add support to load ilastik h5 flavors

volumina/__main__.py

@@ -1,17 +1,26 @@
 import argparse
 import contextlib
+import json
 import signal
 import sys
+from pathlib import Path
+from typing import Tuple, Union
 
+import h5py
 import numpy
 import xarray
 from PyQt5.QtWidgets import QApplication
 
 from volumina import __version__
 from volumina.api import Viewer
 from volumina.colortables import default16_new
-from volumina.pixelpipeline.datasources import ArraySinkSource, ArraySource
 from volumina.layer import ColortableLayer, GrayscaleLayer
+from volumina.pixelpipeline.datasources import ArraySinkSource, ArraySource
+
+
+class _Suffixes:
+    npy = [".npy"]
+    h5 = [".h5", ".hdf5", ".hdf", ".ilp"]
 
 
 @contextlib.contextmanager
@@ -48,9 +57,12 @@
         usage="",
         epilog="",
     )
-    p.add_argument("image", help="Path to .npy image")
+    p.add_argument("image", help="Path to [.npy, .h5] image (in case of h5 with internal path)")
     p.add_argument(
-        "--axistags", help="Strings describing axes in image. Valid values: 'tzyxc'", type=axiorder_type, required=True
+        "--axistags",
+        help="Strings describing axes in image. Valid values: 'tzyxc'. Required for `.npy` files.",
+        type=axiorder_type,
+        required=False,
     )
     p.add_argument("--version", action="version", version=__version__)
 
@@ -67,16 +79,93 @@
     return tagged_data_5d.data
 
 
+def is_npy(path: Path) -> bool:
+    if path.suffix in _Suffixes.npy:
+        return True
+
+    return False
+
+
+def _external_internal_h5(path: Path) -> Tuple[Path, Path]:
+    if path.suffix in _Suffixes.h5:
+        external_path = path
+        # no internal path given!
+        # guess if there is only one ds in file:
+        with h5py.File(external_path, "r") as f:
+            if len(f) == 1:
+                internal_path = next(iter(f.keys()))
+            else:
+                raise ValueError("Could not determine internal path.")
+
+    else:
+        external_path = next(iter(a for a in path.parents if a.suffix in _Suffixes.h5), None)
+
+        if not external_path:
+            raise ValueError("Could not determine external path.")
+
+        internal_path = path.relative_to(external_path)
+
+    return external_path, internal_path
+
+
+def is_h5(path: Path) -> bool:
+    try:
+        _ = _external_internal_h5(path)
+    except ValueError:
+        return False
+
+    return True
+
+
+def load(data_path: str) -> Tuple[numpy.ndarray, Union[str, None]]:
+    p = Path(data_path)
+    if is_npy(p):
+        return load_npy(p), None
+    elif is_h5(p):
+        return load_h5(p)
+    else:
+        raise NotImplementedError("Unsupported file format - sorry.")
+
+
+def load_npy(data_path: Path) -> numpy.ndarray:
+    return numpy.load(data_path)
+
+
+def determine_h5_axistags(ds: h5py.Dataset) -> str:
+    if ds.dims:
+        return "".join([dim.label for dim in ds.dims])
+    elif "axistags" in ds.attrs:
+        return "".join([ax["key"] for ax in json.loads(ds.attrs["axistags"])["axes"]])
+
+    return ""
+
+
+def load_h5(data_path: Path) -> Tuple[numpy.ndarray, str]:
+    external_path, internal_path = _external_internal_h5(data_path)
+    with h5py.File(external_path, "r") as f:
+        ds: h5py.Dataset = f[str(internal_path)]
+        axistags = determine_h5_axistags(ds)
+        data = ds[()]
+        return data, axistags
+
+
 def main():
     args = parse_args()
-    data = numpy.load(args.image)
-    reordered_data = reorder_to_volumina(data, args.axistags)
+    data, axistags = load(args.image)
+
+    if not (args.axistags or axistags):
+        print(f"Axistags required for file {args.image} with shape {data.shape}")
+        return 1
+
+    reordered_data = reorder_to_volumina(data, args.axistags or axistags)
 
     with volumina_viewer() as v:
         v.addGrayscaleLayer(reordered_data, name="raw")
         v.setWindowTitle(f"Volumina - {args.image}-{args.axistags}")
         v.showMaximized()
 
+    return 0
+
 
 if __name__ == "__main__":
-    main()
+    sys.exit(main())