px4-firmware/Tools/msg/px_generate_uorb_topic_help...

344 lines
10 KiB
Python

#!/usr/bin/env python3
#############################################################################
#
# Copyright (C) 2013-2022 PX4 Pro Development Team. 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.
# 3. Neither the name PX4 nor the names of its contributors may be
# used to endorse or promote products derived from this software
# without specific prior written permission.
#
# 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.
#
#############################################################################
'''
Helper methods & common code for the uorb message templates msg.{cpp,h}.em
Another positive effect of having the code here, is that this file will get
precompiled and thus message generation will be much faster
'''
import os
import errno
import re
import genmsg.msgs
type_map = {
'int8': 'int8_t',
'int16': 'int16_t',
'int32': 'int32_t',
'int64': 'int64_t',
'uint8': 'uint8_t',
'uint16': 'uint16_t',
'uint32': 'uint32_t',
'uint64': 'uint64_t',
'float32': 'float',
'float64': 'double',
'bool': 'bool',
'char': 'char',
}
type_map_short = {
# We use some range outside of alpha-numeric and special characters.
# This needs to match with orb_get_c_type()
'int8': '\\x82',
'int16': '\\x83',
'int32': '\\x84',
'int64': '\\x85',
'uint8': '\\x86',
'uint16': '\\x87',
'uint32': '\\x88',
'uint64': '\\x89',
'float32': '\\x8a',
'float64': '\\x8b',
'bool': '\\x8c',
'char': '\\x8d',
}
type_serialize_map = {
'int8': 'int8_t',
'int16': 'int16_t',
'int32': 'int32_t',
'int64': 'int64_t',
'uint8': 'uint8_t',
'uint16': 'uint16_t',
'uint32': 'uint32_t',
'uint64': 'uint64_t',
'float32': 'float',
'float64': 'double',
'bool': 'bool',
'char': 'char',
}
type_idl_map = {
'bool': 'boolean',
'byte': 'octet',
'char': 'char',
'int8': 'octet',
'uint8': 'octet',
'int16': 'short',
'uint16': 'unsigned short',
'int32': 'long',
'uint32': 'unsigned long',
'int64': 'long long',
'uint64': 'unsigned long long',
'float32': 'float',
'float64': 'double',
'string': 'string',
}
msgtype_size_map = {
'int8': 1,
'int16': 2,
'int32': 4,
'int64': 8,
'uint8': 1,
'uint16': 2,
'uint32': 4,
'uint64': 8,
'float32': 4,
'float64': 8,
'bool': 1,
'char': 1,
}
type_printf_map = {
'int8': '%d',
'int16': '%d',
'int32': '%" PRId32 "',
'int64': '%" PRId64 "',
'uint8': '%u',
'uint16': '%u',
'uint32': '%" PRIu32 "',
'uint64': '%" PRIu64 "',
'float32': '%.4f',
'float64': '%.6f',
'bool': '%u',
'char': '%c',
}
def bare_name(msg_type):
"""
Get bare_name from <dir>/<bare_name>[x] format
"""
bare = msg_type
if '/' in msg_type:
# removing prefix
bare = (msg_type.split('/'))[1]
# removing suffix
return bare.split('[')[0]
def sizeof_field_type(field):
"""
Get size of a field, used for sorting
"""
bare_name_str = bare_name(field.type)
if bare_name_str in msgtype_size_map:
return msgtype_size_map[bare_name_str]
return 0 # this is for non-builtin types: sort them at the end
def get_children_fields(base_type, search_path):
(package, name) = genmsg.names.package_resource_name(base_type)
tmp_msg_context = genmsg.msg_loader.MsgContext.create_default()
spec_temp = genmsg.msg_loader.load_msg_by_type(
tmp_msg_context, '%s/%s' % (package, name), search_path)
return spec_temp.parsed_fields()
def get_message_fields_str_for_message_hash(msg_fields, search_path):
"""
Get all fields (including for nested types) in the form of:
'''
uint64 timestamp
uint8 esc_count
uint8 esc_online_flags
EscReport[8] esc
uint64 timestamp
uint32 esc_errorcount
int32 esc_rpm
float32 esc_voltage
uint16 failures
int8 esc_power
'''
"""
all_fields_str = ''
for field in msg_fields:
if field.is_header:
continue
type_name = field.type
# detect embedded types
sl_pos = type_name.find('/')
if sl_pos >= 0:
type_name = type_name[sl_pos + 1:]
all_fields_str += type_name + ' ' + field.name + '\n'
if sl_pos >= 0: # nested type, add all nested fields
children_fields = get_children_fields(field.base_type, search_path)
all_fields_str += get_message_fields_str_for_message_hash(children_fields, search_path)
return all_fields_str
def hash_32_fnv1a(data: str):
hash_val = 0x811c9dc5
prime = 0x1000193
for i in range(len(data)):
value = ord(data[i])
hash_val = hash_val ^ value
hash_val *= prime
hash_val &= 0xffffffff
return hash_val
def get_message_hash(msg_fields, search_path):
"""
Get a 32 bit message hash over all fields
"""
all_fields_str = get_message_fields_str_for_message_hash(msg_fields, search_path)
return hash_32_fnv1a(all_fields_str)
def add_padding_bytes(fields, search_path):
"""
Add padding fields before the embedded types, at the end and calculate the
struct size
returns a tuple with the struct size and padding at the end
"""
struct_size = 0
align_to = 8 # this is always 8, because of the 64bit timestamp
i = 0
padding_idx = 0
while i < len(fields):
field = fields[i]
if not field.is_header:
a_pos = field.type.find('[')
array_size = 1
if field.is_array:
array_size = field.array_len
if field.is_builtin:
field.sizeof_field_type = sizeof_field_type(field)
else:
# embedded type: may need to add padding
num_padding_bytes = align_to - (struct_size % align_to)
if num_padding_bytes != align_to:
padding_field = genmsg.Field('_padding' + str(padding_idx),
'uint8[' + str(num_padding_bytes) + ']')
padding_idx += 1
padding_field.sizeof_field_type = 1
struct_size += num_padding_bytes
fields.insert(i, padding_field)
i += 1
children_fields = get_children_fields(
field.base_type, search_path)
field.sizeof_field_type, unused = add_padding_bytes(children_fields,
search_path)
struct_size += field.sizeof_field_type * array_size
i += 1
# add padding at the end (necessary for embedded types)
num_padding_bytes = align_to - (struct_size % align_to)
if num_padding_bytes == align_to:
num_padding_bytes = 0
else:
padding_field = genmsg.Field('_padding' + str(padding_idx),
'uint8[' + str(num_padding_bytes) + ']')
padding_idx += 1
padding_field.sizeof_field_type = 1
struct_size += num_padding_bytes
fields.append(padding_field)
return (struct_size, num_padding_bytes)
def convert_type(spec_type, use_short_type=False):
"""
Convert from msg type to C type
"""
bare_type = spec_type
if '/' in spec_type:
# removing prefix
bare_type = (spec_type.split('/'))[1]
msg_type, is_array, array_length = genmsg.msgs.parse_type(bare_type)
c_type = msg_type
if use_short_type and msg_type in type_map_short:
c_type = type_map_short[msg_type]
elif msg_type in type_map:
c_type = type_map[msg_type]
else:
c_type = re.sub(r'(?<!^)(?=[A-Z])', '_', c_type).lower() # PascalCase to snake_case
if is_array:
return c_type + "[" + str(array_length) + "]"
return c_type
def print_field_def(field):
"""
Print the C type from a field
"""
# check if there are any upper case letters in the field name
assert not any(a.isupper()
for a in field.name), "%r field contains uppercase letters" % field.name
type_name = field.type
# detect embedded types
sl_pos = type_name.find('/')
type_appendix = ''
type_prefix = ''
if (sl_pos >= 0):
type_name = type_name[sl_pos + 1:]
type_name = re.sub(r'(?<!^)(?=[A-Z])', '_', type_name).lower() # PascalCase to snake_case
type_prefix = 'struct '
type_appendix = '_s'
# detect arrays
a_pos = type_name.find('[')
array_size = ''
if (a_pos >= 0):
# field is array
array_size = type_name[a_pos:]
type_name = type_name[:a_pos]
if type_name in type_map:
# need to add _t: int8 --> int8_t
type_px4 = type_map[type_name]
else:
type_px4 = type_name
comment = ''
if field.name.startswith('_padding'):
comment = ' // required for logger'
print(('\t%s%s%s %s%s;%s' % (type_prefix, type_px4, type_appendix, field.name,
array_size, comment)))