generate_icsneo40_structs.py

# run before: clang-format --style=mozilla -i include\ics\icsnVC40.h
import re
import os.path
from collections import OrderedDict
from subprocess import STDOUT, CalledProcessError, run, PIPE
from enum import Enum, auto
import random
from pathlib import Path
import sys
import ctypes #used with eval...
from ics_utility import GEN_ICS_DIR

debug_print = False

__unique_numbers = []


def get_unique_number():
    "Returns a unique integer"
    global __unique_numbers
    random.seed(1234)
    while True:
        value = random.randint(0, 65535)
        if value not in __unique_numbers:
            __unique_numbers.append(value)
            return value


class DataType(Enum):
    Unknown = auto()
    Union = auto()
    Struct = auto()
    Enum = auto()


class CVariable(object):
    name = ""
    data_type = None
    array_length = None
    bitfield_size = False
    is_enum = False
    enum_value = None

    def __init__(self, name, data_type, array_length, bitfield_size, enum_value=None):
        self.name = name
        self.data_type = data_type
        self.array_length = array_length
        self.bitfield_size = bitfield_size
        self.enum_value = enum_value
        self.is_enum = bool(self.enum_value)

    def __repr__(self):
        return f"<{self.__class__.__name__} {self.name} {self.data_type} @ {hex(id(self))}>"

    def to_ordered_dict(self):
        od = OrderedDict()
        od["name"] = self.name
        od["data_type"] = self.data_type
        od["array_length"] = self.array_length
        od["bitfield_size"] = self.bitfield_size
        od["is_enum"] = self.is_enum
        od["enum_value"] = self.enum_value
        return od


class CEnum(CVariable):
    def __init__(self, name, value):
        super().__init__(name, "int", 0, 0, value)


class CObject(object):
    def __init__(self):
        self.data_type = DataType.Unknown
        self.is_anonymous = False
        self.packing = None
        self.preferred_name = ""
        self.names = []
        # list of CObject and CObjectMember
        self.members = []
        self.enum_last_value = None

    def __repr__(self):
        if self.names:
            name = self.names[-1]
        else:
            name = "Anon"

        if self.data_type == DataType.Struct:
            t_name = "Struct"
        elif self.data_type == DataType.Union:
            t_name = "Union"
        elif self.data_type == DataType.Enum:
            t_name = "Enum"
        else:
            t_name = "Unknown"

        return f"<{self.__class__.__name__} {name} {t_name} {len(self.members)} members @ {hex(id(self))}>"

    def to_ordered_dict(self):
        od = OrderedDict()
        if self.data_type == DataType.Struct:
            t_name = "Struct"
        elif self.data_type == DataType.Union:
            t_name = "Union"
        elif self.data_type == DataType.Enum:
            t_name = "Enum"
        else:
            t_name = "Unknown"

        od["names"] = self.names
        od["data_type"] = t_name
        od["packing"] = self.packing
        od["is_anonymous"] = self.is_anonymous
        od["members"] = []
        for member in self.members:
            od["members"].append(member.to_ordered_dict())
        return od

    def assign_preferred_name(self, is_embedded=False):
        """This assigns a name to the object if we are nameless and anonymous, safe to call multiple times."""
        self.preferred_name = get_preferred_struct_name(self.names)
        # Any object that starts with _ won't be imported with import * statement so lets push it to the end
        if self.preferred_name and self.preferred_name.startswith("_"):
            modified_name = reverse_leading_underscores(self.preferred_name)
            self.names.append(modified_name)
            self.preferred_name = get_preferred_struct_name(self.names)
        if self.is_anonymous and not self.preferred_name:
            self.preferred_name = f"Nameless{get_unique_number()}"
        if self.preferred_name and not is_embedded:
            self.preferred_name = convert_to_snake_case(self.preferred_name)


_start_of_obj_blacklist = (re.compile(r"""struct [aA-zZ\\_]+ [aA-zZ\\_]+;"""),)  # struct some_random_struct_name value;


def is_line_start_of_object(line):
    "Returns True if we are a c object (enum, struct, union)"
    for regex in _start_of_obj_blacklist:
        if bool(regex.search(line)):
            return False
    return bool(re.search(r"\btypedef struct$|struct$|struct \S*$|enum$|enum |union$|union ", line))


# This contains all the objects that don't pass convert_to_ctype_object
NON_CTYPE_OBJ_NAMES = []
# This contains a list of every object we collected
ALL_C_OBJECTS = []


def get_object_from_name(name):
    global ALL_C_OBJECTS
    for obj in ALL_C_OBJECTS:
        if name == obj.preferred_name or name in obj.names:
            return obj
    return None


def parse_object(f, pos=-1, pack_size=None, is_embedded=False):
    """
    takes a file object and returns a class with the object data.
    if pos is -1, don't reset position when finished
    """
    start_pos = f.tell()
    opening_bracket_count = 0
    try:
        # Read the first line and make sure we have a C Object
        line = f.readline()
        if not line:
            raise RuntimeError("Reached end of file when we shouldn't have!")
        # Remove new lines
        line = line.strip()
        if not is_line_start_of_object(line):
            raise RuntimeError("Current line is not the declaration of a C Object (enum/struct/union)")
        # Lets start parsing the lines for our C Object
        new_obj = CObject()
        new_obj.packing = pack_size
        # Grab the object name
        name = re.sub(r"typedef|struct|enum|union|\{|\s*", "", line)
        # Only append the name if its not anonymous
        if name:
            new_obj.names.append(name)
        # Determine the type of object
        if "struct" in line:
            new_obj.data_type = DataType.Struct
        elif "union" in line:
            new_obj.data_type = DataType.Union
        elif "enum" in line:
            new_obj.data_type = DataType.Enum
        else:
            new_obj.data_type = DataType.Unknown
        # Time to parse the rest of the object
        last_pos = f.tell()
        line = f.readline()
        finished = False
        while line:
            try:
                # Remove new lines
                line = line.strip()
                if not line:
                    continue
                # Parse any embedded objects
                if is_line_start_of_object(line):
                    f.seek(last_pos)
                    embedded_object = parse_object(f, -1, pack_size, True)
                    new_obj.members.append(embedded_object)
                    continue
                opening_bracket_count += line.count("{")
                opening_bracket_count -= line.count("}")
                assert opening_bracket_count >= 0
                # Determine if we are at the end of the struct
                if opening_bracket_count == 0 and re.match(r"}.*;", line):
                    extra_names = "".join(line.split()).strip("};").split(",")
                    try:
                        # Remove dangling empty string
                        extra_names.remove("")
                    except ValueError:
                        pass
                    if extra_names:
                        new_obj.names.extend(extra_names)
                    # We are finally done
                    finished = True
                    break
                # Nothing to do with this line anymore
                if re.match(r".*{.*", line):
                    continue
                # Parse the member
                if new_obj.data_type == DataType.Enum:
                    member = CEnum(*parse_enum_member(line))
                elif new_obj.data_type == DataType.Struct:
                    member = CVariable(*parse_struct_member(line))
                elif new_obj.data_type == DataType.Union:
                    member = CVariable(*parse_struct_member(line))
                new_obj.members.append(member)
            finally:
                # Grab the position before we read the next line
                if not finished:
                    last_pos = f.tell()
                    line = f.readline()
        new_obj.is_anonymous = not bool(new_obj.names)
        new_obj.assign_preferred_name(is_embedded)
        # Append the objects to a global list for parsing later
        global ALL_C_OBJECTS
        ALL_C_OBJECTS.append(new_obj)
        return new_obj
    finally:
        if pos != -1:
            f.seek(pos)
    raise RuntimeError("parse_object(): Failure. This is a bug and we shouldn't be here!")


def reverse_leading_underscores(value: str):
    """Take a string starting with underscores and moves them to the end. Returns a string"""
    if value.startswith("_"):
        underscore_count = len(value) - len(value.lstrip("_"))
        return value.lstrip("_") + "_" * underscore_count
    # We didn't have underscores in the value
    return value


def parse_struct_member(buffered_line):
    # unsigned : 31;
    # unsigned char bIPV6_Address[16];
    # unsigned asdf[12];
    # unsigned : 15;
    # unsigned test : 1;
    # uint8_t test;
    # uint8_t test2[12];
    # STRUCT_TYPE asdf;
    # unsigned char test
    #        [16]; // why not make something defined like this?
    # uint8_t data[4 * 1024]; /* The data */

    # remove all unneeded whitespace
    # print("DEBUG BEFORE:", buffered_line)
    # if len(buffered_line.split('\n')):
    #    new_buffered_line = ''
    #    for line in buffered_line.split('\n'):
    #        new_buffered_line += re.sub('{|union|}|;', '', line) + '\n'
    #    buffered_line = new_buffered_line
    buffered_line = re.sub(r"\s*\[", "[", " ".join(buffered_line.split()))
    # print("DEBUG AFTER:", buffered_line)

    # Figure out if we are an array type and get the array length
    array_subsection = re.search(r"\[.*\]", buffered_line)
    is_array = array_subsection is not None
    if is_array:
        array_length = int(eval(array_subsection.group(0).strip("[]")))
    else:
        array_length = 0
    # Remove the array portion
    buffered_line = re.sub(r"\[.*\]", "", buffered_line)
    # split up the remaining
    words = buffered_line.split()
    if not len(words):
        return None, "", 0, 0  # data_type, array_length, bitwise_length
    if ":" in words:
        # we are a bitfield ;(
        try:
            bitwise_length = int(re.search(r"\d*", words[words.index(":") + 1]).group(0))
        except Exception as ex:
            if debug_print:
                print("EXCEPTION:", ex)
            bitwise_length = 0
        # see if we get a valid ctype object before : to check for things like "unsigned : 31;"
        pre_colon_text = " ".join(words[: words.index(":")])
        check_data_type = convert_to_ctype_object(pre_colon_text)
        if check_data_type:
            data_name = ""
            data_type = pre_colon_text
        else:
            data_name_index = words.index(":") - 1
            data_name = words[data_name_index]
            data_type = " ".join(words[0:data_name_index])
    else:
        bitwise_length = 0
        data_name = words[-1]
        data_type = " ".join(words[0 : len(words) - 1])
        if data_name.startswith("*"):
            data_type += " *"
            data_name = data_name.lstrip("*")
    # DEBUG: ['uint8_t', 'MACAddress[6];']
    # print("DEBUG:", words, buffered_line)

    # remove stuff that shouldn't be in the name
    data_name = re.sub(r"{|{|}|\s*", "", data_name)
    data_type = re.sub(r"union|{|{|}|^struct", "", data_type)
    data_type = data_type.strip()
    if not data_type:
        data_name = ""
    # Any object that starts with _ won't be imported with import * statement so lets push it to the end
    if data_type.startswith("_") and not convert_to_ctype_object(data_type):
        data_type = reverse_leading_underscores(data_type)
    return re.sub(r"\[.*\]|;", "", data_name), data_type, array_length, bitwise_length


def parse_enum_member(buffered_line):
    # VARIABLE_NAME,
    # VARIABLE_NAME = 0,

    # remove all unneeded whitespace
    buffered_line = re.sub(r"\s*\[", "[", " ".join(buffered_line.split()))

    if "=" in buffered_line:
        name = buffered_line.split("=")[0]
        value = buffered_line.split("=")[1]
        value = re.sub(r"{|{|}|,|\s*", "", value)
        try:
            value = int(value)
        except ValueError as ex:
            if "0X" in value.upper():
                value = int(value, 16)
            elif "0B" in value.upper():
                value = int(value, 2)
    else:
        name = re.sub(r",", "", buffered_line)
        value = None
    return name, value


def get_struct_name_from_header(line):
    # line.split('struct ')[-1]
    struct_name = re.sub(r"typedef|struct|enum|union|\{|\s*", "", line)
    if not struct_name:  # == 'typedef struct':
        anonymous_struct = True
        struct_name = "anonymous"
    elif "{" == struct_name:
        anonymous_struct = True
        struct_name = "anonymous"
    else:
        anonymous_struct = False
    return struct_name, anonymous_struct


def get_struct_names(data):
    names = OrderedDict()
    # Pack all the names into dictionary
    # key = the "key" in the data
    # values = list of other names
    for name in data.keys():
        if name in ("names", "pack"):
            continue
        if "names" in data[name]:
            names[name] = data[name]["names"]
        else:
            names[name] = [
                name,
            ]
    return names


def get_preferred_struct_name(_names):
    # Remove the reference
    names = _names[:]
    r = re.compile(r"^[sS].*")
    r_underscore = re.compile(r"^(?![_]).*")
    r_end_t = re.compile(r"^.*(_[tT])$")
    # Find any that don't start with _ for _t matching below
    struct_name_no_underscore = list(filter(r_underscore.match, names))
    # Remove all instances of _t at end unless that is all we have
    struct_name_with_t = list(filter(r_end_t.match, struct_name_no_underscore))
    if len(struct_name_with_t):
        if len(struct_name_with_t) != len(struct_name_no_underscore):
            for name in struct_name_with_t:
                names.remove(name)
    # Find any that start with 'S' and prefer that
    struct_name_with_s = list(filter(r.match, names))
    if struct_name_with_s:
        return struct_name_with_s[0]
    # Find any that don't start with _ and prefer that
    struct_name_no_underscore = list(filter(r_underscore.match, names))
    if struct_name_no_underscore:
        return struct_name_no_underscore[0]
    try:
        return names[0]
    except IndexError as _:
        # We didn't find any names to deal with...
        return None


def convert_to_snake_case(name):
    # return name
    s1 = re.sub(r"(.)([A-Z][a-z]+)", r"\1_\2", name)
    s2 = re.sub(r"([a-z0-9])([A-Z])", r"\1_\2", s1).lower()
    return re.sub(r"(_)\1{1,}", "_", s2)


def convert_to_ctype_object(data_type):
    ctype_types = {
        "c_bool": ("_Bool", "bool"),
        "c_char": ("char",),
        "c_wchar": ("wchar_t",),
        "c_byte": (),
        "c_ubyte": ("unsigned char",),
        "c_short": ("short",),
        "c_ushort": ("unsigned short",),
        "c_int": ("int",),
        "c_uint": (
            "unsigned",
            "unsigned int",
        ),
        "c_long": ("long",),
        "c_ulong": ("unsigned long",),
        "c_longlong": (
            "__int64",
            "long long",
        ),
        "c_ulonglong": (
            "unsigned __int64",
            "unsigned long long",
        ),
        "c_size_t": ("size_t",),
        "c_ssize_t": (
            "ssize_t",
            "Py_ssize_t",
        ),
        "c_float": ("float",),
        "c_double": ("double",),
        "c_void_p": (
            "void*",
            "void *",
        ),
    }
    # Add all the intX_t types
    for d in [8, 16, 32, 64]:
        ctype_types[f"c_int{d}"] = (f"int{d}_t",)
        ctype_types[f"c_uint{d}"] = (f"uint{d}_t",)

    # This is dirty but we don't parse typedefs...
    # descIdType is int16_t in icsnVC40.h (EXTERNAL_PROJECT build)
    ctype_types["c_int16"] = ctype_types["c_int16"] + ("descIdType",)
    ctype_types["c_uint64"] = ctype_types["c_uint64"] + ("_clock_identity",)

    is_pointer = "*" in data_type and not "void" in data_type
    for ctype_type, c_types in ctype_types.items():
        for c_type in c_types:
            if is_pointer:
                data_type = data_type.replace("*", "")
            if c_type == data_type:
                try:
                    t = f"ctypes.{ctype_type}"
                    eval(t)
                    if is_pointer:
                        t = f"ctypes.POINTER({t})"
                    return t
                except AttributeError as ex:
                    return None
    return None


def format_file(filename):
    import os

    processed_fname = os.path.basename(filename)
    name, ext = os.path.splitext(processed_fname)
    processed_fname = f"{name}_processed{ext}"
    # processed_fname = "icsnVC40_processed.h"
    # Run it through the preprocessor
    # clang -E -P .\include\ics\icsnVC40.h -o output.h
    result = run(
        ["clang", "-DEXTERNAL_PROJECT", "-E", "-P", filename, "-o", processed_fname], stdout=PIPE, stderr=STDOUT
    )
    try:
        result.check_returncode()
    except CalledProcessError as ex:
        print(f"ERROR: {' '.join(result.args)} failed with error code {ex.returncode}")
        print(ex.stdout.decode("UTF-8"))
        # print(ex.stderr)

    # Format the file
    result = run(
        ["clang-format", "-i", "--style", "{BasedOnStyle: Mozilla, ColumnLimit: '200'}", processed_fname],
        stdout=PIPE,
        stderr=STDOUT,
    )
    try:
        result.check_returncode()
    except CalledProcessError as ex:
        print(f"ERROR: {' '.join(result.args)} failed with error code {ex.returncode}")
        print(ex.stdout.decode("UTF-8"))

    return processed_fname


def parse_header_file(filename):
    with open(filename, "r") as f:
        pack_size = 0
        line_count = 0
        last_pos = f.tell()
        c_objects = []
        enum_objects = []
        pack_size_stack = []
        line = f.readline()
        while line:
            try:
                # Remove newlines
                line = line.strip()
                line_count += 1
                # Remove all unneeded whitespace
                line = " ".join(line.split())
                # Remove preprocessor statements
                if line.startswith("#"):
                    # get the pack size
                    if re.match(r"#pragma.*pack.*.*pop.*", line):
                        if pack_size_stack:
                            pack_size = pack_size_stack.pop()
                        else:
                            pack_size = 0
                    elif re.match(r"#pragma.*pack.*.*push.*", line) or re.match(r"#pragma.*pack\(\d\)", line):
                        pushing = "push" in line
                        try:
                            last = pack_size
                            pack_size = int(re.search(r"\d{1,}", line).group(0))
                            if pushing:
                                pack_size_stack.append(last)
                        except AttributeError:
                            if not pushing:
                                raise
                            pack_size_stack.append(pack_size)
                        if debug_print:
                            print("PACK SIZE:", pack_size)
                    elif (
                        re.match(r"#define.*", line)
                        and len(line.split(" ")) == 3
                        and not "\\" in line
                        and not "sizeof" in line
                    ):
                        # preprocessor define that can be used as a number in array sizes...
                        # we are going to hack the variable into the module's globals so we can eval() it later
                        words = line.split(" ")
                        if debug_print:
                            print("ADDING {} to globals with value of {}".format(words[1], words[2]))
                        try:
                            globals()[words[1]] = eval(words[2])
                        except SyntaxError as ex:
                            # This happens when we have an integer literal
                            # https://en.cppreference.com/w/cpp/language/integer_literal
                            # TODO: 0x1p5
                            literals = ("u", "U", "l", "L", "z", "Z", "f", "F")
                            modified_word = words[2]
                            while modified_word.endswith(literals):
                                for literal in literals:
                                    modified_word = modified_word.rstrip(literal)
                            globals()[words[1]] = eval(modified_word)
                        except NameError as ex:
                            # Python interpreter can't parse this word value so lets convert it to a string and be done...
                            # Line: #57: "#define __INTMAX_C_SUFFIX__ LL"
                            globals()[words[1]] = eval(f'"{modified_word}"')

                    continue
                # Process the object
                if is_line_start_of_object(line):
                    pos = f.tell()
                    f.seek(last_pos)
                    obj = parse_object(f, -1, pack_size)
                    if obj.data_type == DataType.Struct:
                        obj.packing = pack_size
                    if obj.data_type == DataType.Enum:
                        enum_objects.append(obj)
                    else:
                        c_objects.append(obj)
            except Exception as ex:
                print(f'Line: #{line_count}: "{line}"')
                raise ex
            finally:
                last_pos = f.tell()
                line = f.readline()
        return c_objects, enum_objects


def generate(filename="include/ics/icsnVC40.h"):
    import shutil
    import json
    import os

    basename = os.path.basename(filename)
    filename = format_file(filename)
    c_objects, enum_objects = parse_header_file(filename)
    # make the json file
    c_objects_converted = []
    enum_objects_converted = []
    for c_object in c_objects:
        c_objects_converted.append(c_object.to_ordered_dict())
    for enum_object in enum_objects:
        enum_objects_converted.append(enum_object.to_ordered_dict())
    j = json.dumps(c_objects_converted, indent=4, sort_keys=False)
    with open(f"{basename}.json", "w+") as f:
        f.write(j)
    j = json.dumps(enum_objects_converted, indent=4, sort_keys=False)
    with open(f"{basename}.enums.json", "w+") as f:
        f.write(j)
    # generate the python files
    output_dir = GEN_ICS_DIR / "structures"
    print(f"Removing {output_dir}...")
    try:
        shutil.rmtree(output_dir)
    except FileNotFoundError:
        pass
    ignore_names = [
        "fsid_t__",
        "__fsid_t",
        "__darwin_pthread_handler_rec",
        "_mbstate_t",
        "mbstate_t_",
        "mbstatet_",
        "Mbstatet_",
        "_Mbstatet",
        "_LONGDOUBLE",
        "LONGDOUBLE_",
        "_opaque_pthread_attr_t",
        "_opaque_pthread_cond_t",
        "_opaque_pthread_condattr_t",
        "_opaque_pthread_mutex_t",
        "_opaque_pthread_mutexattr_t",
        "_opaque_pthread_once_t",
        "_opaque_pthread_rwlock_t",
        "_opaque_pthread_rwlockattr_t",
        "_opaque_pthread_t",
        "crt_locale_pointers_",
        "crt_locale_data_public_",
        "__crt_locale_data_public",
        "ldiv_t",
        "lldiv_t",
        "ldouble_",
        "ldbl12_",
        "div_t",
        "crt_float_",
        "crt_double_",
        "ndis_adapter_information",
        #"NeoDevice",
        #"neo_device",
        #"NeoDeviceEx",
        #"neo_device_ex",
        "icsSpyMessage",
        "icsSpyMessageJ1850",
        "ics_spy_message",
        "ics_spy_message_j1850",
    ]
    file_names = []
    prefered_names = []
    all_objects = c_objects + enum_objects
    print(f"Generating python files objects...")
    ignored_enum_count = 0
    for c_object in all_objects:
        # TODO: Bypass all anonymous enums
        if c_object.data_type == DataType.Enum:
            if c_object.is_anonymous:
                ignored_enum_count += 1
                continue
        # Bypass all the ignore_names above
        names = c_object.names
        names.append(c_object.preferred_name)
        if set(names) & set(ignore_names):
            continue
        # Generate the python file for the c_object
        file_name, file_path = generate_pyfile(c_object, output_dir)
        if debug_print:
            print(file_name, file_path)
        file_names.append(file_name)

    # Verify we don't have any unknown data types here
    global NON_CTYPE_OBJ_NAMES
    global ALL_C_OBJECTS
    errors = False
    for obj_name in NON_CTYPE_OBJ_NAMES:
        obj = get_object_from_name(obj_name)
        if not obj:
            print(f"Warning: Not a valid object: {obj_name}")
            errors = False
    if errors:
        raise RuntimeError("Failed to parse all objects properly")
    else:
        print(f"Generated all python {len(all_objects)-ignored_enum_count} files.")

    # Generate __init__.py and add all the modules to __all__
    with open(os.path.join(output_dir, "__init__.py"), "w+") as f:
        f.write("__all__ = [\n")
        for file_name in file_names:
            fname = re.sub(r"(\.py)", "", file_name)
            r = re.compile(r"(" + fname + ")")
            if list(filter(r.match, ignore_names)):
                # print("IGNORING:", fname)
                continue
            f.write('    "')
            f.write(fname)
            f.write('",\n')
        f.write("]\n")
    # write a hidden_import python file for pyinstaller
    hidden_imports_path = GEN_ICS_DIR / "hiddenimports.py"
    with open(hidden_imports_path, "w+") as f:
        f.write("hidden_imports = [\n")
        for file_name in file_names:
            fname = re.sub(r"(\.py)", "", file_name)
            r = re.compile(r"(" + fname + ")")
            if list(filter(r.match, ignore_names)):
                # print("IGNORING:", fname)
                continue
            f.write(f'    "ics.structures.{fname}",\n')
        f.write("]\n\n")

    validate_generated_modules(file_names)
    print("\nDone.")


def validate_generated_modules(file_names):
    """Check imports without keeping an existing native extension loaded in setup."""
    import json

    # Importing ics also loads ics.ics when it exists. Windows cannot replace a
    # loaded .pyd, so wait for a separate interpreter to exit before building.
    # Pass paths and names as data, including paths containing spaces/backslashes.
    validation_code = """
import importlib
import json
import sys

package_path, file_names = json.load(sys.stdin)
sys.path.insert(0, package_path)
for file_name in file_names:
    if file_name.startswith("__"):
        continue
    module_name = "ics.structures." + file_name.removesuffix(".py")
    print(f"Importing / Verifying {module_name}...", flush=True)
    importlib.import_module(module_name)
"""
    run(
        [sys.executable, "-c", validation_code],
        input=json.dumps([str(GEN_ICS_DIR.parent.resolve()), file_names]),
        text=True,
        check=True,
    )


def _write_c_object(f, c_object):
    # Write the header
    if c_object.data_type == DataType.Struct:
        f.write(f"class {c_object.preferred_name}(ctypes.Structure):\n")
    elif c_object.data_type == DataType.Union:
        f.write(f"class {c_object.preferred_name}(ctypes.Union):\n")
    elif c_object.data_type == DataType.Enum:
        f.write(f"class {c_object.preferred_name}(enum.IntEnum):\n")
        f.write('    """A ctypes-compatible IntEnum superclass."""\n')
        f.write("    @classmethod\n")
        f.write("    def from_param(cls, obj):\n")
        f.write("        return int(obj)\n")
        f.write("\n")
    else:
        raise ValueError("CObject is not of a known data type!")

    # Write the rest of the header for structs/unions
    if c_object.data_type in (DataType.Struct, DataType.Union):
        # Setup the packing
        if c_object.packing:
            f.write(f"    _pack_ = {c_object.packing}\n")
        # Grab all the anonymous names
        anonymous_names = []
        for member in c_object.members:
            if isinstance(member, CObject) and member.is_anonymous:
                anonymous_names.append(member.preferred_name)
        if anonymous_names:
            f.write(f"    _anonymous_  = {str(tuple(anonymous_names))}\n")
        f.write(f"    _fields_ = [\n")

    # Write the members
    for member in c_object.members:
        if c_object.data_type == DataType.Enum:
            enum_value = member.enum_value
            if enum_value == None:
                if c_object.enum_last_value != None:
                    enum_value = "enum.auto()"
                    c_object.enum_last_value += 1
                else:
                    # https://docs.python.org/3/library/enum.html#enum.auto
                    # By default, the initial value starts at 1.
                    c_object.enum_last_value = 0
                    enum_value = "0"
            elif c_object.enum_last_value == None:
                c_object.enum_last_value = 0
            f.write(f"    {member.name} = {enum_value}\n")
        else:
            # Struct/Union
            def _write_member(f, member, is_struct_or_union=False):
                # Get the ctypes data type
                if not is_struct_or_union:
                    data_type = convert_to_ctype_object(member.data_type)
                    # If we aren't a valid ctypes data type we are probably a struct
                    if not data_type:
                        # print(f"Warning: Couldn't find a valid ctype type for '{member.data_type}' in '{member.name}'")
                        global NON_CTYPE_OBJ_NAMES
                        NON_CTYPE_OBJ_NAMES.append(member.data_type)
                        data_type = member.data_type
                    obj = get_object_from_name(member.data_type)
                    if obj and obj.data_type == DataType.Enum:
                        # C enum types can be char, unsigned, signed but seem to default to
                        # 4 byte integer on most systems (even 64-bit)
                        # This is a potential hole but nothing we can do here
                        data_type = "ctypes.c_int32"
                        # Use the unsigned variant when any enumerator needs the
                        # full 32 bits (e.g. DeviceSettingsNone = 0xFFFFFFFF) so
                        # the value round-trips through the field.
                        for enum_member in obj.members:
                            if isinstance(enum_member.enum_value, int) and enum_member.enum_value > 0x7FFFFFFF:
                                data_type = "ctypes.c_uint32"
                                break
                else:
                    data_type = member.data_type
                if member.bitfield_size:
                    f.write(f"        ('{member.name}', {data_type}, {member.bitfield_size}),\n")
                elif member.array_length:
                    f.write(f"        ('{member.name}', {data_type} * {member.array_length}),\n")
                else:
                    f.write(f"        ('{member.name}', {data_type}),\n")

            if isinstance(member, CVariable):
                _write_member(f, member, False)
            elif isinstance(member, CObject):
                _name = member.preferred_name
                _created_member = CVariable(_name, _name, 0, 0)
                _write_member(f, _created_member, True)

    # Finalize the _fields_ attribute and extra names
    if c_object.data_type in (DataType.Struct, DataType.Union):
        f.write(f"    ]\n")
    # Extra names here
    f.write("\n\n")
    for name in c_object.names:
        # Ignore the actual object name
        if name == c_object.preferred_name:
            continue
        f.write("{} = {}\n".format(re.sub(r"^\*", "", name), c_object.preferred_name))
    f.write("\n")


def generate_pyfile(c_object, path):
    # name = get_preferred_struct_name(c_object.names)
    # c_object.preferred_name = convert_to_snake_case(name)
    # Make the fname and the path
    fname = f"{c_object.preferred_name}.py"
    fname_with_path = os.path.normpath(os.path.join(path, fname))
    # create the needed directories
    if not os.path.exists(os.path.split(fname_with_path)[0]):
        try:
            os.makedirs(os.path.split(fname_with_path)[0])
        except OSError as ex:
            import errno

            if ex.errno != errno.EEXIST:
                # Race condition handler, if someone else makes a directory after we check if exists
                raise
    with open(fname_with_path, "w+") as f:
        # Write the boiler plate code
        f.write("# This file was auto generated; Do not modify, if you value your sanity!\n")
        f.write("import ctypes\n")
        f.write("import enum\n")
        f.write("\n")
        # Generate all the imports

        def get_c_object_imports(c_object):
            import_names = []
            for member in c_object.members:
                # If CVariable and not a datatype then its a custom struct/union
                if isinstance(member, CVariable):
                    is_ctype_type = bool(convert_to_ctype_object(member.data_type))
                    if not is_ctype_type:
                        # Attempt to get the preferred name here.
                        obj = get_object_from_name(member.data_type)
                        if obj:
                            actual_data_type = obj.preferred_name
                        else:
                            actual_data_type = member.data_type
                        import_names.append(convert_to_snake_case(actual_data_type))
                elif isinstance(member, CObject):
                    import_names.extend(get_c_object_imports(member))
            # anonymous/nameless objects put an empty string in the list, lets remove it here
            import_names = [name for name in import_names if name]
            return sorted(set(import_names))

        import_names = get_c_object_imports(c_object)
        for import_name in import_names:
            f.write(f"from ics.structures.{import_name} import *\n")
        f.write("\n\n")

        def _generate_inner_objects(f, c_object):
            for member in c_object.members:
                if isinstance(member, CObject):
                    _generate_inner_objects(f, member)
                    _write_c_object(f, member)

        # generate all the inner objects
        _generate_inner_objects(f, c_object)
        # Finally write our main object
        _write_c_object(f, c_object)
    return fname, fname_with_path

def generate_all_files():
    import sys
    import os
    import pathlib

    print(f"Creating directory '{GEN_ICS_DIR}'...")
    GEN_ICS_DIR.mkdir(parents=True, exist_ok=True)
    filenames = ("icsnVC40.h", "icsnVC40Internal.h")
    for filename in filenames:
        path = pathlib.Path("include/ics/")
        path = path.joinpath(filename)
        if path.exists():
            if "Internal" in str(filename):
                print("WARNING: Generating internal header!")
            print(f"Parsing {str(path)}...")
            generate(str(path))

if __name__ == "__main__":
    generate_all_files()