Reader Name

Do not use the data constructors of RdrName directly: prefer the family of functions that creates them, such as mkRdrUnqual

• Note: A Located RdrName will only have API Annotations if it is a compound one, e.g.

`bar`
( ~ )

• AnnKeywordId : AnnType, AnnOpen '(' or '[' or '[:', AnnClose ')' or ']' or ':]',, AnnBackquote '`', AnnVal AnnTilde,

Make a qualified RdrName in the given namespace and where the ModuleName and the OccName are taken from the first and second elements of the tuple respectively

Local Reader Environment

This environment is used to store local bindings (let, where, lambda, case). It is keyed by OccName, because we never use it for qualified names We keep the current mapping, *and* the set of all Names in scope Reason: see Note [Splicing Exact names] in RnEnv

Global Reader Environment

Keyed by OccName; when looking up a qualified name we look up the OccName part, and then check the Provenance to see if the appropriate qualification is valid. This saves routinely doubling the size of the env by adding both qualified and unqualified names to the domain.

The list in the codomain is required because there may be name clashes These only get reported on lookup, not on construction

INVARIANT 1: All the members of the list have distinct gre_name fields; that is, no duplicate Names

INVARIANT 2: Imported provenance => Name is an ExternalName However LocalDefs can have an InternalName. This happens only when type-checking a [d| ... |] Template Haskell quotation; see this note in RnNames Note [Top-level Names in Template Haskell decl quotes]

INVARIANT 3: If the GlobalRdrEnv maps [occ -> gre], then greOccName gre = occ

NB: greOccName gre is usually the same as nameOccName (gre_name gre), but not always in the case of record seectors; see greOccName

Look for precisely this Name in the environment. This tests whether it is in scope, ignoring anything else that might be in scope with the same OccName.

Look for a particular record field selector in the environment, where the selector name and field label may be different: the GlobalRdrEnv is keyed on the label. See Note [Parents for record fields] for why this happens.

Look for precisely this Name in the environment, but with an OccName that might differ from that of the Name. See lookupGRE_FieldLabel and Note [Parents for record fields].

Apply a transformation function to the GREs for these OccNames

Takes a list of GREs which have the right OccName x Pick those GREs that are in scope * Qualified, as x if want_qual is Qual M _ * Unqualified, as x if want_unqual is Unqual _

Return each such GRE, with its ImportSpecs filtered, to reflect how it is in scope qualified or unqualified respectively. See Note [GRE filtering]

Pick GREs that are in scope *both* qualified *and* unqualified Return each GRE that is, as a pair (qual_gre, unqual_gre) These two GREs are the original GRE with imports filtered to express how it is in scope qualified an unqualified respectively

Used only for the 'module M' item in export list; see RnNames.exports_from_avail

make a GlobalRdrEnv where all the elements point to the same Provenance (useful for "hiding" imports, or imports with no details).

Takes a list of distinct GREs and folds them into AvailInfos. This is more efficient than mapping each individual GRE to an AvailInfo and the folding using plusAvail but needs the uniqueness assumption.

Global Reader Element

An element of the GlobalRdrEnv

Is this a record field defined with DuplicateRecordFields? (See Note [Parents for record fields])

Test if an unqualified version of this thing would be in scope

Is in scope qualified with the given module?

Is in scope unqualified?

Print out one place where the name was define/imported (With -dppr-debug, print them all)

The children of a Name are the things that are abbreviated by the ".." notation in export lists. See Note [Parents]

Import Specification

The ImportSpec of something says how it came to be imported It's quite elaborate so that we can give accurate unused-name warnings.

Import Declaration Specification

Describes a particular import declaration and is shared among all the Provenances for that decl

Import Item Specification

Describes import info a particular Name

Display info about the treatment of * under NoStarIsType.

With StarIsType, three properties of * hold:

(a) it is not an infix operator (b) it is always in scope (c) it is a synonym for Data.Kind.Type

However, the user might not know that he's working on a module with NoStarIsType and write code that still assumes (a), (b), and (c), which actually do not hold in that module.

Violation of (a) shows up in the parser. For instance, in the following examples, we have * not applied to enough arguments:

data A :: * data F :: * -> *

Violation of (b) or (c) show up in the renamer and the typechecker respectively. For instance:

type K = Either * Bool

This will parse differently depending on whether StarIsType is enabled, but it will parse nonetheless. With NoStarIsType it is parsed as a type operator, thus we have ((*) Either Bool). Now there are two cases to consider:

1. There is no definition of (*) in scope. In this case the renamer will fail to look it up. This is a violation of assumption (b).
2. There is a definition of the (*) type operator in scope (for example coming from GHC.TypeNats). In this case the user will get a kind mismatch error. This is a violation of assumption (c).

The user might unknowingly be working on a module with NoStarIsType or use * as Type out of habit. So it is important to give a hint whenever an assumption about * is violated. Unfortunately, it is somewhat difficult to deal with (c), so we limit ourselves to (a) and (b).

starInfo generates an appropriate hint to the user depending on the extensions enabled in the module and the name that triggered the error. That is, if we have NoStarIsType and the error is related to * or its Unicode variant, the resulting SDoc will contain a helpful suggestion. Otherwise it is empty.