Added buffer of bytes round-tripping (#8)

* Added buffer of bytes round-tripping

* Skip upload for PRs

[ci skip]

* Fix condition

[ci skip]

Author: TomAugspurger

.appveyor.yml

@@ -84,6 +84,4 @@ test_script:
   - cmd: conda build conda-recipes/cyberpandas --python=${PYTHON}
 
 on_success:
-  - cmd: anaconda -t %UPLOAD_KEY% upload -u intake --force %CONDA_ROOT%\\conda-bld\\*\\*.tar.bz2
-    on:
-      branch: master
+  - cmd: if "%APPVEYOR_PULL_REQUEST_NUMBER%"=="" anaconda -t %UPLOAD_KEY% upload -u intake --force %CONDA_ROOT%\\conda-bld\\*\\*.tar.bz2

cyberpandas/ip_array.py

@@ -35,7 +35,7 @@ class IPType(ExtensionDtype):
     name = 'ip'
     type = IPv4v6Base
     kind = 'O'
-    mybase = np.dtype([('hi', '>u8'), ('lo', '>u8')])
+    _record_type = np.dtype([('hi', '>u8'), ('lo', '>u8')])
     na_value = ipaddress.IPv4Address(0)
 
     @classmethod
@@ -142,6 +142,17 @@ def to_pyipaddress(self):
     def to_pyints(self):
         return [combine(*map(int, x)) for x in self.data]
 
+    def to_bytes(self):
+        """Serialize the IPArray as a Python bytestring.
+
+        Examples
+        --------
+        >>> arr = IPArray([10, 20])
+        >>> arr.to_bytes()
+        b'\x00\x00\...x00\x02'
+        """
+        return self.data.tobytes()
+
     def __repr__(self):
         formatted = self._format_values()
         return "IPArray({!r})".format(formatted)
@@ -173,6 +184,59 @@ def from_pyints(cls, values):
         # type: T.Sequence[int]) -> 'IPArray'
         return cls(_to_ipaddress_pyint(values))
 
+    @classmethod
+    def from_bytes(cls, bytestring):
+        """Create an IPArray from a bytestring.
+
+        Parameters
+        ----------
+        bytestring : bytes
+            Note that bytestring is a Python 3-style string of bytes,
+            not a sequences of bytes where each element represents an
+            IPAddress.
+
+        Returns
+        -------
+        IPArray
+
+        Examples
+        --------
+        >>> arr = IPArray([10, 20])
+        >>> buf = arr.to_bytes()
+        >>> buf
+        b'\x00\x00\...x00\x02'
+        >>> IPArray.from_bytes(buf)
+        IPArray(['0.0.0.10', '0.0.0.20'])
+
+        See Also
+        --------
+        to_bytes
+        from_pyints
+        """
+        data = np.frombuffer(bytestring, dtype=IPType._record_type)
+        return cls._from_ndarray(data)
+
+    @classmethod
+    def _from_ndarray(cls, data, copy=False):
+        """Zero-copy construction of an IPArray from an ndarray.
+
+        Parameters
+        ----------
+        data : ndarray
+            This should have IPType._record_type dtype
+        copy : bool, default False
+            Whether to copy the data.
+
+        Returns
+        -------
+        ExtensionArray
+        """
+        if copy:
+            data = data.copy()
+        new = IPArray([])
+        new.data = data
+        return new
+
     def __eq__(self, other):
         # TDOO: scalar ipaddress
         if not isinstance(other, IPArray):

cyberpandas/parser.py

@@ -47,9 +47,10 @@ def _to_ip_array(values):
 
     if isinstance(values, IPArray):
         return values.data
-    if not (isinstance(values, np.ndarray) and values.dtype == IPType.mybase):
+    if not (isinstance(values, np.ndarray) and
+            values.dtype == IPType._record_type):
         values = _to_int_pairs(values)
-    return np.atleast_1d(np.asarray(values, dtype=IPType.mybase))
+    return np.atleast_1d(np.asarray(values, dtype=IPType._record_type))
 
 
 def _to_int_pairs(values):
@@ -80,7 +81,7 @@ def _to_ipaddress_pyint(values):
     from .ip_array import IPType
 
     values2 = [unpack(pack(x)) for x in values]
-    return np.atleast_1d(np.asarray(values2, dtype=IPType.mybase))
+    return np.atleast_1d(np.asarray(values2, dtype=IPType._record_type))
 
 
 def _as_ip_object(val):

cyberpandas/test_ip.py

@@ -11,6 +11,7 @@
 import pandas as pd
 import cyberpandas as ip
 import pandas.util.testing as tm
+from cyberpandas.common import _U8_MAX
 
 
 def test_make_container():
@@ -19,7 +20,7 @@ def test_make_container():
         values.data,
         np.array([(0, 1),
                   (0, 2),
-                  (0, 3)], dtype=values.dtype.mybase)
+                  (0, 3)], dtype=values.dtype._record_type)
     )
 
 
@@ -268,3 +269,12 @@ def test_setitem_array():
     ser[[1, 2]] = [10, 20]
     expected = ip.IPArray([0, 10, 20])
     assert ser.equals(expected)
+
+
+def test_bytes_roundtrip():
+    arr = ip.IPArray([1, 2, 3, _U8_MAX + 10])
+    bytestring = arr.to_bytes()
+    assert isinstance(bytestring, bytes)
+
+    result = ip.IPArray.from_bytes(bytestring)
+    assert result.equals(arr)