New file writer image dimensions (cctbx#676)

New file writer image dimensions

Reported now as slow, fast i.e. correct rather than historical fast, slow
which we have always had to invert

Fixes cctbx#657

Much more sensible fix suggested by @biochem-fan - checks firmware version
using packaging.version as a sane comparitor. Thanks to @spmeisburger for pointing out errors.

Co-authored-by: Takanori Nakane <nakane.t@gmail.com>

newsfragments/676.bugfix

@@ -0,0 +1,6 @@
+Eiger support: following bug fix in DECTRIS file writer, invert the module 
+dimensions only for older firmware versions. See
+
+https://media.dectris.com/230203-Release_Notes-DECTRIS_EIGER2.pdf
+
+for reference (Corrected data size orientation)

src/dxtbx/format/FormatNXmxEigerFilewriter.py

@@ -4,6 +4,7 @@
 
 import h5py
 import nxmx
+from packaging import version
 
 from scitbx.array_family import flex
 
@@ -14,7 +15,6 @@
 
 
 class FormatNXmxEigerFilewriter(FormatNXmx):
-
     _cached_file_handle = None
 
     @staticmethod
@@ -49,14 +49,14 @@ def _get_nxmx(self, fh: h5py.File):
         if nxdetector.underload_value is None:
             nxdetector.underload_value = 0
 
-        # data_size is reversed - we should probably be more specific in when
-        # we do this, i.e. check data_size is in a list of known reversed
-        # values
-        known_safe = [
-            (1082, 1035),
-        ]
-        for module in nxdetector.modules:
-            if not tuple(module.data_size) in known_safe:
+        # older firmware versions had the detector dimensions inverted
+        fw_version_string = (
+            fh["/entry/instrument/detector/detectorSpecific/eiger_fw_version"][()]
+            .decode()
+            .replace("release-", "")
+        )
+        if version.parse("2022.1.2") > version.parse(fw_version_string):
+            for module in nxdetector.modules:
                 module.data_size = module.data_size[::-1]
         return nxmx_obj