Enforce sorting of FOVs for ROI table creation

src/fractal_faim_ipa/roi_tables.py

@@ -1,3 +1,5 @@
+import re
+
 import anndata as ad
 import numpy as np
 import pandas as pd
@@ -76,7 +78,12 @@ def create_fov_ROI_table(
     min_z = tile.position.z * pixel_size_zyx[0]
     max_z = (tile.position.z + 1) * pixel_size_zyx[0]
     fov_counter = 1
-    for tile in tiles:
+    # Initial robust sort
+    sorted_tiles = sorted(tiles)
+    # Sort again by trying to find site information as usually contained in
+    # MD filenames
+    sorted_tiles = sorted(sorted_tiles, key=_extract_fov_sort_key)
+    for tile in sorted_tiles:
         z_start = tile.position.z * pixel_size_zyx[0]
         z_end = (tile.position.z + 1) * pixel_size_zyx[0]
         if z_start < min_z:
@@ -105,3 +112,13 @@ def create_fov_ROI_table(
     # Cast the values to float to avoid anndata type issues
     roi_table = roi_table.astype(np.float32)
     return ad.AnnData(roi_table)
+
+
+def _extract_fov_sort_key(tile):
+    """Extract the FOV site information from an MD filename."""
+    filename = tile.path.split("/")[-1]
+    match = re.search(r"_s(\d+)_", filename)
+    if match:
+        return int(match.group(1)), filename
+    # Use a large number to push files without 's' to the end
+    return float("inf"), filename

tests/test_roi_tables.py

@@ -5,7 +5,7 @@
 import pytest
 from faim_ipa.hcs.acquisition import TileAlignmentOptions
 from fractal_faim_ipa.md_converter_utils import ModeEnum
-from fractal_faim_ipa.roi_tables import create_ROI_tables
+from fractal_faim_ipa.roi_tables import _extract_fov_sort_key, create_ROI_tables
 
 
 @pytest.mark.parametrize(
@@ -128,3 +128,25 @@ def test_roi_table_overlaps(alignment):
             target_values_well[alignment],
         )
     )
+
+
+def test_roi_sorting():
+    ROOT_DIR = Path(__file__).parent
+    image_dir = str(join(ROOT_DIR.parent, "resources", "Projection-Mix"))
+
+    mode = ModeEnum("MD Stack Acquisition")
+    plate_acquisition = mode.get_plate_acquisition(
+        acquisition_dir=image_dir,
+        alignment=TileAlignmentOptions.GRID,
+    )
+    well_acquisition = plate_acquisition.get_well_acquisitions(selection=["E07"])[0]
+    tiles = well_acquisition.get_tiles()
+    sorted_tiles = sorted(tiles, key=_extract_fov_sort_key)
+    assert (
+        sorted_tiles[0].path.split("/")[-1]
+        == "Projection-Mix_E07_s1_w1091EB8A5-272A-466D-B8A0-7547C6BA392B.tif"
+    )
+    assert (
+        sorted_tiles[-1].path.split("/")[-1]
+        == "Projection-Mix_E07_s2_w4F95A8A9F-0939-47C2-8D3E-F6E91AF0C4ED.tif"
+    )