc.abstractmethod
+    def __init__(self, template: str | None):
+        if template is None:
+            template = self.default_template
+
+        self.template = template
+        self.figure: Any = None
+
+    @abc.abstractmethod
+    def add_lines(self, lines: list[LineData]) -> None:
+        pass
+
+    @abc.abstractmethod
+    def set_labels(self, xlabel: str | None = None, ylabel: str | None = None) -> None:
+        pass
+
+    @abc.abstractmethod
+    def set_legend_properties(self, legend_properties: dict[str, Any]) -> None:
+        pass
+
+
+class PlotlyBackend(PlotBackend):
+    is_available: bool = True
+    default_template: str = "simple_white"
+
+    @classmethod
+    def get_default_palette(cls) -> list:
+        return px.colors.qualitative.Set2
+
+    def __init__(self, template: str | None):
+        super().__init__(template)
+        self._fig = go.Figure()
+        self._fig.update_layout(template=self.template)
+        self.figure = self._fig
+
+    def add_lines(self, lines: list[LineData]) -> None:
+        for line in lines:
+            trace = go.Scatter(
+                x=line.x,
+                y=line.y,
+                name=line.name,
+                mode="lines",
+                line_color=line.color,
+                showlegend=line.show_in_legend,
+                connectgaps=True,
+            )
+            self._fig.add_trace(trace)
+
+    def set_labels(self, xlabel: str | None = None, ylabel: str | None = None) -> None:
+        self._fig.update_layout(xaxis_title_text=xlabel, yaxis_title_text=ylabel)
+
+    def set_legend_properties(self, legend_properties: dict[str, Any]) -> None:
+        self._fig.update_layout(legend=legend_properties)
+
+
+class MatplotlibBackend(PlotBackend):
+    is_available: bool = IS_MATPLOTLIB_INSTALLED
+    default_template: str = "default"
+
+    @classmethod
+    def get_default_palette(cls) -> list:
+        return [mpl.colormaps["Set2"](i) for i in range(mpl.colormaps["Set2"].N)]
+
+    def __init__(self, template: str | None):
+        super().__init__(template)
+        plt.style.use(self.template)
+        self._fig, self._ax = plt.subplots()
+        self.figure = self._fig
+
+    def add_lines(self, lines: list[LineData]) -> None:
+        for line in lines:
+            self._ax.plot(
+                line.x,
+                line.y,
+                color=line.color,
+                label=line.name if line.show_in_legend else None,
+            )
+
+    def set_labels(self, xlabel: str | None = None, ylabel: str | None = None) -> None:
+        self._ax.set(xlabel=xlabel, ylabel=ylabel)
+
+    def set_legend_properties(self, legend_properties: dict[str, Any]) -> None:
+        self._ax.legend(**legend_properties)
+
+
+PLOT_BACKEND_CLASSES = {
+    "plotly": PlotlyBackend,
+    "matplotlib": MatplotlibBackend,
+}
+
+
+def get_plot_backend_class(backend_name: str) -> type[PlotBackend]:
+    if backend_name not in PLOT_BACKEND_CLASSES:
+        msg = (
+            f"Invalid backend name '{backend_name}'. "
+            f"Supported backends are: {', '.join(PLOT_BACKEND_CLASSES.keys())}."
+        )
+        raise InvalidPlottingBackendError(msg)
+
+    return _get_backend_if_installed(backend_name)
+
+
+def _get_backend_if_installed(backend_name: str) -> type[PlotBackend]:
+    plot_cls = PLOT_BACKEND_CLASSES[backend_name]
+
+    if not plot_cls.is_available:
+        msg = (
+            f"The '{backend_name}' backend is not installed. "
+            f"Install the package using either 'pip install {backend_name}' or "
+            f"'conda install -c conda-forge {backend_name}'"
+        )
+        raise NotInstalledError(msg)
+
+    return plot_cls