diff options
Diffstat (limited to 'numpy/lib/recfunctions.py')
| -rw-r--r-- | numpy/lib/recfunctions.py | 152 |
1 files changed, 91 insertions, 61 deletions
diff --git a/numpy/lib/recfunctions.py b/numpy/lib/recfunctions.py index fabb509ab..40060b41a 100644 --- a/numpy/lib/recfunctions.py +++ b/numpy/lib/recfunctions.py @@ -26,10 +26,13 @@ _check_fill_value = np.ma.core._check_fill_value __all__ = [ - 'append_fields', 'drop_fields', 'find_duplicates', - 'get_fieldstructure', 'join_by', 'merge_arrays', - 'rec_append_fields', 'rec_drop_fields', 'rec_join', - 'recursive_fill_fields', 'rename_fields', 'stack_arrays', + 'append_fields', 'apply_along_fields', 'assign_fields_by_name', + 'drop_fields', 'find_duplicates', 'flatten_descr', + 'get_fieldstructure', 'get_names', 'get_names_flat', + 'join_by', 'merge_arrays', 'rec_append_fields', + 'rec_drop_fields', 'rec_join', 'recursive_fill_fields', + 'rename_fields', 'repack_fields', 'require_fields', + 'stack_arrays', 'structured_to_unstructured', 'unstructured_to_structured', ] @@ -69,14 +72,14 @@ def recursive_fill_fields(input, output): current = input[field] except ValueError: continue - if current.dtype.names: + if current.dtype.names is not None: recursive_fill_fields(current, output[field]) else: output[field][:len(current)] = current return output -def get_fieldspec(dtype): +def _get_fieldspec(dtype): """ Produce a list of name/dtype pairs corresponding to the dtype fields @@ -91,7 +94,7 @@ def get_fieldspec(dtype): >>> dt = np.dtype([(('a', 'A'), np.int64), ('b', np.double, 3)]) >>> dt.descr [(('a', 'A'), '<i8'), ('b', '<f8', (3,))] - >>> get_fieldspec(dt) + >>> _get_fieldspec(dt) [(('a', 'A'), dtype('int64')), ('b', dtype(('<f8', (3,))))] """ @@ -136,11 +139,11 @@ def get_names(adtype): names = adtype.names for name in names: current = adtype[name] - if current.names: + if current.names is not None: listnames.append((name, tuple(get_names(current)))) else: listnames.append(name) - return tuple(listnames) or None + return tuple(listnames) def get_names_flat(adtype): @@ -173,9 +176,9 @@ def get_names_flat(adtype): for name in names: listnames.append(name) current = adtype[name] - if current.names: + if current.names is not None: listnames.extend(get_names_flat(current)) - return tuple(listnames) or None + return tuple(listnames) def flatten_descr(ndtype): @@ -204,12 +207,7 @@ def flatten_descr(ndtype): return tuple(descr) -def _zip_dtype_dispatcher(seqarrays, flatten=None): - return seqarrays - - -@array_function_dispatch(_zip_dtype_dispatcher) -def zip_dtype(seqarrays, flatten=False): +def _zip_dtype(seqarrays, flatten=False): newdtype = [] if flatten: for a in seqarrays: @@ -217,16 +215,15 @@ def zip_dtype(seqarrays, flatten=False): else: for a in seqarrays: current = a.dtype - if current.names and len(current.names) <= 1: - # special case - dtypes of 0 or 1 field are flattened - newdtype.extend(get_fieldspec(current)) + if current.names is not None and len(current.names) == 1: + # special case - dtypes of 1 field are flattened + newdtype.extend(_get_fieldspec(current)) else: newdtype.append(('', current)) return np.dtype(newdtype) -@array_function_dispatch(_zip_dtype_dispatcher) -def zip_descr(seqarrays, flatten=False): +def _zip_descr(seqarrays, flatten=False): """ Combine the dtype description of a series of arrays. @@ -237,7 +234,7 @@ def zip_descr(seqarrays, flatten=False): flatten : {boolean}, optional Whether to collapse nested descriptions. """ - return zip_dtype(seqarrays, flatten=flatten).descr + return _zip_dtype(seqarrays, flatten=flatten).descr def get_fieldstructure(adtype, lastname=None, parents=None,): @@ -271,7 +268,7 @@ def get_fieldstructure(adtype, lastname=None, parents=None,): names = adtype.names for name in names: current = adtype[name] - if current.names: + if current.names is not None: if lastname: parents[name] = [lastname, ] else: @@ -284,7 +281,7 @@ def get_fieldstructure(adtype, lastname=None, parents=None,): elif lastname: lastparent = [lastname, ] parents[name] = lastparent or [] - return parents or None + return parents def _izip_fields_flat(iterable): @@ -318,12 +315,7 @@ def _izip_fields(iterable): yield element -def _izip_records_dispatcher(seqarrays, fill_value=None, flatten=None): - return seqarrays - - -@array_function_dispatch(_izip_records_dispatcher) -def izip_records(seqarrays, fill_value=None, flatten=True): +def _izip_records(seqarrays, fill_value=None, flatten=True): """ Returns an iterator of concatenated items from a sequence of arrays. @@ -443,9 +435,9 @@ def merge_arrays(seqarrays, fill_value=-1, flatten=False, if isinstance(seqarrays, (ndarray, np.void)): seqdtype = seqarrays.dtype # Make sure we have named fields - if not seqdtype.names: + if seqdtype.names is None: seqdtype = np.dtype([('', seqdtype)]) - if not flatten or zip_dtype((seqarrays,), flatten=True) == seqdtype: + if not flatten or _zip_dtype((seqarrays,), flatten=True) == seqdtype: # Minimal processing needed: just make sure everythng's a-ok seqarrays = seqarrays.ravel() # Find what type of array we must return @@ -468,7 +460,7 @@ def merge_arrays(seqarrays, fill_value=-1, flatten=False, sizes = tuple(a.size for a in seqarrays) maxlength = max(sizes) # Get the dtype of the output (flattening if needed) - newdtype = zip_dtype(seqarrays, flatten=flatten) + newdtype = _zip_dtype(seqarrays, flatten=flatten) # Initialize the sequences for data and mask seqdata = [] seqmask = [] @@ -496,9 +488,9 @@ def merge_arrays(seqarrays, fill_value=-1, flatten=False, seqdata.append(itertools.chain(data, [fval] * nbmissing)) seqmask.append(itertools.chain(mask, [fmsk] * nbmissing)) # Create an iterator for the data - data = tuple(izip_records(seqdata, flatten=flatten)) + data = tuple(_izip_records(seqdata, flatten=flatten)) output = ma.array(np.fromiter(data, dtype=newdtype, count=maxlength), - mask=list(izip_records(seqmask, flatten=flatten))) + mask=list(_izip_records(seqmask, flatten=flatten))) if asrecarray: output = output.view(MaskedRecords) else: @@ -516,7 +508,7 @@ def merge_arrays(seqarrays, fill_value=-1, flatten=False, else: fval = None seqdata.append(itertools.chain(data, [fval] * nbmissing)) - output = np.fromiter(tuple(izip_records(seqdata, flatten=flatten)), + output = np.fromiter(tuple(_izip_records(seqdata, flatten=flatten)), dtype=newdtype, count=maxlength) if asrecarray: output = output.view(recarray) @@ -661,7 +653,7 @@ def rename_fields(base, namemapper): for name in ndtype.names: newname = namemapper.get(name, name) current = ndtype[name] - if current.names: + if current.names is not None: newdtype.append( (newname, _recursive_rename_fields(current, namemapper)) ) @@ -743,7 +735,7 @@ def append_fields(base, names, data, dtypes=None, # output = ma.masked_all( max(len(base), len(data)), - dtype=get_fieldspec(base.dtype) + get_fieldspec(data.dtype)) + dtype=_get_fieldspec(base.dtype) + _get_fieldspec(data.dtype)) output = recursive_fill_fields(base, output) output = recursive_fill_fields(data, output) # @@ -834,17 +826,18 @@ def repack_fields(a, align=False, recurse=False): Examples -------- + >>> from numpy.lib import recfunctions as rfn >>> def print_offsets(d): ... print("offsets:", [d.fields[name][1] for name in d.names]) ... print("itemsize:", d.itemsize) ... - >>> dt = np.dtype('u1,<i4,<f4', align=True) + >>> dt = np.dtype('u1, <i8, <f8', align=True) >>> dt dtype({'names':['f0','f1','f2'], 'formats':['u1','<i8','<f8'], 'offsets':[0,8,16], 'itemsize':24}, align=True) >>> print_offsets(dt) offsets: [0, 8, 16] itemsize: 24 - >>> packed_dt = repack_fields(dt) + >>> packed_dt = rfn.repack_fields(dt) >>> packed_dt dtype([('f0', 'u1'), ('f1', '<i8'), ('f2', '<f8')]) >>> print_offsets(packed_dt) @@ -881,16 +874,35 @@ def _get_fields_and_offsets(dt, offset=0): scalar fields in the dtype "dt", including nested fields, in left to right order. """ + + # counts up elements in subarrays, including nested subarrays, and returns + # base dtype and count + def count_elem(dt): + count = 1 + while dt.shape != (): + for size in dt.shape: + count *= size + dt = dt.base + return dt, count + fields = [] for name in dt.names: field = dt.fields[name] - if field[0].names is None: - count = 1 - for size in field[0].shape: - count *= size - fields.append((field[0], count, field[1] + offset)) + f_dt, f_offset = field[0], field[1] + f_dt, n = count_elem(f_dt) + + if f_dt.names is None: + fields.append((np.dtype((f_dt, (n,))), n, f_offset + offset)) else: - fields.extend(_get_fields_and_offsets(field[0], field[1] + offset)) + subfields = _get_fields_and_offsets(f_dt, f_offset + offset) + size = f_dt.itemsize + + for i in range(n): + if i == 0: + # optimization: avoid list comprehension if no subarray + fields.extend(subfields) + else: + fields.extend([(d, c, o + i*size) for d, c, o in subfields]) return fields @@ -932,12 +944,13 @@ def structured_to_unstructured(arr, dtype=None, copy=False, casting='unsafe'): Examples -------- + >>> from numpy.lib import recfunctions as rfn >>> a = np.zeros(4, dtype=[('a', 'i4'), ('b', 'f4,u2'), ('c', 'f4', 2)]) >>> a array([(0, (0., 0), [0., 0.]), (0, (0., 0), [0., 0.]), (0, (0., 0), [0., 0.]), (0, (0., 0), [0., 0.])], dtype=[('a', '<i4'), ('b', [('f0', '<f4'), ('f1', '<u2')]), ('c', '<f4', (2,))]) - >>> structured_to_unstructured(arr) + >>> rfn.structured_to_unstructured(a) array([[0., 0., 0., 0., 0.], [0., 0., 0., 0., 0.], [0., 0., 0., 0., 0.], @@ -945,7 +958,7 @@ def structured_to_unstructured(arr, dtype=None, copy=False, casting='unsafe'): >>> b = np.array([(1, 2, 5), (4, 5, 7), (7, 8 ,11), (10, 11, 12)], ... dtype=[('x', 'i4'), ('y', 'f4'), ('z', 'f8')]) - >>> np.mean(structured_to_unstructured(b[['x', 'z']]), axis=-1) + >>> np.mean(rfn.structured_to_unstructured(b[['x', 'z']]), axis=-1) array([ 3. , 5.5, 9. , 11. ]) """ @@ -954,6 +967,12 @@ def structured_to_unstructured(arr, dtype=None, copy=False, casting='unsafe'): fields = _get_fields_and_offsets(arr.dtype) n_fields = len(fields) + if n_fields == 0 and dtype is None: + raise ValueError("arr has no fields. Unable to guess dtype") + elif n_fields == 0: + # too many bugs elsewhere for this to work now + raise NotImplementedError("arr with no fields is not supported") + dts, counts, offsets = zip(*fields) names = ['f{}'.format(n) for n in range(n_fields)] @@ -982,6 +1001,7 @@ def structured_to_unstructured(arr, dtype=None, copy=False, casting='unsafe'): # finally is it safe to view the packed fields as the unstructured type return arr.view((out_dtype, (sum(counts),))) + def _unstructured_to_structured_dispatcher(arr, dtype=None, names=None, align=None, copy=None, casting=None): return (arr,) @@ -1027,6 +1047,7 @@ def unstructured_to_structured(arr, dtype=None, names=None, align=False, Examples -------- + >>> from numpy.lib import recfunctions as rfn >>> dt = np.dtype([('a', 'i4'), ('b', 'f4,u2'), ('c', 'f4', 2)]) >>> a = np.arange(20).reshape((4,5)) >>> a @@ -1034,7 +1055,7 @@ def unstructured_to_structured(arr, dtype=None, names=None, align=False, [ 5, 6, 7, 8, 9], [10, 11, 12, 13, 14], [15, 16, 17, 18, 19]]) - >>> unstructured_to_structured(a, dt) + >>> rfn.unstructured_to_structured(a, dt) array([( 0, ( 1., 2), [ 3., 4.]), ( 5, ( 6., 7), [ 8., 9.]), (10, (11., 12), [13., 14.]), (15, (16., 17), [18., 19.])], dtype=[('a', '<i4'), ('b', [('f0', '<f4'), ('f1', '<u2')]), ('c', '<f4', (2,))]) @@ -1043,6 +1064,9 @@ def unstructured_to_structured(arr, dtype=None, names=None, align=False, if arr.shape == (): raise ValueError('arr must have at least one dimension') n_elem = arr.shape[-1] + if n_elem == 0: + # too many bugs elsewhere for this to work now + raise NotImplementedError("last axis with size 0 is not supported") if dtype is None: if names is None: @@ -1055,7 +1079,11 @@ def unstructured_to_structured(arr, dtype=None, names=None, align=False, raise ValueError("don't supply both dtype and names") # sanity check of the input dtype fields = _get_fields_and_offsets(dtype) - dts, counts, offsets = zip(*fields) + if len(fields) == 0: + dts, counts, offsets = [], [], [] + else: + dts, counts, offsets = zip(*fields) + if n_elem != sum(counts): raise ValueError('The length of the last dimension of arr must ' 'be equal to the number of fields in dtype') @@ -1111,11 +1139,12 @@ def apply_along_fields(func, arr): Examples -------- + >>> from numpy.lib import recfunctions as rfn >>> b = np.array([(1, 2, 5), (4, 5, 7), (7, 8 ,11), (10, 11, 12)], ... dtype=[('x', 'i4'), ('y', 'f4'), ('z', 'f8')]) - >>> apply_along_fields(np.mean, b) + >>> rfn.apply_along_fields(np.mean, b) array([ 2.66666667, 5.33333333, 8.66666667, 11. ]) - >>> apply_along_fields(np.mean, b[['x', 'z']]) + >>> rfn.apply_along_fields(np.mean, b[['x', 'z']]) array([ 3. , 5.5, 9. , 11. ]) """ @@ -1200,14 +1229,15 @@ def require_fields(array, required_dtype): Examples -------- + >>> from numpy.lib import recfunctions as rfn >>> a = np.ones(4, dtype=[('a', 'i4'), ('b', 'f8'), ('c', 'u1')]) - >>> require_fields(a, [('b', 'f4'), ('c', 'u1')]) + >>> rfn.require_fields(a, [('b', 'f4'), ('c', 'u1')]) array([(1., 1), (1., 1), (1., 1), (1., 1)], dtype=[('b', '<f4'), ('c', 'u1')]) - >>> require_fields(a, [('b', 'f4'), ('newf', 'u1')]) + >>> rfn.require_fields(a, [('b', 'f4'), ('newf', 'u1')]) array([(1., 0), (1., 0), (1., 0), (1., 0)], dtype=[('b', '<f4'), ('newf', 'u1')]) - + """ out = np.empty(array.shape, dtype=required_dtype) assign_fields_by_name(out, array) @@ -1270,10 +1300,10 @@ def stack_arrays(arrays, defaults=None, usemask=True, asrecarray=False, fldnames = [d.names for d in ndtype] # dtype_l = ndtype[0] - newdescr = get_fieldspec(dtype_l) + newdescr = _get_fieldspec(dtype_l) names = [n for n, d in newdescr] for dtype_n in ndtype[1:]: - for fname, fdtype in get_fieldspec(dtype_n): + for fname, fdtype in _get_fieldspec(dtype_n): if fname not in names: newdescr.append((fname, fdtype)) names.append(fname) @@ -1496,15 +1526,15 @@ def join_by(key, r1, r2, jointype='inner', r1postfix='1', r2postfix='2', # # Build the new description of the output array ....... # Start with the key fields - ndtype = get_fieldspec(r1k.dtype) + ndtype = _get_fieldspec(r1k.dtype) # Add the fields from r1 - for fname, fdtype in get_fieldspec(r1.dtype): + for fname, fdtype in _get_fieldspec(r1.dtype): if fname not in key: ndtype.append((fname, fdtype)) # Add the fields from r2 - for fname, fdtype in get_fieldspec(r2.dtype): + for fname, fdtype in _get_fieldspec(r2.dtype): # Have we seen the current name already ? # we need to rebuild this list every time names = list(name for name, dtype in ndtype) |