Python API: Add enum and union support.

As well as a rewrite of the encoders/decoders to make it more readable and extensible.
(Re-commit after fix to verify build.)

Change-Id: Ic244d3cebe070bb2570491f8a24f4a1e203f889a
Signed-off-by: Ole Troan <ot@cisco.com>

3 files changed, 1023 insertions, 0 deletions

diff --git a/src/vpp-api/python/vpp_papi/__init__.py b/src/vpp-api/python/vpp_papi/__init__.py
new file mode 100644
index 00000000000..f9afcf17f29
--- /dev/null
+++ b/src/vpp-api/python/vpp_papi/__init__.py
@@ -0,0 +1 @@
+from .vpp_papi import *
diff --git a/src/vpp-api/python/vpp_papi/vpp_papi.py b/src/vpp-api/python/vpp_papi/vpp_papi.py
new file mode 100644
index 00000000000..edc0c6ab364
--- /dev/null
+++ b/src/vpp-api/python/vpp_papi/vpp_papi.py
@@ -0,0 +1,679 @@
+#!/usr/bin/env python
+#
+# Copyright (c) 2016 Cisco and/or its affiliates.
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at:
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+
+from __future__ import print_function
+from __future__ import absolute_import
+import sys
+import os
+import logging
+import collections
+import struct
+import json
+import threading
+import fnmatch
+import weakref
+import atexit
+from cffi import FFI
+import cffi
+from . vpp_serializer import VPPType, VPPEnumType, VPPUnionType, BaseTypes
+from . vpp_serializer import VPPMessage
+
+if sys.version[0] == '2':
+    import Queue as queue
+else:
+    import queue as queue
+
+ffi = FFI()
+ffi.cdef("""
+typedef void (*vac_callback_t)(unsigned char * data, int len);
+typedef void (*vac_error_callback_t)(void *, unsigned char *, int);
+int vac_connect(char * name, char * chroot_prefix, vac_callback_t cb,
+    int rx_qlen);
+int vac_disconnect(void);
+int vac_read(char **data, int *l, unsigned short timeout);
+int vac_write(char *data, int len);
+void vac_free(void * msg);
+
+int vac_get_msg_index(unsigned char * name);
+int vac_msg_table_size(void);
+int vac_msg_table_max_index(void);
+
+void vac_rx_suspend (void);
+void vac_rx_resume (void);
+void vac_set_error_handler(vac_error_callback_t);
+ """)
+
+# Barfs on failure, no need to check success.
+vpp_api = ffi.dlopen('libvppapiclient.so')
+
+
+def vpp_atexit(vpp_weakref):
+    """Clean up VPP connection on shutdown."""
+    vpp_instance = vpp_weakref()
+    if vpp_instance.connected:
+        vpp_instance.logger.debug('Cleaning up VPP on exit')
+        vpp_instance.disconnect()
+
+
+vpp_object = None
+
+
+def vpp_iterator(d):
+    if sys.version[0] == '2':
+        return d.iteritems()
+    else:
+        return d.items()
+
+
+@ffi.callback("void(unsigned char *, int)")
+def vac_callback_sync(data, len):
+    vpp_object.msg_handler_sync(ffi.buffer(data, len))
+
+
+@ffi.callback("void(unsigned char *, int)")
+def vac_callback_async(data, len):
+    vpp_object.msg_handler_async(ffi.buffer(data, len))
+
+
+@ffi.callback("void(void *, unsigned char *, int)")
+def vac_error_handler(arg, msg, msg_len):
+    vpp_object.logger.warning("VPP API client:: %s", ffi.string(msg, msg_len))
+
+
+class Empty(object):
+    pass
+
+
+class FuncWrapper(object):
+    def __init__(self, func):
+        self._func = func
+        self.__name__ = func.__name__
+
+    def __call__(self, **kwargs):
+        return self._func(**kwargs)
+
+
+class VPP():
+    """VPP interface.
+
+    This class provides the APIs to VPP.  The APIs are loaded
+    from provided .api.json files and makes functions accordingly.
+    These functions are documented in the VPP .api files, as they
+    are dynamically created.
+
+    Additionally, VPP can send callback messages; this class
+    provides a means to register a callback function to receive
+    these messages in a background thread.
+    """
+
+    def process_json_file(self, apidef_file):
+        api = json.load(apidef_file)
+        types = {}
+        for t in api['enums']:
+            t[0] = 'vl_api_' + t[0] + '_t'
+            types[t[0]] = {'type': 'enum', 'data': t}
+        for t in api['unions']:
+            t[0] = 'vl_api_' + t[0] + '_t'
+            types[t[0]] = {'type': 'union', 'data': t}
+        for t in api['types']:
+            t[0] = 'vl_api_' + t[0] + '_t'
+            types[t[0]] = {'type': 'type', 'data': t}
+
+        i = 0
+        while True:
+            unresolved = {}
+            for k, v in types.items():
+                t = v['data']
+                if v['type'] == 'enum':
+                    try:
+                        VPPEnumType(t[0], t[1:])
+                    except ValueError:
+                        unresolved[k] = v
+                elif v['type'] == 'union':
+                    try:
+                        VPPUnionType(t[0], t[1:])
+                    except ValueError:
+                        unresolved[k] = v
+                elif v['type'] == 'type':
+                    try:
+                        VPPType(t[0], t[1:])
+                    except ValueError:
+                        unresolved[k] = v
+            if len(unresolved) == 0:
+                break
+            if i > 3:
+                raise ValueError('Unresolved type definitions {}'
+                                 .format(unresolved))
+            types = unresolved
+            i += 1
+
+        for m in api['messages']:
+            try:
+                self.messages[m[0]] = VPPMessage(m[0], m[1:])
+            except NotImplementedError:
+                self.logger.error('Not implemented error for {}'.format(m[0]))
+
+    def __init__(self, apifiles=None, testmode=False, async_thread=True,
+                 logger=logging.getLogger('vpp_papi'), loglevel='debug',
+                 read_timeout=0):
+        """Create a VPP API object.
+
+        apifiles is a list of files containing API
+        descriptions that will be loaded - methods will be
+        dynamically created reflecting these APIs.  If not
+        provided this will load the API files from VPP's
+        default install location.
+
+        logger, if supplied, is the logging logger object to log to.
+        loglevel, if supplied, is the log level this logger is set
+        to report at (from the loglevels in the logging module).
+        """
+        global vpp_object
+        vpp_object = self
+
+        if logger is None:
+            logger = logging.getLogger(__name__)
+            if loglevel is not None:
+                logger.setLevel(loglevel)
+        self.logger = logger
+
+        self.messages = {}
+        self.id_names = []
+        self.id_msgdef = []
+        self.connected = False
+        self.header = VPPType('header', [['u16', 'msgid'],
+                                         ['u32', 'client_index']])
+        self.apifiles = []
+        self.event_callback = None
+        self.message_queue = queue.Queue()
+        self.read_timeout = read_timeout
+        self.vpp_api = vpp_api
+        self.async_thread = async_thread
+
+        if not apifiles:
+            # Pick up API definitions from default directory
+            try:
+                apifiles = self.find_api_files()
+            except RuntimeError:
+                # In test mode we don't care that we can't find the API files
+                if testmode:
+                    apifiles = []
+                else:
+                    raise
+
+        for file in apifiles:
+            with open(file) as apidef_file:
+                self.process_json_file(apidef_file)
+
+        self.apifiles = apifiles
+
+        # Basic sanity check
+        if len(self.messages) == 0 and not testmode:
+            raise ValueError(1, 'Missing JSON message definitions')
+
+        # Make sure we allow VPP to clean up the message rings.
+        atexit.register(vpp_atexit, weakref.ref(self))
+
+        # Register error handler
+        if not testmode:
+            vpp_api.vac_set_error_handler(vac_error_handler)
+
+        # Support legacy CFFI
+        # from_buffer supported from 1.8.0
+        (major, minor, patch) = [int(s) for s in
+                                 cffi.__version__.split('.', 3)]
+        if major >= 1 and minor >= 8:
+            self._write = self._write_new_cffi
+        else:
+            self._write = self._write_legacy_cffi
+
+    class ContextId(object):
+        """Thread-safe provider of unique context IDs."""
+        def __init__(self):
+            self.context = 0
+            self.lock = threading.Lock()
+
+        def __call__(self):
+            """Get a new unique (or, at least, not recently used) context."""
+            with self.lock:
+                self.context += 1
+                return self.context
+    get_context = ContextId()
+
+    @classmethod
+    def find_api_dir(cls):
+        """Attempt to find the best directory in which API definition
+        files may reside. If the value VPP_API_DIR exists in the environment
+        then it is first on the search list. If we're inside a recognized
+        location in a VPP source tree (src/scripts and src/vpp-api/python)
+        then entries from there to the likely locations in build-root are
+        added. Finally the location used by system packages is added.
+
+        :returns: A single directory name, or None if no such directory
+            could be found.
+        """
+        dirs = []
+
+        if 'VPP_API_DIR' in os.environ:
+            dirs.append(os.environ['VPP_API_DIR'])
+
+        # perhaps we're in the 'src/scripts' or 'src/vpp-api/python' dir;
+        # in which case, plot a course to likely places in the src tree
+        import __main__ as main
+        if hasattr(main, '__file__'):
+            # get the path of the calling script
+            localdir = os.path.dirname(os.path.realpath(main.__file__))
+        else:
+            # use cwd if there is no calling script
+            localdir = os.getcwd()
+        localdir_s = localdir.split(os.path.sep)
+
+        def dmatch(dir):
+            """Match dir against right-hand components of the script dir"""
+            d = dir.split('/')  # param 'dir' assumes a / separator
+            length = len(d)
+            return len(localdir_s) > length and localdir_s[-length:] == d
+
+        def sdir(srcdir, variant):
+            """Build a path from srcdir to the staged API files of
+            'variant'  (typically '' or '_debug')"""
+            # Since 'core' and 'plugin' files are staged
+            # in separate directories, we target the parent dir.
+            return os.path.sep.join((
+                srcdir,
+                'build-root',
+                'install-vpp%s-native' % variant,
+                'vpp',
+                'share',
+                'vpp',
+                'api',
+            ))
+
+        srcdir = None
+        if dmatch('src/scripts'):
+            srcdir = os.path.sep.join(localdir_s[:-2])
+        elif dmatch('src/vpp-api/python'):
+            srcdir = os.path.sep.join(localdir_s[:-3])
+        elif dmatch('test'):
+            # we're apparently running tests
+            srcdir = os.path.sep.join(localdir_s[:-1])
+
+        if srcdir:
+            # we're in the source tree, try both the debug and release
+            # variants.
+            dirs.append(sdir(srcdir, '_debug'))
+            dirs.append(sdir(srcdir, ''))
+
+        # Test for staged copies of the scripts
+        # For these, since we explicitly know if we're running a debug versus
+        # release variant, target only the relevant directory
+        if dmatch('build-root/install-vpp_debug-native/vpp/bin'):
+            srcdir = os.path.sep.join(localdir_s[:-4])
+            dirs.append(sdir(srcdir, '_debug'))
+        if dmatch('build-root/install-vpp-native/vpp/bin'):
+            srcdir = os.path.sep.join(localdir_s[:-4])
+            dirs.append(sdir(srcdir, ''))
+
+        # finally, try the location system packages typically install into
+        dirs.append(os.path.sep.join(('', 'usr', 'share', 'vpp', 'api')))
+
+        # check the directories for existance; first one wins
+        for dir in dirs:
+            if os.path.isdir(dir):
+                return dir
+
+        return None
+
+    @classmethod
+    def find_api_files(cls, api_dir=None, patterns='*'):
+        """Find API definition files from the given directory tree with the
+        given pattern. If no directory is given then find_api_dir() is used
+        to locate one. If no pattern is given then all definition files found
+        in the directory tree are used.
+
+        :param api_dir: A directory tree in which to locate API definition
+            files; subdirectories are descended into.
+            If this is None then find_api_dir() is called to discover it.
+        :param patterns: A list of patterns to use in each visited directory
+            when looking for files.
+            This can be a list/tuple object or a comma-separated string of
+            patterns. Each value in the list will have leading/trialing
+            whitespace stripped.
+            The pattern specifies the first part of the filename, '.api.json'
+            is appended.
+            The results are de-duplicated, thus overlapping patterns are fine.
+            If this is None it defaults to '*' meaning "all API files".
+        :returns: A list of file paths for the API files found.
+        """
+        if api_dir is None:
+            api_dir = cls.find_api_dir()
+            if api_dir is None:
+                raise RuntimeError("api_dir cannot be located")
+
+        if isinstance(patterns, list) or isinstance(patterns, tuple):
+            patterns = [p.strip() + '.api.json' for p in patterns]
+        else:
+            patterns = [p.strip() + '.api.json' for p in patterns.split(",")]
+
+        api_files = []
+        for root, dirnames, files in os.walk(api_dir):
+            # iterate all given patterns and de-dup the result
+            files = set(sum([fnmatch.filter(files, p) for p in patterns], []))
+            for filename in files:
+                api_files.append(os.path.join(root, filename))
+
+        return api_files
+
+    def status(self):
+        """Debug function: report current VPP API status to stdout."""
+        print('Connected') if self.connected else print('Not Connected')
+        print('Read API definitions from', ', '.join(self.apifiles))
+
+    @property
+    def api(self):
+        if not hasattr(self, "_api"):
+            raise Exception("Not connected, api definitions not available")
+        return self._api
+
+    def make_function(self, msg, i, multipart, async):
+        if (async):
+            def f(**kwargs):
+                return self._call_vpp_async(i, msg, **kwargs)
+        else:
+            def f(**kwargs):
+                return self._call_vpp(i, msg, multipart, **kwargs)
+
+        f.__name__ = str(msg.name)
+        f.__doc__ = ", ".join(["%s %s" %
+                               (msg.fieldtypes[j], k)
+                               for j, k in enumerate(msg.fields)])
+        return f
+
+    def _register_functions(self, async=False):
+        self.id_names = [None] * (self.vpp_dictionary_maxid + 1)
+        self.id_msgdef = [None] * (self.vpp_dictionary_maxid + 1)
+        self._api = Empty()
+        for name, msg in vpp_iterator(self.messages):
+            n = name + '_' + msg.crc[2:]
+            i = vpp_api.vac_get_msg_index(n.encode())
+            if i > 0:
+                self.id_msgdef[i] = msg
+                self.id_names[i] = name
+                # TODO: Fix multipart (use services)
+                multipart = True if name.find('_dump') > 0 else False
+                f = self.make_function(msg, i, multipart, async)
+                setattr(self._api, name, FuncWrapper(f))
+            else:
+                self.logger.debug(
+                    'No such message type or failed CRC checksum: %s', n)
+
+    def _write_new_cffi(self, buf):
+        """Send a binary-packed message to VPP."""
+        if not self.connected:
+            raise IOError(1, 'Not connected')
+        return vpp_api.vac_write(ffi.from_buffer(buf), len(buf))
+
+    def _write_legacy_cffi(self, buf):
+        """Send a binary-packed message to VPP."""
+        if not self.connected:
+            raise IOError(1, 'Not connected')
+        return vpp_api.vac_write(bytes(buf), len(buf))
+
+    def _read(self):
+        if not self.connected:
+            raise IOError(1, 'Not connected')
+        mem = ffi.new("char **")
+        size = ffi.new("int *")
+        rv = vpp_api.vac_read(mem, size, self.read_timeout)
+        if rv:
+            raise IOError(rv, 'vac_read failed')
+        msg = bytes(ffi.buffer(mem[0], size[0]))
+        vpp_api.vac_free(mem[0])
+        return msg
+
+    def connect_internal(self, name, msg_handler, chroot_prefix, rx_qlen,
+                         async):
+        pfx = chroot_prefix.encode() if chroot_prefix else ffi.NULL
+        rv = vpp_api.vac_connect(name.encode(), pfx, msg_handler, rx_qlen)
+        if rv != 0:
+            raise IOError(2, 'Connect failed')
+        self.connected = True
+        self.vpp_dictionary_maxid = vpp_api.vac_msg_table_max_index()
+        self._register_functions(async=async)
+
+        # Initialise control ping
+        crc = self.messages['control_ping'].crc
+        self.control_ping_index = vpp_api.vac_get_msg_index(
+            ('control_ping' + '_' + crc[2:]).encode())
+        self.control_ping_msgdef = self.messages['control_ping']
+        if self.async_thread:
+            self.event_thread = threading.Thread(
+                target=self.thread_msg_handler)
+            self.event_thread.daemon = True
+            self.event_thread.start()
+        return rv
+
+    def connect(self, name, chroot_prefix=None, async=False, rx_qlen=32):
+        """Attach to VPP.
+
+        name - the name of the client.
+        chroot_prefix - if VPP is chroot'ed, the prefix of the jail
+        async - if true, messages are sent without waiting for a reply
+        rx_qlen - the length of the VPP message receive queue between
+        client and server.
+        """
+        msg_handler = vac_callback_sync if not async else vac_callback_async
+        return self.connect_internal(name, msg_handler, chroot_prefix, rx_qlen,
+                                     async)
+
+    def connect_sync(self, name, chroot_prefix=None, rx_qlen=32):
+        """Attach to VPP in synchronous mode. Application must poll for events.
+
+        name - the name of the client.
+        chroot_prefix - if VPP is chroot'ed, the prefix of the jail
+        rx_qlen - the length of the VPP message receive queue between
+        client and server.
+        """
+
+        return self.connect_internal(name, ffi.NULL, chroot_prefix, rx_qlen,
+                                     async=False)
+
+    def disconnect(self):
+        """Detach from VPP."""
+        rv = vpp_api.vac_disconnect()
+        self.connected = False
+        self.message_queue.put("terminate event thread")
+        return rv
+
+    def msg_handler_sync(self, msg):
+        """Process an incoming message from VPP in sync mode.
+
+        The message may be a reply or it may be an async notification.
+        """
+        r = self.decode_incoming_msg(msg)
+        if r is None:
+            return
+
+        # If we have a context, then use the context to find any
+        # request waiting for a reply
+        context = 0
+        if hasattr(r, 'context') and r.context > 0:
+            context = r.context
+
+        if context == 0:
+            # No context -> async notification that we feed to the callback
+            self.message_queue.put_nowait(r)
+        else:
+            raise IOError(2, 'RPC reply message received in event handler')
+
+    def decode_incoming_msg(self, msg):
+        if not msg:
+            self.logger.warning('vpp_api.read failed')
+            return
+
+        i, ci = self.header.unpack(msg, 0)
+        if self.id_names[i] == 'rx_thread_exit':
+            return
+
+        #
+        # Decode message and returns a tuple.
+        #
+        msgobj = self.id_msgdef[i]
+        if not msgobj:
+            raise IOError(2, 'Reply message undefined')
+
+        r = msgobj.unpack(msg)
+
+        return r
+
+    def msg_handler_async(self, msg):
+        """Process a message from VPP in async mode.
+
+        In async mode, all messages are returned to the callback.
+        """
+        r = self.decode_incoming_msg(msg)
+        if r is None:
+            return
+
+        msgname = type(r).__name__
+
+        if self.event_callback:
+            self.event_callback(msgname, r)
+
+    def _control_ping(self, context):
+        """Send a ping command."""
+        self._call_vpp_async(self.control_ping_index,
+                             self.control_ping_msgdef,
+                             context=context)
+
+    def validate_args(self, msg, kwargs):
+        d = set(kwargs.keys()) - set(msg.field_by_name.keys())
+        if d:
+            raise ValueError('Invalid argument {} to {}'
+                             .format(list(d), msg.name))
+
+    def _call_vpp(self, i, msg, multipart, **kwargs):
+        """Given a message, send the message and await a reply.
+
+        msgdef - the message packing definition
+        i - the message type index
+        multipart - True if the message returns multiple
+        messages in return.
+        context - context number - chosen at random if not
+        supplied.
+        The remainder of the kwargs are the arguments to the API call.
+
+        The return value is the message or message array containing
+        the response.  It will raise an IOError exception if there was
+        no response within the timeout window.
+        """
+
+        if 'context' not in kwargs:
+            context = self.get_context()
+            kwargs['context'] = context
+        else:
+            context = kwargs['context']
+        kwargs['_vl_msg_id'] = i
+
+        self.validate_args(msg, kwargs)
+        b = msg.pack(kwargs)
+        vpp_api.vac_rx_suspend()
+        self._write(b)
+
+        if multipart:
+            # Send a ping after the request - we use its response
+            # to detect that we have seen all results.
+            self._control_ping(context)
+
+        # Block until we get a reply.
+        rl = []
+        while (True):
+            msg = self._read()
+            if not msg:
+                raise IOError(2, 'VPP API client: read failed')
+            r = self.decode_incoming_msg(msg)
+            msgname = type(r).__name__
+            if context not in r or r.context == 0 or context != r.context:
+                self.message_queue.put_nowait(r)
+                continue
+
+            if not multipart:
+                rl = r
+                break
+            if msgname == 'control_ping_reply':
+                break
+
+            rl.append(r)
+
+        vpp_api.vac_rx_resume()
+
+        return rl
+
+    def _call_vpp_async(self, i, msg, **kwargs):
+        """Given a message, send the message and await a reply.
+
+        msgdef - the message packing definition
+        i - the message type index
+        context - context number - chosen at random if not
+        supplied.
+        The remainder of the kwargs are the arguments to the API call.
+        """
+        if 'context' not in kwargs:
+            context = self.get_context()
+            kwargs['context'] = context
+        else:
+            context = kwargs['context']
+        kwargs['client_index'] = 0
+        kwargs['_vl_msg_id'] = i
+        b = msg.pack(kwargs)
+
+        self._write(b)
+
+    def register_event_callback(self, callback):
+        """Register a callback for async messages.
+
+        This will be called for async notifications in sync mode,
+        and all messages in async mode.  In sync mode, replies to
+        requests will not come here.
+
+        callback is a fn(msg_type_name, msg_type) that will be
+        called when a message comes in.  While this function is
+        executing, note that (a) you are in a background thread and
+        may wish to use threading.Lock to protect your datastructures,
+        and (b) message processing from VPP will stop (so if you take
+        a long while about it you may provoke reply timeouts or cause
+        VPP to fill the RX buffer).  Passing None will disable the
+        callback.
+        """
+        self.event_callback = callback
+
+    def thread_msg_handler(self):
+        """Python thread calling the user registerd message handler.
+
+        This is to emulate the old style event callback scheme. Modern
+        clients should provide their own thread to poll the event
+        queue.
+        """
+        while True:
+            r = self.message_queue.get()
+            if r == "terminate event thread":
+                break
+            msgname = type(r).__name__
+            if self.event_callback:
+                self.event_callback(msgname, r)
+
+
+# vim: tabstop=8 expandtab shiftwidth=4 softtabstop=4
diff --git a/src/vpp-api/python/vpp_papi/vpp_serializer.py b/src/vpp-api/python/vpp_papi/vpp_serializer.py
new file mode 100644
index 00000000000..7f5c5ace2dc
--- /dev/null
+++ b/src/vpp-api/python/vpp_papi/vpp_serializer.py
@@ -0,0 +1,343 @@
+#
+# Copyright (c) 2018 Cisco and/or its affiliates.
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at:
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+
+import struct
+import collections
+from enum import IntEnum
+import logging
+
+#
+# Set log-level in application by doing e.g.:
+# logger = logging.getLogger('vpp_serializer')
+# logger.setLevel(logging.DEBUG)
+#
+logger = logging.getLogger(__name__)
+
+
+class BaseTypes():
+    def __init__(self, type, elements=0):
+        base_types = {'u8': '>B',
+                      'u16': '>H',
+                      'u32': '>I',
+                      'i32': '>i',
+                      'u64': '>Q',
+                      'f64': '>d',
+                      'header': '>HI'}
+
+        if elements > 0 and type == 'u8':
+            self.packer = struct.Struct('>%ss' % elements)
+        else:
+            self.packer = struct.Struct(base_types[type])
+        self.size = self.packer.size
+        logger.debug('Adding {} with format: {}'
+                     .format(type, base_types[type]))
+
+    def pack(self, data, kwargs=None):
+        logger.debug("Data: {} Format: {}".format(data, self.packer.format))
+        return self.packer.pack(data)
+
+    def unpack(self, data, offset, result=None):
+        logger.debug("@ {} Format: {}".format(offset, self.packer.format))
+        return self.packer.unpack_from(data, offset)[0]
+
+
+types = {}
+types['u8'] = BaseTypes('u8')
+types['u16'] = BaseTypes('u16')
+types['u32'] = BaseTypes('u32')
+types['i32'] = BaseTypes('i32')
+types['u64'] = BaseTypes('u64')
+types['f64'] = BaseTypes('f64')
+
+
+class FixedList_u8():
+    def __init__(self, name, field_type, num):
+        self.name = name
+        self.num = num
+        self.packer = BaseTypes(field_type, num)
+        self.size = self.packer.size
+
+    def pack(self, list, kwargs):
+        """Packs a fixed length bytestring. Left-pads with zeros
+        if input data is too short."""
+        logger.debug("Data: {}".format(list))
+
+        if len(list) > self.num:
+            raise ValueError('Fixed list length error for "{}", got: {}'
+                             ' expected: {}'
+                             .format(self.name, len(list), self.num))
+        return self.packer.pack(list)
+
+    def unpack(self, data, offset=0, result=None):
+        if len(data[offset:]) < self.num:
+            raise ValueError('Invalid array length for "{}" got {}'
+                             ' expected {}'
+                             .format(self.name, len(data), self.num))
+        return self.packer.unpack(data, offset)
+
+
+class FixedList():
+    def __init__(self, name, field_type, num):
+        self.num = num
+        self.packer = types[field_type]
+        self.size = self.packer.size * num
+
+    def pack(self, list, kwargs):
+        logger.debug("Data: {}".format(list))
+
+        if len(list) != self.num:
+            raise ValueError('Fixed list length error, got: {} expected: {}'
+                             .format(len(list), self.num))
+        b = bytes()
+        for e in list:
+            b += self.packer.pack(e)
+        return b
+
+    def unpack(self, data, offset=0, result=None):
+        # Return a list of arguments
+        result = []
+        for e in range(self.num):
+            x = self.packer.unpack(data, offset)
+            result.append(x)
+            offset += self.packer.size
+        return result
+
+
+class VLAList():
+    def __init__(self, name, field_type, len_field_name, index):
+        self.index = index
+        self.packer = types[field_type]
+        self.size = self.packer.size
+        self.length_field = len_field_name
+
+    def pack(self, list, kwargs=None):
+        logger.debug("Data: {}".format(list))
+        if len(list) != kwargs[self.length_field]:
+            raise ValueError('Variable length error, got: {} expected: {}'
+                             .format(len(list), kwargs[self.length_field]))
+        b = bytes()
+
+        # u8 array
+        if self.packer.size == 1:
+            return bytearray(list)
+
+        for e in list:
+            b += self.packer.pack(e)
+        return b
+
+    def unpack(self, data, offset=0, result=None):
+        logger.debug("Data: {} @ {} Result: {}"
+                     .format(list, offset, result[self.index]))
+        # Return a list of arguments
+
+        # u8 array
+        if self.packer.size == 1:
+            if result[self.index] == 0:
+                return b''
+            p = BaseTypes('u8', result[self.index])
+            r = p.unpack(data, offset)
+            return r
+
+        r = []
+        for e in range(result[self.index]):
+            x = self.packer.unpack(data, offset)
+            r.append(x)
+            offset += self.packer.size
+        return r
+
+
+class VLAList_legacy():
+    def __init__(self, name, field_type):
+        self.packer = types[field_type]
+        self.size = self.packer.size
+
+    def pack(self, list, kwargs=None):
+        logger.debug("Data: {}".format(list))
+
+        if self.packer.size == 1:
+            return bytes(list)
+
+        b = bytes()
+        for e in list:
+            b += self.packer.pack(e)
+        return b
+
+    def unpack(self, data, offset=0, result=None):
+        # Return a list of arguments
+        if (len(data) - offset) % self.packer.size:
+            raise ValueError('Legacy Variable Length Array length mismatch.')
+        elements = int((len(data) - offset) / self.packer.size)
+        r = []
+        logger.debug("Legacy VLA: {} elements of size {}"
+                     .format(elements, self.packer.size))
+        for e in range(elements):
+            x = self.packer.unpack(data, offset)
+            r.append(x)
+            offset += self.packer.size
+        return r
+
+
+class VPPEnumType():
+    def __init__(self, name, msgdef):
+        self.size = types['u32'].size
+        e_hash = {}
+        for f in msgdef:
+            if type(f) is dict and 'enumtype' in f:
+                if f['enumtype'] != 'u32':
+                    raise NotImplementedError
+                continue
+            ename, evalue = f
+            e_hash[ename] = evalue
+        self.enum = IntEnum(name, e_hash)
+        types[name] = self
+        logger.debug('Adding enum {}'.format(name))
+
+    def __getattr__(self, name):
+        return self.enum[name]
+
+    def pack(self, data, kwargs=None):
+        logger.debug("Data: {}".format(data))
+        return types['u32'].pack(data, kwargs)
+
+    def unpack(self, data, offset=0, result=None):
+        x = types['u32'].unpack(data, offset)
+        return self.enum(x)
+
+
+class VPPUnionType():
+    def __init__(self, name, msgdef):
+        self.name = name
+        self.size = 0
+        self.maxindex = 0
+        fields = []
+        self.packers = collections.OrderedDict()
+        for i, f in enumerate(msgdef):
+            if type(f) is dict and 'crc' in f:
+                self.crc = f['crc']
+                continue
+            f_type, f_name = f
+            if f_type not in types:
+                logger.debug('Unknown union type {}'.format(f_type))
+                raise ValueError('Unknown message type {}'.format(f_type))
+            fields.append(f_name)
+            size = types[f_type].size
+            self.packers[f_name] = types[f_type]
+            if size > self.size:
+                self.size = size
+                self.maxindex = i
+
+        types[name] = self
+        self.tuple = collections.namedtuple(name, fields, rename=True)
+        logger.debug('Adding union {}'.format(name))
+
+    def pack(self, data, kwargs=None):
+        logger.debug("Data: {}".format(data))
+        for k, v in data.items():
+            logger.debug("Key: {} Value: {}".format(k, v))
+            b = self.packers[k].pack(v, kwargs)
+            offset = self.size - self.packers[k].size
+            break
+        r = bytearray(self.size)
+        r[offset:] = b
+        return r
+
+    def unpack(self, data, offset=0, result=None):
+        r = []
+        for k, p in self.packers.items():
+            union_offset = self.size - p.size
+            r.append(p.unpack(data, offset + union_offset))
+        return self.tuple._make(r)
+
+
+class VPPType():
+    # Set everything up to be able to pack / unpack
+    def __init__(self, name, msgdef):
+        self.name = name
+        self.msgdef = msgdef
+        self.packers = []
+        self.fields = []
+        self.fieldtypes = []
+        self.field_by_name = {}
+        size = 0
+        for i, f in enumerate(msgdef):
+            if type(f) is dict and 'crc' in f:
+                self.crc = f['crc']
+                continue
+            f_type, f_name = f[:2]
+            self.fields.append(f_name)
+            self.field_by_name[f_name] = None
+            self.fieldtypes.append(f_type)
+            if f_type not in types:
+                logger.debug('Unknown type {}'.format(f_type))
+                raise ValueError('Unknown message type {}'.format(f_type))
+            if len(f) == 3:  # list
+                list_elements = f[2]
+                if list_elements == 0:
+                    p = VLAList_legacy(f_name, f_type)
+                    self.packers.append(p)
+                elif f_type == 'u8':
+                    p = FixedList_u8(f_name, f_type, list_elements)
+                    self.packers.append(p)
+                    size += p.size
+                else:
+                    p = FixedList(f_name, f_type, list_elements)
+                    self.packers.append(p)
+                    size += p.size
+            elif len(f) == 4:  # Variable length list
+                    # Find index of length field
+                    length_index = self.fields.index(f[3])
+                    p = VLAList(f_name, f_type, f[3], length_index)
+                    self.packers.append(p)
+            else:
+                self.packers.append(types[f_type])
+                size += types[f_type].size
+
+        self.size = size
+        self.tuple = collections.namedtuple(name, self.fields, rename=True)
+        types[name] = self
+        logger.debug('Adding type {}'.format(name))
+
+    def pack(self, data, kwargs=None):
+        if not kwargs:
+            kwargs = data
+        logger.debug("Data: {}".format(data))
+        b = bytes()
+        for i, a in enumerate(self.fields):
+            if a not in data:
+                logger.debug("Argument {} not given, defaulting to 0"
+                             .format(a))
+                b += b'\x00' * self.packers[i].size
+                continue
+
+            if isinstance(self.packers[i], VPPType):
+                b += self.packers[i].pack(data[a], kwargs[a])
+            else:
+                b += self.packers[i].pack(data[a], kwargs)
+        return b
+
+    def unpack(self, data, offset=0, result=None):
+        # Return a list of arguments
+        result = []
+        for p in self.packers:
+            x = p.unpack(data, offset, result)
+            if type(x) is tuple and len(x) == 1:
+                x = x[0]
+            result.append(x)
+            offset += p.size
+        return self.tuple._make(result)
+
+
+class VPPMessage(VPPType):
+    pass