CTI resurrection Phase 4 companions: factor-graph runtime fixes

autocti/aggregator/imaging_ci.py

@@ -89,6 +89,7 @@ def values_from(hdu: int) -> aa.Array2D:
             cosmic_ray_map=cosmic_ray_map,
             settings_dict=settings_dict,
             layout=layout,
+            check_noise_map=False,
         )
 
         dataset_list.append(dataset.apply_mask(mask=mask))

autocti/charge_injection/imaging/imaging.py

@@ -24,8 +24,11 @@ def __init__(
         noise_scaling_map_dict: Optional[Dict] = None,
         fpr_value: Optional[float] = None,
         settings_dict: Optional[Dict] = None,
+        check_noise_map: bool = True,
     ):
-        super().__init__(data=data, noise_map=noise_map)
+        super().__init__(
+            data=data, noise_map=noise_map, check_noise_map=check_noise_map
+        )
 
         self.data = self.data.native
         self.noise_map = self.noise_map.native
@@ -319,7 +322,9 @@ def output_to_fits(
             exception is raised.
         """
         fitsable.output_to_fits(
-            values=np.asarray(self.data.native), file_path=data_path, overwrite=overwrite
+            values=np.asarray(self.data.native),
+            file_path=data_path,
+            overwrite=overwrite,
         )
 
         if noise_map_path is not None:

autocti/charge_injection/model/plotter.py

@@ -7,6 +7,8 @@
 from autocti.charge_injection.plot import fit_ci_plots
 from autocti.model.plotter import Plotter, plot_setting
 
+from autoarray import exc as aa_exc
+
 from autocti import exc
 
 logger = logging.getLogger(__name__)
@@ -85,7 +87,16 @@ def should_plot(name):
                     output_format=self.fmt,
                     title_prefix=self.title_prefix,
                 )
-            except (exc.PlottingException, exc.RegionException, TypeError, ValueError):
+            except (
+                exc.PlottingException,
+                exc.RegionException,
+                aa_exc.ArrayException,
+                TypeError,
+                ValueError,
+            ):
+                # Trimmed datasets (e.g. via `apply_settings`) have layouts whose
+                # extraction regions no longer match the array shape, making the
+                # binned-FPR diagnostic ill-defined.
                 logger.info(
                     "VISUALIZATION - Could not visualize the ImagingCI binned data"
                 )

autocti/charge_injection/model/result.py

@@ -5,17 +5,17 @@
 class ResultImagingCI(ResultDataset):
     @property
     def max_log_likelihood_full_fit(self) -> FitImagingCI:
-        return self.analysis.fit_via_instance_and_dataset_from(
+        return self.analysis_unwrapped.fit_via_instance_and_dataset_from(
             instance=self.instance,
-            dataset=self.analysis.dataset_full,
+            dataset=self.analysis_unwrapped.dataset_full,
             hyper_noise_scale=True,
         )
 
     @property
     def max_log_likelihood_full_fit_no_hyper_scaling(self):
-        return self.analysis.fit_via_instance_and_dataset_from(
+        return self.analysis_unwrapped.fit_via_instance_and_dataset_from(
             instance=self.instance,
-            dataset=self.analysis.dataset_full,
+            dataset=self.analysis_unwrapped.dataset_full,
             hyper_noise_scale=False,
         )
 

autocti/charge_injection/model/visualizer.py

@@ -165,12 +165,15 @@ def visualize_combined(
         paths: af.DirectoryPaths,
         instance: af.ModelInstance,
         during_analysis: bool,
+        quick_update: bool = False,
     ):
         if analyses is None:
             return
 
         fit_list = [
-            analysis.fit_via_instance_from(instance=instance) for analysis in analyses
+            # The factor graph passes one instance per analysis factor.
+            analysis.fit_via_instance_from(instance=instance_single)
+            for analysis, instance_single in zip(analyses, instance)
         ]
 
         fpr_value_list = [fit.dataset.fpr_value for fit in fit_list]
@@ -180,7 +183,9 @@ def visualize_combined(
             fpr_value_list=fpr_value_list,
         )
 
-        region_list = analyses[0].region_list_from(model=instance)
+        # The factor graph passes one instance per analysis factor; the region
+        # list is derived from the first (the CTI model is shared across factors).
+        region_list = analyses[0].region_list_from(model=instance[0])
 
         visualizer = PlotterImagingCI(image_path=paths.image_path)
         visualizer.fit_combined(fit_list=fit_list, during_analysis=during_analysis)
@@ -193,9 +198,9 @@ def visualize_combined(
         if analyses[0].dataset_full is not None:
             fit_full_list = [
                 analysis.fit_via_instance_and_dataset_from(
-                    instance=instance, dataset=analysis.dataset_full
+                    instance=instance_single, dataset=analysis.dataset_full
                 )
-                for analysis in analyses
+                for analysis, instance_single in zip(analyses, instance)
             ]
 
             fit_full_list = analyses[0].in_ascending_fpr_order_from(

autocti/charge_injection/plot/fit_ci_plots.py

@@ -226,6 +226,9 @@ def subplot_fit_list(
     _pf = (lambda t: f"{title_prefix.rstrip()} {t}") if title_prefix else (lambda t: t)
 
     n = len(fit_list)
+    if n == 0:
+        raise ValueError("An empty list was passed to a *_list plot function.")
+
     cols = min(n, 3)
     rows = (n + cols - 1) // cols
 
@@ -279,6 +282,9 @@ def subplot_fit_region_list(
         output_format = output_format[0]
 
     n = len(fit_list)
+    if n == 0:
+        raise ValueError("An empty list was passed to a *_list plot function.")
+
     cols = min(n, 3)
     rows = (n + cols - 1) // cols
 

autocti/charge_injection/plot/imaging_ci_plots.py

@@ -328,6 +328,9 @@ def subplot_data_region_list(
         output_format = output_format[0]
 
     n = len(dataset_list)
+    if n == 0:
+        raise ValueError("An empty list was passed to a *_list plot function.")
+
     cols = min(n, 3)
     rows = (n + cols - 1) // cols
 

autocti/dataset_1d/dataset_1d/dataset_1d.py

@@ -182,7 +182,9 @@ def output_to_fits(
             exception is raised.
         """
         fitsable.output_to_fits(
-            values=np.asarray(self.data.native), file_path=data_path, overwrite=overwrite
+            values=np.asarray(self.data.native),
+            file_path=data_path,
+            overwrite=overwrite,
         )
         fitsable.output_to_fits(
             values=np.asarray(self.noise_map.native),

autocti/dataset_1d/model/visualizer.py

@@ -139,12 +139,15 @@ def visualize_combined(
         paths: af.DirectoryPaths,
         instance: af.ModelInstance,
         during_analysis: bool,
+        quick_update: bool = False,
     ):
         if analyses is None:
             return
 
         fit_list = [
-            analysis.fit_via_instance_from(instance=instance) for analysis in analyses
+            # The factor graph passes one instance per analysis factor.
+            analysis.fit_via_instance_from(instance=instance_single)
+            for analysis, instance_single in zip(analyses, instance)
         ]
 
         fpr_value_list = [fit.dataset.fpr_value for fit in fit_list]
@@ -167,9 +170,9 @@ def visualize_combined(
         if analyses[0].dataset_full is not None:
             fit_full_list = [
                 analysis.fit_via_instance_and_dataset_from(
-                    instance=instance, dataset=analysis.dataset_full
+                    instance=instance_single, dataset=analysis.dataset_full
                 )
-                for analysis in analyses
+                for analysis, instance_single in zip(analyses, instance)
             ]
 
             fit_full_list = analyses[0].in_ascending_fpr_order_from(

autocti/dataset_1d/plot/dataset_1d_plots.py

@@ -141,6 +141,9 @@ def subplot_dataset_list(
     suffix = f"_{region}" if region is not None else ""
 
     n = len(dataset_list)
+    if n == 0:
+        raise ValueError("An empty list was passed to a *_list plot function.")
+
     cols = min(n, 3)
     rows = (n + cols - 1) // cols
 

autocti/dataset_1d/plot/fit_plots.py

@@ -163,6 +163,9 @@ def subplot_fit_list(
     suffix = f"_{region}" if region is not None else ""
 
     n = len(fit_list)
+    if n == 0:
+        raise ValueError("An empty list was passed to a *_list plot function.")
+
     cols = min(n, 3)
     rows = (n + cols - 1) // cols
 

autocti/extract/two_d/abstract.py

@@ -498,10 +498,10 @@ def add_gaussian_noise_to(
         array = array.native
 
         for arr, region in zip(array_2d_list, region_list):
-            array[
-                region.y0 : region.y1, region.x0 : region.x1
-            ] = aa.preprocess.data_with_gaussian_noise_added(
-                data=arr, sigma=noise_sigma, seed=noise_seed
+            array[region.y0 : region.y1, region.x0 : region.x1] = (
+                aa.preprocess.data_with_gaussian_noise_added(
+                    data=arr, sigma=noise_sigma, seed=noise_seed
+                )
             )
 
         return array

autocti/instruments/acs/image.py

@@ -102,12 +102,8 @@ def from_fits(
                 file_path=bias_file_path, hdu=hdu, do_not_scale_image_data=True
             )
 
-            header_sci_obj = fitsable.header_obj_from(
-                file_path=bias_file_path, hdu=0
-            )
-            header_hdu_obj = fitsable.header_obj_from(
-                file_path=bias_file_path, hdu=hdu
-            )
+            header_sci_obj = fitsable.header_obj_from(file_path=bias_file_path, hdu=0)
+            header_hdu_obj = fitsable.header_obj_from(file_path=bias_file_path, hdu=hdu)
 
             bias_header = HeaderACS(
                 header_sci_obj=header_sci_obj,

autocti/mask/mask_2d.py

@@ -227,9 +227,7 @@ def from_fits(
                 pixel_scales = (float(pixel_scales), float(pixel_scales))
 
         mask = cls.manual(
-            mask=fitsable.ndarray_via_fits_from(
-                file_path=file_path, hdu=hdu
-            ),
+            mask=fitsable.ndarray_via_fits_from(file_path=file_path, hdu=hdu),
             pixel_scales=pixel_scales,
             origin=origin,
         )

autocti/model/result.py

@@ -1,36 +1,49 @@
-from autofit.non_linear import result
-
-
-class Result(result.Result):
-    def __init__(
-        self,
-        samples_summary,
-        paths=None,
-        samples=None,
-        analysis=None,
-        search_internal=None,
-    ):
-        """
-        The result of a phase
-        """
-        super().__init__(
-            samples_summary=samples_summary,
-            paths=paths,
-            samples=samples,
-            search_internal=search_internal,
-            analysis=analysis,
-        )
-
-    @property
-    def clocker(self):
-        return self.analysis.clocker
-
-
-class ResultDataset(Result):
-    @property
-    def max_log_likelihood_fit(self):
-        return self.analysis.fit_via_instance_from(instance=self.instance)
-
-    @property
-    def mask(self):
-        return self.max_log_likelihood_fit.mask
+from autofit.non_linear import result
+
+
+class Result(result.Result):
+    def __init__(
+        self,
+        samples_summary,
+        paths=None,
+        samples=None,
+        analysis=None,
+        search_internal=None,
+    ):
+        """
+        The result of a phase
+        """
+        super().__init__(
+            samples_summary=samples_summary,
+            paths=paths,
+            samples=samples,
+            search_internal=search_internal,
+            analysis=analysis,
+        )
+
+    @property
+    def analysis_unwrapped(self):
+        """
+        The CTI analysis this result was inferred from.
+
+        A multi-dataset fit via a factor graph gives each child result an
+        `AnalysisFactor` wrapper as its analysis, which does not delegate
+        attribute access to the analysis it wraps — so it is unwrapped here
+        (the same unwrap PyAutoFit performs when dispatching combined
+        visualization).
+        """
+        return getattr(self.analysis, "analysis", self.analysis)
+
+    @property
+    def clocker(self):
+        return self.analysis_unwrapped.clocker
+
+
+class ResultDataset(Result):
+    @property
+    def max_log_likelihood_fit(self):
+        return self.analysis_unwrapped.fit_via_instance_from(instance=self.instance)
+
+    @property
+    def mask(self):
+        return self.max_log_likelihood_fit.mask

autocti/util/plot_utils.py

@@ -128,16 +128,23 @@ def fpr_mask_from(dataset) -> Mask2D:
         pixel_scales=dataset.pixel_scales,
     )
 
+    # A layout may legitimately lack prescan / overscan regions (e.g. after the
+    # dataset is trimmed via `apply_settings`), in which case there is nothing
+    # to mask for that region.
     serial_prescan = dataset.layout.extract.serial_prescan.serial_prescan
-    fpr_mask[
-        serial_prescan.y0 : serial_prescan.y1, serial_prescan.x0 : serial_prescan.x1
-    ] = True
+
+    if serial_prescan is not None:
+        fpr_mask[
+            serial_prescan.y0 : serial_prescan.y1, serial_prescan.x0 : serial_prescan.x1
+        ] = True
 
     serial_overscan = dataset.layout.extract.serial_overscan.serial_overscan
-    fpr_mask[
-        serial_overscan.y0 : serial_overscan.y1,
-        serial_overscan.x0 : serial_overscan.x1,
-    ] = True
+
+    if serial_overscan is not None:
+        fpr_mask[
+            serial_overscan.y0 : serial_overscan.y1,
+            serial_overscan.x0 : serial_overscan.x1,
+        ] = True
 
     return fpr_mask