Updatest to resolve conflicts with changes related to floating point …

…headers for indexing.

src/mdio/segy/_workers.py

@@ -45,7 +45,9 @@ def header_scan_worker(
         segy_endian: Endianness of the input SEG-Y. Rev.2 allows little endian
 
     Returns:
-        Numpy array of parsed headers for the current block.
+        dictionary with headers:  keys are the index names, values are numpy
+            arrays of parsed headers for the current block. Array is of type
+            byte_type with the exception of IBM32 which is mapped to FLOAT32.
 
     Raises:
         TypeError: if segy_path_or_handle is incorrect / unsupported.
@@ -59,7 +61,6 @@ def header_scan_worker(
             ignore_geometry=True,
             endian=segy_endian,
         ) as segy_handle:
-
             block_headers = [
                 segy_handle.header[trc_idx] for trc_idx in range(start, stop)
             ]
@@ -109,7 +110,8 @@ def header_scan_worker(
 
         out_dtype.append((name, native_dtype))
 
-    out_array = np.empty(n_traces, out_dtype)
+    # out_array = np.empty(n_traces, out_dtype)
+    out_array = {}
 
     # TODO: Add strict=True and remove noqa when minimum Python is 3.10
     for name, loc, type_ in zip(index_names, byte_locs, byte_types):  # noqa: B905

src/mdio/segy/parsers.py

@@ -94,7 +94,10 @@ def parse_trace_headers(
         progress_bar: Enable or disable progress bar. Default is True.
 
     Returns:
-        Numpy array of parsed trace headers.
+        dictionary with headers:  keys are the index names, values are numpy
+            arrays of parsed headers for the current block. Array is of type
+            byte_type with the exception of IBM32 which is mapped to FLOAT32.
+
     """
     trace_count = get_trace_count(segy_path, segy_endian)
     n_blocks = int(ceil(trace_count / block_size))
@@ -133,9 +136,13 @@ def parse_trace_headers(
 
         # This executes the lazy work.
         headers = list(lazy_work)
-
+    final_headers = {}
+    for header_name in index_names:
+        final_headers[header_name] = np.concatenate(
+            [header[header_name] for header in headers]
+        )
     # Merge blocks before return
-    return np.concatenate(headers)
+    return final_headers
 
 
 def parse_sample_axis(binary_header: dict) -> Dimension:

src/mdio/segy/utilities.py

@@ -126,11 +126,11 @@ def get_grid_plan(  # noqa:  C901
         # to prevent so many lookups
         if geom_type == GeometryTemplateType.STREAMER_B:
             for idx, cable in enumerate(unique_cables):
-                cable_idxs = np.where(index_headers[:, cable_idx] == cable)
+                cable_idxs = np.where(index_headers["cable"][:] == cable)
                 cc_min = cable_chan_min[idx]
                 # print(f"idx = {idx}  cable = {cable} cc_min={cc_min}")
-                index_headers[cable_idxs, chan_idx] = (
-                    index_headers[cable_idxs, chan_idx] - cc_min + 1
+                index_headers["channel"][cable_idxs] = (
+                    index_headers["channel"][cable_idxs] - cc_min + 1
                 )
 
     if "CalculateCable" in grid_overrides:
@@ -187,25 +187,23 @@ def qc_index_headers(
             cable_chan_max: Array containing the max channel number for each cable,
             geom_type:  Whether type a or b (wrapped or unwrapped chans)
     """
-    if trace_qc_count is None:
-        trace_qc_count = index_headers.shape[0]
-    if trace_qc_count > index_headers.shape[0]:
-        trace_qc_count = index_headers.shape[0]
-
     if "cable" in index_names and "channel" in index_names and "shot" in index_names:
-        cable_idx = index_names.index("cable")
-        channel_idx = index_names.index("channel")
+        if trace_qc_count is None:
+            trace_qc_count = index_headers["cable"].shape[0]
+        if trace_qc_count > index_headers["cable"].shape[0]:
+            trace_qc_count = index_headers["cable"].shape[0]
+
         # Find unique cable ids
-        unique_cables = np.sort(np.unique(index_headers[0:trace_qc_count, cable_idx]))
+        unique_cables = np.sort(np.unique(index_headers["cable"][0:trace_qc_count]))
 
         # Find channel min and max values for each cable
         cable_chan_min = np.empty(unique_cables.shape)
         cable_chan_max = np.empty(unique_cables.shape)
 
         for idx, cable in enumerate(unique_cables):
             my_chan = np.take(
-                index_headers[0:trace_qc_count, channel_idx],
-                np.where(index_headers[0:trace_qc_count, cable_idx] == cable),
+                index_headers["channel"][0:trace_qc_count],
+                np.where(index_headers["cable"][0:trace_qc_count] == cable),
             )
             cable_chan_min[idx] = np.min(my_chan)
             cable_chan_max[idx] = np.max(my_chan)

tests/integration/test_segy_import_export.py

@@ -82,7 +82,7 @@ def create_4d_segy(
 
 @pytest.mark.parametrize("header_locations", [(17, 137, 13)])
 @pytest.mark.parametrize("header_names", [("shot", "cable", "channel")])
-@pytest.mark.parametrize("header_lengths", [(4, 2, 4)])
+@pytest.mark.parametrize("header_types", [("int32", "int16", "int32")])
 @pytest.mark.parametrize("endian", ["big"])
 @pytest.mark.parametrize(
     "grid_overrides", [{"AutoChannelWrap": True, "AutoChannelTraceQC": 100000}, None]
@@ -97,7 +97,7 @@ def test_import_4d_segy(
         zarr_tmp,
         header_locations,
         header_names,
-        header_lengths,
+        header_types,
         endian,
         grid_overrides,
         chan_header_type,
@@ -121,7 +121,7 @@ def test_import_4d_segy(
             mdio_path_or_buffer=zarr_tmp.__str__(),
             index_bytes=header_locations,
             index_names=header_names,
-            index_lengths=header_lengths,
+            index_types=header_types,
             chunksize=(8, 2, 128, 1024),
             overwrite=True,
             endian=endian,