meson/docs/refman/generatormd.py

# SPDX-License-Identifier: Apache-2.0
# Copyright 2021 The Meson development team

from .generatorbase import GeneratorBase
import re
import json

from .model import (
    ReferenceManual,
    Function,
    Method,
    Object,
    ObjectType,
    Type,
    DataTypeInfo,
    ArgBase,
    PosArg,
    VarArgs,
    Kwarg,
)

from pathlib import Path
from textwrap import dedent
import typing as T

from mesonbuild import mlog

PlaceholderTypes = T.Union[None, str, bool]
FunctionDictType = T.Dict[
    str,
    T.Union[
        PlaceholderTypes,
        T.Dict[str, PlaceholderTypes],
        T.Dict[str, T.Dict[str, PlaceholderTypes]],
        T.Dict[str, T.List[T.Dict[str, PlaceholderTypes]]],
        T.List[T.Dict[str, PlaceholderTypes]],
        T.List[str],
    ]
]

_ROOT_BASENAME = 'Reference-manual'

_OBJ_ID_MAP = {
    ObjectType.ELEMENTARY: 'elementary',
    ObjectType.BUILTIN: 'builtin',
    ObjectType.MODULE: 'module',
    ObjectType.RETURNED: 'returned',
}

# Indent all but the first line with 4*depth spaces.
# This function is designed to be used with `dedent`
# and fstrings where multiline strings are used during
# the string interpolation.
def smart_indent(raw: str, depth: int = 3) -> str:
    lines = raw.split('\n')
    first_line = lines[0]
    lines = [' ' * (4 * depth) + x for x in lines]
    lines[0] = first_line  # Do not indent the first line
    return '\n'.join(lines)

def code_block(code: str) -> str:
    code = dedent(code)
    return f'<pre><code class="language-meson">{code}</code></pre>'

class GeneratorMD(GeneratorBase):
    def __init__(self, manual: ReferenceManual, sitemap_out: Path, sitemap_in: Path, link_def_out: Path, enable_modules: bool) -> None:
        super().__init__(manual)
        self.sitemap_out = sitemap_out.resolve()
        self.sitemap_in = sitemap_in.resolve()
        self.link_def_out = link_def_out.resolve()
        self.out_dir = self.sitemap_out.parent
        self.enable_modules = enable_modules
        self.generated_files: T.Dict[str, str] = {}

    # Utility functions
    def _gen_filename(self, file_id: str, *, extension: str = 'md') -> str:
        parts = file_id.split('.')
        assert parts[0] == 'root'
        assert all([x for x in parts])
        parts[0] = _ROOT_BASENAME
        parts = [re.sub(r'[0-9]+_', '', x) for x in parts]
        return f'{"_".join(parts)}.{extension}'

    def _gen_object_file_id(self, obj: Object) -> str:
        '''
            Deterministically generate a unique file ID for the Object.

            This ID determines where the object will be inserted in the sitemap.
        '''
        if obj.obj_type == ObjectType.RETURNED and obj.defined_by_module is not None:
            base = self._gen_object_file_id(obj.defined_by_module)
            return f'{base}.{obj.name}'
        return f'root.{_OBJ_ID_MAP[obj.obj_type]}.{obj.name}'

    def _link_to_object(self, obj: T.Union[Function, Object], in_code_block: bool = False) -> str:
        '''
            Generate a palaceholder tag for the function/method/object documentation.
            This tag is then replaced in the custom hotdoc plugin.
        '''
        prefix = '#' if in_code_block else ''
        if isinstance(obj, Object):
            return f'[[{prefix}@{obj.name}]]'
        elif isinstance(obj, Method):
            return f'[[{prefix}{obj.obj.name}.{obj.name}]]'
        elif isinstance(obj, Function):
            return f'[[{prefix}{obj.name}]]'
        else:
            raise RuntimeError(f'Invalid argument {obj}')

    def _write_file(self, data: str, file_id: str) -> None:#
        ''' Write the data to disk and store the id for the generated data '''

        self.generated_files[file_id] = self._gen_filename(file_id)
        out_file = self.out_dir / self.generated_files[file_id]
        out_file.write_text(data, encoding='ascii')
        mlog.log('Generated', mlog.bold(out_file.name))

    def _write_template(self, data: T.Dict[str, T.Any], file_id: str, template_name: T.Optional[str] = None) -> None:
        ''' Render the template mustache files and write the result '''
        template_dir = Path(__file__).resolve().parent / 'templates'
        template_name = template_name or file_id
        template_name = f'{template_name}.mustache'
        template_file = template_dir / template_name

        # Import here, so that other generators don't also depend on it
        import chevron
        result = chevron.render(
            template=template_file.read_text(encoding='utf-8'),
            data=data,
            partials_path=template_dir.as_posix(),
            warn=True,
        )

        self._write_file(result, file_id)


    # Actual generator functions
    def _gen_func_or_method(self, func: Function) -> FunctionDictType:
        def render_type(typ: Type, in_code_block: bool = False) -> str:
            def data_type_to_str(dt: DataTypeInfo) -> str:
                base = self._link_to_object(dt.data_type, in_code_block)
                if dt.holds:
                    return f'{base}[{render_type(dt.holds, in_code_block)}]'
                return base
            assert typ.resolved
            return ' | '.join([data_type_to_str(x) for x in typ.resolved])

        def len_stripped(s: str) -> int:
            s = s.replace(']]', '')
            # I know, this regex is ugly but it works.
            return len(re.sub(r'\[\[(#|@)*([^\[])', r'\2', s))

        def arg_anchor(arg: ArgBase) -> str:
            return f'{func.name}_{arg.name.replace("<", "_").replace(">", "_")}'

        def render_signature() -> str:
            # Skip a lot of computations if the function does not take any arguments
            if not any([func.posargs, func.optargs, func.kwargs, func.varargs]):
                return f'{render_type(func.returns, True)} {func.name}()'

            signature = dedent(f'''\
                # {self.brief(func)}
                {render_type(func.returns, True)} {func.name}(
            ''')

            # Calculate maximum lengths of the type and name
            all_args: T.List[ArgBase] = []
            all_args += func.posargs
            all_args += func.optargs
            all_args += [func.varargs] if func.varargs else []

            max_type_len = 0
            max_name_len = 0
            if all_args:
                max_type_len = max([len_stripped(render_type(x.type)) for x in all_args])
                max_name_len = max([len(x.name) for x in all_args])

            # Generate some common strings
            def prepare(arg: ArgBase, link: bool = True) -> T.Tuple[str, str, str, str]:
                type_str = render_type(arg.type, True)
                type_len = len_stripped(type_str)
                type_space = ' ' * (max_type_len - type_len)
                name_space = ' ' * (max_name_len - len(arg.name))
                name_str = f'<b>{arg.name.replace("<", "&lt;").replace(">", "&gt;")}</b>'
                if link:
                    name_str = f'<a href="#{arg_anchor(arg)}">{name_str}</a>'

                return type_str, type_space, name_str, name_space

            for i in func.posargs:
                type_str, type_space, name_str, name_space = prepare(i)
                signature += f'  {type_str}{type_space} {name_str},{name_space}     # {self.brief(i)}\n'

            for i in func.optargs:
                type_str, type_space, name_str, name_space = prepare(i)
                signature += f'  {type_str}{type_space} [{name_str}],{name_space}   # {self.brief(i)}\n'

            if func.varargs:
                type_str, type_space, name_str, name_space = prepare(func.varargs, link=False)
                signature += f'  {type_str}{type_space} {name_str}...,{name_space}  # {self.brief(func.varargs)}\n'

            # Abort if there are no kwargs
            if not func.kwargs:
                return signature + ')'

            # Only add this separator if there are any posargs
            if all_args:
                signature += '\n  # Keyword arguments:\n'

            # Recalculate lengths for kwargs
            all_args = list(func.kwargs.values())
            max_type_len = max([len_stripped(render_type(x.type)) for x in all_args])
            max_name_len = max([len(x.name) for x in all_args])

            for kwarg in self.sorted_and_filtered(list(func.kwargs.values())):
                type_str, type_space, name_str, name_space = prepare(kwarg)
                required = ' <i>[required]</i> ' if kwarg.required else '            '
                required = required if any([x.required for x in func.kwargs.values()]) else ''
                signature += f'  {name_str}{name_space} : {type_str}{type_space} {required} # {self.brief(kwarg)}\n'

            return signature + ')'

        def gen_arg_data(arg: T.Union[PosArg, Kwarg, VarArgs], *, optional: bool = False) -> T.Dict[str, PlaceholderTypes]:
            data: T.Dict[str, PlaceholderTypes] = {
                'row-id': arg_anchor(arg),
                'name': arg.name,
                'type': render_type(arg.type),
                'description': arg.description,
                'since': arg.since or None,
                'deprecated': arg.deprecated or None,
                'optional': optional,
                'default': None,
            }

            if isinstance(arg, VarArgs):
                data.update({
                    'min': str(arg.min_varargs) if arg.min_varargs > 0 else '0',
                    'max': str(arg.max_varargs) if arg.max_varargs > 0 else 'infinity',
                })
            if isinstance(arg, (Kwarg, PosArg)):
                data.update({'default': arg.default or None})
            if isinstance(arg, Kwarg):
                data.update({'required': arg.required})
            return data

        mname = f'\\{func.name}' if func.name == '[index]' else func.name

        data: FunctionDictType = {
            'name': f'{func.obj.name}.{mname}' if isinstance(func, Method) else func.name,
            'base_level': '##' if isinstance(func, Method) else '#',
            'type_name_upper': 'Method' if isinstance(func, Method) else 'Function',
            'type_name': 'method' if isinstance(func, Method) else 'function',
            'description': func.description,
            'notes': func.notes,
            'warnings': func.warnings,
            'example': func.example or None,
            'signature_level': 'h4' if isinstance(func, Method) else 'h3',
            'signature': render_signature(),
            'has_args': bool(func.posargs or func.optargs or func.kwargs or func.varargs),
            # Merge posargs and optargs by generating the *[optional]* tag for optargs
            'posargs': {
                'args': [gen_arg_data(x) for x in func.posargs] + [gen_arg_data(x, optional=True) for x in func.optargs]
            } if func.posargs or func.optargs else None,
            'kwargs':  {'args': [gen_arg_data(x) for x in self.sorted_and_filtered(list(func.kwargs.values()))]} if func.kwargs else None,
            'varargs': gen_arg_data(func.varargs) if func.varargs else None,
            'arg_flattening': func.arg_flattening,

            # For the feature taggs template
            'since': func.since or None,
            'deprecated': func.deprecated or None,
            'optional': False,
            'default': None
        }

        return data

    def _write_object(self, obj: Object) -> None:
        data = {
            'name': obj.name,
            'title': obj.long_name if obj.obj_type == ObjectType.RETURNED else obj.name,
            'description': obj.description,
            'notes': obj.notes,
            'warnings': obj.warnings,
            'long_name': obj.long_name,
            'obj_type_name': _OBJ_ID_MAP[obj.obj_type].capitalize(),
            'example': obj.example or None,
            'has_methods': bool(obj.methods),
            'has_inherited_methods': bool(obj.inherited_methods),
            'has_subclasses': bool(obj.extended_by),
            'is_returned': bool(obj.returned_by),
            'extends': obj.extends_obj.name if obj.extends_obj else None,
            'returned_by': [self._link_to_object(x) for x in self.sorted_and_filtered(obj.returned_by)],
            'extended_by': [self._link_to_object(x) for x in self.sorted_and_filtered(obj.extended_by)],
            'methods': [self._gen_func_or_method(m) for m in self.sorted_and_filtered(obj.methods)],
            'inherited_methods': [self._gen_func_or_method(m) for m in self.sorted_and_filtered(obj.inherited_methods)],
        }

        self._write_template(data, self._gen_object_file_id(obj), 'object')

    def _write_functions(self) -> None:
        data = {'functions': [self._gen_func_or_method(x) for x in self.functions]}
        self._write_template(data, 'root.functions')

    def _root_refman_docs(self) -> None:
        def gen_obj_links(objs: T.List[Object]) -> T.List[T.Dict[str, str]]:
            ret: T.List[T.Dict[str, str]] = []
            for o in objs:
                ret += [{'indent': '', 'link': self._link_to_object(o), 'brief': self.brief(o)}]
                for m in self.sorted_and_filtered(o.methods):
                    ret += [{'indent': '  ', 'link': self._link_to_object(m), 'brief': self.brief(m)}]
                if o.obj_type == ObjectType.MODULE and self.extract_returned_by_module(o):
                    tmp = gen_obj_links(self.extract_returned_by_module(o))
                    tmp = [{**x, 'indent': '  ' + x['indent']} for x in tmp]
                    ret += [{'indent': '  ', 'link': '**New objects:**', 'brief': ''}]
                    ret += [*tmp]
            return ret

        data = {
            'root': self._gen_filename('root'),
            'elementary': gen_obj_links(self.elementary),
            'returned': gen_obj_links(self.returned),
            'builtins': gen_obj_links(self.builtins),
            'modules': gen_obj_links(self.modules),
            'functions': [{'indent': '', 'link': self._link_to_object(x), 'brief': self.brief(x)} for x in self.functions],
            'enable_modules': self.enable_modules,
        }

        dummy = {'root': self._gen_filename('root')}

        self._write_template(data, 'root')
        self._write_template({**dummy, 'name': 'Elementary types'}, f'root.{_OBJ_ID_MAP[ObjectType.ELEMENTARY]}', 'dummy')
        self._write_template({**dummy, 'name': 'Builtin objects'},  f'root.{_OBJ_ID_MAP[ObjectType.BUILTIN]}',    'dummy')
        self._write_template({**dummy, 'name': 'Returned objects'}, f'root.{_OBJ_ID_MAP[ObjectType.RETURNED]}',   'dummy')

        if self.enable_modules:
            self._write_template({**dummy, 'name': 'Modules'},          f'root.{_OBJ_ID_MAP[ObjectType.MODULE]}',     'dummy')


    def generate(self) -> None:
        mlog.log('Generating markdown files...')
        with mlog.nested():
            self._write_functions()
            for obj in self.objects:
                if not self.enable_modules and (obj.obj_type == ObjectType.MODULE or obj.defined_by_module is not None):
                    continue
                self._write_object(obj)
            self._root_refman_docs()
            self._configure_sitemap()
            self._generate_link_def()

    def _configure_sitemap(self) -> None:
        '''
            Replaces the `@REFMAN_PLACEHOLDER@` placeholder with the reference
            manual sitemap. The structure of the sitemap is derived from the
            file IDs.
        '''
        raw = self.sitemap_in.read_text(encoding='utf-8')
        out = ''
        for l in raw.split('\n'):
            if '@REFMAN_PLACEHOLDER@' not in l:
                out += f'{l}\n'
                continue
            mlog.log('Generating', mlog.bold(self.sitemap_out.as_posix()))
            base_indent = l.replace('@REFMAN_PLACEHOLDER@', '')
            for k in sorted(self.generated_files.keys()):
                indent = base_indent + '\t' * k.count('.')
                out += f'{indent}{self.generated_files[k]}\n'
        self.sitemap_out.write_text(out, encoding='utf-8')

    def _generate_link_def(self) -> None:
        '''
            Generate the link definition file for the refman_links hotdoc
            plugin. The plugin is then responsible for replacing the [[tag]]
            tags with custom HTML elements.
        '''
        data: T.Dict[str, str] = {}

        # Objects and methods
        for obj in self.objects:
            obj_file = self._gen_filename(self._gen_object_file_id(obj), extension='html')
            data[f'@{obj.name}'] = obj_file
            for m in obj.methods:
                data[f'{obj.name}.{m.name}'] = f'{obj_file}#{obj.name}{m.name}'

        # Functions
        funcs_file = self._gen_filename('root.functions', extension='html')
        for fn in self.functions:
            data[fn.name] = f'{funcs_file}#{fn.name}'

        self.link_def_out.write_text(json.dumps(data, indent=2), encoding='utf-8')