Fix: update "One curve per result title" mode to respect ROIs in plot analysis

Fix #132

Author: PierreRaybaut

CHANGELOG.md

@@ -10,6 +10,9 @@ See DataLab [roadmap page](https://datalab-platform.com/en/contributing/roadmap.
 
 🛠️ Bug fixes:
 
+* Fixed [Issue #132](https://github.com/DataLab-Platform/DataLab/issues/132) - Plot analysis results: "One curve per result title" mode ignores ROIs
+  * Before this fix, the "One curve per result title" mode was ignoring ROIs, and was plotting the selected result for all objects (signals or images) without taking into account the ROI defined on the objects
+  * This is now fixed: the "One curve per result title" mode now takes into account the ROI defined on the objects, and plots the selected result for each object (signal or image) and for each ROI defined on the object
 * Fixed [Issue #128](https://github.com/DataLab-Platform/DataLab/issues/128) - Support long object titles in Signal and Image panels
 * Fixed [Issue #133](https://github.com/DataLab-Platform/DataLab/issues/133) - Remove specific analysis results from metadata clipboard during copy operation
 * Image ROI features:

cdl/core/gui/panel/base.py

@@ -1286,33 +1286,56 @@ class PlotResultParam(gds.DataSet):
 
                 i_yaxis = rdata.xlabels.index(param.yaxis)
                 if param.kind == "one_curve_per_title":
-                    # One curve per result title:
-                    for title, results in grouped_results.items():  # title
-                        x, y = [], []
-                        for index, result in enumerate(results):  # object
-                            if param.xaxis == "indices":
-                                x.append(index)
-                            else:
-                                i_xaxis = rdata.xlabels.index(param.xaxis)
-                                x.append(result.shown_array[0][i_xaxis])
-                            y.append(result.shown_array[0][i_yaxis])
-                        self.__add_result_signal(x, y, title, param.xaxis, param.yaxis)
+                    # One curve per ROI (if any) and per result title
+                    # ------------------------------------------------------------------
+                    # Begin by checking if all results have the same number of ROIs:
+                    # for simplicity, let's check the number of unique ROI indices.
+                    all_roi_indexes = [
+                        np.unique(result.array[:, 0]) for result in rdata.results
+                    ]
+                    # Check if all roi_indexes are the same:
+                    if len(set(map(tuple, all_roi_indexes))) > 1:
+                        QW.QMessageBox.warning(
+                            self,
+                            _("Plot results"),
+                            _(
+                                "All objects associated with results must have the "
+                                "same regions of interest (ROIs) to plot results "
+                                "together."
+                            ),
+                        )
+                        return
+                    for i_roi in all_roi_indexes[0]:
+                        roi_suffix = f"|ROI{int(i_roi+1)}" if i_roi >= 0 else ""
+                        for title, results in grouped_results.items():  # title
+                            x, y = [], []
+                            for index, result in enumerate(results):
+                                mask = result.array[:, 0] == i_roi
+                                if param.xaxis == "indices":
+                                    x.append(index)
+                                else:
+                                    i_xaxis = rdata.xlabels.index(param.xaxis)
+                                    x.append(float(result.shown_array[mask, i_xaxis]))
+                                y.append(float(result.shown_array[mask, i_yaxis]))
+                            self.__add_result_signal(
+                                x, y, f"{title}{roi_suffix}", param.xaxis, param.yaxis
+                            )
                 else:
-                    # One curve per result title, per object and per ROI:
+                    # One curve per result title, per object and per ROI
+                    # ------------------------------------------------------------------
                     for title, results in grouped_results.items():  # title
                         for index, result in enumerate(results):  # object
                             roi_idx = np.array(np.unique(result.array[:, 0]), dtype=int)
                             for i_roi in roi_idx:  # ROI
+                                roi_suffix = f"|ROI{int(i_roi+1)}" if i_roi >= 0 else ""
                                 mask = result.array[:, 0] == i_roi
                                 if param.xaxis == "indices":
                                     x = np.arange(result.array.shape[0])[mask]
                                 else:
                                     i_xaxis = rdata.xlabels.index(param.xaxis)
                                     x = result.shown_array[mask, i_xaxis]
                                 y = result.shown_array[mask, i_yaxis]
-                                stitle = f"{title} ({objs[index].short_id})"
-                                if len(roi_idx) > 1:
-                                    stitle += f"|ROI{i_roi}"
+                                stitle = f"{title} ({objs[index].short_id}){roi_suffix}"
                                 self.__add_result_signal(
                                     x, y, stitle, param.xaxis, param.yaxis
                                 )