This is the final refactoring for extracting the bultin object
logic out of Interpreterbase. I decided to do both arrays and
dicts in one go since splitting it would have been a lot more
confusing.
Another commit in my quest to rid InterpreterBase from all higher
level object processing logic.
Additionally, there is a a logic change here, since `str.join` now
uses varargs and can now accept more than one argument (and supports
list flattening).
Since it cannot resolve `import typing as T` in order to figure out that
T.* is doing annotation-worthy stuff.
Since T.cast('Foo') is not actually using Foo except in an annotation
context (due to being a string) it requires extra work to resolve, and
the only thing that would currently work is actually using
'typing.cast'. However, we have decided to not use it except as T...
Since this import is only imported during mypy it's not so bad to noqa
it.
We have a lot of these. Some of them are harmless, if unidiomatic, such
as `if (condition)`, others are potentially dangerous `assert(...)`, as
`assert(condtion)` works as expected, but `assert(condition, message)`
will result in an assertion that never triggers, as what you're actually
asserting is `bool(tuple[2])`, which will always be true.
Follow-up on commit 5a7b8d86d0
Sometimes, we find a parent meson.build which is also malformed, and we
shouldn't suggest that maybe the user meant to use that, if it isn't a
valid project() either. Do a rough and dirty check to see if the very
first line is a project() declaration, and if not, don't try to be
clever and suggest using it.
The "invalid source tree" error suffices here, since we're not
absolutely sure meson can be successfully run in that parent directory
and actually advising people about the wrong location is a lot more
confusing than just saying "please figure this out yourself, here is
what to look for".
Granted, we miss cases where project() comes after blank lines and/or
comments, but doing a full AST parse here is excessively overkill and
probably too painful to do, and we don't need to be *that* clever. So
let's be content with merely going above and beyond the call of duty.
Explicitly mention that the project definition is invalid, and clarify
that project is `project()` -- a function.
Also try to walk the directory tree upward, and if there are parent
meson.build files, just say this isn't the project root, and "maybe you
meant to run meson there instead?"
This won't catch calls to subdir('foo/bar') but we can't be perfect,
only better than before and catch the *majority* of such cases, and
hopefully it's a lot more clear if meson protests that the project is
"invalid, there is no project() function", where the user should look
for a potential solution.
Fixes#3426
This commit introduces a new type of `HoldableObject`: The
`SecondLevelHolder`. The primary purpose of this class is
to handle cases where two (or more) `HoldableObject`s are
stored at the same time (with one default object). The
best (and currently only) example here is the `BothLibraries`
class.
This is meant to allow simple type conversions to happen before the
interpreter function is called. This should simplify some cases like the
"native" keyword arugment that are booleans in the Meson DSL, but an
Enum in the implementation
This attribute is a callable that returns a string if the value is
invalid, otherwise None. This intended for cases like the `install_*`
function's `install_mode` paramater, which is either an int or the
string "preserve", which allows us to do nice things like:
```python
class Kwargs(TypedDict):
install_mode: T.Union[int, T.Literal['preserve']]
@typed_kwargs(
'foo', KwargInfo('install_mode', ...,
validator=lambda x: None if isinstance(x, int) or x == 'preserve' else 'must be the literal "preserve"),
)
def install_data(self, node, args, kwargs: 'Kwargs'):
...
```
In this case mypy *knows* that the string is preserve, as do we, and we
can simply do tests like:
```python
if kwargs['install_mode'] == 'preserve':
...
else:
# this is an int
```
It's really inconvenient to want a thing that is always a list, but not
be able to provide a default value of a list because of mutation. To
that end the typed_kwargs method now makes a shallow copy of the default
when using a `ContainerTypeInfo` as the type. This mean that using a
default of `[]` is perfectly safe.
The only advantage they have is they have the interpreter in arguments,
but it's already available as self.interpreter. We should discourage
usage of the interpreter API and rely on ModuleState object instead in
the future.
This also lift the restriction that a module method cannot add build
targets, but that was not enforced for snippet methods anyway (and some
modules were doing it) and it's really loose restriction as it should
check for many other things if we wanted to make it consistent.
Daniel Mensinger
Eli Schwartz
Dylan Baker
Xavier Claessens
Tristan Partin
Laurin-Luis Lehning