diff --git a/lib/freetypy/subset.py b/lib/freetypy/subset.py
new file mode 100644
index 0000000..1a19ded
--- /dev/null
+++ b/lib/freetypy/subset.py
@@ -0,0 +1,343 @@
+# -*- coding: utf-8 -*-
+
+# Copyright (c) 2015, Michael Droettboom All rights reserved.
+
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+
+# 1. Redistributions of source code must retain the above copyright
+#    notice, this list of conditions and the following disclaimer.
+# 2. Redistributions in binary form must reproduce the above copyright
+#    notice, this list of conditions and the following disclaimer in
+#    the documentation and/or other materials provided with the
+#    distribution.
+
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
+# FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
+# COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
+# INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
+# BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+# LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+# CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+# LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
+# ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+# POSSIBILITY OF SUCH DAMAGE.
+
+# The views and conclusions contained in the software and
+# documentation are those of the authors and should not be interpreted
+# as representing official policies, either expressed or implied, of
+# the FreeBSD Project.
+
+"""
+A library to subset SFNT-style fonts.
+"""
+
+from __future__ import absolute_import, division, unicode_literals, print_function
+
+
+__all__ = ['subset_font']
+
+
+from collections import OrderedDict
+import struct
+
+
+from freetypy import Face
+
+
+UNDERSTOOD_VERSIONS = (0x00010000, 0x4f54544f)
+
+
+class _BinaryStruct(object):
+    """
+    A wrapper around the Python stdlib struct module to define a
+    binary struct more like a dictionary than a tuple.
+    """
+    def __init__(self, descr, endian='>'):
+        """
+        Parameters
+        ----------
+        descr : list of tuple
+            Each entry is a pair ``(name, format)``, where ``format``
+            is one of the format types understood by `struct`.
+        endian : str, optional
+            The endianness of the struct.  Must be ``>`` or ``<``.
+        """
+        self._fmt = [endian]
+        self._offsets = {}
+        self._names = []
+        i = 0
+        for name, fmt in descr:
+            self._fmt.append(fmt)
+            self._offsets[name] = (i, (endian + fmt).encode('ascii'))
+            self._names.append(name)
+            i += struct.calcsize(fmt.encode('ascii'))
+        self._fmt = ''.join(self._fmt).encode('ascii')
+        self._size = struct.calcsize(self._fmt)
+
+    @property
+    def size(self):
+        """
+        Return the size of the struct.
+        """
+        return self._size
+
+    def pack(self, **kwargs):
+        """
+        Pack the given arguments, which are given as kwargs, and
+        return the binary struct.
+        """
+        fields = [0] * len(self._names)
+        for key, val in kwargs.items():
+            if key not in self._offsets:
+                raise KeyError("No header field '{0}'".format(key))
+            i = self._names.index(key)
+            fields[i] = val
+        return struct.pack(self._fmt, *fields)
+
+    def unpack(self, buff):
+        """
+        Unpack the given binary buffer into the fields.  The result
+        is a dictionary mapping field names to values.
+        """
+        args = struct.unpack_from(self._fmt, buff[:self._size])
+        return dict(zip(self._names, args))
+
+    def read(self, fd):
+        """
+        Read a struct from the current location in the file.
+        """
+        buff = fd.read(self.size)
+        return self.unpack(buff)
+
+    def write(self, fd, data):
+        """
+        Write a struct to the current location in the file.
+        """
+        buff = self.pack(**data)
+        fd.write(buff)
+
+
+class Table(object):
+    header_struct = _BinaryStruct([
+        ('tag', '4s'),
+        ('checkSum', 'I'),
+        ('offset', 'I'),
+        ('length', 'I')
+    ])
+
+    def __init__(self, header, content):
+        self._header = header
+        self._content = content
+
+    def __repr__(self):
+        return "<Table '{0}'>".format(
+            self._header['tag'].decode('ascii'))
+
+    def _calc_checksum(self, content):
+        end = ((len(content) + 3) & ~3) - 4
+        sum = 0
+        for i in range(0, end, 4):
+            sum += struct.unpack('I', content[i:i+4])[0]
+            sum = sum & 0xffffffff
+        return sum
+
+    @classmethod
+    def read(cls, fd):
+        header = Table.header_struct.read(fd)
+        content = fd.read(header['length'])
+
+        return cls(header, content)
+
+    @property
+    def header(self):
+        return self._header
+
+    @property
+    def content(self):
+        return self._content
+    @content.setter
+    def content(self, new_content):
+        checksum = self._calc_checksum(new_content)
+        self._header['checkSum'] = checksum
+        self._header['length'] = len(new_content)
+        self._content = new_content
+
+
+class HeadTable(Table):
+    head_table_struct = _BinaryStruct([
+        ('version', 'I'),
+        ('fontRevision', 'I'),
+        ('checkSumAdjustment', 'I'),
+        ('magicNumber', 'I'),
+        ('flags', 'H'),
+        ('unitsPerEm', 'H'),
+        ('created', 'q'),
+        ('modified', 'q'),
+        ('xMin', 'h'),
+        ('yMin', 'h'),
+        ('xMax', 'h'),
+        ('yMax', 'h'),
+        ('macStyle', 'H'),
+        ('lowestRecPPEM', 'H'),
+        ('fontDirectionHint', 'h'),
+        ('indexToLocFormat', 'h'),
+        ('glyphDataFormat', 'h')])
+
+    def __init__(self, header, content):
+        super(HeadTable, self).__init__(header, content)
+
+        self.__dict__.update(self.head_table_struct.unpack(content))
+
+        if self.version not in UNDERSTOOD_VERSIONS:
+            raise ValueError("Not a TrueType or OpenType file")
+
+        if self.magicNumber != 0x5F0F3CF5:
+            raise ValueError("Bad magic number")
+
+
+class LocaTable(Table):
+    def _get_formats(self, fontfile):
+        if fontfile[b'head'].indexToLocFormat == 0:  # short offsets
+            return '>H', 2, 2
+        else:
+            return '>I', 4, 1
+
+    def get_offsets(self, fontfile):
+        entry_format, entry_size, scale = self._get_formats(fontfile)
+
+        content = self._content
+        offsets = []
+        for i in range(0, len(content), entry_size):
+            value = struct.unpack(
+                entry_format, content[i:i+entry_size])[0] * scale
+            offsets.append(value)
+
+        return offsets
+
+    def subset(self, fontfile, glyphs, offsets):
+        new_offsets = []
+        offset = 0
+        j = 0
+        for i in range(len(offsets) - 1):
+            new_offsets.append(offset)
+            if j < len(glyphs) and i == glyphs[j]:
+                offset += offsets[i+1] - offsets[i]
+                j += 1
+        new_offsets.append(offset)
+
+        entry_format, entry_size, scale = self._get_formats(fontfile)
+        new_content = []
+        for value in new_offsets:
+            new_content.append(struct.pack(entry_format, value // scale))
+        self.content = b''.join(new_content)
+
+
+class GlyfTable(Table):
+    def subset(self, glyphs, offsets):
+        content = self.content
+        new_content = []
+        for gind in glyphs:
+            new_content.append(content[offsets[gind]:offsets[gind+1]])
+        self.content = b''.join(new_content)
+
+
+SPECIAL_TABLES = {
+    b'head': HeadTable,
+    b'loca': LocaTable,
+    b'glyf': GlyfTable
+}
+
+
+class FontFile(object):
+    """
+    A class to subset SFNT-style fonts (TrueType and OpenType).
+    """
+
+    header_struct = _BinaryStruct([
+        ('version', 'I'),
+        ('numTables', 'H'),
+        ('searchRange', 'H'),
+        ('entrySelector', 'H'),
+        ('rangeShift', 'H')
+    ])
+
+    def __init__(self, face, header, tables):
+        self._face = face
+        self._header = header
+        self._tables = tables
+
+    def __getitem__(self, tag):
+        return self._tables[tag]
+
+    @classmethod
+    def read(cls, fd):
+        header = cls.header_struct.read(fd)
+
+        if header['version'] not in UNDERSTOOD_VERSIONS:
+            raise ValueError("Not a TrueType or OpenType file")
+
+        table_dir = []
+        for i in range(header['numTables']):
+            table_dir.append(Table.header_struct.read(fd))
+
+        tables = OrderedDict()
+        for table_header in table_dir:
+            fd.seek(table_header['offset'])
+            content = fd.read(table_header['length'])
+            table_cls = SPECIAL_TABLES.get(table_header['tag'], Table)
+            tables[table_header['tag']] = table_cls(table_header, content)
+
+        fd.seek(0)
+        face = Face(fd)
+
+        return cls(face, header, tables)
+
+    def subset(self, ccodes):
+        glyphs = []
+        for ccode in ccodes:
+            glyphs.append(self._face.get_char_index_unicode(ccode))
+        glyphs.sort()
+
+        offsets = self[b'loca'].get_offsets(self)
+        self[b'glyf'].subset(glyphs, offsets)
+        self[b'loca'].subset(self, glyphs, offsets)
+
+    def write(self, fd):
+        self.header_struct.write(fd, self._header)
+
+        offset = (self.header_struct.size +
+                  Table.header_struct.size * len(self._tables))
+
+        for table in self._tables.values():
+            table.header['offset'] = offset
+            offset += len(table.content)
+
+        for table in self._tables.values():
+            Table.header_struct.write(fd, table.header)
+
+        for table in self._tables.values():
+            fd.write(table._content)
+
+
+def subset_font(input_fd, output_fd, charcodes):
+    """
+    Subset a SFNT-style (TrueType or OpenType) font.
+
+    Parameters
+    ----------
+    input_fd : readable file-like object, for bytes
+        The font file to read.
+
+    output_fd : writable file-like object, for bytes
+        The file to write a subsetted font file to.
+
+    charcodes : list of int or unicode string
+        The character codes to include in the output font file.
+    """
+    fontfile = FontFile.read(input_fd)
+    fontfile.subset(charcodes)
+    fontfile.write(output_fd)
diff --git a/lib/freetypy/tests/test_subset.py b/lib/freetypy/tests/test_subset.py
new file mode 100644
index 0000000..37b0585
--- /dev/null
+++ b/lib/freetypy/tests/test_subset.py
@@ -0,0 +1,64 @@
+# -*- coding: utf-8 -*-
+
+# Copyright (c) 2015, Michael Droettboom All rights reserved.
+
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+
+# 1. Redistributions of source code must retain the above copyright
+#    notice, this list of conditions and the following disclaimer.
+# 2. Redistributions in binary form must reproduce the above copyright
+#    notice, this list of conditions and the following disclaimer in
+#    the documentation and/or other materials provided with the
+#    distribution.
+
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
+# FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
+# COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
+# INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
+# BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+# LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+# CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+# LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
+# ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+# POSSIBILITY OF SUCH DAMAGE.
+
+# The views and conclusions contained in the software and
+# documentation are those of the authors and should not be interpreted
+# as representing official policies, either expressed or implied, of
+# the FreeBSD Project.
+
+from __future__ import print_function, unicode_literals, absolute_import
+
+import freetypy as ft
+from freetypy import subset
+from .util import *
+
+import io
+
+
+def test_subset():
+    face = ft.Face(vera_path())
+    face.set_char_size(12, 12, 300, 300)
+    glyph = face.load_char_unicode('B')
+    original_B = glyph.outline.to_string(' M ', ' L ', ' C ', ' Q ')
+
+    with open(vera_path(), 'rb') as input_fd:
+        output_fd = io.BytesIO()
+        subset.subset_font(input_fd, output_fd, 'ABCD')
+
+    output_fd.seek(0)
+
+    face = ft.Face(output_fd)
+    face.set_char_size(12, 12, 300, 300)
+    glyph = face.load_char_unicode('B')
+    s = glyph.outline.to_string(' M ', ' L ', ' C ', ' Q ')
+    assert original_B == s
+
+    # Not here
+    glyph = face.load_char_unicode('X')
+    s = glyph.outline.to_string(' M ', ' L ', ' C ', ' Q ')
+    assert len(s) == 0
diff --git a/lib/freetypy/util.py b/lib/freetypy/util.py
index 6a88627..362c59b 100644
--- a/lib/freetypy/util.py
+++ b/lib/freetypy/util.py
@@ -58,8 +58,10 @@ def bitmap_to_ascii(a):
 
     lines = []
     for row in a.to_list():
+        line = []
         for col in row:
             col = int(float(col) / 255. * 4.)
             c = shades[col]
-            lines.append(c)
+            line.append(c)
+        lines.append(''.join(line))
     return '\n'.join(lines)
diff --git a/src/outline.c b/src/outline.c
index 78cd3d5..6a4d861 100644
--- a/src/outline.c
+++ b/src/outline.c
@@ -863,6 +863,7 @@ Py_Outline_to_string(Py_Outline* self, PyObject* args, PyObject* kwds)
         "move_command", "line_command", "cubic_command", "conic_command",
         "prefix", NULL};
 
+    data.prefix = 0;
     data.conic_command = NULL;
 
     if (!PyArg_ParseTupleAndKeywords(