Remove the `NonMatchesExplode` variant now that no deriving impl uses it.
Removed `EnumNonMatching` entirely.
Remove now irrelevant `on_matching` field and `HandleNonMatchingEnums` type.
Removed unused `EnumNonMatchFunc` type def.
Drive-by: revise `EnumNonMatchCollapsedFunc` doc.
Made all calls to `expand_enum_method_body` go directly to
`build_enum_match_tuple`.
Alpha-rename `enum_nonmatch_g` back to `enum_nonmatch_f` to reduce overall diff noise.
Inline sole call of `some_ordering_const`.
Inline sole call of `ordering_const`.
Removed a bunch of code that became dead after the above changes.
In the above formulas, `n` is the number of variants, and `k` is the
number of self-args fed into deriving. In the particular case of
interest (namely `PartialOrd` and `Ord`), `k` is always 2, so we are
basically comparing `O(n)` versus `O(n^2)`.
Also, the stage is set for having *all* enum deriving codes go through
`build_enum_match_tuple` and getting rid of `build_enum_match`.
Also, seriously attempted to clean up the code itself. Added a bunch
of comments attempting to document what I learned as I worked through
the original code and adapted it to this new strategy.
In particular, I want authors of deriving modes to understand what
they are opting into (namely quadratic code size or worse) when they
select NonMatchesExplode.
The old method of building up a list of items and threading it through
all of the decorators was unwieldy and not really scalable as
non-deriving ItemDecorators become possible. The API is now that the
decorator gets an immutable reference to the item it's attached to, and
a callback that it can pass new items to. If we want to add syntax
extensions that can modify the item they're attached to, we can add that
later, but I think it'll have to be separate from ItemDecorator to avoid
strange ordering issues.
This makes error messages about (e.g.) `#[deriving(Clone)] struct Foo {
x: Type }` point at `x: Type` rather than `Clone` in the header (while
still referring to the `#[deriving(Clone)]` in the expansion info).
This means that compilation continues for longer, and so we can see more
errors per compile. This is mildly more user-friendly because it stops
users having to run rustc n times to see n macro errors: just run it
once to see all of them.
ToStr, Encodable and Decodable are not marked as such, since they're
already expensive, and lead to large methods, so inlining will bloat the
metadata & the binaries.
This means that something like
#[deriving(Eq)]
struct A { x: int }
creates an instance like
#[doc = "Automatically derived."]
impl ::std::cmp::Eq for A {
#[inline]
fn eq(&self, __arg_0: &A) -> ::bool {
match *__arg_0 {
A{x: ref __self_1_0} =>
match *self {
A{x: ref __self_0_0} => true && __self_0_0.eq(__self_1_0)
}
}
}
#[inline]
fn ne(&self, __arg_0: &A) -> ::bool {
match *__arg_0 {
A{x: ref __self_1_0} =>
match *self {
A{x: ref __self_0_0} => false || __self_0_0.ne(__self_1_0)
}
}
}
}
(The change being the `#[inline]` attributes.)
This rearranges the deriving code so that #[deriving] a trait on a field
that doesn't implement that trait will point to the field in question,
e.g.
struct NotEq; // doesn't implement Eq
#[deriving(Eq)]
struct Foo {
ok: int,
also_ok: ~str,
bad: NotEq // error points here.
}
Unfortunately, this means the error is disconnected from the `deriving`
itself but there's no current way to pass that information through to
rustc except via the spans, at the moment.
Fixes#7724.
This does a number of things, but especially dramatically reduce the
number of allocations performed for operations involving attributes/
meta items:
- Converts ast::meta_item & ast::attribute and other associated enums
to CamelCase.
- Converts several standalone functions in syntax::attr into methods,
defined on two traits AttrMetaMethods & AttributeMethods. The former
is common to both MetaItem and Attribute since the latter is a thin
wrapper around the former.
- Deletes functions that are unnecessary due to iterators.
- Converts other standalone functions to use iterators and the generic
AttrMetaMethods rather than allocating a lot of new vectors (e.g. the
old code would have to allocate a new vector to use functions that
operated on &[meta_item] on &[attribute].)
- Moves the core algorithm of the #[cfg] matching to syntax::attr,
similar to find_inline_attr and find_linkage_metas.
This doesn't have much of an effect on the speed of #[cfg] stripping,
despite hugely reducing the number of allocations performed; presumably
most of the time is spent in the ast folder rather than doing attribute
checks.
Also fixes the Eq instance of MetaItem_ to correctly ignore spaces, so
that `rustc --cfg 'foo(bar)'` now works.
