libadalang/extensions/python

## vim: filetype=makopython

def token_match(self, other):
    """
    Helper for the finditer/find/findall methods, so that a token matches
    another token even if they are not strictly equivalent.
    """
    return self == other or self.text == other


def _coerce_bytes(label, value, what='a bytes string', or_none=False):
    """
    Take bytes (forwarded as-is to C) but also accept text (encoded using
    the system encoding).
    """
    if value is None and or_none:
        return None
    elif isinstance(value, bytes):
        return value
    elif isinstance(value, str):
        return value.encode()
    else:
        raise TypeError('`{}` argument must be {} (got {})'
                        .format(label, what, _type_fullname(type(value))))


_set_config_pragmas_mapping = _import_func(
    "ada_set_config_pragmas_mapping",
    [AnalysisContext._c_type,
     AnalysisUnit._c_type,
     ctypes.POINTER(AnalysisUnit._c_type)],
    None
)


class TargetInformation:
    ${py_doc("libadalang.target_info_type", 4)}

    def __init__(self, filename: str):
        ${py_doc("libadalang.target_info_load", 8)}

        # First, define this attribute so that __del__ work even if the
        # constructor aborts later on because of an exception.
        self._c_value = None

        c_filename = _coerce_bytes("filename", filename)
        self._c_value = self._c_load(c_filename)

    def __del__(self):
        if self._c_value is not None:
            self._c_free(self._c_value)

    _c_type = _hashable_c_pointer()

    _c_load = staticmethod(_import_func(
        "ada_target_info_load", [ctypes.c_char_p], _c_type,
    ))
    _c_free = staticmethod(
        _import_func("ada_target_info_free", [_c_type], None)
    )
    _c_set = staticmethod(
        _import_func(
            "ada_target_info_set", [AnalysisContext._c_type, _c_type], None
        )
    )


## Handling of string arrays

class _c_string_array(ctypes.Structure):
    _fields_ = [
        ("length", ctypes.c_int),
        ("c_ptr", ctypes.POINTER(ctypes.c_char_p)),
        # Omit the "items" field: it has variable size and is not necessary
        # to just read the items.
    ]

    @property
    def wrap(self) -> List[str]:
        return [self.c_ptr[i] for i in range(self.length)]

_c_string_array_ptr = ctypes.POINTER(_c_string_array)

_c_free_string_array = _import_func(
    "ada_free_string_array", [_c_string_array_ptr], None,
)


@property
def doc_name(n):
    """
    Format this name to be a readable qualified name for the entity designated
    by it. Meant to be used in documentation context.

    If the entity is local, it will return the relative name. If it is
    non-local, return the shortest qualified name not taking use clauses into
    account.

    .. WARNING:: This is an EXPERIMENTAL feature. This is a python specific
        method, because for the moment this is not conveniently implementable
        directly as a libadalang property.  Consider it an experimental API
        endpoint, and use it at your own risk.
    """
    if n.p_is_defining and not n.is_a(DefiningName):
        n = n.p_enclosing_defining_name

    ref_decl = n.p_basic_decl if n.p_is_defining else n.p_referenced_decl()
    ref_decl_fqn = ref_decl.p_fully_qualified_name

    enclosing_package = next(
        (p for p in n.parents() if p.is_a(BasePackageDecl)),
        None
    )

    if enclosing_package is None or enclosing_package == ref_decl:
        return ref_decl_fqn

    enclosing_decl_fqn = enclosing_package.p_fully_qualified_name

    if ref_decl_fqn.lower().startswith(enclosing_decl_fqn.lower()):
        return ref_decl_fqn[len(enclosing_decl_fqn):].strip(".")
    else:
        return ref_decl_fqn

Token.match = token_match
Name.doc_name = doc_name


import enum
class SourceFilesMode(enum.Enum):
    """
    Mode to get a list of source files from a project file.

    See ``SourceFiles.for_project``.
    """
    default = 0
    root_project = 1
    whole_project = 2
    whole_project_with_runtime = 3


class GPRProject:
    """
    Load a GPR project file.
    """

    class _UnitProvider(UnitProvider):
        def __init__(self, project: GPRProject, c_value: Any):
            super().__init__(c_value)

            # Keep a reference on the GPRProject instance that was used to
            # create this unit provider so that the project lives at least as
            # long as the unit provider.
            self._project = project

    def __init__(self,
                 project_file: str,
                 scenario_vars: Dict[str, str] = {},
                 target: Opt[str] = None,
                 runtime: Opt[str] = None,
                 config_file: Opt[str] = None,
                 ada_only: bool = False,
                 print_errors: bool = True):
        """
        Load a GPR project file.

        This may raise an ``InvalidProjectError`` exception if an error occurs
        when loading the project.

        :param project_file: Filename for the project to load.
        :param screnario_vars: External variables for the project to load.
        :param target: Name of the target for the project to load. Assume the
            native platform if left to None.
        :param runtime: Name of the runtime for the project to load. Use the
            default runtime for the selected target if left to None.
        :param config_file: Name of the GPR configuration file for the project
            to load. Use autoconfiguration if left to None.
        :param ada_only: Whether to restrict GPR file loading to the Ada
            language.
        :param print_errors: Whether to print non-critical error messages
            emitted during project loading on the standard error stream. See
            the ``errors`` method to have programmatic access to the list of
            errors.
        """
        # First, define this attribute so that __del__ work even if the
        # constructor aborts later on because of an exception.
        self._c_value = None

        # Turn arguments into C API values
        c_project_file = _coerce_bytes('project_file', project_file)
        c_target = _coerce_bytes('target', target, or_none=True)
        c_runtime = _coerce_bytes('runtime', runtime, or_none=True)
        c_config_file = _coerce_bytes('config_file', config_file, or_none=True)

        if scenario_vars:
            items = scenario_vars.items()
            scn_vars_array_type = (
                self._c_scenario_variable * (len(items) + 1)
            )
            c_scenario_vars = scn_vars_array_type()
            for i, (name, value) in enumerate(items):
                what = 'a dict mapping bytes strings to bytes strings'
                name = _coerce_bytes('scenario_vars', name, what)
                value = _coerce_bytes('scenario_vars', value, what)
                c_scenario_vars[i] = self._c_scenario_variable(
                    name, value
                )
            c_scenario_vars[-1] = self._c_scenario_variable(None, None)
        else:
            c_scenario_vars = None

        # Load the project
        c_errors = _c_string_array_ptr()
        c_project = self._c_type()
        self._c_load(
            c_project_file,
            c_scenario_vars,
            c_target,
            c_runtime,
            c_config_file,
            int(ada_only),
            ctypes.byref(c_project),
            ctypes.byref(c_errors),
        )
        self._c_value = c_project

        # Extract the possibly empty list of error messages and print it if
        # requested. For user convenience, convert error messages to unicode
        # strings and discard decoding errors.
        self._errors = [
            msg.decode("utf-8", "replace") for msg in c_errors.contents.wrap
        ]
        _c_free_string_array(c_errors)
        if print_errors and self.errors:
            print(f"Errors while loading {project_file}:", file=sys.stderr)
            for e in self.errors:
                print(e, file=sys.stderr)

    def __del__(self):
        if self._c_value is not None:
            self._c_free(self._c_value)

    @property
    def errors(self) -> List[str]:
        """
        Possibly empty list of non-critical error messages emitted during
        project loading.
        """
        return list(self._errors)

    def create_unit_provider(self, project: Opt[str] = None) -> UnitProvider:
        """
        Return a unit provider that uses this GPR project.

        :param project: If None, let the unit provider use the whole project
            tree. Otherwise, restrict the unit provider to the project with the
            given name in the project tree.

            As unit providers must guarantee that there exists at most one
            source file for each couple (unit name, unit kind), aggregate
            projects that contains several conflicting units are not supported:
            trying to use one will yield an ``InvalidProjectError`` exception.
        """
        c_project = _coerce_bytes('project', project, or_none=True)
        c_value = self._c_create_unit_provider(self._c_value, c_project)
        return self._UnitProvider(self, c_value)

    def source_files(
        self,
        mode: SourceFilesMode = SourceFilesMode.default,
        projects: List[str] = [],
    ):
        """
        Return the list of source files in this project according to ``mode``:

        * ``default``: sources in the root project and its non-externally built
          dependencies;

        * ``root_project``: sources in the root project only;

        * ``whole_project``: sources in the whole project tree (i.e. including
          externally built dependencies);

        * ``whole_project_with_runtime``: sources in the whole project tree
          plus runtime sources.

        If ``projects`` is not empty, return instead the list for the sources
        in all the sub-projects that ``projects`` designates, still applying
        the given mode to the search.
        """

        assert isinstance(mode, SourceFilesMode)
        c_mode = mode.value

        projects_type = ctypes.c_char_p * len(projects)
        projects_c_strings = [
            ctypes.c_char_p(_coerce_bytes("projects", p)) for p in projects
        ]
        c_projects = projects_type(*projects_c_strings)

        # Compute the list of source files, extract it (no error expected there
        # unless we have a bug) and free the resources.
        c_value = self._c_source_files(
            self._c_value, c_mode, c_projects, len(projects)
        )
        assert c_value
        result = c_value.contents.wrap
        _c_free_string_array(c_value)

        # Now convert filenames to Unicode strings using the system default
        # encoding, to be more consistent with other Python APIs.
        return [f.decode() for f in result]

    def default_charset(self, project: Opt[str] = None) -> str:
        """
        Try to detect the default charset to use for the given project.

        Restrict the detection to the subproject ``project``, or to ``self``'s
        root project if ``project`` is left to ``None``.

        Note that, as of today, this detection only looks for the ``-gnatW8``
        compiler switch: other charsets are not supported.
        """
        c_project = _coerce_bytes('project', project, or_none=True)
        result = self._c_default_charset(self._c_value, c_project)
        return _unwrap_str(result)

    def create_preprocessor(
        self,
        project: Opt[str] = None,
        line_mode: Optional[FileReader.LineMode] = None,
    ) -> FileReader:
        """
        Create preprocessor data from compiler arguments found in the given GPR
        project (``-gnatep`` and ``-gnateD`` arguments), or from the
        ``project`` sub-project (if the argument is passed).

        Note that this function collects all arguments and returns an
        approximation from them: it does not replicates exactly gprbuild's
        behavior. This may raise a ``File_Read_Error`` exception if this fails
        to read a preprocessor data file and a ``Syntax_Error`` exception if
        one such file has invalid syntax.
        """
        c_project = _coerce_bytes('project', project, or_none=True)

        if line_mode is None:
            c_line_mode_ref = None
        else:
            c_line_mode = ctypes.c_int(FileReader.LineMode._unwrap(line_mode))
            c_line_mode_ref = ctypes.byref(c_line_mode)

        return FileReader(
            self._c_create_preprocessor(
                self._c_value, c_project, c_line_mode_ref
            ),
        )

    def create_context(
        self,
        project: Opt[str] = None,
        event_handler: Opt[EventHandler] = None,
        with_trivia: bool = True,
        tab_stop: int = 8,
    ) -> AnalysisContext:
        ${py_doc("libadalang.gpr_project_initialize_context", 8)}
        c_project = _coerce_bytes("project", project, or_none=True)

        event_handler_wrapper, c_event_handler = (
            _EventHandlerWrapper.create(event_handler)
        )

        # Manually allocate a C-level analysis context so that we can
        # initialize it ourselves.
        c_value = _allocate_analysis_context()

        # Create the Python wrapper, so that we have one ready for event
        # handler callbacks triggered during context initialization.
        result = AnalysisContext(_c_value=c_value)

        # The wrapper created its own ownership share: release ours
        _context_decref(c_value)

        # Attach extra wrappers to the analysis context wrapper so that
        # wrappers live at least as long as the analysis context.
        #
        # HACK: use the "_unit_provider" slot for the GPRProject instance
        # itself. The initialization procedure will take care of creating a
        # unit provider and make the context actually use it: there is no need
        # for a UnitProvider wrapper here.
        result._event_handler_wrapper = event_handler_wrapper
        result._unit_provider = self

        # Finally, initialize the analysis context. Note that this step may
        # raise an exception: in that case, the analysis context is considered
        # not initialized, but we are not returning "result", so it will become
        # unreachable from the Python world, and garbage collection will take
        # care of releasing it.
        self._c_initialize_context(
            self._c_value,
            c_value,
            c_project,
            c_event_handler,
            with_trivia,
            tab_stop,
        )

        return result

    _c_type = _hashable_c_pointer()

    class _c_scenario_variable(ctypes.Structure):
        _fields_ = [('name', ctypes.c_char_p),
                    ('value', ctypes.c_char_p)]

    _c_load = staticmethod(_import_func(
        "ada_gpr_project_load",
        [ctypes.c_char_p,
         ctypes.POINTER(_c_scenario_variable),
         ctypes.c_char_p,
         ctypes.c_char_p,
         ctypes.c_char_p,
         ctypes.c_int,
         ctypes.POINTER(_c_type),
         ctypes.POINTER(_c_string_array_ptr)],
        None,
    ))

    _c_free = staticmethod(
        _import_func("ada_gpr_project_free", [_c_type], None)
    )

    _c_create_unit_provider = staticmethod(_import_func(
        "ada_gpr_project_create_unit_provider",
        [_c_type, ctypes.c_char_p],
        _unit_provider,
    ))

    _c_source_files = staticmethod(_import_func(
        "ada_gpr_project_source_files",
        [_c_type, ctypes.c_int, ctypes.POINTER(ctypes.c_char_p), ctypes.c_int],
        _c_string_array_ptr,
    ))

    _c_default_charset = staticmethod(_import_func(
        "ada_gpr_project_default_charset",
        [_c_type, ctypes.c_char_p], ctypes.POINTER(ctypes.c_char),
    ))

    _c_create_preprocessor = staticmethod(_import_func(
        "ada_gpr_project_create_preprocessor",
        [_c_type, ctypes.c_char_p, ctypes.POINTER(ctypes.c_int)],
        _file_reader,
    ))

    _c_initialize_context = staticmethod(_import_func(
        "ada_gpr_project_initialize_context",
        [_c_type,                  # gpr_project
         AnalysisContext._c_type, # context
         ctypes.c_char_p,         # project
         _event_handler,          # event_handler
         ctypes.c_int,            # with_trivia
         ctypes.c_int],           # tab_stop
        None,
    ))