From 5009e062a6f7d4e074cab0fcb42a780ac2b1d7d4 Mon Sep 17 00:00:00 2001 From: James Tocknell Date: Thu, 28 Dec 2017 20:55:55 +1100 Subject: [PATCH] FIX: Don't reorder compound types, breaks on numpy 1.14 --- h5py/h5t.pyx | 25 +++++++------------------ setup.py | 2 +- tox.ini | 4 ++-- 3 files changed, 10 insertions(+), 21 deletions(-) diff --git a/h5py/h5t.pyx b/h5py/h5t.pyx index cc2344e1..7445e9eb 100644 --- a/h5py/h5t.pyx +++ b/h5py/h5t.pyx @@ -1136,12 +1136,6 @@ cdef class TypeCompoundID(TypeCompositeID): else: if sys.version[0] == '3': field_names = [x.decode('utf8') for x in field_names] - if len(field_names) > 0: - collated_fields = zip(field_names, field_types, field_offsets) - ordered_fields = sorted( - collated_fields, key=operator.itemgetter(2)) - field_names, field_types, field_offsets = \ - map(list, zip(*ordered_fields)) typeobj = dtype({ 'names': field_names, 'formats': field_types, @@ -1458,8 +1452,7 @@ cdef TypeCompoundID _c_compound(dtype dt, int logical, int aligned): cdef dtype member_dt cdef size_t member_offset = 0 - cdef dict offsets = {} - cdef list fields = [] + cdef dict fields = {} # The challenge with correctly converting a numpy/h5py dtype to a HDF5 type # which is composed of subtypes has three aspects we must consider @@ -1468,19 +1461,14 @@ cdef TypeCompoundID _c_compound(dtype dt, int logical, int aligned): # 2. For correct round-tripping of aligned dtypes, we need to consider how # much padding we need by looking at the field offsets # 3. There is no requirement that the offsets be monotonically increasing - # (so we start by sorting the names as a function of increasing offset) # # The code below tries to cover these aspects - # Get offsets for each compound member - for name, field in dt.fields.items(): - offsets[name] = field[1] - # Build list of names, offsets, and types, sorted by increasing offset # (i.e. the position of the member in the struct) - for name in sorted(dt.names, key=offsets.__getitem__): + for name in sorted(dt.names, key=(lambda n: dt.fields[n][1])): field = dt.fields[name] - name = name.encode('utf8') if isinstance(name, unicode) else name + h5_name = name.encode('utf8') if isinstance(name, unicode) else name # Get HDF5 data types and set the offset for each member member_dt = field[0] @@ -1489,7 +1477,7 @@ cdef TypeCompoundID _c_compound(dtype dt, int logical, int aligned): if aligned and (member_offset > field[1] or member_dt.itemsize != member_type.get_size()): raise TypeError("Enforced alignment not compatible with HDF5 type") - fields.append((name, member_offset, member_type)) + fields[name] = (h5_name, member_offset, member_type) # Update member offset based on the HDF5 type size member_offset += member_type.get_size() @@ -1500,8 +1488,9 @@ cdef TypeCompoundID _c_compound(dtype dt, int logical, int aligned): # Create compound with the necessary size, and insert its members tid = H5Tcreate(H5T_COMPOUND, member_offset) - for (name, member_offset, member_type) in fields: - H5Tinsert(tid, name, member_offset, member_type.id) + for name in dt.names: + h5_name, member_offset, member_type = fields[name] + H5Tinsert(tid, h5_name, member_offset, member_type.id) return TypeCompoundID(tid) diff --git a/setup.py b/setup.py index ec2a78a7..bbb086f6 100755 --- a/setup.py +++ b/setup.py @@ -32,7 +32,7 @@ # these are required to build h5py # RUN_REQUIRES is included as setup.py test needs RUN_REQUIRES for testing # RUN_REQUIRES can be removed when setup.py test is removed -SETUP_REQUIRES = RUN_REQUIRES + [NUMPY_DEP, 'Cython>=0.19', 'pkgconfig'] +SETUP_REQUIRES = RUN_REQUIRES + [NUMPY_DEP, 'Cython>=0.23', 'pkgconfig'] # Needed to avoid trying to install numpy/cython on pythons which the latest # versions don't support