Documentation

Transforms are organised by the function being applied on the lhs of the transform, to make looking up appropriate transforms quicker

Add transformation rules from a definition, which will reverse the definition for an error to make it more readable

Say that the name should always be reduced in error messages, to help readability/error reflection

get the accessibility of a name outside this module

Adds error handlers for a particular function and argument. If names are ambiguous, all matching handlers are updated.

Trace all the names in a call graph starting at the given name

Add an interface implementation function.

Precondition: the implementation should have the correct type.

Dodgy hack 1: Put integer implementations first in the list so they are resolved by default.

Dodgy hack 2: put constraint chasers (ParentN) last

Add a privileged implementation - one which implementation search will happily resolve immediately if it is type correct This is used for naming parent implementations when defining an implementation with constraints. Returns the old list, so we can revert easily at the end of a block

Used to preserve sharing of names

A version of liftIO that puts errors into the exception type of the Idris monad

InternalApp keeps track of the real function application for making case splits from, not the application the programmer wrote, which doesn't have the whole context in any case other than top level definitions

Pattern definitions are only used for coverage checking, so erase them when we're done

Erase types from Ps in the context (basically ending up with what's in the .ibc, which is all we need after all the analysis is done)

Save information about a name that is not yet defined

Tell clients how much was parsed and loaded

Set appropriate optimisation set for the given level. We only have one optimisation that is configurable at the moment, however!

Log an action of the parser

Log an action of the elaborator.

Log an action of the compiler.

Log aspect of Idris execution

An empty set of logging levels is used to denote all categories.

@TODO update IDE protocol

Calculate a priority for a type, for deciding elaboration order * if it's just a type variable or concrete type, do it early (0) * if there's only type variables and injective constructors, do it next (1) * if there's a function type, next (2) * finally, everything else (3)

Add implicit bindings from using block, and bind any missing names

Even if auto_implicits is off, we need to call this so we record which arguments are implicit

Add implicit arguments in function calls

Add the implicit arguments to applications in the term [Name] gives the names to always expend, even when under a binder of that name (this is to expand methods with implicit arguments in dependent interfaces).

return True if the second argument is an implicit argument which is expected in the implicits, or if it's not an implicit

Remove functions which aren't applied to anything, which must then be resolved by unification. Assume names resolved and alternatives filled in (so no ambiguity).

Find static argument positions (the declared ones, plus any names in argument position in the declared statics) FIXME: It's possible that this really has to happen after elaboration

Syntactic match of a against b, returning pair of variables in a and what they match. Returns the pair that failed if not a match.

Rename any binders which are repeated (so that we don't have to mess about with shadowing anywhere else).