Extend igor.struct.Structure and .Field to support nesting.
authorW. Trevor King <wking@tremily.us>
Wed, 18 Jul 2012 20:34:05 +0000 (16:34 -0400)
committerW. Trevor King <wking@tremily.us>
Wed, 18 Jul 2012 20:34:05 +0000 (16:34 -0400)
igor/binarywave.py
igor/packed.py
igor/struct.py

index f9c4e18a2a87e681c1aff6eb61bab5ff11d04a70..4fa5e4a68a276eec0a3ed6a2bcd3b5743f4c20ae 100644 (file)
@@ -260,13 +260,13 @@ def load(filename, strict=True):
     try:
         BinHeaderCommon.set_byte_order('=')
         b = buffer(f.read(BinHeaderCommon.size))
-        version = BinHeaderCommon.unpack_dict_from(b)['version']
+        version = BinHeaderCommon.unpack_from(b)['version']
         needToReorderBytes = _need_to_reorder_bytes(version)
         byteOrder = _byte_order(needToReorderBytes)
 
         if needToReorderBytes:
             BinHeaderCommon.set_byte_order(byteOrder)
-            version = BinHeaderCommon.unpack_dict_from(b)['version']
+            version = BinHeaderCommon.unpack_from(b)['version']
         bin_struct,wave_struct,checkSumSize = _version_structs(
             version, byteOrder)
 
@@ -276,8 +276,8 @@ def load(filename, strict=True):
             raise ValueError(
                 ('This does not appear to be a valid Igor binary wave file.  '
                  'Error in checksum: should be 0, is {}.').format(c))
-        bin_info = bin_struct.unpack_dict_from(b)
-        wave_info = wave_struct.unpack_dict_from(b, offset=bin_struct.size)
+        bin_info = bin_struct.unpack_from(b)
+        wave_info = wave_struct.unpack_from(b, offset=bin_struct.size)
         if version in [1,2,3]:
             tail = 16  # 16 = size of wData field in WaveHeader2 structure
             waveDataSize = bin_info['wfmSize'] - wave_struct.size
index ebcbdfc4764ea96a26752649851944cea211c663..69d73536ee279b48f9574601132ec281e14ab170 100644 (file)
@@ -124,7 +124,7 @@ def load(filename, strict=True, ignore_unknown=True):
             b = buffer(f.read(PackedFileRecordHeader.size))
             if not b:
                 break
-            header = PackedFileRecordHeader.unpack_dict_from(b)
+            header = PackedFileRecordHeader.unpack_from(b)
             data = buffer(f.read(header['numDataBytes']))
             record_type = RECORD_TYPE.get(
                 header['recordType'] & PACKEDRECTYPE_MASK, UnknownRecord)
index d16ca8a4703bf1a7b87596ecc0ea144bfe2b03a3..32c2b9386af0ee62ef9c55bcdf0616b2bc14d442 100644 (file)
@@ -1,6 +1,12 @@
 # Copyright
 
-"Structure and Field classes for declaring structures "
+"""Structure and Field classes for declaring structures
+
+There are a few formats that can be used to represent the same data, a
+binary packed format with all the data in a buffer, a linearized
+format with each field in a single Python list, and a nested format
+with each field in a hierarchy of Python dictionaries.
+"""
 
 from __future__ import absolute_import
 import struct as _struct
@@ -14,21 +20,249 @@ _buffer = buffer  # save builtin buffer for clobbered situations
 class Field (object):
     """Represent a Structure field.
 
+    The format argument can be a format character from the ``struct``
+    documentation (e.g., ``c`` for ``char``, ``h`` for ``short``, ...)
+    or ``Structure`` instance (for building nested structures).
+
+    Examples
+    --------
+
+    >>> from pprint import pprint
+    >>> import numpy
+
+    Example of an unsigned short integer field:
+
+    >>> time = Field(
+    ...     'I', 'time', default=0, help='POSIX time')
+    >>> time.total_count
+    1
+    >>> list(time.pack_data(1))
+    [1]
+    >>> list(time.pack_item(2))
+    [2]
+    >>> time.unpack_data([3])
+    3
+    >>> time.unpack_item([4])
+    4
+
+    Example of a multi-dimensional float field:
+
+    >>> data = Field(
+    ...     'f', 'data', help='example data', count=(2,3,4))
+    >>> data.total_count
+    24
+    >>> list(data.indexes())  # doctest: +ELLIPSIS
+    [[0, 0, 0], [0, 0, 1], [0, 0, 2], [0, 0, 3], [0, 1, 0], ..., [1, 2, 3]]
+    >>> list(data.pack_data(
+    ...     [[[0, 1, 2, 3], [4, 5, 6, 7], [8, 9, 10, 11]],
+    ...      [[12, 13, 14, 15], [16, 17, 18, 19], [20, 21, 22, 23]]])
+    ...     )  # doctest: +ELLIPSIS
+    [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, ..., 19, 20, 21, 22, 23]
+    >>> list(data.pack_item(3))
+    [3]
+    >>> data.unpack_data(range(data.total_count))
+    array([[[ 0,  1,  2,  3],
+            [ 4,  5,  6,  7],
+            [ 8,  9, 10, 11]],
+    <BLANKLINE>
+           [[12, 13, 14, 15],
+            [16, 17, 18, 19],
+            [20, 21, 22, 23]]])
+    >>> data.unpack_item([3])
+    3
+
+    Example of a nested structure field:
+
+    >>> run = Structure('run', fields=[time, data])
+    >>> runs = Field(run, 'runs', help='pair of runs', count=2)
+    >>> runs.total_count  # = 2 * (1 + 24)
+    50
+    >>> data1 = numpy.arange(data.total_count).reshape(data.count)
+    >>> data2 = data1 + data.total_count
+    >>> list(runs.pack_data(
+    ...     [{'time': 100, 'data': data1},
+    ...      {'time': 101, 'data': data2}])
+    ...     )  # doctest: +ELLIPSIS
+    [100, 0, 1, 2, ..., 22, 23, 101, 24, 25, ..., 46, 47]
+    >>> list(runs.pack_item({'time': 100, 'data': data1})
+    ...     )  # doctest: +ELLIPSIS
+    [100, 0, 1, 2, ..., 22, 23]
+    >>> pprint(runs.unpack_data(range(runs.total_count)))
+    [{'data': array([[[ 1,  2,  3,  4],
+            [ 5,  6,  7,  8],
+            [ 9, 10, 11, 12]],
+    <BLANKLINE>
+           [[13, 14, 15, 16],
+            [17, 18, 19, 20],
+            [21, 22, 23, 24]]]),
+      'time': 0},
+     {'data': array([[[26, 27, 28, 29],
+            [30, 31, 32, 33],
+            [34, 35, 36, 37]],
+    <BLANKLINE>
+           [[38, 39, 40, 41],
+            [42, 43, 44, 45],
+            [46, 47, 48, 49]]]),
+      'time': 25}]
+    >>> pprint(runs.unpack_item(range(runs.structure_count)))
+    {'data': array([[[ 1,  2,  3,  4],
+            [ 5,  6,  7,  8],
+            [ 9, 10, 11, 12]],
+    <BLANKLINE>
+           [[13, 14, 15, 16],
+            [17, 18, 19, 20],
+            [21, 22, 23, 24]]]),
+     'time': 0}
+
+    If you don't give enough values for an array field, the remaining
+    values are filled in with their defaults.
+
+    >>> list(data.pack_data(
+    ...     [[[0, 1, 2, 3], [4, 5, 6]], [[10]]]))  # doctest: +ELLIPSIS
+    Traceback (most recent call last):
+      ...
+    ValueError: no default for <Field data ...>
+    >>> data.default = 0
+    >>> list(data.pack_data(
+    ...     [[[0, 1, 2, 3], [4, 5, 6]], [[10]]]))
+    [0, 1, 2, 3, 4, 5, 6, 0, 0, 0, 0, 0, 10, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+
     See Also
     --------
     Structure
     """
     def __init__(self, format, name, default=None, help=None, count=1):
-        self.format = format # See the struct documentation
+        self.format = format
         self.name = name
-        self.default = None
+        self.default = default
         self.help = help
         self.count = count
-        self.total_count = _numpy.prod(count)
+        self.item_count = _numpy.prod(count)  # number of item repeats
+        if isinstance(self.format, Structure):
+            self.structure_count = sum(f.total_count for f in format.fields)
+            self.total_count = self.item_count * self.structure_count
+        else:
+            self.total_count = self.item_count  # struct.Struct format chars
+
+    def __str__(self):
+        return self.__repr__()
+
+    def __repr__(self):
+        return '<{} {} {}>'.format(
+            self.__class__.__name__, self.name, id(self))
+
+    def indexes(self):
+        """Iterate through indexes to a possibly multi-dimensional array"""
+        assert self.item_count > 1, self
+        try:
+            i = [0] * len(self.count)
+        except TypeError:  # non-iterable count
+            for i in range(self.count):
+                yield i
+        else:
+            for i in range(self.item_count):
+                index = []
+                for j,c in enumerate(reversed(self.count)):
+                    index.insert(0, i % c)
+                    i /= c
+                yield index
+
+    def pack_data(self, data=None):
+        """Linearize a single field's data to a flat list.
+
+        If the field is repeated (count > 1), the incoming data should
+        be iterable with each iteration returning a single item.
+        """
+        if self.item_count > 1:
+            if data is None:
+                data = []
+            if hasattr(data, 'flat'):  # take advantage of numpy's ndarray.flat
+                items = 0
+                for item in data.flat:
+                    items += 1
+                    for arg in self.pack_item(item):
+                        yield arg
+                if items < self.item_count:
+                    if f.default is None:
+                        raise ValueError(
+                            'no default for {}.{}'.format(self, f))
+                    for i in range(self.item_count - items):
+                        yield f.default
+            else:
+                for index in self.indexes():
+                    try:
+                        if isinstance(index, int):
+                            item = data[index]
+                        else:
+                            item = data
+                            for i in index:
+                                item = item[i]
+                    except IndexError:
+                        item = None
+                    for arg in self.pack_item(item):
+                        yield arg
+        else:
+            for arg in self.pack_item(data):
+                yield arg
+
+    def pack_item(self, item=None):
+        """Linearize a single count of the field's data to a flat iterable
+        """
+        if isinstance(self.format, Structure):
+            for i in self.format._pack_item(item):
+                yield i
+        elif item is None:
+            if self.default is None:
+                raise ValueError('no default for {}'.format(self))
+            yield self.default
+        else:
+            yield item
+
+    def unpack_data(self, data):
+        """Inverse of .pack_data"""
+        iterator = iter(data)
+        try:
+            items = [iterator.next() for i in range(self.total_count)]
+        except StopIteration:
+            raise ValueError('not enough data to unpack {}'.format(self))
+        try:
+            iterator.next()
+        except StopIteration:
+            pass
+        else:
+            raise ValueError('too much data to unpack {}'.format(self))
+        if isinstance(self.format, Structure):
+            # break into per-structure clumps
+            s = self.structure_count
+            items = zip(*[items[i::s] for i in range(s)])
+        else:
+            items = [[i] for i in items]
+        unpacked = [self.unpack_item(i) for i in items]
+        if self.count == 1:
+            return unpacked[0]
+        if isinstance(self.format, Structure):
+            try:
+                len(self.count)
+            except TypeError:
+                pass
+            else:
+                raise NotImplementedError('reshape Structure field')
+        else:
+            unpacked = _numpy.array(unpacked)
+            unpacked = unpacked.reshape(self.count)
+        return unpacked
+
+    def unpack_item(self, item): 
+        """Inverse of .unpack_item"""
+        if isinstance(self.format, Structure):
+            return self.format._unpack_item(item)
+        else:
+            assert len(item) == 1, item
+            return item[0]
 
 
 class Structure (_struct.Struct):
-    """Represent a C structure.
+    r"""Represent a C structure.
 
     A convenient wrapper around struct.Struct that uses Fields and
     adds dict-handling methods for transparent name assignment.
@@ -40,41 +274,86 @@ class Structure (_struct.Struct):
     Examples
     --------
 
-    Represent the C structure::
+    >>> import array
+    >>> from pprint import pprint
+
+    Represent the C structures::
+
+        struct run {
+          unsigned int time;
+          short data[2][3];
+        }
 
-        struct thing {
-          short version;
-          long size[3];
+        struct experiment {
+          unsigned short version;
+          struct run runs[2];
         }
 
     As
 
-    >>> import array
-    >>> from pprint import pprint
-    >>> thing = Structure(name='thing',
-    ...     fields=[Field('h', 'version'), Field('l', 'size', count=3)])
-    >>> thing.set_byte_order('>')
-    >>> b = array.array('b', range(2+4*3))
-    >>> d = thing.unpack_dict_from(buffer=b)
+    >>> time = Field('I', 'time', default=0, help='POSIX time')
+    >>> data = Field(
+    ...     'h', 'data', default=0, help='example data', count=(2,3))
+    >>> run = Structure('run', fields=[time, data])
+    >>> version = Field(
+    ...     'H', 'version', default=1, help='example version')
+    >>> runs = Field(run, 'runs', help='pair of runs', count=2)
+    >>> experiment = Structure('experiment', fields=[version, runs])
+
+    The structures automatically calculate the flattened data format:
+
+    >>> run.format
+    '=Ihhhhhh'
+    >>> run.size  # 4 + 2*3*2
+    16
+    >>> experiment.format
+    '=HIhhhhhhIhhhhhh'
+    >>> experiment.size  # 2 + 2*(4 + 2*3*2)
+    34
+
+    You can read data out of any object supporting the buffer
+    interface:
+
+    >>> b = array.array('B', range(experiment.size))
+    >>> experiment.set_byte_order('>')
+    >>> d = experiment.unpack_from(buffer=b)
     >>> pprint(d)
-    {'size': array([ 33752069, 101124105, 168496141]), 'version': 1}
-    >>> [hex(x) for x in d['size']]
-    ['0x2030405L', '0x6070809L', '0xa0b0c0dL']
-
-    You can even get fancy with multi-dimensional arrays.
-
-    >>> thing = Structure(name='thing',
-    ...     fields=[Field('h', 'version'), Field('l', 'size', count=(3,2))])
-    >>> thing.set_byte_order('>')
-    >>> b = array.array('b', range(2+4*3*2))
-    >>> d = thing.unpack_dict_from(buffer=b)
-    >>> d['size'].shape
-    (3, 2)
+    {'runs': [{'data': array([[1543, 2057, 2571],
+           [3085, 3599, 4113]]),
+               'time': 33752069},
+              {'data': array([[5655, 6169, 6683],
+           [7197, 7711, 8225]]),
+               'time': 303240213}],
+     'version': 1}
+    >>> [hex(x) for x in d['runs'][0]['data'].flat]
+    ['0x607L', '0x809L', '0xa0bL', '0xc0dL', '0xe0fL', '0x1011L']
+
+    You can also read out from strings:
+
+    >>> d = experiment.unpack(b.tostring())
     >>> pprint(d)
-    {'size': array([[ 33752069, 101124105],
-           [168496141, 235868177],
-           [303240213, 370612249]]),
+    {'runs': [{'data': array([[1543, 2057, 2571],
+           [3085, 3599, 4113]]),
+               'time': 33752069},
+              {'data': array([[5655, 6169, 6683],
+           [7197, 7711, 8225]]),
+               'time': 303240213}],
      'version': 1}
+
+    If you don't give enough values for an array field, the remaining
+    values are filled in with their defaults.
+
+    >>> experiment.pack_into(buffer=b, data=d)
+    >>> b.tostring()[:17]
+    '\x00\x01\x02\x03\x04\x05\x06\x07\x08\t\n\x0b\x0c\r\x0e\x0f\x10'
+    >>> b.tostring()[17:]
+    '\x11\x12\x13\x14\x15\x16\x17\x18\x19\x1a\x1b\x1c\x1d\x1e\x1f !'
+    >>> run0 = d['runs'].pop(0)
+    >>> b = experiment.pack(data=d)
+    >>> b[:17]
+    '\x00\x01\x12\x13\x14\x15\x16\x17\x18\x19\x1a\x1b\x1c\x1d\x1e\x1f '
+    >>> b[17:]
+    '!\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00'
     """
     def __init__(self, name, fields, byte_order='='):
         # '=' for native byte order, standard size and alignment
@@ -86,6 +365,10 @@ class Structure (_struct.Struct):
     def __str__(self):
         return self.name
 
+    def __repr__(self):
+        return '<{} {} {}>'.format(
+            self.__class__.__name__, self.name, id(self))
+
     def set_byte_order(self, byte_order):
         """Allow changing the format byte_order on the fly.
         """
@@ -94,67 +377,67 @@ class Structure (_struct.Struct):
             return  # no need to change anything
         format = []
         for field in self.fields:
-            format.extend([field.format]*field.total_count)
+            if isinstance(field.format, Structure):
+                field_format = field.format.sub_format(
+                    ) * field.item_count
+            else:
+                field_format = [field.format]*field.item_count
+            format.extend(field_format)
         super(Structure, self).__init__(
             format=byte_order+''.join(format).replace('P', 'L'))
 
-    def _flatten_args(self, args):
-        # handle Field.count > 0
-        flat_args = []
-        for a,f in zip(args, self.fields):
-            if f.total_count > 1:
-                flat_args.extend(a)
-            else:
-                flat_args.append(a)
-        return flat_args
+    def sub_format(self):
+        return self.format.lstrip('=<>')  # byte order handled by parent
 
-    def _unflatten_args(self, args):
-        # handle Field.count > 0
-        unflat_args = []
-        i = 0
+    def _pack_item(self, item=None):
+        """Linearize a single count of the structure's data to a flat iterable
+        """
+        if item is None:
+            item = {}
         for f in self.fields:
-            if f.total_count > 1:
-                data = _numpy.array(args[i:i+f.total_count])
-                data = data.reshape(f.count)
-                unflat_args.append(data)
-            else:
-                unflat_args.append(args[i])
-            i += f.total_count
-        return unflat_args
-        
-    def pack(self, *args):
-        return super(Structure, self)(*self._flatten_args(args))
-
-    def pack_into(self, buffer, offset, *args):
-        return super(Structure, self).pack_into(
-            buffer, offset, *self._flatten_args(args))
+            try:
+                data = item[f.name]
+            except KeyError:
+                data = None
+            for arg in f.pack_data(data):
+                yield arg
 
-    def _clean_dict(self, dict):
+    def _unpack_item(self, args):
+        """Inverse of ._unpack_item"""
+        data = {}
+        iterator = iter(args)
         for f in self.fields:
-            if f.name not in dict:
-                if f.default != None:
-                    dict[f.name] = f.default
-                else:
-                    raise ValueError('{} field not set for {}'.format(
-                            f.name, self.__class__.__name__))
-        return dict
-
-    def pack_dict(self, dict):
-        dict = self._clean_dict(dict)
-        return self.pack(*[dict[f.name] for f in self.fields])
-
-    def pack_dict_into(self, buffer, offset, dict={}):
-        dict = self._clean_dict(dict)
-        return self.pack_into(buffer, offset,
-                              *[dict[f.name] for f in self.fields])
-
-    def unpack(self, string):
-        return self._unflatten_args(
-            super(Structure, self).unpack(string))
-
-    def unpack_from(self, buffer, offset=0):
+            try:
+                items = [iterator.next() for i in range(f.total_count)]
+            except StopIteration:
+                raise ValueError('not enough data to unpack {}.{}'.format(
+                        self, f))
+            data[f.name] = f.unpack_data(items)
         try:
-            args = super(Structure, self).unpack_from(buffer, offset)
+            iterator.next()
+        except StopIteration:
+            pass
+        else:
+            raise ValueError('too much data to unpack {}'.format(self))
+        return data
+
+    def pack(self, data):
+        args = list(self._pack_item(data))
+        return super(Structure, self).pack(*args)
+
+    def pack_into(self, buffer, offset=0, data={}):
+        args = list(self._pack_item(data))
+        return super(Structure, self).pack_into(
+            buffer, offset, *args)
+
+    def unpack(self, *args, **kwargs):
+        args = super(Structure, self).unpack(*args, **kwargs)
+        return self._unpack_item(args)
+
+    def unpack_from(self, buffer, offset=0, *args, **kwargs):
+        try:
+            args = super(Structure, self).unpack_from(
+                buffer, offset, *args, **kwargs)
         except _struct.error as e:
             if not self.name in ('WaveHeader2', 'WaveHeader5'):
                 raise
@@ -166,16 +449,6 @@ class Structure (_struct.Struct):
                 # missing wData?  Pad with zeros
                 buffer += _buffer('\x00'*(self.size + offset - len(buffer)))
             args = super(Structure, self).unpack_from(buffer, offset)
-            unpacked = self._unflatten_args(args)
-            data = dict(zip([f.name for f in self.fields],
-                            unpacked))
+            data = self._unpack_item(args)
             assert data['npnts'] == 0, data['npnts']
-        return self._unflatten_args(args)
-
-    def unpack_dict(self, string):
-        return dict(zip([f.name for f in self.fields],
-                        self.unpack(string)))
-
-    def unpack_dict_from(self, buffer, offset=0):
-        return dict(zip([f.name for f in self.fields],
-                        self.unpack_from(buffer, offset)))
+        return self._unpack_item(args)