virtualx-engine/doc/tools/make_rst.py

#!/usr/bin/env python3

# This script makes RST files from the XML class reference for use with the online docs.

import argparse
import os
import platform
import re
import sys
import xml.etree.ElementTree as ET
from collections import OrderedDict

# Uncomment to do type checks. I have it commented out so it works below Python 3.5
# from typing import List, Dict, TextIO, Tuple, Iterable, Optional, DefaultDict, Any, Union

# $DOCS_URL/path/to/page.html(#fragment-tag)
GODOT_DOCS_PATTERN = re.compile(r"^\$DOCS_URL/(.*)\.html(#.*)?$")

# Based on reStructedText inline markup recognition rules
# https://docutils.sourceforge.io/docs/ref/rst/restructuredtext.html#inline-markup-recognition-rules
MARKUP_ALLOWED_PRECEDENT = " -:/'\"<([{"
MARKUP_ALLOWED_SUBSEQUENT = " -.,:;!?\\/'\")]}>"

# Used to translate section headings and other hardcoded strings when required with
# the --lang argument. The BASE_STRINGS list should be synced with what we actually
# write in this script (check `translate()` uses), and also hardcoded in
# `doc/translations/extract.py` to include them in the source POT file.
BASE_STRINGS = [
    "Description",
    "Tutorials",
    "Properties",
    "Constructors",
    "Methods",
    "Operators",
    "Theme Properties",
    "Signals",
    "Enumerations",
    "Constants",
    "Annotations",
    "Property Descriptions",
    "Constructor Descriptions",
    "Method Descriptions",
    "Operator Descriptions",
    "Theme Property Descriptions",
    "Inherits:",
    "Inherited By:",
    "(overrides %s)",
    "Default",
    "Setter",
    "value",
    "Getter",
    "This method should typically be overridden by the user to have any effect.",
    "This method has no side effects. It doesn't modify any of the instance's member variables.",
    "This method accepts any number of arguments after the ones described here.",
    "This method is used to construct a type.",
    "This method doesn't need an instance to be called, so it can be called directly using the class name.",
    "This method describes a valid operator to use with this type as left-hand operand.",
]
strings_l10n = {}

STYLES = {}


def print_error(error, state):  # type: (str, State) -> None
    print("{}{}ERROR:{} {}{}".format(STYLES["red"], STYLES["bold"], STYLES["regular"], error, STYLES["reset"]))
    state.num_errors += 1


class TypeName:
    def __init__(self, type_name, enum=None):  # type: (str, Optional[str]) -> None
        self.type_name = type_name
        self.enum = enum

    def to_rst(self, state):  # type: ("State") -> str
        if self.enum is not None:
            return make_enum(self.enum, state)
        elif self.type_name == "void":
            return "void"
        else:
            return make_type(self.type_name, state)

    @classmethod
    def from_element(cls, element):  # type: (ET.Element) -> "TypeName"
        return cls(element.attrib["type"], element.get("enum"))


class PropertyDef:
    def __init__(
        self, name, type_name, setter, getter, text, default_value, overrides
    ):  # type: (str, TypeName, Optional[str], Optional[str], Optional[str], Optional[str], Optional[str]) -> None
        self.name = name
        self.type_name = type_name
        self.setter = setter
        self.getter = getter
        self.text = text
        self.default_value = default_value
        self.overrides = overrides


class ParameterDef:
    def __init__(self, name, type_name, default_value):  # type: (str, TypeName, Optional[str]) -> None
        self.name = name
        self.type_name = type_name
        self.default_value = default_value


class SignalDef:
    def __init__(self, name, parameters, description):  # type: (str, List[ParameterDef], Optional[str]) -> None
        self.name = name
        self.parameters = parameters
        self.description = description


class MethodDef:
    def __init__(
        self, name, return_type, parameters, description, qualifiers
    ):  # type: (str, TypeName, List[ParameterDef], Optional[str], Optional[str]) -> None
        self.name = name
        self.return_type = return_type
        self.parameters = parameters
        self.description = description
        self.qualifiers = qualifiers


class ConstantDef:
    def __init__(self, name, value, text, bitfield):  # type: (str, str, Optional[str], Optional[bool]) -> None
        self.name = name
        self.value = value
        self.text = text
        self.is_bitfield = bitfield


class EnumDef:
    def __init__(self, name, bitfield):  # type: (str, Optional[bool]) -> None
        self.name = name
        self.values = OrderedDict()  # type: OrderedDict[str, ConstantDef]
        self.is_bitfield = bitfield


class ThemeItemDef:
    def __init__(
        self, name, type_name, data_name, text, default_value
    ):  # type: (str, TypeName, str, Optional[str], Optional[str]) -> None
        self.name = name
        self.type_name = type_name
        self.data_name = data_name
        self.text = text
        self.default_value = default_value


class ClassDef:
    def __init__(self, name):  # type: (str) -> None
        self.name = name
        self.constants = OrderedDict()  # type: OrderedDict[str, ConstantDef]
        self.enums = OrderedDict()  # type: OrderedDict[str, EnumDef]
        self.properties = OrderedDict()  # type: OrderedDict[str, PropertyDef]
        self.constructors = OrderedDict()  # type: OrderedDict[str, List[MethodDef]]
        self.methods = OrderedDict()  # type: OrderedDict[str, List[MethodDef]]
        self.operators = OrderedDict()  # type: OrderedDict[str, List[MethodDef]]
        self.signals = OrderedDict()  # type: OrderedDict[str, SignalDef]
        self.annotations = OrderedDict()  # type: OrderedDict[str, List[MethodDef]]
        self.theme_items = OrderedDict()  # type: OrderedDict[str, ThemeItemDef]
        self.inherits = None  # type: Optional[str]
        self.brief_description = None  # type: Optional[str]
        self.description = None  # type: Optional[str]
        self.tutorials = []  # type: List[Tuple[str, str]]

        # Used to match the class with XML source for output filtering purposes.
        self.filepath = ""  # type: str


class State:
    def __init__(self):  # type: () -> None
        self.num_errors = 0
        self.classes = OrderedDict()  # type: OrderedDict[str, ClassDef]
        self.current_class = ""  # type: str

    def parse_class(self, class_root, filepath):  # type: (ET.Element, str) -> None
        class_name = class_root.attrib["name"]

        class_def = ClassDef(class_name)
        self.classes[class_name] = class_def
        class_def.filepath = filepath

        inherits = class_root.get("inherits")
        if inherits is not None:
            class_def.inherits = inherits

        brief_desc = class_root.find("brief_description")
        if brief_desc is not None and brief_desc.text:
            class_def.brief_description = brief_desc.text

        desc = class_root.find("description")
        if desc is not None and desc.text:
            class_def.description = desc.text

        properties = class_root.find("members")
        if properties is not None:
            for property in properties:
                assert property.tag == "member"

                property_name = property.attrib["name"]
                if property_name in class_def.properties:
                    print_error('{}.xml: Duplicate property "{}".'.format(class_name, property_name), self)
                    continue

                type_name = TypeName.from_element(property)
                setter = property.get("setter") or None  # Use or None so '' gets turned into None.
                getter = property.get("getter") or None
                default_value = property.get("default") or None
                if default_value is not None:
                    default_value = "``{}``".format(default_value)
                overrides = property.get("overrides") or None

                property_def = PropertyDef(
                    property_name, type_name, setter, getter, property.text, default_value, overrides
                )
                class_def.properties[property_name] = property_def

        constructors = class_root.find("constructors")
        if constructors is not None:
            for constructor in constructors:
                assert constructor.tag == "constructor"

                method_name = constructor.attrib["name"]
                qualifiers = constructor.get("qualifiers")

                return_element = constructor.find("return")
                if return_element is not None:
                    return_type = TypeName.from_element(return_element)

                else:
                    return_type = TypeName("void")

                params = parse_arguments(constructor)

                desc_element = constructor.find("description")
                method_desc = None
                if desc_element is not None:
                    method_desc = desc_element.text

                method_def = MethodDef(method_name, return_type, params, method_desc, qualifiers)
                if method_name not in class_def.constructors:
                    class_def.constructors[method_name] = []

                class_def.constructors[method_name].append(method_def)

        methods = class_root.find("methods")
        if methods is not None:
            for method in methods:
                assert method.tag == "method"

                method_name = method.attrib["name"]
                qualifiers = method.get("qualifiers")

                return_element = method.find("return")
                if return_element is not None:
                    return_type = TypeName.from_element(return_element)

                else:
                    return_type = TypeName("void")

                params = parse_arguments(method)

                desc_element = method.find("description")
                method_desc = None
                if desc_element is not None:
                    method_desc = desc_element.text

                method_def = MethodDef(method_name, return_type, params, method_desc, qualifiers)
                if method_name not in class_def.methods:
                    class_def.methods[method_name] = []

                class_def.methods[method_name].append(method_def)

        operators = class_root.find("operators")
        if operators is not None:
            for operator in operators:
                assert operator.tag == "operator"

                method_name = operator.attrib["name"]
                qualifiers = operator.get("qualifiers")

                return_element = operator.find("return")
                if return_element is not None:
                    return_type = TypeName.from_element(return_element)

                else:
                    return_type = TypeName("void")

                params = parse_arguments(operator)

                desc_element = operator.find("description")
                method_desc = None
                if desc_element is not None:
                    method_desc = desc_element.text

                method_def = MethodDef(method_name, return_type, params, method_desc, qualifiers)
                if method_name not in class_def.operators:
                    class_def.operators[method_name] = []

                class_def.operators[method_name].append(method_def)

        constants = class_root.find("constants")
        if constants is not None:
            for constant in constants:
                assert constant.tag == "constant"

                constant_name = constant.attrib["name"]
                value = constant.attrib["value"]
                enum = constant.get("enum")
                is_bitfield = constant.get("is_bitfield") or False
                constant_def = ConstantDef(constant_name, value, constant.text, is_bitfield)
                if enum is None:
                    if constant_name in class_def.constants:
                        print_error('{}.xml: Duplicate constant "{}".'.format(class_name, constant_name), self)
                        continue

                    class_def.constants[constant_name] = constant_def

                else:
                    if enum in class_def.enums:
                        enum_def = class_def.enums[enum]

                    else:
                        enum_def = EnumDef(enum, is_bitfield)
                        class_def.enums[enum] = enum_def

                    enum_def.values[constant_name] = constant_def

        annotations = class_root.find("annotations")
        if annotations is not None:
            for annotation in annotations:
                assert annotation.tag == "annotation"

                annotation_name = annotation.attrib["name"]
                qualifiers = annotation.get("qualifiers")

                params = parse_arguments(annotation)

                desc_element = annotation.find("description")
                annotation_desc = None
                if desc_element is not None:
                    annotation_desc = desc_element.text

                annotation_def = MethodDef(annotation_name, return_type, params, annotation_desc, qualifiers)
                if annotation_name not in class_def.annotations:
                    class_def.annotations[annotation_name] = []

                class_def.annotations[annotation_name].append(annotation_def)

        signals = class_root.find("signals")
        if signals is not None:
            for signal in signals:
                assert signal.tag == "signal"

                signal_name = signal.attrib["name"]

                if signal_name in class_def.signals:
                    print_error('{}.xml: Duplicate signal "{}".'.format(class_name, signal_name), self)
                    continue

                params = parse_arguments(signal)

                desc_element = signal.find("description")
                signal_desc = None
                if desc_element is not None:
                    signal_desc = desc_element.text

                signal_def = SignalDef(signal_name, params, signal_desc)
                class_def.signals[signal_name] = signal_def

        theme_items = class_root.find("theme_items")
        if theme_items is not None:
            for theme_item in theme_items:
                assert theme_item.tag == "theme_item"

                theme_item_name = theme_item.attrib["name"]
                theme_item_data_name = theme_item.attrib["data_type"]
                theme_item_id = "{}_{}".format(theme_item_data_name, theme_item_name)
                if theme_item_id in class_def.theme_items:
                    print_error(
                        '{}.xml: Duplicate theme item "{}" of type "{}".'.format(
                            class_name, theme_item_name, theme_item_data_name
                        ),
                        self,
                    )
                    continue

                default_value = theme_item.get("default") or None
                if default_value is not None:
                    default_value = "``{}``".format(default_value)

                theme_item_def = ThemeItemDef(
                    theme_item_name,
                    TypeName.from_element(theme_item),
                    theme_item_data_name,
                    theme_item.text,
                    default_value,
                )
                class_def.theme_items[theme_item_name] = theme_item_def

        tutorials = class_root.find("tutorials")
        if tutorials is not None:
            for link in tutorials:
                assert link.tag == "link"

                if link.text is not None:
                    class_def.tutorials.append((link.text.strip(), link.get("title", "")))

    def sort_classes(self):  # type: () -> None
        self.classes = OrderedDict(sorted(self.classes.items(), key=lambda t: t[0]))


def parse_arguments(root):  # type: (ET.Element) -> List[ParameterDef]
    param_elements = root.findall("argument")
    params = [None] * len(param_elements)  # type: Any
    for param_element in param_elements:
        param_name = param_element.attrib["name"]
        index = int(param_element.attrib["index"])
        type_name = TypeName.from_element(param_element)
        default = param_element.get("default")

        params[index] = ParameterDef(param_name, type_name, default)

    cast = params  # type: List[ParameterDef]

    return cast


def main():  # type: () -> None
    # Enable ANSI escape code support on Windows 10 and later (for colored console output).
    # <https://bugs.python.org/issue29059>
    if platform.system().lower() == "windows":
        from ctypes import windll, c_int, byref

        stdout_handle = windll.kernel32.GetStdHandle(c_int(-11))
        mode = c_int(0)
        windll.kernel32.GetConsoleMode(c_int(stdout_handle), byref(mode))
        mode = c_int(mode.value | 4)
        windll.kernel32.SetConsoleMode(c_int(stdout_handle), mode)

    parser = argparse.ArgumentParser()
    parser.add_argument("path", nargs="+", help="A path to an XML file or a directory containing XML files to parse.")
    parser.add_argument("--filter", default="", help="The filepath pattern for XML files to filter.")
    parser.add_argument("--lang", "-l", default="en", help="Language to use for section headings.")
    parser.add_argument(
        "--color",
        action="store_true",
        help="If passed, force colored output even if stdout is not a TTY (useful for continuous integration).",
    )
    group = parser.add_mutually_exclusive_group()
    group.add_argument("--output", "-o", default=".", help="The directory to save output .rst files in.")
    group.add_argument(
        "--dry-run",
        action="store_true",
        help="If passed, no output will be generated and XML files are only checked for errors.",
    )
    args = parser.parse_args()

    should_color = args.color or (hasattr(sys.stdout, "isatty") and sys.stdout.isatty())
    STYLES["red"] = "\x1b[91m" if should_color else ""
    STYLES["green"] = "\x1b[92m" if should_color else ""
    STYLES["bold"] = "\x1b[1m" if should_color else ""
    STYLES["regular"] = "\x1b[22m" if should_color else ""
    STYLES["reset"] = "\x1b[0m" if should_color else ""

    # Retrieve heading translations for the given language.
    if not args.dry_run and args.lang != "en":
        lang_file = os.path.join(
            os.path.dirname(os.path.realpath(__file__)), "..", "translations", "{}.po".format(args.lang)
        )
        if os.path.exists(lang_file):
            try:
                import polib
            except ImportError:
                print("Base template strings localization requires `polib`.")
                exit(1)

            pofile = polib.pofile(lang_file)
            for entry in pofile.translated_entries():
                if entry.msgid in BASE_STRINGS:
                    strings_l10n[entry.msgid] = entry.msgstr
        else:
            print('No PO file at "{}" for language "{}".'.format(lang_file, args.lang))

    print("Checking for errors in the XML class reference...")

    file_list = []  # type: List[str]

    for path in args.path:
        # Cut off trailing slashes so os.path.basename doesn't choke.
        if path.endswith("/") or path.endswith("\\"):
            path = path[:-1]

        if os.path.basename(path) == "modules":
            for subdir, dirs, _ in os.walk(path):
                if "doc_classes" in dirs:
                    doc_dir = os.path.join(subdir, "doc_classes")
                    class_file_names = (f for f in os.listdir(doc_dir) if f.endswith(".xml"))
                    file_list += (os.path.join(doc_dir, f) for f in class_file_names)

        elif os.path.isdir(path):
            file_list += (os.path.join(path, f) for f in os.listdir(path) if f.endswith(".xml"))

        elif os.path.isfile(path):
            if not path.endswith(".xml"):
                print('Got non-.xml file "{}" in input, skipping.'.format(path))
                continue

            file_list.append(path)

    classes = {}  # type: Dict[str, ET.Element]
    state = State()

    for cur_file in file_list:
        try:
            tree = ET.parse(cur_file)
        except ET.ParseError as e:
            print_error("{}: Parse error while reading the file: {}".format(cur_file, e), state)
            continue
        doc = tree.getroot()

        if "version" not in doc.attrib:
            print_error('{}: "version" attribute missing from "doc".'.format(cur_file), state)
            continue

        name = doc.attrib["name"]
        if name in classes:
            print_error('{}: Duplicate class "{}".'.format(cur_file, name), state)
            continue

        classes[name] = (doc, cur_file)

    for name, data in classes.items():
        try:
            state.parse_class(data[0], data[1])
        except Exception as e:
            print_error("{}.xml: Exception while parsing class: {}".format(name, e), state)

    state.sort_classes()

    pattern = re.compile(args.filter)

    # Create the output folder recursively if it doesn't already exist.
    os.makedirs(args.output, exist_ok=True)

    for class_name, class_def in state.classes.items():
        if args.filter and not pattern.search(class_def.filepath):
            continue
        state.current_class = class_name
        make_rst_class(class_def, state, args.dry_run, args.output)

    if state.num_errors == 0:
        print("{}No errors found in the class reference XML.{}".format(STYLES["green"], STYLES["reset"]))
        if not args.dry_run:
            print("Wrote reStructuredText files for each class to: %s" % args.output)
    else:
        if state.num_errors >= 2:
            print(
                "{}{} errors were found in the class reference XML. Please check the messages above.{}".format(
                    STYLES["red"], state.num_errors, STYLES["reset"]
                )
            )
        else:
            print(
                "{}1 error was found in the class reference XML. Please check the messages above.{}".format(
                    STYLES["red"], STYLES["reset"]
                )
            )
        exit(1)


def translate(string):  # type: (str) -> str
    """Translate a string based on translations sourced from `doc/translations/*.po`
    for a language if defined via the --lang command line argument.
    Returns the original string if no translation exists.
    """
    return strings_l10n.get(string, string)


def make_rst_class(class_def, state, dry_run, output_dir):  # type: (ClassDef, State, bool, str) -> None
    class_name = class_def.name

    if dry_run:
        f = open(os.devnull, "w", encoding="utf-8")
    else:
        f = open(os.path.join(output_dir, "class_" + class_name.lower() + ".rst"), "w", encoding="utf-8")

    # Warn contributors not to edit this file directly
    f.write(":github_url: hide\n\n")
    f.write(".. Generated automatically by doc/tools/make_rst.py in Godot's source tree.\n")
    f.write(".. DO NOT EDIT THIS FILE, but the " + class_name + ".xml source instead.\n")
    f.write(".. The source is found in doc/classes or modules/<name>/doc_classes.\n\n")

    f.write(".. _class_" + class_name + ":\n\n")
    f.write(make_heading(class_name, "=", False))

    # Inheritance tree
    # Ascendants
    if class_def.inherits:
        inherits = class_def.inherits.strip()
        f.write("**" + translate("Inherits:") + "** ")
        first = True
        while inherits in state.classes:
            if not first:
                f.write(" **<** ")
            else:
                first = False

            f.write(make_type(inherits, state))
            inode = state.classes[inherits].inherits
            if inode:
                inherits = inode.strip()
            else:
                break
        f.write("\n\n")

    # Descendents
    inherited = []
    for c in state.classes.values():
        if c.inherits and c.inherits.strip() == class_name:
            inherited.append(c.name)

    if len(inherited):
        f.write("**" + translate("Inherited By:") + "** ")
        for i, child in enumerate(inherited):
            if i > 0:
                f.write(", ")
            f.write(make_type(child, state))
        f.write("\n\n")

    # Brief description
    if class_def.brief_description is not None:
        f.write(rstize_text(class_def.brief_description.strip(), state) + "\n\n")

    # Class description
    if class_def.description is not None and class_def.description.strip() != "":
        f.write(make_heading("Description", "-"))
        f.write(rstize_text(class_def.description.strip(), state) + "\n\n")

    # Online tutorials
    if len(class_def.tutorials) > 0:
        f.write(make_heading("Tutorials", "-"))
        for url, title in class_def.tutorials:
            f.write("- " + make_link(url, title) + "\n\n")

    # Properties overview
    if len(class_def.properties) > 0:
        f.write(make_heading("Properties", "-"))
        ml = []  # type: List[Tuple[str, str, str]]
        for property_def in class_def.properties.values():
            type_rst = property_def.type_name.to_rst(state)
            default = property_def.default_value
            if default is not None and property_def.overrides:
                ref = ":ref:`{1}<class_{1}_property_{0}>`".format(property_def.name, property_def.overrides)
                # Not using translate() for now as it breaks table formatting.
                ml.append((type_rst, property_def.name, default + " " + "(overrides %s)" % ref))
            else:
                ref = ":ref:`{0}<class_{1}_property_{0}>`".format(property_def.name, class_name)
                ml.append((type_rst, ref, default))
        format_table(f, ml, True)

    # Constructors, Methods, Operators overview
    if len(class_def.constructors) > 0:
        f.write(make_heading("Constructors", "-"))
        ml = []
        for method_list in class_def.constructors.values():
            for m in method_list:
                ml.append(make_method_signature(class_def, m, "constructor", state))
        format_table(f, ml)

    if len(class_def.methods) > 0:
        f.write(make_heading("Methods", "-"))
        ml = []
        for method_list in class_def.methods.values():
            for m in method_list:
                ml.append(make_method_signature(class_def, m, "method", state))
        format_table(f, ml)

    if len(class_def.operators) > 0:
        f.write(make_heading("Operators", "-"))
        ml = []
        for method_list in class_def.operators.values():
            for m in method_list:
                ml.append(make_method_signature(class_def, m, "operator", state))
        format_table(f, ml)

    # Theme properties
    if len(class_def.theme_items) > 0:
        f.write(make_heading("Theme Properties", "-"))
        pl = []
        for theme_item_def in class_def.theme_items.values():
            ref = ":ref:`{0}<class_{2}_theme_{1}_{0}>`".format(
                theme_item_def.name, theme_item_def.data_name, class_name
            )
            pl.append((theme_item_def.type_name.to_rst(state), ref, theme_item_def.default_value))
        format_table(f, pl, True)

    # Signals
    if len(class_def.signals) > 0:
        f.write(make_heading("Signals", "-"))
        index = 0

        for signal in class_def.signals.values():
            if index != 0:
                f.write("----\n\n")

            f.write(".. _class_{}_signal_{}:\n\n".format(class_name, signal.name))
            _, signature = make_method_signature(class_def, signal, "", state)
            f.write("- {}\n\n".format(signature))

            if signal.description is not None and signal.description.strip() != "":
                f.write(rstize_text(signal.description.strip(), state) + "\n\n")

            index += 1

    # Enums
    if len(class_def.enums) > 0:
        f.write(make_heading("Enumerations", "-"))
        index = 0

        for e in class_def.enums.values():
            if index != 0:
                f.write("----\n\n")

            f.write(".. _enum_{}_{}:\n\n".format(class_name, e.name))
            # Sphinx seems to divide the bullet list into individual <ul> tags if we weave the labels into it.
            # As such I'll put them all above the list. Won't be perfect but better than making the list visually broken.
            # As to why I'm not modifying the reference parser to directly link to the _enum label:
            # If somebody gets annoyed enough to fix it, all existing references will magically improve.
            for value in e.values.values():
                f.write(".. _class_{}_constant_{}:\n\n".format(class_name, value.name))

            if e.is_bitfield:
                f.write("flags **{}**:\n\n".format(e.name))
            else:
                f.write("enum **{}**:\n\n".format(e.name))

            for value in e.values.values():
                f.write("- **{}** = **{}**".format(value.name, value.value))
                if value.text is not None and value.text.strip() != "":
                    # If value.text contains a bullet point list, each entry needs additional indentation
                    f.write(" --- " + indent_bullets(rstize_text(value.text.strip(), state)))

                f.write("\n\n")

            index += 1

    # Constants
    if len(class_def.constants) > 0:
        f.write(make_heading("Constants", "-"))
        # Sphinx seems to divide the bullet list into individual <ul> tags if we weave the labels into it.
        # As such I'll put them all above the list. Won't be perfect but better than making the list visually broken.
        for constant in class_def.constants.values():
            f.write(".. _class_{}_constant_{}:\n\n".format(class_name, constant.name))

        for constant in class_def.constants.values():
            f.write("- **{}** = **{}**".format(constant.name, constant.value))
            if constant.text is not None and constant.text.strip() != "":
                f.write(" --- " + rstize_text(constant.text.strip(), state))

            f.write("\n\n")

    if len(class_def.annotations) > 0:
        f.write(make_heading("Annotations", "-"))
        index = 0

        for method_list in class_def.annotations.values():
            for i, m in enumerate(method_list):
                if index != 0:
                    f.write("----\n\n")

                if i == 0:
                    f.write(".. _class_{}_annotation_{}:\n\n".format(class_name, m.name.strip("@")))

                ret_type, signature = make_method_signature(class_def, m, "", state)
                f.write("- {} {}\n\n".format(ret_type, signature))

                if m.description is not None and m.description.strip() != "":
                    f.write(rstize_text(m.description.strip(), state) + "\n\n")

                index += 1

    # Property descriptions
    if any(not p.overrides for p in class_def.properties.values()) > 0:
        f.write(make_heading("Property Descriptions", "-"))
        index = 0

        for property_def in class_def.properties.values():
            if property_def.overrides:
                continue

            if index != 0:
                f.write("----\n\n")

            f.write(".. _class_{}_property_{}:\n\n".format(class_name, property_def.name))
            f.write("- {} **{}**\n\n".format(property_def.type_name.to_rst(state), property_def.name))

            info = []
            # Not using translate() for now as it breaks table formatting.
            if property_def.default_value is not None:
                info.append(("*" + "Default" + "*", property_def.default_value))
            if property_def.setter is not None and not property_def.setter.startswith("_"):
                info.append(("*" + "Setter" + "*", property_def.setter + "(" + "value" + ")"))
            if property_def.getter is not None and not property_def.getter.startswith("_"):
                info.append(("*" + "Getter" + "*", property_def.getter + "()"))

            if len(info) > 0:
                format_table(f, info)

            if property_def.text is not None and property_def.text.strip() != "":
                f.write(rstize_text(property_def.text.strip(), state) + "\n\n")

            index += 1

    # Constructor, Method, Operator descriptions
    if len(class_def.constructors) > 0:
        f.write(make_heading("Constructor Descriptions", "-"))
        index = 0

        for method_list in class_def.constructors.values():
            for i, m in enumerate(method_list):
                if index != 0:
                    f.write("----\n\n")

                if i == 0:
                    f.write(".. _class_{}_constructor_{}:\n\n".format(class_name, m.name))

                ret_type, signature = make_method_signature(class_def, m, "", state)
                f.write("- {} {}\n\n".format(ret_type, signature))

                if m.description is not None and m.description.strip() != "":
                    f.write(rstize_text(m.description.strip(), state) + "\n\n")

                index += 1

    if len(class_def.methods) > 0:
        f.write(make_heading("Method Descriptions", "-"))
        index = 0

        for method_list in class_def.methods.values():
            for i, m in enumerate(method_list):
                if index != 0:
                    f.write("----\n\n")

                if i == 0:
                    f.write(".. _class_{}_method_{}:\n\n".format(class_name, m.name))

                ret_type, signature = make_method_signature(class_def, m, "", state)
                f.write("- {} {}\n\n".format(ret_type, signature))

                if m.description is not None and m.description.strip() != "":
                    f.write(rstize_text(m.description.strip(), state) + "\n\n")

                index += 1

    if len(class_def.operators) > 0:
        f.write(make_heading("Operator Descriptions", "-"))
        index = 0

        for method_list in class_def.operators.values():
            for i, m in enumerate(method_list):
                if index != 0:
                    f.write("----\n\n")

                if i == 0:
                    f.write(
                        ".. _class_{}_operator_{}_{}:\n\n".format(
                            class_name, sanitize_operator_name(m.name, state), m.return_type.type_name
                        )
                    )

                ret_type, signature = make_method_signature(class_def, m, "", state)
                f.write("- {} {}\n\n".format(ret_type, signature))

                if m.description is not None and m.description.strip() != "":
                    f.write(rstize_text(m.description.strip(), state) + "\n\n")

                index += 1

    # Theme property descriptions
    if len(class_def.theme_items) > 0:
        f.write(make_heading("Theme Property Descriptions", "-"))
        index = 0

        for theme_item_def in class_def.theme_items.values():
            if index != 0:
                f.write("----\n\n")

            f.write(".. _class_{}_theme_{}_{}:\n\n".format(class_name, theme_item_def.data_name, theme_item_def.name))
            f.write("- {} **{}**\n\n".format(theme_item_def.type_name.to_rst(state), theme_item_def.name))

            info = []
            if theme_item_def.default_value is not None:
                # Not using translate() for now as it breaks table formatting.
                info.append(("*" + "Default" + "*", theme_item_def.default_value))

            if len(info) > 0:
                format_table(f, info)

            if theme_item_def.text is not None and theme_item_def.text.strip() != "":
                f.write(rstize_text(theme_item_def.text.strip(), state) + "\n\n")

            index += 1

    f.write(make_footer())


def escape_rst(text, until_pos=-1):  # type: (str) -> str
    # Escape \ character, otherwise it ends up as an escape character in rst
    pos = 0
    while True:
        pos = text.find("\\", pos, until_pos)
        if pos == -1:
            break
        text = text[:pos] + "\\\\" + text[pos + 1 :]
        pos += 2

    # Escape * character to avoid interpreting it as emphasis
    pos = 0
    while True:
        pos = text.find("*", pos, until_pos)
        if pos == -1:
            break
        text = text[:pos] + "\*" + text[pos + 1 :]
        pos += 2

    # Escape _ character at the end of a word to avoid interpreting it as an inline hyperlink
    pos = 0
    while True:
        pos = text.find("_", pos, until_pos)
        if pos == -1:
            break
        if not text[pos + 1].isalnum():  # don't escape within a snake_case word
            text = text[:pos] + "\_" + text[pos + 1 :]
            pos += 2
        else:
            pos += 1

    return text


def format_codeblock(code_type, post_text, indent_level, state):  # types: str, str, int, state
    end_pos = post_text.find("[/" + code_type + "]")
    if end_pos == -1:
        print_error("{}.xml: [" + code_type + "] without a closing tag.".format(state.current_class), state)
        return None

    code_text = post_text[len("[" + code_type + "]") : end_pos]
    post_text = post_text[end_pos:]

    # Remove extraneous tabs
    code_pos = 0
    while True:
        code_pos = code_text.find("\n", code_pos)
        if code_pos == -1:
            break

        to_skip = 0
        while code_pos + to_skip + 1 < len(code_text) and code_text[code_pos + to_skip + 1] == "\t":
            to_skip += 1

        if to_skip > indent_level:
            print_error(
                "{}.xml: Four spaces should be used for indentation within [{}].".format(
                    state.current_class, code_type
                ),
                state,
            )

        if len(code_text[code_pos + to_skip + 1 :]) == 0:
            code_text = code_text[:code_pos] + "\n"
            code_pos += 1
        else:
            code_text = code_text[:code_pos] + "\n    " + code_text[code_pos + to_skip + 1 :]
            code_pos += 5 - to_skip
    return ["\n[" + code_type + "]" + code_text + post_text, len("\n[" + code_type + "]" + code_text)]


def rstize_text(text, state):  # type: (str, State) -> str
    # Linebreak + tabs in the XML should become two line breaks unless in a "codeblock"
    pos = 0
    while True:
        pos = text.find("\n", pos)
        if pos == -1:
            break

        pre_text = text[:pos]
        indent_level = 0
        while pos + 1 < len(text) and text[pos + 1] == "\t":
            pos += 1
            indent_level += 1
        post_text = text[pos + 1 :]

        # Handle codeblocks
        if (
            post_text.startswith("[codeblock]")
            or post_text.startswith("[gdscript]")
            or post_text.startswith("[csharp]")
        ):
            block_type = post_text[1:].split("]")[0]
            result = format_codeblock(block_type, post_text, indent_level, state)
            if result is None:
                return ""
            text = pre_text + result[0]
            pos += result[1] - indent_level

        # Handle normal text
        else:
            text = pre_text + "\n\n" + post_text
            pos += 2 - indent_level

    next_brac_pos = text.find("[")
    text = escape_rst(text, next_brac_pos)

    # Handle [tags]
    inside_code = False
    pos = 0
    tag_depth = 0
    previous_pos = 0
    while True:
        pos = text.find("[", pos)
        if pos == -1:
            break

        endq_pos = text.find("]", pos + 1)
        if endq_pos == -1:
            break

        pre_text = text[:pos]
        post_text = text[endq_pos + 1 :]
        tag_text = text[pos + 1 : endq_pos]

        escape_pre = False
        escape_post = False

        if tag_text in state.classes:
            if tag_text == state.current_class:
                # We don't want references to the same class
                tag_text = "``{}``".format(tag_text)
            else:
                tag_text = make_type(tag_text, state)
            escape_pre = True
            escape_post = True
        else:  # command
            cmd = tag_text
            space_pos = tag_text.find(" ")
            if cmd == "/codeblock" or cmd == "/gdscript" or cmd == "/csharp":
                tag_text = ""
                tag_depth -= 1
                inside_code = False
                # Strip newline if the tag was alone on one
                if pre_text[-1] == "\n":
                    pre_text = pre_text[:-1]
            elif cmd == "/code":
                tag_text = "``"
                tag_depth -= 1
                inside_code = False
                escape_post = True
            elif inside_code:
                tag_text = "[" + tag_text + "]"
            elif cmd.find("html") == 0:
                param = tag_text[space_pos + 1 :]
                tag_text = param
            elif (
                cmd.startswith("method")
                or cmd.startswith("member")
                or cmd.startswith("signal")
                or cmd.startswith("constant")
                or cmd.startswith("theme_item")
            ):
                param = tag_text[space_pos + 1 :]

                if param.find(".") != -1:
                    ss = param.split(".")
                    if len(ss) > 2:
                        print_error('{}.xml: Bad reference: "{}".'.format(state.current_class, param), state)
                    class_param, method_param = ss

                else:
                    class_param = state.current_class
                    method_param = param

                ref_type = ""
                if class_param in state.classes:
                    class_def = state.classes[class_param]
                    if cmd.startswith("constructor"):
                        if method_param not in class_def.constructors:
                            print_error(
                                '{}.xml: Unresolved constructor "{}".'.format(state.current_class, param), state
                            )
                        ref_type = "_constructor"
                    if cmd.startswith("method"):
                        if method_param not in class_def.methods:
                            print_error('{}.xml: Unresolved method "{}".'.format(state.current_class, param), state)
                        ref_type = "_method"
                    if cmd.startswith("operator"):
                        if method_param not in class_def.operators:
                            print_error('{}.xml: Unresolved operator "{}".'.format(state.current_class, param), state)
                        ref_type = "_operator"

                    elif cmd.startswith("member"):
                        if method_param not in class_def.properties:
                            print_error('{}.xml: Unresolved member "{}".'.format(state.current_class, param), state)
                        ref_type = "_property"

                    elif cmd.startswith("theme_item"):
                        if method_param not in class_def.theme_items:
                            print_error('{}.xml: Unresolved theme item "{}".'.format(state.current_class, param), state)
                        ref_type = "_theme_{}".format(class_def.theme_items[method_param].data_name)

                    elif cmd.startswith("signal"):
                        if method_param not in class_def.signals:
                            print_error('{}.xml: Unresolved signal "{}".'.format(state.current_class, param), state)
                        ref_type = "_signal"

                    elif cmd.startswith("annotation"):
                        if method_param not in class_def.annotations:
                            print_error('{}.xml: Unresolved annotation "{}".'.format(state.current_class, param), state)
                        ref_type = "_annotation"

                    elif cmd.startswith("constant"):
                        found = False

                        # Search in the current class
                        search_class_defs = [class_def]

                        if param.find(".") == -1:
                            # Also search in @GlobalScope as a last resort if no class was specified
                            search_class_defs.append(state.classes["@GlobalScope"])

                        for search_class_def in search_class_defs:
                            if method_param in search_class_def.constants:
                                class_param = search_class_def.name
                                found = True

                            else:
                                for enum in search_class_def.enums.values():
                                    if method_param in enum.values:
                                        class_param = search_class_def.name
                                        found = True
                                        break

                        if not found:
                            print_error('{}.xml: Unresolved constant "{}".'.format(state.current_class, param), state)
                        ref_type = "_constant"

                else:
                    print_error(
                        '{}.xml: Unresolved type reference "{}" in method reference "{}".'.format(
                            state.current_class, class_param, param
                        ),
                        state,
                    )

                repl_text = method_param
                if class_param != state.current_class:
                    repl_text = "{}.{}".format(class_param, method_param)
                tag_text = ":ref:`{}<class_{}{}_{}>`".format(repl_text, class_param, ref_type, method_param)
                escape_pre = True
                escape_post = True
            elif cmd.find("image=") == 0:
                tag_text = ""  # '![](' + cmd[6:] + ')'
            elif cmd.find("url=") == 0:
                # URLs are handled in full here as we need to extract the optional link
                # title to use `make_link`.
                link_url = cmd[4:]
                endurl_pos = text.find("[/url]", endq_pos + 1)
                if endurl_pos == -1:
                    print_error(
                        "{}.xml: Tag depth mismatch for [url]: no closing [/url]".format(state.current_class), state
                    )
                    break
                link_title = text[endq_pos + 1 : endurl_pos]
                tag_text = make_link(link_url, link_title)

                pre_text = text[:pos]
                post_text = text[endurl_pos + 6 :]

                if pre_text and pre_text[-1] not in MARKUP_ALLOWED_PRECEDENT:
                    pre_text += "\ "
                if post_text and post_text[0] not in MARKUP_ALLOWED_SUBSEQUENT:
                    post_text = "\ " + post_text

                text = pre_text + tag_text + post_text
                pos = len(pre_text) + len(tag_text)
                previous_pos = pos
                continue
            elif cmd == "center":
                tag_depth += 1
                tag_text = ""
            elif cmd == "/center":
                tag_depth -= 1
                tag_text = ""
            elif cmd == "codeblock":
                tag_depth += 1
                tag_text = "\n::\n"
                inside_code = True
            elif cmd == "gdscript":
                tag_depth += 1
                tag_text = "\n .. code-tab:: gdscript\n"
                inside_code = True
            elif cmd == "csharp":
                tag_depth += 1
                tag_text = "\n .. code-tab:: csharp\n"
                inside_code = True
            elif cmd == "codeblocks":
                tag_depth += 1
                tag_text = "\n.. tabs::"
            elif cmd == "/codeblocks":
                tag_depth -= 1
                tag_text = ""
            elif cmd == "br":
                # Make a new paragraph instead of a linebreak, rst is not so linebreak friendly
                tag_text = "\n\n"
                # Strip potential leading spaces
                while post_text[0] == " ":
                    post_text = post_text[1:]
            elif cmd == "i" or cmd == "/i":
                if cmd == "/i":
                    tag_depth -= 1
                    escape_post = True
                else:
                    tag_depth += 1
                    escape_pre = True
                tag_text = "*"
            elif cmd == "b" or cmd == "/b":
                if cmd == "/b":
                    tag_depth -= 1
                    escape_post = True
                else:
                    tag_depth += 1
                    escape_pre = True
                tag_text = "**"
            elif cmd == "u" or cmd == "/u":
                if cmd == "/u":
                    tag_depth -= 1
                    escape_post = True
                else:
                    tag_depth += 1
                    escape_pre = True
                tag_text = ""
            elif cmd == "code":
                tag_text = "``"
                tag_depth += 1
                inside_code = True
                escape_pre = True
            elif cmd == "kbd":
                tag_text = ":kbd:`"
                tag_depth += 1
                escape_pre = True
            elif cmd == "/kbd":
                tag_text = "`"
                tag_depth -= 1
                escape_post = True
            elif cmd.startswith("enum "):
                tag_text = make_enum(cmd[5:], state)
                escape_pre = True
                escape_post = True
            else:
                tag_text = make_type(tag_text, state)
                escape_pre = True
                escape_post = True

        # Properly escape things like `[Node]s`
        if escape_pre and pre_text and pre_text[-1] not in MARKUP_ALLOWED_PRECEDENT:
            pre_text += "\ "
        if escape_post and post_text and post_text[0] not in MARKUP_ALLOWED_SUBSEQUENT:
            post_text = "\ " + post_text

        next_brac_pos = post_text.find("[", 0)
        iter_pos = 0
        while not inside_code:
            iter_pos = post_text.find("*", iter_pos, next_brac_pos)
            if iter_pos == -1:
                break
            post_text = post_text[:iter_pos] + "\*" + post_text[iter_pos + 1 :]
            iter_pos += 2

        iter_pos = 0
        while not inside_code:
            iter_pos = post_text.find("_", iter_pos, next_brac_pos)
            if iter_pos == -1:
                break
            if not post_text[iter_pos + 1].isalnum():  # don't escape within a snake_case word
                post_text = post_text[:iter_pos] + "\_" + post_text[iter_pos + 1 :]
                iter_pos += 2
            else:
                iter_pos += 1

        text = pre_text + tag_text + post_text
        pos = len(pre_text) + len(tag_text)
        previous_pos = pos

    if tag_depth > 0:
        print_error(
            "{}.xml: Tag depth mismatch: too many (or too little) open/close tags.".format(state.current_class), state
        )

    return text


def format_table(f, data, remove_empty_columns=False):  # type: (TextIO, Iterable[Tuple[str, ...]]) -> None
    if len(data) == 0:
        return

    column_sizes = [0] * len(data[0])
    for row in data:
        for i, text in enumerate(row):
            text_length = len(text or "")
            if text_length > column_sizes[i]:
                column_sizes[i] = text_length

    sep = ""
    for size in column_sizes:
        if size == 0 and remove_empty_columns:
            continue
        sep += "+" + "-" * (size + 2)
    sep += "+\n"
    f.write(sep)

    for row in data:
        row_text = "|"
        for i, text in enumerate(row):
            if column_sizes[i] == 0 and remove_empty_columns:
                continue
            row_text += " " + (text or "").ljust(column_sizes[i]) + " |"
        row_text += "\n"
        f.write(row_text)
        f.write(sep)
    f.write("\n")


def make_type(klass, state):  # type: (str, State) -> str
    if klass.find("*") != -1:  # Pointer, ignore
        return klass
    link_type = klass
    if link_type.endswith("[]"):  # Typed array, strip [] to link to contained type.
        link_type = link_type[:-2]
    if link_type in state.classes:
        return ":ref:`{}<class_{}>`".format(klass, link_type)
    print_error('{}.xml: Unresolved type "{}".'.format(state.current_class, klass), state)
    return klass


def make_enum(t, state):  # type: (str, State) -> str
    p = t.find(".")
    if p >= 0:
        c = t[0:p]
        e = t[p + 1 :]
        # Variant enums live in GlobalScope but still use periods.
        if c == "Variant":
            c = "@GlobalScope"
            e = "Variant." + e
    else:
        c = state.current_class
        e = t
        if c in state.classes and e not in state.classes[c].enums:
            c = "@GlobalScope"

    if c in state.classes and e in state.classes[c].enums:
        return ":ref:`{0}<enum_{1}_{0}>`".format(e, c)

    # Don't fail for `Vector3.Axis`, as this enum is a special case which is expected not to be resolved.
    if "{}.{}".format(c, e) != "Vector3.Axis":
        print_error('{}.xml: Unresolved enum "{}".'.format(state.current_class, t), state)

    return t


def make_method_signature(
    class_def, method_def, ref_type, state
):  # type: (ClassDef, Union[MethodDef, SignalDef], str, State) -> Tuple[str, str]
    ret_type = " "

    if isinstance(method_def, MethodDef):
        ret_type = method_def.return_type.to_rst(state)

    out = ""

    if ref_type != "":
        if ref_type == "operator":
            out += ":ref:`{0}<class_{1}_{2}_{3}_{4}>` ".format(
                method_def.name,
                class_def.name,
                ref_type,
                sanitize_operator_name(method_def.name, state),
                method_def.return_type.type_name,
            )
        else:
            out += ":ref:`{0}<class_{1}_{2}_{0}>` ".format(method_def.name, class_def.name, ref_type)
    else:
        out += "**{}** ".format(method_def.name)

    out += "**(**"
    for i, arg in enumerate(method_def.parameters):
        if i > 0:
            out += ", "
        else:
            out += " "

        out += "{} {}".format(arg.type_name.to_rst(state), arg.name)

        if arg.default_value is not None:
            out += "=" + arg.default_value

    if isinstance(method_def, MethodDef) and method_def.qualifiers is not None and "vararg" in method_def.qualifiers:
        if len(method_def.parameters) > 0:
            out += ", ..."
        else:
            out += " ..."

    out += " **)**"

    if isinstance(method_def, MethodDef) and method_def.qualifiers is not None:
        # Use substitutions for abbreviations. This is used to display tooltips on hover.
        # See `make_footer()` for descriptions.
        for qualifier in method_def.qualifiers.split():
            out += " |" + qualifier + "|"

    return ret_type, out


def make_heading(title, underline, l10n=True):  # type: (str, str, bool) -> str
    if l10n:
        new_title = translate(title)
        if new_title != title:
            title = new_title
            underline *= 2  # Double length to handle wide chars.
    return title + "\n" + (underline * len(title)) + "\n\n"


def make_footer():  # type: () -> str
    # Generate reusable abbreviation substitutions.
    # This way, we avoid bloating the generated rST with duplicate abbreviations.
    # fmt: off
    return (
        ".. |virtual| replace:: :abbr:`virtual (" + translate("This method should typically be overridden by the user to have any effect.") + ")`\n"
        ".. |const| replace:: :abbr:`const (" + translate("This method has no side effects. It doesn't modify any of the instance's member variables.") + ")`\n"
        ".. |vararg| replace:: :abbr:`vararg (" + translate("This method accepts any number of arguments after the ones described here.") + ")`\n"
        ".. |constructor| replace:: :abbr:`constructor (" + translate("This method is used to construct a type.") + ")`\n"
        ".. |static| replace:: :abbr:`static (" + translate("This method doesn't need an instance to be called, so it can be called directly using the class name.") + ")`\n"
        ".. |operator| replace:: :abbr:`operator (" + translate("This method describes a valid operator to use with this type as left-hand operand.") + ")`\n"
    )
    # fmt: on


def make_link(url, title):  # type: (str, str) -> str
    match = GODOT_DOCS_PATTERN.search(url)
    if match:
        groups = match.groups()
        if match.lastindex == 2:
            # Doc reference with fragment identifier: emit direct link to section with reference to page, for example:
            # `#calling-javascript-from-script in Exporting For Web`
            # Or use the title if provided.
            if title != "":
                return "`" + title + " <../" + groups[0] + ".html" + groups[1] + ">`__"
            return "`" + groups[1] + " <../" + groups[0] + ".html" + groups[1] + ">`__ in :doc:`../" + groups[0] + "`"
        elif match.lastindex == 1:
            # Doc reference, for example:
            # `Math`
            if title != "":
                return ":doc:`" + title + " <../" + groups[0] + ">`"
            return ":doc:`../" + groups[0] + "`"
    else:
        # External link, for example:
        # `http://enet.bespin.org/usergroup0.html`
        if title != "":
            return "`" + title + " <" + url + ">`__"
        return "`" + url + " <" + url + ">`__"


def sanitize_operator_name(dirty_name, state):  # type: (str, State) -> str
    clear_name = dirty_name.replace("operator ", "")

    if clear_name == "!=":
        clear_name = "neq"
    elif clear_name == "==":
        clear_name = "eq"

    elif clear_name == "<":
        clear_name = "lt"
    elif clear_name == "<=":
        clear_name = "lte"
    elif clear_name == ">":
        clear_name = "gt"
    elif clear_name == ">=":
        clear_name = "gte"

    elif clear_name == "+":
        clear_name = "sum"
    elif clear_name == "-":
        clear_name = "dif"
    elif clear_name == "*":
        clear_name = "mul"
    elif clear_name == "/":
        clear_name = "div"
    elif clear_name == "%":
        clear_name = "mod"
    elif clear_name == "**":
        clear_name = "pow"

    elif clear_name == "unary+":
        clear_name = "unplus"
    elif clear_name == "unary-":
        clear_name = "unminus"

    elif clear_name == "<<":
        clear_name = "bwsl"
    elif clear_name == ">>":
        clear_name = "bwsr"
    elif clear_name == "&":
        clear_name = "bwand"
    elif clear_name == "|":
        clear_name = "bwor"
    elif clear_name == "^":
        clear_name = "bwxor"
    elif clear_name == "~":
        clear_name = "bwnot"

    elif clear_name == "[]":
        clear_name = "idx"

    else:
        clear_name = "xxx"
        print_error('Unsupported operator type "{}", please add the missing rule.'.format(dirty_name), state)

    return clear_name


def indent_bullets(text):  # type: (str) -> str
    # Take the text and check each line for a bullet point represented by "-".
    # Where found, indent the given line by a further "\t".
    # Used to properly indent bullet points contained in the description for enum values.
    # Ignore the first line - text will be prepended to it so bullet points wouldn't work anyway.
    bullet_points = "-"

    lines = text.splitlines(keepends=True)
    for line_index, line in enumerate(lines[1:], start=1):
        pos = 0
        while pos < len(line) and line[pos] == "\t":
            pos += 1

        if pos < len(line) and line[pos] in bullet_points:
            lines[line_index] = line[:pos] + "\t" + line[pos:]

    return "".join(lines)


if __name__ == "__main__":
    main()