fix(convert): 🐛 say which setting is missing for a glob input

A glob input without sequence_pattern fell through to bioio, which reads a
single file and could only report "BioImage does not support the image:
'.../*tif'" -- naming the path rather than the config key that was missing.

to_ome_zarr now recognises a multi-file input and names sequence_pattern,
with an example regex. run_multi checks it during its pre-flight too, so a
cluster run fails before submitting anything instead of after a queue wait
and a conversion attempt.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

Author: lguerard

src/patchworks/plugins/ome_zarr.py

@@ -45,6 +45,7 @@
 
 from __future__ import annotations
 
+import glob
 import logging
 import math
 from concurrent.futures import ThreadPoolExecutor
@@ -1201,6 +1202,21 @@ def _to_dask(
     if sequence_pattern is not None:
         arr, detected, ps = _open_tiff_sequence(path, sequence_pattern)
         return arr, axes or detected, ps
+    if any(ch in path for ch in "*?[") or (
+        not Path(path).exists() and glob.glob(path)
+    ):
+        # Without sequence_pattern this would be handed to bioio, which reads
+        # single files and reports only that it cannot handle the "format" --
+        # pointing at the glob rather than the missing setting.
+        raise ValueError(
+            f"{path!r} looks like a glob over several files, but "
+            "sequence_pattern is not set, so there is no way to tell which "
+            "part of each filename is Z, C or T.\n"
+            "Set it to a regex with named groups, e.g.\n"
+            "    sequence_pattern: '_Z(?P<Z>\\d+)_C(?P<C>\\d+)_V\\d+'\n"
+            "for files like sample_Z001_C0_V0.tif. Pass a single file "
+            "instead if you did not mean a sequence."
+        )
     if path.endswith(".zarr"):
         arr = load_ome_zarr(source, channel=None)
         ax = axes or _default_axes(arr.ndim)

tests/test_ome_zarr.py

@@ -399,3 +399,19 @@ def test_sharding(tmp_path):
     assert (
         getattr(_zarr.open_array(f"{out3}/0", mode="r"), "shards", None) is None
     )
+
+
+def test_glob_without_sequence_pattern_says_so(tmp_path):
+    """A glob input must name the missing setting, not blame the format.
+
+    Without sequence_pattern the glob fell through to bioio, which reads one
+    file and could only report "does not support the image" -- pointing at the
+    path instead of at the config key that was missing.
+    """
+    with pytest.raises(ValueError, match="sequence_pattern is not set"):
+        to_ome_zarr(str(tmp_path / "*.tif"), tmp_path / "out.zarr")
+
+    # A real single file must still reach the normal readers.
+    with pytest.raises(Exception) as exc:
+        to_ome_zarr(str(tmp_path / "scan.ims"), tmp_path / "o2.zarr")
+    assert "sequence_pattern" not in str(exc.value)

workflow/scripts/run_multi.py

@@ -114,6 +114,18 @@ def _spread(key):
                 f"share a chunk layout; got {_spread(key)}"
             )
 
+    for path, cfg in zip(paths, cfgs):
+        source = str(cfg.get("input", ""))
+        if any(ch in source for ch in "*?[") and not cfg.get(
+            "sequence_pattern"
+        ):
+            problems.append(
+                f"{path.name}: input {source!r} is a glob over several files "
+                "but sequence_pattern is unset, so nothing says which part of "
+                "each filename is Z/C/T. Set e.g. "
+                r"sequence_pattern: '_Z(?P<Z>\d+)_C(?P<C>\d+)_V\d+'"
+            )
+
     names = [cfg.get("label_name") for cfg in cfgs]
     duplicates = {n for n in names if names.count(n) > 1}
     if duplicates: