In order to desugar `dbg` in a pipeline we need to allow a bare `dbg`
node in desugaring and only report it as an error if the bare node
survives to the next step of canonicalization. This means we move the
error code out of `desugar_expr` and into `canonicalize_expr`. This is
much simpler to do now that these functions use the same `env` struct,
since previously we would have had to pass down extra args to
`canonicalize_expr`. Sharing the `env` struct means that we also don't
have to worry about calculating `line_info` more than once.
This refactor simplifies the desugar pass by reducing the number of
arguments threaded through each recursive function call.
- Add the module src string to `Env`.
- Add `line_info` to `Env` as a lazy-evaled function.
- Refactor desugar functions to take the `can::Env` struct in place of a
number of params. This is mostly a find-and-replace, but in a few
places `Vec::from_iter_in` was changed to `Vec::with_capacity_in`
followed by a `for` loop in order to avoid lifetime issues.
- Remove unnecessary linter annotations for `clippy::too_many_arguments`
When an error message reports on a symbol that was generated during
canonicalization, use text like "This value" instead of "This `123`
value". Generated symbols use the identifier index as the symbol name,
since valid Roc variables cannot begin with a number so there's no
chance of collision. We don't want to display generated symbols to the
user, so when building the error message we check if the symbol's name
starts with a digit.
Fix a bug in `dbg` expression desugaring by using the module scope to
generate unique identifiers instead of the variable store.
In the initial implementation of `dbg` expressions we used the
`VarStore` to generate unique identifiers for new variables created
during desugaring. We should have instead used the current module's
`Scope`, which handles identifiers within the module. Each scope has its
own incrementing variable count which is independent of the shared
variable store. The scope is used to generate new identifiers at other
points in canonicalization, such as when assigning a global identifier
to closures and `expect`s. It's possible that the identifier generated
for `dbg` could conflict with an identifier generated by the scope,
resulting in a confusing error.
Instead of parsing dbg with an expression block, parse the dbg keyword
with no additional arguments. This way the parser treats dbg just like a
variable in function application. We desugar by pattern matching on
`Apply(Dbg, args, called_via)` nodes. This changes the output of syntax
tests since the initial AST is different, but does not change the output
of can or mono.
Add two new errors for dbg in expression position with either no args or
too many args. This is similar to the error behavior of `crash`.
Continue to parse dbg statements with an expression block, as before.
Increase from 36 levels of indentation to 80. This is excessive, but
it's confusing to have a parser test crash just because
parse_debug_trace is enabled.
Add dbg parsing logic everywhere we parse normal expressions. Add
special case to statement parsing to handle a series of statements
ending in a `dbg` in expression position.
Rename existing `dbg_help` function to `dbg_stmt_help`, similarly rename
syntax snapshot test files to specify which ones are for dbg statements.
Make `Expr::Dbg` node specific to dbg expressions, add `Expr::DbgStmt`
for dbg statements. Debug expressions don't have a continuation branch.
Support formatting both kinds of `Dbg` nodes.
