This adds the ability to match on metadata for a node as if it was an attribute on the node. It also opens up the ability to match on node attributes via metadata only. It allows for metadata match specifications with similar flexibility to standard matchers.
This makes sure we always wrap elements in a SubscriptElement, even when there
is only one element. This makes things more regular while still being backwards
compatible with existing creation. The meat of this is in two halves, which can't
be split due to not wanting to break the build between commits. The first half
is just the changes to the parser and updates to tests. This includes a test to
be sure we can still render code that uses old construction types. The second half
is changes to codegen which made assumptions about `Subscript` and demonstrates
the need to make this change in the first place. This includes a fix to
`CSTNode.with_deep_changes` type to make it more correct and also more usable in
transforms without additional type assertions.
This is somewhat complicated by the fact that we need to not just allow
construction of nodes/matchers using `ExtSlice` still for backwards compatibility,
but we also need to be able to call `visit_ExtSlice` and `leave_ExtSlice` on
old visitors even though the new node is named `SubscriptElement`. The
construction/instance check/matching side of things will work since internally we
refer to everything as `SubscriptElement` and alias `ExtSlice` to this everywhere,
but for string-based function lookup, we need to get a little more clever and make
the default `visit_SubscriptElement` delegate onward to `visit_ExtSlice` so that
either form works.
This can all be removed again once we're past the deprecation period for ExtSlice.
I noticed that the typed visitors codegen was creating messy types such as Union[SingleType]. We have a clean-up for Union[SingleType] snuck into gen_matcher_classes. So, lets remove that snuck-in clean-up from gen_matcher_classes and apply it globally to all codegen. This makes its purpose a lot more obvious, while helping decouple necessary codegen from prettifying/simplifying transforms. It also cleans up typed visitors, so that's a bonus.
This makes matcher equality and hash equivalent to the way LibCST nodes behave. Not only does this make us more consistent, but it also fixes a bug where matcher decorators could not be used with a matcher that initialized a sequence type as a list.
Add a RemoveFromParent() function as a convenience to returning RemovalSentinel.REMOVE.
Introduce a `deep_remove()` on CSTNode analogous to `deep_replace()` but for removing.
There are a lot of nodes that cannot be removed or converted to maybes, such as
most of the Op tokens. It would be a bit of a lie to codegen leave_* methods
that allow these nodes to be converted, only to throw a runtime error later. So,
upgrade the codegen to allow us to see whether certain nodes are used in conjunction
with a MaybeSentinel/None, or inside a Sequence, to inform ourselves as to when to
allow MaybeSentinel or RemovalSentinel.
We want to make sure that the generated function stubs stay in sync with
the node definitions. So, make a unit test that fails if codegen generates
a different file than the existing file, so that somebody modifying code
knows they need to re-run codegen.
