Files
mercury/compiler/prog_item.m
2023-01-22 19:37:20 +11:00

3257 lines
141 KiB
Mathematica

%---------------------------------------------------------------------------%
% vim: ft=mercury ts=4 sw=4 et
%---------------------------------------------------------------------------%
% Copyright (C) 1996-2011 The University of Melbourne.
% Copyright (C) 2014-2021 The Mercury team.
% This file may only be copied under the terms of the GNU General
% Public License - see the file COPYING in the Mercury distribution.
%---------------------------------------------------------------------------%
%
% File: prog_item.m.
% Original author: fjh.
%
% This module, together with prog_data*.m, defines a data structure for
% representing Mercury programs.
%
% This data structure specifies basically the same information as is
% contained in the source code, but in a parse tree rather than a flat file.
% This module defines the parts of the parse tree that are *not* needed
% by the various compiler backends; parts of the parse tree that
% are needed by the backends are contained in prog_data*.m.
%
%---------------------------------------------------------------------------%
%
% One important consideration in the design of the parse trees is that
% they have two different use cases:
%
% - to represent files being read in, and
% - to represent files being written out.
%
% The two have slightly different requirements, which is why in several
% kinds of parse trees seemingly the same information is present in
% more than one set of fields. This is because while we will never
% knowingly write out erroneous Mercury code, we know that we *will*
% read in some. An example is import_module and use_module declarations.
% Each parse_tree_module_src contains four fields that respectively specify
%
% - the locations where a module has an import_module in the interface
% - the locations where a module has an use_module in the interface
% - the locations where a module has an import_module in the implementation
% - the locations where a module has an use_module in the implementation
%
% It is an error if a module has an entry in more than one of these maps,
% with the sole exception being the use_module in interface and import_module
% in implementation combination (because each grants a permission that the
% other does not). Yet we want the ability to represent even invalid
% combinations of these declarations, so that we can wait to generate
% the appropriate error messages until we know all the relevant facts.
% And in the process of checking for and reporting errors, we build up
% another data structure, the import_and_or_use_map, which contains
% a record of how the rest of the compiler should view, not just the
% import_module and use_module declarations explicitly present in the
% source code, but also the ones that get made available to it implicitly.
% (Examples include builtin and private_builtin, which are implicitly available
% to every module, and table_builtin, which is implicitly available
% to modules that do certain kinds of tabling.)
%
%---------------------------------------------------------------------------%
:- module parse_tree.prog_item.
:- interface.
:- import_module libs.
:- import_module libs.globals.
:- import_module mdbcomp.
:- import_module mdbcomp.prim_data.
:- import_module mdbcomp.sym_name.
:- import_module parse_tree.error_spec.
:- import_module parse_tree.maybe_error.
:- import_module parse_tree.prog_data.
:- import_module parse_tree.prog_data_foreign.
:- import_module parse_tree.prog_data_pragma.
:- import_module recompilation. % XXX undesirable dependency
:- import_module assoc_list.
:- import_module cord.
:- import_module list.
:- import_module map.
:- import_module maybe.
:- import_module one_or_more.
:- import_module one_or_more_map.
:- import_module pair.
:- import_module set.
%---------------------------------------------------------------------------%
:- type include_module_map == map(module_name, include_module_info).
:- type int_include_module_map == map(module_name, int_include_module_info).
:- type include_module_info
---> include_module_info(module_section, prog_context).
% The "include_module" declaration occurs in the given section
% of the relevant file, and at the given context.
:- type int_include_module_info =< include_module_info
---> include_module_info(int_module_section, prog_context).
:- type int_module_section =< module_section
---> ms_interface.
:- type module_name_context == map(module_name, prog_context).
:- type module_names_contexts == one_or_more_map(module_name, prog_context).
% Maps from module names to the includes, imports or uses
% in the named section. The code creating these maps will have
% detected and diagnosed any duplicate entries of the same kind
% of declaration for the same module in the same section.
% However, unlike include_module_maps or import_and_or_use_maps,
% which summarize the information in the first two of the maps below
% (for include_module_map) or the last four (for import_and_or_use_map),
% these maps may contain redundant entries as long as they are all
% in *different* maps (such as the module name A occurring in both
% the int_import_context_map and the int_use_context_map of module B).
:- type int_incl_context_map
---> int_incl_context_map(module_name_context).
:- type imp_incl_context_map
---> imp_incl_context_map(module_name_context).
:- type int_import_context_map
---> int_import_context_map(module_name_context).
:- type int_use_context_map
---> int_use_context_map(module_name_context).
:- type imp_import_context_map
---> imp_import_context_map(module_name_context).
:- type imp_use_context_map
---> imp_use_context_map(module_name_context).
%---------------------%
% The module being compiled can have another module made available to it
% either explicitly or implicitly.
%
% An explicit availability can happen either through an `:- import_module'
% or a `:- use_module' declaration (import or use, for short), and these
% declarations can occur in either the interface section or in the
% implementation section.
%
% The values of section_import_and_or_use specify the possible valid ways
% that a module may be made available explicitly. The first four specify
% the usual ways: an import or use in either section, with the one context.
% The last one says that the module was named in an use_module declaration
% in the interface section and in an import_module declaration
% in the implementation section, and give the two contexts respectively.
%
% The values of the implicit_import_or_use type specify the possible ways
% that a module may be made available implicitly. Most implicit availability
% is to give the compiler access to the declarations of predicates and
% functions that the compiler will automatically insert calls to as part of
% the implementation of some language feature, such as table resets for
% memoed procedures. Since these automatically generated references
% will be created fully module qualified, an import of the target modules
% is not needed; a use is enough. However, the public builtin module
% is implicitly imported into every Mercury module, and this one
% does get imported, not used.
%
% Note that we do not record contexts for implicit uses. In general,
% there is no *single specific* context that makes an implicit use needed,
% and we don't need contexts for any error messages about implicit imports,
% since we don't want to require Mercury programmers to have to know
% such details of the Mercury implementation. We *could* collect the set of
% contexts that make a given implicit use needed, for internal compiler
% purposes, but we do not (yet) have any need for that information.
:- type section_import_and_or_use
---> int_import(prog_context)
; int_use(prog_context)
; imp_import(prog_context)
; imp_use(prog_context)
; int_use_imp_import(prog_context, prog_context).
:- type section_use =< section_import_and_or_use
---> int_use(prog_context)
; imp_use(prog_context).
:- type implicit_import_or_use
---> implicit_int_import
; implicit_int_use
; implicit_imp_use.
:- type maybe_implicit_import_and_or_use
---> explicit_avail(
section_import_and_or_use
)
; implicit_avail(
implicit_import_or_use,
maybe(section_import_and_or_use)
).
% Values of this type specify how each module we have available
% was *made* available. If a module had redundant import_module
% and/or use_module declarations, each of these has had a warning
% generated for it and was then discarded. One of these declarations
% can be made redundant redundant not only by another declaration
% of the same kind in the same section, but also by more permissive
% declarations; import_module declarations grant more permissions
% than use_module declarations, and declarations in the interface
% give more permissions than the declarations of the same kind
% in the implementation section.
:- type section_import_and_or_use_map ==
map(module_name, section_import_and_or_use).
:- type section_use_map ==
map(module_name, section_use).
:- type import_and_or_use_map ==
map(module_name, maybe_implicit_import_and_or_use).
%---------------------------------------------------------------------------%
%
% The parse_tree_{src,int,opt} types define the ASTs we use for source files,
% interface files and optimization files respectively.
%
% Nested submodules may appear in source files, but not in interface files
% or optimization files.
%
% We use cords of items instead of lists of items where we may need to add
% items to an already-existing partial parse tree.
%
% The contexts of module declarations below may be term_context.dummy_context
% if the actual context isn't known, but if the recorded context is
% not term_context.dummy_context, then it is valid.
:- type parse_tree_src
---> parse_tree_src(
pts_module_name :: module_name,
% The context of the `:- module' declaration.
pts_module_name_context :: prog_context,
% The contents of the module.
pts_components :: cord(module_component)
).
:- type module_component
---> mc_section(
mcs_module_name :: module_name,
mcs_section_kind :: module_section,
% The context of the `:- interface' or `:- implementation'
% declaration.
mcs_section_context :: prog_context,
mcs_includes :: cord(item_include),
mcs_avails :: cord(item_avail),
pti_fims :: cord(item_fim),
mcs_items :: cord(item)
)
; mc_nested_submodule(
% The name of the *including* module.
mcns_module_name :: module_name,
% What kind of section is the submodule in?
mcns_in_section_kind :: module_section,
% The context of the section that the submodule is in.
mcns_in_section_context :: prog_context,
% The submodule itself.
mcns_submodule :: parse_tree_src
).
:- type parse_tree_module_src
---> parse_tree_module_src(
ptms_module_name :: module_name,
% The context of the `:- module' declaration.
ptms_module_name_context :: prog_context,
% The set of modules mentioned in `:- include_module'
% declarations in the interface and in the implementation,
% and their locations. If a module has been included N times,
% which is an error, it will appear in (one or both of)
% these maps N times.
ptms_int_includes :: module_names_contexts,
ptms_imp_includes :: module_names_contexts,
% A cleaned-up version of the above two fields,
% which maps each included module to *one* effective
% section of inclusion (which will be the interface section
% if the module is ever included in the interface)
% and *one* effective context. The process of filling in
% this field will generate error messages for any duplicate
% inclusions.
ptms_include_map :: include_module_map,
% A specification of the set of modules mentioned in
% `:- import_module' and/or `:- use_module' declarations
% in each section, mapped to their location(s).
% Again, any module imported and/or used N times
% will appear in these maps N times.
ptms_int_imports :: module_names_contexts,
ptms_int_uses :: module_names_contexts,
ptms_imp_imports :: module_names_contexts,
ptms_imp_uses :: module_names_contexts,
% A cleaned-up and extended version of the above four fields.
%
% The cleaned-up part means that each module is mapped
% to exactly *one* section_import_and_or_use,
% reporting any invalid duplicate availability in the process.
% (Having a use_module in the interface section and an
% import_module in the implementation section is the only
% allowed situation in which a module may have more than one
% import or use declaration.)
%
% The extended part means that this field contains information
% about implicit availability of builtin modules as well,
% in a form that allows explicit vs implicit availability
% to be clearly distinguished from each other.
ptms_import_use_map :: import_and_or_use_map,
% A cleaned-up version of the set of explicit
% `:- pragma foreign_import_module' declarations
% in the interface and in the implementation.
% The cleaned-up part means that we we have reported both
%
% - FIMs that occur more than once in a given section, and
% - FIMs that occur in both sections.
%
% We keep the context of only the first FIM for a given
% fim_spec in each section, and if a fim_spec occurs in
% both sections, we keep only the (first) occurrence in the
% interface section.
%
% We don't have a field containing the original, non-cleaned-up
% data, since no part of the compiler (yet) need this.
ptms_int_fims :: map(fim_spec, prog_context),
ptms_imp_fims :: map(fim_spec, prog_context),
% The set of foreign languages for which this module
% should have implicit foreign_import_module declaration
% for itself, in the interface and implementation respectively.
ptms_int_self_fim_langs :: set(foreign_language),
ptms_imp_self_fim_langs :: set(foreign_language),
ptms_type_defns :: type_ctor_checked_map,
ptms_inst_defns :: inst_ctor_checked_map,
ptms_mode_defns :: mode_ctor_checked_map,
% The error messages generated during the construction
% of ptms_type_defns. We have found some invalid types if
% some of these error_specs (a) are severity_error, and
% (b) are phase_type_inst_mode_check_invalid_type.
ptms_type_specs :: list(error_spec),
% The error messages generated during the construction
% of ptms_inst_defns and ptms_mode_defns. We have found
% some invalid insts and/or more if some of these error_specs
% (a) are severity_error, and (b) are
% phase_type_inst_mode_check_invalid_inst_mode.
ptms_inst_mode_specs :: list(error_spec),
% Items of various kinds in the interface.
% All these items are to be treated as being in the
% interface section, with one exception.
% If this module has some submodules, i.e. if the
% ptms_include_map field above is nonempty, then we handle
% any nonabstract instance items in the interface by
% - treating only an abstract version of the item as being
% in the interface, and
% - treating the original version as being in the
% implementation section, but exported to submodules.
% (For abstract instances, there is no point in adding them
% twice, once in each section, so we treat them as only
% being in the interface.)
ptms_int_typeclasses :: list(item_typeclass_info),
ptms_int_instances :: list(item_instance_info),
ptms_int_pred_decls :: list(item_pred_decl_info),
ptms_int_mode_decls :: list(item_mode_decl_info),
ptms_int_decl_pragmas :: list(item_decl_pragma_info),
ptms_int_promises :: list(item_promise_info),
% The set of predicate names for which the interface contains
% either attempts at a definition (i.e. a clause or a
% foreign_proc), or something else that tells us that
% generating a warning about a lack of a definition
% in the implementation section (if in fact there is
% no definition there) would be more misleading than useful.
ptms_int_bad_clauses :: set(pred_pf_name_arity),
% A repeat of everything above, but in the implementation
% section, with the addition of some item kinds that may occur
% *only* in implementation sections.
%
% However, note that the conversion process we now use
% to generate parse_tree_module_srcs will put any impl pragmas,
% initialises, finalises and mutables that were wrongly placed
% in the interface section into their fields below, so that
% if there is something wrong with them *beyond* their
% location, the compiler can detect and report it in the
% same compiler invocation. It would be easy to put these
% misplaced items into separate fields of their own,
% but so far there has been no need for that.
%
% If this module has no submodules, i.e. if the
% ptms_include_map field above is empty, then all the items
% in these fields are to be treated as in being in the
% implementation section. However, if this module HAS
% at least one submodule (in either section), then only
% the following kinds of items are to be treated as being
% private to this module:
%
% clauses
% foreign_export_enums
% impl_pragmas
% initialises
% finalises
%
% All the other kinds of items are to be treated as being
% exported to submodules.
ptms_imp_typeclasses :: list(item_typeclass_info),
ptms_imp_instances :: list(item_instance_info),
ptms_imp_pred_decls :: list(item_pred_decl_info),
ptms_imp_mode_decls :: list(item_mode_decl_info),
ptms_imp_clauses :: list(item_clause_info),
ptms_imp_foreign_export_enums ::
list(item_foreign_export_enum_info),
ptms_imp_decl_pragmas :: list(item_decl_pragma_info),
ptms_imp_impl_pragmas :: list(item_impl_pragma_info),
ptms_imp_promises :: list(item_promise_info),
ptms_imp_initialises :: list(item_initialise_info),
ptms_imp_finalises :: list(item_finalise_info),
ptms_imp_mutables :: list(item_mutable_info)
).
% When comp_unit_interface.m creates the contents of an interface file,
% it will always set the maybe_version_numbers field of that interface file
% to `no_version_numbers'. If the value of that field is needed,
% it will be filled in by the actually_write_interface_file predicate
% in write_module_interface_files.m, which (unlike comp_unit_interface.m)
% has access to the I/O state to read in the *previous* version
% of that interface file.
:- type maybe_version_numbers
---> no_version_numbers
; version_numbers(module_item_version_numbers).
% The representations specific to .int0, .int, .int2 and .int3 files.
% XXX We should replace the lists of items of various kinds with data
% structures that encode uniqueness properties, such as "each type constructor
% may be defined only once". Maps from primary keys such as type_ctors,
% or symnames/arity pairs in general, would work for this.
% A representation of the contents of .int0 files.
:- type parse_tree_int0
---> parse_tree_int0(
pti0_module_name :: module_name,
% The context of the `:- module' declaration.
pti0_module_name_context :: prog_context,
pti0_maybe_version_numbers :: maybe_version_numbers,
% The set of modules mentioned in `:- include_module'
% declarations in the interface and implementation,
% and their locations.
pti0_include_map :: include_module_map,
% The set of modules mentioned in `:- import_module'
% declarations in the interface and implementation,
% and their locations.
pti0_import_use_map :: section_import_and_or_use_map,
% `:- pragma foreign_import_module' declarations
% in the interface and in the implementation.
pti0_int_fims :: set(fim_spec),
pti0_imp_fims :: set(fim_spec),
% Type, inst and mode definitions from both
% the interface and implementation sections.
pti0_type_defns :: type_ctor_checked_map,
pti0_inst_defns :: inst_ctor_checked_map,
pti1_mode_defns :: mode_ctor_checked_map,
% Items of various kinds in the interface.
% XXX For the consumers of the .int0 file, in most cases
% it makes no difference whether an item was in the parent's
% interface or implementation section. We should make that
% distinction here ONLY when we have to.
pti0_int_typeclasses :: list(item_typeclass_info),
pti0_int_instances :: list(item_instance_info),
pti0_int_pred_decls :: list(item_pred_decl_info),
pti0_int_mode_decls :: list(item_mode_decl_info),
pti0_int_decl_pragmas :: list(item_decl_pragma_info),
pti0_int_promises :: list(item_promise_info),
% Items of various kinds in the implementation section.
pti0_imp_typeclasses :: list(item_typeclass_info),
pti0_imp_instances :: list(item_instance_info),
pti0_imp_pred_decls :: list(item_pred_decl_info),
pti0_imp_mode_decls :: list(item_mode_decl_info),
pti0_imp_decl_pragmas :: list(item_decl_pragma_info),
pti0_imp_promises :: list(item_promise_info)
).
% A representation of the contents of .int files.
:- type parse_tree_int1
---> parse_tree_int1(
pti1_module_name :: module_name,
% The context of the `:- module' declaration.
pti1_module_name_context :: prog_context,
pti1_maybe_version_numbers :: maybe_version_numbers,
% The set of modules mentioned in `:- include_module'
% declarations in the interface and implementation,
% and their contexts.
pti1_include_map :: include_module_map,
% The set of modules mentioned in `:- use_module'
% declarations in the interface and implementation,
% and their locations.
pti1_use_map :: section_use_map,
% `:- pragma foreign_import_module' declarations
% in the interface and in the implementation.
pti1_int_fims :: set(fim_spec),
pti1_imp_fims :: set(fim_spec),
% Type, inst and mode definitions, all of which are
% in the interface, with the exception of some type
% definitions from the implementation section
% (which should not be needed after we start actually
% *using* type_repn items).
pti1_type_defns :: type_ctor_checked_map,
pti1_inst_defns :: inst_ctor_checked_map,
pti1_mode_defns :: mode_ctor_checked_map,
% Items of various kinds in the interface.
pti1_int_typeclasses :: list(item_typeclass_info),
pti1_int_instances :: list(item_instance_info),
pti1_int_pred_decls :: list(item_pred_decl_info),
pti1_int_mode_decls :: list(item_mode_decl_info),
pti1_int_decl_pragmas :: list(item_decl_pragma_info),
pti1_int_promises :: list(item_promise_info),
% The representations of all types defined in the module,
% whether exported or not.
pti1_type_repns :: type_ctor_repn_map,
% Items of various kinds in the implementation.
pti1_imp_typeclasses :: list(item_typeclass_info)
).
% A representation of the contents of .int2 files.
:- type parse_tree_int2
---> parse_tree_int2(
pti2_module_name :: module_name,
% The context of the `:- module' declaration.
pti2_module_name_context :: prog_context,
% XXX While it is clear that .int files need version number
% fields while .int3 files do not, I (zs) don't see any
% clear argument either way for .int2 files. Having
% the field here preserves old behavior.
pti2_maybe_version_numbers :: maybe_version_numbers,
% The set of modules mentioned in `:- include_module'
% declarations in the interface, and their locations.
pti3_int_includes :: int_include_module_map,
% The set of modules mentioned in `:- use_module'
% declarations in the interface, and their locations.
pti2_use_map :: section_use_map,
% `:- pragma foreign_import_module' declarations
% in the interface and in the implementation.
pti2_int_fims :: set(fim_spec),
pti2_imp_fims :: set(fim_spec),
% Type, inst and mode definitions, all of which are
% in the interface, with the exception of some type
% definitions from the implementation section
% (which should not be needed after we start actually
% *using* type_repn items).
pti2_type_defns :: type_ctor_checked_map,
pti2_inst_defns :: inst_ctor_checked_map,
pti2_mode_defns :: mode_ctor_checked_map,
% Items of various kinds in the interface.
pti2_int_typeclasses :: list(item_typeclass_info),
pti2_int_instances :: list(item_instance_info),
% The representations of all types defined in the module,
% whether exported or not.
pti2_type_repns :: type_ctor_repn_map
).
% A representation of the contents of .int3 files.
:- type parse_tree_int3
---> parse_tree_int3(
pti3_module_name :: module_name,
% The context of the `:- module' declaration.
pti3_module_name_context :: prog_context,
% The set of modules mentioned in `:- include_module'
% declarations in the interface, and their locations.
pti3_int_includes :: int_incl_context_map,
% The set of modules mentioned in `:- import_module'
% declarations in the interface, and their locations.
pti3_int_imports :: int_import_context_map,
% Type, inst and mode definitions, all of which are
% in the interface.
pti3_type_defns :: type_ctor_checked_map,
pti3_inst_defns :: inst_ctor_checked_map,
pti3_mode_defns :: mode_ctor_checked_map,
% Items of various kinds in the interface.
pti3_int_typeclasses :: list(item_typeclass_info),
pti3_int_instances :: list(item_instance_info),
pti3_int_type_repns :: type_ctor_repn_map
).
%---------------------------------------------------------------------------%
%
% The intended semantics of a type_ctor_defn_map is a map of
% all the type constructors defined in a given SECTION of a given
% interface file to all its definitions in that section.
%
% There are four intended uses of a type_ctor_defn_map. The most
% important is the fourth one.
%
% One is to eliminate unnecessary items from interface files.
% For example, library/set.m contains two definitions of the set
% type constructor: an abstract definition in the publicly documented
% interface section, and an actual du definition in another interface
% section that we do not include in the automatically generated
% documentation but we *do* export to other modules. In situations
% like this, the abstract definition is redundant. Never including it
% in an interface file lets that interface file to remain unchanged
% in the event that the user deletes the abstract definition from
% the source file as well.
%
% The second use is to canonicalize the parts of interface files
% containing type definitions.
%
% The third use is to help deal with sets of definitions that
% don't make sense. There are many rules that a set of definitions
% for a given type constructor must meet (such as "there may be at most one
% definition for a type constructor that is a du, equivalence or solver
% definition), and bugs may manifest themselves as violations of these rules.
%
% We have a choice in when these violations are detected.
%
% - If we allow the inclusion of inconsistent sets of type definitions
% in interface files, then we must detect and handle these
% inconsistencies every time a compiler invocation reads that interface
% file. These invocations won't generate error messages for these
% inconsistencies since the type constructor won't be local, but
% they may generate messages for other "errors" that look like errors
% only because the compiler's resolution of the inconsistency (i.e.
% its choice of which type definitions to keep and which to throw out)
% differs from the programmer's choice.
%
% - If we do NOT allow the inclusion of inconsistent sets of type
% definitions in interface files, then we must report any violations
% at interface file construction time, and make them cause that
% construction to fail. Printing such error messages to stdout
% instead of the module's .err file is less than ideal, but
% this early detection can avoid avalanches of misleading diagnostics
% of the kind mentioned in the previous point. It can also save
% recompilations. If a module's source file contains inconsistent
% definitions for a type constructor, then the programmer will
% have to delete the unintended ones. Once this is done, the
% interface file will have to be rebuilt. If we allow inconsistent
% definitions in the interface file, its new contents will differ
% from its old contents, which means that all the compilations
% of *other* modules that read the old contents will have been wasted.
% If we cause the construction of the interface file to fail instead,
% those compilations won't have taken place.
%
% We implement the first choice by checking whether each entry in
% a type_ctor_defn_map makes sense, and generating error messages
% when they don't. This is done by code in check_type_inst_mode_defns.m.
%
% The fourth and most motivating use is that having all the definitions
% of a type_ctor, *and* all the foreign_enum pragmas that apply to that
% type_ctor, all together at once will make the code that decides
% the proper representation of that type significantly simpler.
%
% Everything above except the fourth use also applies to the inst_
% and mode_ctor_defn_maps, though for those, the consistency rules are
% much simpler: that each inst and mode constructor must have at most one
% non-abstract definition.
%
:- type type_ctor_defn_map == map(type_ctor, type_ctor_all_defns).
:- type type_ctor_all_defns
---> type_ctor_all_defns(
% Abstract and nonabstract solver type definitions.
tcad_abstract_solver :: list(item_type_defn_info_abstract),
tcad_solver :: list(item_type_defn_info_solver),
% Abstract and nonabstract nonsolver type definitions.
tcad_abstract_std :: list(item_type_defn_info_abstract),
tcad_eqv :: list(item_type_defn_info_eqv),
tcad_du :: list(item_type_defn_info_du),
tcad_sub :: list(item_type_defn_info_sub),
tcad_foreign :: c_j_cs_defns
).
:- type type_ctor_maybe_defn
---> type_ctor_maybe_defn(
% Abstract and nonabstract solver type definitions.
tcmd_abstract_solver :: maybe(item_type_defn_info_abstract),
tcmd_solver :: maybe(item_type_defn_info_solver),
% Abstract and nonabstract nonsolver type definitions.
tcmd_abstract_std :: maybe(item_type_defn_info_abstract),
tcmd_eqv :: maybe(item_type_defn_info_eqv),
tcmd_du :: maybe(item_type_defn_info_du),
tcmd_sub :: maybe(item_type_defn_info_sub),
tcmd_foreign :: c_j_cs_maybe_defn
).
% We support foreign type definitions in all three of our target languages,
% C, Java and C#. Likewise, we allow foreign enum declarations
% in these three languages.
%
% There are several kinds of info that we may want to store for every
% one of these foreign languages. This can be done in instances
% of this type, whose fields always contain the info for C, Java and C#
% (in that order).
:- type c_java_csharp(T)
---> c_java_csharp(T, T, T).
:- type c_j_cs_defns ==
c_java_csharp(list(item_type_defn_info_foreign)).
:- type c_j_cs_maybe_defn ==
c_java_csharp(maybe(item_type_defn_info_foreign)).
:- type c_j_cs_enums ==
c_java_csharp(list(item_foreign_enum_info)).
:- type c_j_cs_maybe_enum ==
c_java_csharp(maybe(item_foreign_enum_info)).
:- type c_j_cs_repn ==
c_java_csharp(maybe(foreign_type_repn)).
:- type c_j_cs_enum_repn ==
c_java_csharp(maybe(enum_foreign_repn)).
:- type inst_ctor_defn_map == map(inst_ctor, inst_ctor_all_defns).
:- type inst_ctor_all_defns
---> inst_ctor_all_defns(
icad_abstract :: list(item_inst_defn_info_abstract),
icad_eqv :: list(item_inst_defn_info_eqv)
).
:- type mode_ctor_defn_map == map(mode_ctor, mode_ctor_all_defns).
:- type mode_ctor_all_defns
---> mode_ctor_all_defns(
mcad_abstract :: list(item_mode_defn_info_abstract),
mcad_eqv :: list(item_mode_defn_info_eqv)
).
:- type type_ctor_foreign_enum_map == map(type_ctor, c_j_cs_enums).
:- type type_ctor_repn_map == map(type_ctor, item_type_repn_info).
%---------------------------------------------------------------------------%
:- type parse_tree_plain_opt
---> parse_tree_plain_opt(
ptpo_module_name :: module_name,
% The context of the `:- module' declaration.
ptpo_module_name_context :: prog_context,
% `:- use_module' (not `:- import_module') declarations.
ptpo_uses :: module_names_contexts,
ptpo_fims :: set(fim_spec),
ptpo_type_defns :: list(item_type_defn_info),
ptpo_foreign_enums :: list(item_foreign_enum_info),
ptpo_inst_defns :: list(item_inst_defn_info),
ptpo_mode_defns :: list(item_mode_defn_info),
ptpo_typeclasses :: list(item_typeclass_info),
ptpo_instances :: list(item_instance_info),
ptpo_pred_decls :: list(item_pred_decl_info),
ptpo_mode_decls :: list(item_mode_decl_info),
ptpo_clauses :: list(item_clause_info),
ptpo_foreign_procs :: list(item_foreign_proc),
ptpo_promises :: list(item_promise_info),
ptpo_pred_marker_pragmas :: list(item_pred_marker),
ptpo_type_spec_pragmas :: list(item_type_spec),
ptpo_unused_args :: list(item_unused_args),
ptpo_termination :: list(item_termination),
ptpo_termination2 :: list(item_termination2),
ptpo_exceptions :: list(item_exceptions),
ptpo_trailing :: list(item_trailing),
ptpo_mm_tabling :: list(item_mm_tabling),
ptpo_struct_sharing :: list(item_struct_sharing),
ptpo_struct_reuse :: list(item_struct_reuse)
).
:- type parse_tree_trans_opt
---> parse_tree_trans_opt(
ptto_module_name :: module_name,
% The context of the `:- module' declaration.
ptto_module_name_context :: prog_context,
ptto_termination :: list(item_termination),
ptto_termination2 :: list(item_termination2),
ptto_exceptions :: list(item_exceptions),
ptto_trailing :: list(item_trailing),
ptto_mm_tabling :: list(item_mm_tabling),
ptto_struct_sharing :: list(item_struct_sharing),
ptto_struct_reuse :: list(item_struct_reuse)
).
%---------------------------------------------------------------------------%
%
% A parse_tree_module_src is one module to be compiled. A parse_tree_src that
% contains N nested submodules corresponds to 1 + N parse_tree_module_srcs,
% one for the top level module, and one for each (possibly deeply) nested
% submodule.
%
% A raw compilation unit consists of some raw item blocks, with each raw
% item block containing the items in an interface or implementation section
% of its module.
%
% Before we convert a parse_tree_module_src into the HLDS, we augment it
% with the contents of the interface files of the modules it imports
% (directly or indirectly), and if requested, with the contents of the
% optimization files of those modules as well. The augmented compilation unit
% will consist of the following for compiler invocations that generate
% target language code. (Compiler invocations that generate .int and .int2
% files will construct an aug_make_int_unit, not an aug_compilation_unit.)
%
% - The module_src field contains the original parse_tree_module_src.
%
% - The ancestor_int_specs field contains the .int0 interface files of
% the ancestors of this module, which are always implicitly imported.
%
% - The direct_int_specs field contains the .int files of the modules
% directly imported or used by this module, with the "override" exception
% noted below.
%
% - The indirect_int_specs field contains the .int2 files of the modules
% indirectly imported or used by this module, again with the "override"
% exception noted below.
%
% In this case, module A "indirectly imports or uses" module C if
% module A imports or uses a module B whose .int file uses module C.
% (.int files only use modules; they do not import them.)
%
% The exceptions above are that
%
% o if a module's .int0 file is in the ancestor_int_specs field,
% we don't include its .int1 file in the direct_int_specs field,
% or its .int2 file in the indirect_int_specs field. In effect,
% the appearance of a module in the ancestor_int_specs field
% overrides (i.e. prevents) its appearance in the direct_int_specs
% or the indirect_int_specs fields.
%
% o if a module's .int file is in the direct_int_specs field,
% we don't include its .int2 file in the indirect_int_specs field.
% Again, the appearance of a module in the direct_int_specs field
% overrides its appearance in the indirect_int_specs field.
%
% The reason for the exceptions is that an .int0 file contains (or at least
% is intended to contain, which *may* be different) every item that
% the .int file for the same module contains, and the same relationship
% holds between .int and .int2 files. The exceptions thus save the compiler
% from doing work that (a) is unnecessary, and (b) would lead things
% being declared or defined more than once.
%
% - Provided intermodule optimization is enabled, the plain_opts field
% will contain
%
% o the .opt files of the modules whose .int0, .int or .int2 files
% are in the ancestor_int_sprcs, direct_int_specs and indirect_int_specs
% fields above, and
%
% o unless the compiler is invoked with --no-read-opt-files-transitively,
% the .opt files of every other module the .opt files specified
% by either the previous bullet point or *this* bullet point
% import either explicitly or implicitly.
%
% These .opt files are supposed to contain more information about
% the ancestor-, direct- or indirect-imported modules than their
% .int0, .int or .int2 files do. Unfortunately, they often also
% *duplicate* items in those interface files, which leads to
% double definitions, which the submodules of make_hlds.m have to
% be prepared to detect and ignore.
%
% - Provided transitive intermodule optimization is enabled, the trans_opts
% field will contain the .trans_opt files of the modules named in
% the module's .d file as the module's trans_opt dependencies.
% XXX This seems to me (zs) a bit too indirect.
%
% - If intermodule optimization is enabled, the int_for_opt_specs field
% will contain
%
% o the .int0 files of the ancestor modules of the modules whose .opt files
% are in the plain_opts field,
%
% o the .int files of the modules imported or used either explicitly
% or implicitly by the modules whose .opt files are in the plain_opts
% field, or by their ancestors, and
%
% o the .int2 files of the modules used by the .int files in the previous
% bullet point.
%
% The idea is that these interface files may in general be needed to define
% entities (such as types, insts or modes) that the .opt files in the
% plain_opts field may need.
%
% XXX There is a problem here, which is that override exception does *not*
% apply to the int_for_opt_specs field. It is possible for e.g. a module's
% .int2 file to appear in the indirect_int_specs field, but its .int0 or
% .int file to appear in the int_for_opt_specs field. This may also lead
% to double definitions of e.g. types, insts or modes. The compiler does
% ignore such double definitions, under the principle of generating error
% messages for double definitions *only* when the entity being double-defined
% has the module currently being compiled as its module qualifier.
% Nevertheless, including more than one interface file for any given module
% in the augmented compilation unit will lead to wasted work, which means
% that we should avoid doing that if possible.
%
:- type aug_compilation_unit
---> aug_compilation_unit(
% The source code of the module.
acu_module_src :: parse_tree_module_src,
% The interface files of the ancestors of this module.
% (If we have e.g. module foo.bar among the modules
% we import int_for_opt, we also need to grab its ancestor foo,
% but such .int0 files also go into the int_for_opt field.
acu_ancestor_int_specs :: map(module_name,
ancestor_int_spec),
% The interface files of directly imported modules.
acu_direct_int1_specs :: map(module_name,
direct_int1_spec),
% The interface files of indirectly imported modules.
acu_indirect_int2_specs :: map(module_name,
indirect_int2_spec),
% The optimization files of directly or indirectly
% imported modules.
acu_plain_opts :: map(module_name,
parse_tree_plain_opt),
acu_trans_opts :: map(module_name,
parse_tree_trans_opt),
% The interface files needed to make sense
% of those optimization files.
acu_int_for_opt_specs :: map(module_name,
int_for_opt_spec),
% Interface files that we read in only for the type
% representation information they contain
acu_type_repn_specs :: map(module_name,
type_repn_spec),
% The module_version_numbers records in all the imported
% interface files.
acu_module_item_version_numbers_map ::
module_item_version_numbers_map
).
:- type aug_make_int_unit
---> aug_make_int_unit(
% The source code of the module.
amiu_module_src :: parse_tree_module_src,
% The interface files of the ancestors of this module.
% (The read_why_int0 is always implicitly rwi0_section.)
amiu_ancestor_int_specs :: map(module_name,
parse_tree_int0),
% The interface files of directly imported modules.
amiu_direct_int3_specs :: map(module_name,
direct_int3_spec),
% The interface files of indirectly imported modules.
amiu_indirect_int3_specs :: map(module_name,
indirect_int3_spec),
% The module_version_numbers records in all the imported
% interface files.
amiu_module_item_version_numbers_map ::
module_item_version_numbers_map
).
% init_aug_compilation_unit(ParseTreeModuleSrc, AugCompUnit):
%
% Initialize an augmented compilation unit structure. Put the given
% ParseTreeModuleSrc into it, and leave the rest of the structure empty.
% Our caller is the expected to fill in (i.e. augment) the structure
% by calling the aug_compilation_unit_add_X predicates in grab_modules.
% to add the parse trees of the interface and optimization files needed
% to compile ParseTreeModuleSrc.
%
:- pred init_aug_compilation_unit(parse_tree_module_src::in,
aug_compilation_unit::out) is det.
:- type ancestor_int_spec
---> ancestor_int0(parse_tree_int0, read_why_int0).
:- type direct_int1_spec
---> direct_int1(parse_tree_int1, read_why_int1).
:- type direct_int3_spec
---> direct_int3(parse_tree_int3, read_why_int3).
:- type indirect_int2_spec
---> indirect_int2(parse_tree_int2, read_why_int2).
:- type indirect_int3_spec
---> indirect_int3(parse_tree_int3, read_why_int3).
:- type int_for_opt_spec
---> for_opt_int0(parse_tree_int0, read_why_int0)
; for_opt_int1(parse_tree_int1, read_why_int1)
; for_opt_int2(parse_tree_int2, read_why_int2).
:- type type_repn_spec
---> type_repn_spec_int1(parse_tree_int1).
% All these record recomp_avail_int_import as recompilation reason.
% (Since there is no recomp_avail_ancestor_import, yet).
:- type read_why_int0
---> rwi0_section
% Add the parse tree to the set of directly-read interfaces.
; rwi0_opt.
% Add the parse tree to the set of read-int-for-opt interfaces.
:- type read_why_int1
---> rwi1_int_import
% Add the parse tree to the set of directly-read interfaces.
%
% Record recomp_avail_int_import as recompilation reason.
; rwi1_int_use
% Add the parse tree to the set of directly-read interfaces.
%
% Record recomp_avail_int_use as recompilation reason.
; rwi1_imp_import
% Add the parse tree to the set of directly-read interfaces.
%
% Record recomp_avail_imp_import as recompilation reason.
; rwi1_imp_use
% Add the parse tree to the set of directly-read interfaces.
%
% Record recomp_avail_imp_use as recompilation reason.
; rwi1_int_use_imp_import
% Add the parse tree to the set of directly-read interfaces.
%
% Record recomp_avail_int_use_imp_import as recompilation reason.
; rwi1_opt
% Add the parse tree to the set of read-int-for-opt interfaces.
%
% Record recomp_avail_imp_use as recompilation reason.
; rwi1_type_repn.
% The only items that should be paid attention to from this
% .int file are the type_repn items. They don't need any
% section markers.
%
% Add the parse tree to the type-repn interfaces.
%
% Record recomp_avail_int_import as recompilation reason.
% XXX TYPE_REPN This is a lie, but it is the best we can do now,
% because smart recompilation "cannot handle the truth",
% due to not yet having been adapted to handle dependencies
% on interface files that are needed only for type representation
% information.
% All these record recomp_avail_imp_use as recompilation reason.
:- type read_why_int2
---> rwi2_int_use
% Add the parse tree to the set of indirectly-read interfaces.
; rwi2_imp_use
% Add the parse tree to the set of indirectly-read interfaces.
; rwi2_abstract
% Add the parse tree to the set of indirectly-read interfaces.
; rwi2_opt.
% Add the parse tree to the set of read-int-for-opt interfaces.
% XXX TYPE_REPN Do we need a rwi2_type_repn?
:- type read_why_int3
---> rwi3_direct_ancestor_import
% Add the parse tree to the set of directly-read interfaces.
%
% Record recomp_avail_int_import as recompilation reason.
% (Since there is no recomp_avail_ancestor_import, yet).
; rwi3_direct_int_import
% Add the parse tree to the set of directly-read interfaces.
%
% Record recomp_avail_int_import as recompilation reason.
; rwi3_direct_imp_import
% Add the parse tree to the set of directly-read interfaces.
%
% Record recomp_avail_imp_import as recompilation reason.
; rwi3_direct_ancestor_use
% Add the parse tree to the set of directly-read interfaces.
%
% Record recomp_avail_int_use as recompilation reason.
% (Since there is no recomp_avail_ancestor_use, yet).
; rwi3_direct_int_use
% Add the parse tree to the set of directly-read interfaces.
%
% Record recomp_avail_int_use as recompilation reason.
; rwi3_direct_imp_use
% Add the parse tree to the set of directly-read interfaces.
%
% Record recomp_avail_imp_use as recompilation reason.
; rwi3_direct_int_use_imp_import
% Add the parse tree to the set of directly-read interfaces.
%
% Record recomp_avail_int_use_imp_import as recompilation reason.
; rwi3_indirect_int_use
% Add the parse tree to the set of indirectly-read interfaces.
%
% Record recomp_avail_int_use as recompilation reason.
% (Since there is no recomp_avail_indirect_use_int, yet).
; rwi3_indirect_imp_use.
% Add the parse tree to the set of indirectly-read interfaces.
%
% Record recomp_avail_imp_use as recompilation reason.
% (Since there is no recomp_avail_indirect_use_imp, yet).
%---------------------------------------------------------------------------%
:- type module_section
---> ms_interface
; ms_implementation.
%---------------------------------------------------------------------------%
% An import_locn is used to describe the place where an item was
% imported from.
:- type import_locn
---> import_locn_implementation
% The item is from a module imported in the implementation.
; import_locn_interface
% The item is from a module imported in the interface.
; import_locn_import_by_ancestor
% The item is from a module imported by an ancestor.
% XXX Did the ancestor do the import in its interface, or not?
; import_locn_ancestor_int0_interface
; import_locn_ancestor_int0_implementation.
% The item is from the interface or implementation section
% of the .int0 file of an ancestor module.
%---------------------------------------------------------------------------%
%
% The main parts of parse trees are items. There are many kinds of items,
% and most of those kinds have their own item-kind-specific type that stores
% all the information the parse tree has about an item of that kind.
%
% The sequence number fields in the item-kind-specific types are intended to
% allow the recreation of the original item sequence after we have processed
% it into more complex data structures. Negative sequence numbers represent
% items that were not in the original read-in sequence, but which were added
% by the compiler. It is possible for two items to have the same sequence
% number if one original term (e.g. one that imports two or more modules)
% is split apart (e.g. into several items that each import only one module).
%
% When we create interface files, we print out selected items in the module.
% If the sequence of items printed changes, all the other modules depending
% on that interface file will be recompiled.
%
% A nontrivial fraction of changes to a module affect only the *order*
% of the items included in the interface, not their *content*. To minimize
% the amount of recompilation we have to do, we sort (most of the kinds of)
% items in the interface file, so that a change in the item order in the
% source file does not change the order of the items in the interface file.
% To make this sorting effective, we put the fields we prefer to use as
% the sort keys at the start of the item-kind-specific types. These are
% usually those that define the name of the entity, and if it makes sense
% to have more than item with that name, the main fields that distinguish
% items of the same name from each other.
%
% Did an item originate in user code or was it added by the compiler
% as part of a source-to-source transformation, e.g. the initialise
% declarations? If the latter, specify the information that the
% make_hlds pass may need to answer questions about the item.
%
:- type item_maybe_attrs
---> item_origin_user
; item_origin_compiler(item_compiler_attributes).
:- type item_compiler_attributes
---> item_compiler_attributes(
compiler_origin
).
:- type compiler_origin
---> compiler_origin_initialise
; compiler_origin_finalise
; compiler_origin_class_method(
cm_class_id :: class_id,
cm_method :: pred_pf_name_arity
)
; compiler_origin_solver_repn(
cosr_type_ctor :: type_ctor,
cosr_aux_pred_kind :: solver_type_pred_kind
)
; compiler_origin_mutable(
com_module_name :: module_name,
com_mutable_name :: string,
com_aux_pred_kind :: mutable_pred_kind
)
; compiler_origin_tabling(
cot_pred_spec :: pred_pf_name_arity,
cot_aux_pred_kind :: tabling_aux_pred_kind
).
:- type item
---> item_clause(item_clause_info)
; item_type_defn(item_type_defn_info)
; item_inst_defn(item_inst_defn_info)
; item_mode_defn(item_mode_defn_info)
; item_pred_decl(item_pred_decl_info)
; item_mode_decl(item_mode_decl_info)
; item_foreign_enum(item_foreign_enum_info)
; item_foreign_export_enum(item_foreign_export_enum_info)
; item_decl_pragma(item_decl_pragma_info)
; item_impl_pragma(item_impl_pragma_info)
; item_generated_pragma(item_generated_pragma_info)
; item_promise(item_promise_info)
; item_typeclass(item_typeclass_info)
; item_instance(item_instance_info)
; item_initialise(item_initialise_info)
; item_finalise(item_finalise_info)
; item_mutable(item_mutable_info)
; item_type_repn(item_type_repn_info).
:- type item_clause_info
---> item_clause_info(
cl_pred_or_func :: pred_or_func,
cl_predname :: sym_name,
cl_head_args :: list(prog_term),
cl_varset :: prog_varset,
cl_body :: maybe2(goal,
list(warning_spec)),
cl_context :: prog_context,
cl_seq_num :: item_seq_num
).
:- type item_type_defn_info == item_type_defn_info_general(type_defn).
:- type item_type_defn_info_abstract
== item_type_defn_info_general(type_details_abstract).
:- type item_type_defn_info_solver
== item_type_defn_info_general(type_details_solver).
:- type item_type_defn_info_eqv
== item_type_defn_info_general(type_details_eqv).
:- type item_type_defn_info_du
== item_type_defn_info_general(type_details_du).
:- type item_type_defn_info_sub
== item_type_defn_info_general(type_details_sub).
:- type item_type_defn_info_foreign
== item_type_defn_info_general(type_details_foreign_generic).
:- type item_type_defn_info_general(T)
---> item_type_defn_info(
% `:- type ...':
% a definition of a type, or a declaration of an abstract type.
td_ctor_name :: sym_name,
td_ctor_args :: list(type_param),
td_ctor_defn :: T,
td_tvarset :: tvarset,
td_context :: prog_context,
td_seq_num :: item_seq_num
).
:- type item_inst_defn_info
== item_inst_defn_info_general(maybe_abstract_inst_defn).
:- type item_inst_defn_info_abstract
== item_inst_defn_info_general(no_inst_defn).
:- type item_inst_defn_info_eqv
== item_inst_defn_info_general(inst_defn).
:- type item_inst_defn_info_general(T)
---> item_inst_defn_info(
% `:- inst ... = ...':
% a definition of an inst.
id_inst_name :: sym_name,
id_inst_args :: list(inst_var),
id_maybe_for_type :: maybe(type_ctor),
id_inst_defn :: T,
id_varset :: inst_varset,
id_context :: prog_context,
id_seq_num :: item_seq_num
).
:- type no_inst_defn
---> no_inst_defn.
:- type maybe_abstract_inst_defn
---> abstract_inst_defn
; nonabstract_inst_defn(inst_defn).
:- type item_mode_defn_info
== item_mode_defn_info_general(maybe_abstract_mode_defn).
:- type item_mode_defn_info_abstract
== item_mode_defn_info_general(no_mode_defn).
:- type item_mode_defn_info_eqv
== item_mode_defn_info_general(mode_defn).
:- type item_mode_defn_info_general(T)
---> item_mode_defn_info(
% `:- mode ... = ...':
% a definition of a mode.
md_mode_name :: sym_name,
md_mode_args :: list(inst_var),
md_mode_defn :: T,
md_varset :: inst_varset,
md_context :: prog_context,
md_seq_num :: item_seq_num
).
:- type no_mode_defn
---> no_mode_defn.
:- type maybe_abstract_mode_defn
---> abstract_mode_defn
; nonabstract_mode_defn(mode_defn).
:- type item_pred_decl_info
---> item_pred_decl_info(
% `:- pred ...' or `:- func ...':
% a predicate or function declaration.
% This specifies the type of the predicate or function,
% and it may optionally also specify the mode and determinism.
pf_name :: sym_name,
pf_p_or_f :: pred_or_func,
pf_arg_decls :: list(type_and_mode),
% The next two fields hold the `with_type` and `with_inst`
% annotations. This syntactic sugar is expanded out by
% equiv_type.m, which will then set these fields to `no'.
pf_maybe_with_type :: maybe(mer_type),
pf_maybe_with_inst :: maybe(mer_inst),
pf_maybe_detism :: maybe(determinism),
pf_maybe_attrs :: item_maybe_attrs,
pf_tvarset :: tvarset,
pf_instvarset :: inst_varset,
pf_existqvars :: existq_tvars,
pf_purity :: purity,
pf_constraints :: prog_constraints,
pf_context :: prog_context,
pf_seq_num :: item_seq_num
).
:- type item_mode_decl_info
---> item_mode_decl_info(
% `:- mode ...':
% a mode declaration for a predicate or function.
pfm_name :: sym_name,
pfm_p_or_f :: maybe(pred_or_func),
pfm_arg_modes :: list(mer_mode),
% The next field holds the `with_inst` annotation. This
% syntactic sugar is expanded by equiv_type.m, which will
% then set the field to `no'.
pfm_maybe_with_inst :: maybe(mer_inst),
pfm_maybe_detism :: maybe(determinism),
pfm_instvarset :: inst_varset,
pfm_context :: prog_context,
pfm_seq_num :: item_seq_num
).
:- type item_foreign_enum_info
---> item_foreign_enum_info(
fe_language :: foreign_language,
fe_type_ctor :: type_ctor,
fe_values :: one_or_more(
pair(sym_name, string)),
fe_context :: prog_context,
fe_seq_num :: item_seq_num
).
:- type foreign_enum_spec
---> foreign_enum_spec(
foreign_language,
type_ctor,
one_or_more(pair(sym_name, string))
).
:- type item_foreign_export_enum_info
---> item_foreign_export_enum_info(
fee_language :: foreign_language,
fee_type_ctor :: type_ctor,
fee_attributes :: export_enum_attributes,
fee_overrides :: assoc_list(sym_name,
string),
fee_context :: prog_context,
fee_seq_num :: item_seq_num
).
:- type item_decl_pragma_info == item_pragma_info(decl_pragma).
:- type item_impl_pragma_info == item_pragma_info(impl_pragma).
:- type item_generated_pragma_info == item_pragma_info(generated_pragma).
:- type item_fproc_export == item_pragma_info(pragma_info_foreign_proc_export).
:- type item_pragma_info(T)
---> item_pragma_info(
prag_type :: T,
prag_context :: prog_context,
prag_seq_num :: item_seq_num
).
:- type item_promise_info
---> item_promise_info(
prom_type :: promise_type,
prom_clause :: goal,
prom_varset :: prog_varset,
prom_univ_quant_vars :: list(prog_var),
prom_context :: prog_context,
prom_seq_num :: item_seq_num
).
:- type item_typeclass_info
---> item_typeclass_info(
tc_class_name :: class_name,
tc_class_params :: list(tvar),
tc_superclasses :: list(prog_constraint),
tc_fundeps :: list(prog_fundep),
tc_class_methods :: class_interface,
tc_varset :: tvarset,
tc_context :: prog_context,
tc_seq_num :: item_seq_num
).
:- type item_instance_info
---> item_instance_info(
% The original types field preserves the types in the instance
% declaration as written by the programmer. The types field
% is subject to the expansion of equivalence types.
ci_class_name :: class_name,
ci_types :: list(mer_type),
ci_original_types :: list(mer_type),
ci_deriving_class :: list(prog_constraint),
ci_method_instances :: instance_body,
ci_varset :: tvarset,
ci_module_containing_instance :: module_name,
ci_context :: prog_context,
ci_seq_num :: item_seq_num
).
:- type item_initialise_info
---> item_initialise_info(
% :- initialise pred_name.
init_name :: sym_name,
init_arity :: user_arity,
init_maybe_attrs :: item_maybe_attrs,
init_context :: prog_context,
init_seq_num :: item_seq_num
).
:- type item_finalise_info
---> item_finalise_info(
% :- finalise pred_name.
final_name :: sym_name,
final_arity :: user_arity,
final_maybe_attrs :: item_maybe_attrs,
final_context :: prog_context,
final_seq_num :: item_seq_num
).
:- type item_mutable_info
---> item_mutable_info(
% :- mutable(var_name, type, inst, value, attrs).
mut_name :: string,
% The mut_type and mut_inst fields are subject to expansion
% in equiv_type.m; the mut_orig_type and mut_orig_inst fields
% are not. The latter are used to improve error reporting.
mut_orig_type :: mer_type,
mut_type :: mer_type,
mut_orig_inst :: mer_inst,
mut_inst :: mer_inst,
mut_init_value :: prog_term,
mut_init_value_varset :: prog_varset,
mut_attrs :: mutable_var_attributes,
mut_context :: prog_context,
mut_seq_num :: item_seq_num
).
:- type item_type_repn_info_eqv
== item_type_repn_info_general(mer_type).
:- type item_type_repn_info_subtype
== item_type_repn_info_general(type_ctor).
:- type item_type_repn_info
== item_type_repn_info_general(type_ctor_repn_info).
:- type item_type_repn_info_general(T)
---> item_type_repn_info(
% `:- type_representation ...':
% An item added by the compiler to a .int3 file
% to tell readers of that file the information they need
% to correctly reconstruct the representation of the given
% type constructor, even when that information is supposed
% to be invisible to them semantically.
% There should be at most one such item for any type_ctor
% in the .int3 file of its defining module.
% The sym_name should be fully qualified.
tr_ctor :: sym_name,
tr_ctor_arg_tvars :: list(tvar),
tr_ctor_repn_info :: T,
tr_tvarset :: tvarset,
tr_context :: prog_context,
tr_seq_num :: item_seq_num
).
%---------------------------------------------------------------------------%
%
% Declarations of relationships between modules.
%
:- type item_include
---> item_include(
% The representation of an `:- include_module' declaration
% is a list of one or more item_includes, each of which
% declares the named module to be a submodule of the
% current module,
%
% If this item_include occurs in module x.y, then
% the module_name here is guaranteed to have the form x.y.z.
% In other words, the included module is guaranteed to be
% an immediate descendant of the including module.
% Any attempt to include a non-descendant module or a
% non-immediate descendant module will be caught and
% diagnosed by the parser.
incl_module :: module_name,
% The context and item sequence number of the declaration.
incl_context :: prog_context,
incl_seq_num :: item_seq_num
).
:- type import_or_use
---> import_decl
; use_decl.
% The representation of an `:- import_module' or an `:- use_module'
% declaration is a list of one or more item_avails, each of which
% makes available to the current module the entities in the interface
% of the module named in the declaration.
%
% With avail_use, references to these entities must be module qualified;
% with avail_import, they don't have to be.
:- type item_avail
---> avail_import(avail_import_info)
; avail_use(avail_use_info).
% The structures of avail_import_info and avail_use_info are the same,
% with the first argument being the name of the module that is the subject
% of the import_module or use_module declaration, and the second and third
% being the context and item sequence number of the declaration.
%
% The two types are separate to allow parse_tree_opts to contain only
% values of a type that makes it clear that they contain information
% ONLY about use_module declarations, not import_module declarations.
:- type avail_import_info
---> avail_import_info(
aii_module_name :: module_name,
aii_context :: prog_context,
aii_seq_num :: item_seq_num
).
:- type avail_use_info
---> avail_use_info(
aui_module_name :: module_name,
aui_context :: prog_context,
aui_seq_num :: item_seq_num
).
:- type item_fim
---> item_fim(
% A `:- pragma foreign_import_module(Lang, ModuleName)'
% declaration, which tells the compiler to include the
% header file we automatically generate for Module
% in the target language Lang when we compile this module
% to that language, and, if this occurs in the interface,
% when we compile the modules importing this one
% to that same target language.
%
% Equivalent to
% `:- pragma foreign_decl(Lang, "#include <module>.h")',
% except that the name of the header file is not hard-coded,
% and mmake can use the dependency information.
%
% Throughout most parts of the compiler, we use "FIM"
% as shorthand for foreign_import_module.
fim_lang :: foreign_language,
fim_module_name :: module_name,
fim_context :: prog_context,
fim_seq_num :: item_seq_num
).
%---------------------------------------------------------------------------%
%
% Type classes.
%
% The class_decl type represents any declaration that occurs
% in the body of a type class definition.
%
% Such declarations may either declare class methods, or they may declare
% the modes of class methods.
%
:- type class_decl
---> class_decl_pred_or_func(class_pred_or_func_info)
; class_decl_mode(class_mode_info).
:- type class_pred_or_func_info
---> class_pred_or_func_info(
% This is a `pred ...' or `func ...' declaration in a
% type class body, which declares a predicate or function
% method. Such declarations specify the types of the
% arguments, and may optionally also specify argument modes
% and the determinism.
% The name of the predicate or function.
sym_name,
pred_or_func,
% The arguments' types, and maybe modes.
list(type_and_mode),
% Any `with_type` and/or `with_inst` annotation.
maybe(mer_type),
maybe(mer_inst),
% The determinism declaration, if any.
maybe(determinism),
% The varsets of the type and inst variables.
tvarset,
inst_varset,
% The existentially quantified type variables, if any.
existq_tvars,
% Any purity annotation.
purity,
% The typeclass constraints on the declaration.
prog_constraints,
prog_context
).
:- type class_mode_info
---> class_mode_info(
% This is a `mode ...' declaration in a type class body.
% Such a declaration declares a mode for one of the methods
% of the type class.
% The name of the predicate or function.
sym_name,
% Whether the method is a predicate or a function.
% For declarations using `with_inst`, we don't know
% which it is until we have expanded the inst.
maybe(pred_or_func),
% The arguments' modes.
list(mer_mode),
% Any `with_inst` annotation.
maybe(mer_inst),
% Any determinism declaration.
maybe(determinism),
% The varset of the inst variables.
inst_varset,
prog_context
).
%---------------------------------------------------------------------------%
%
% Mutable variables.
%
% Indicates if updates to the mutable are trailed or untrailed.
%
:- type mutable_trailed
---> mutable_untrailed
; mutable_trailed.
% Indicates if a mutable is attached to the I/O state or not.
%
:- type mutable_attach_to_io_state
---> mutable_dont_attach_to_io_state
; mutable_attach_to_io_state.
% Indicates if a mutable is constant or not.
%
:- type mutable_constant
---> mutable_not_constant
; mutable_constant.
% Indicates if a mutable is thread-local or not.
%
:- type mutable_thread_local
---> mutable_not_thread_local
; mutable_thread_local.
% Attributes for mutable variables.
%
:- type mutable_var_attributes
---> mutable_var_attributes(
mutable_foreign_names :: map(foreign_language, string),
mutable_constant :: mutable_maybe_constant
).
:- type mutable_maybe_constant
---> mutable_is_constant
% implies mutable_dont_attach_to_io_state
% implies mutable_untrailed
% implies mutable_not_thread_local
; mutable_is_not_constant(
mutable_attach_to_io_state,
mutable_maybe_thread_local
).
:- type mutable_maybe_thread_local
---> mutable_is_not_thread_local(
mutable_trailed
)
; mutable_is_thread_local.
% implies mutable_untrailed
:- func mutable_var_thread_local(mutable_maybe_constant)
= mutable_thread_local.
:- func mutable_thread_local_trailed(mutable_maybe_thread_local)
= mutable_trailed.
%---------------------------------------------------------------------------%
%
% The representation of a checked-to-be-consistent set of type and
% foreign enum definitions for every type constructor defined in a module.
%
:- type type_ctor_checked_map == map(type_ctor, type_ctor_checked_defn).
% A type is either a solver type, or not.
:- type type_ctor_checked_defn
---> checked_defn_solver(solver_type_defn, src_defns_solver)
; checked_defn_std(std_type_defn, src_defns_std).
%---------------------%
% Replace this one general type with one type for each function symbol
% in solver_type_defn.
:- type src_defns_solver
---> src_defns_solver(
% The item_type_defn_info (if any) in the interface section.
maybe(item_type_defn_info),
% The item_type_defn_info (if any) in the impl section.
maybe(item_type_defn_info)
).
% Replace this one general type with one type for each function symbol
% in std_type_defn.
:- type src_defns_std
---> src_defns_std(
% The item_type_defn_infos in the interface section.
list(item_type_defn_info),
% The item_type_defn_infos and item_foreign_enum_infos
% in the implementation section.
list(item_type_defn_info),
list(item_foreign_enum_info)
).
%---------------------%
:- type solver_type_defn
---> solver_type_abstract(
abstract_solver_type_status,
% The abstract definition. It may be in either section;
% the status specifies the section.
item_type_defn_info_abstract
)
; solver_type_full(
% The abstract definition in the interface section,
% if one exists.
maybe(item_type_defn_info_abstract),
% The full solver type definition, which must be in the
% implementation section.
item_type_defn_info_solver
).
:- type abstract_solver_type_status
---> abstract_solver_type_exported
% The type name is exported. The abstract definition
% is in the interface section.
; abstract_solver_type_private.
% The type name is not exported. The abstract definition
% is in the implementation section.
%---------------------%
:- type std_type_defn
---> std_mer_type_eqv(
std_eqv_type_status,
% The equivalence type definition.
item_type_defn_info_eqv
)
; std_mer_type_subtype(
std_subtype_status,
% The subtype definition.
item_type_defn_info_sub
)
; std_mer_type_du_all_plain_constants(
std_du_type_status,
% The discriminated union type definition which represents
% either a direct dummy type or an enum.
item_type_defn_info_du,
% The first functor name in the type, and any later functor
% names. If there are no later functor names, then the type
% is a direct dummy type, and must satisfy the requirements
% of non_sub_du_type_is_dummy; if there are, then the type
% is an enum type, and must satisfy the requirements of
% non_sub_du_type_is_enum. (Function symbols that do not meet
% the relevant requirements may be constants, but we
% don't consider them *plain* constants.)
string,
list(string),
% For each of our target foreign languages, this field
% specifies whether we have either a foreign language
% definition for this type, or a foreign enum definition.
%
% While the Mercury representation uses small integers
% allocated consecutively from 0 to represent function symbols,
% this is not true even for foreign enum definitions,
% much less foreign type definitions.
c_j_cs_maybe_defn_or_enum
)
; std_mer_type_du_not_all_plain_constants(
std_du_type_status,
% The discriminated union type definition which represents
% a type *other* than a direct dummy type or an enum.
item_type_defn_info_du,
% For each of our target foreign languages, this field
% specifies whether we have a foreign language type definition
% for this type.
c_j_cs_maybe_defn
)
; std_mer_type_abstract(
std_abs_type_status,
% The abstract declaration of the type (not a subtype).
item_type_defn_info_abstract,
% For each of our target foreign languages, this field
% specifies whether we have a foreign language type definition
% for this type.
c_j_cs_maybe_defn
).
:- type maybe_only_constants
---> not_only_plain_constants
; only_plain_constants(
% The names of the constants, in the order of declaration.
opc_head_name :: string,
opc_tail_names :: list(string)
).
:- type std_eqv_type_status
---> std_eqv_type_mer_exported
% The Mercury definition (i.e. the equivalence) is exported.
; std_eqv_type_abstract_exported
% Only the type name is exported. The Mercury definition
% is private.
; std_eqv_type_all_private.
% Everything about the type is private.
:- type std_du_type_status
---> std_du_type_mer_ft_exported
% Both the Mercury and any foreign type definitions are exported.
% Any foreign enum definitions are private, as they have to be.
% This status is not applicable to equivalence types or subtypes,
% since they may not have foreign type definitions.
; std_du_type_mer_exported
% The Mercury definition is exported. Any foreign type definitions
% and/or foreign enum definitions are private.
; std_du_type_abstract_exported
% Only the type name is exported. The Mercury definition and
% any foreign type definitions and/or foreign enum definitions
% are private.
; std_du_type_all_private.
% Everything about the type is private.
% A version of std_du_type_status for subtypes, which may not have
% any foreign type definitions, and for which therefore the question of
% whether any foreign type definitions are exported is moot.
:- type std_subtype_status
---> std_sub_type_mer_exported
; std_sub_type_abstract_exported
; std_sub_type_all_private.
:- type std_abs_type_status
---> std_abs_type_ft_exported
% The type has foreign type definitions that are exported.
% Any foreign enum definitions are private, as they have to be.
; std_abs_type_abstract_exported
% Only the type name is exported. Any foreign type definitions
% and/or foreign enum definitions are private.
; std_abs_type_all_private.
% Everything about the type is private.
%---------------------%
:- type c_j_cs_maybe_defn_or_enum ==
c_java_csharp(maybe(foreign_type_or_enum)).
:- type foreign_type_or_enum
---> foreign_type_or_enum_type(item_type_defn_info_foreign)
; foreign_type_or_enum_enum(checked_foreign_enum).
% Part of checking a foreign enum definition is checking whether
% the correspondence it describes between the Mercury functors
% of the type on the one hand and their foreign language counterparts
% on the other hand is a bijection. If it is, then the second argument
% of the checked_foreign_enum we construct gives the foreign language
% counterpart of each Mercury function symbol in the type in the order
% in which the Mercury function symbols are defined.
%
% For example, given
%
% :- type t ---> m1 ; m2 ; m3.
%
% and a foreign enum definition that gives the correspondence correctly
% but in a different order, such as
%
% :- pragma foreign_enum("C", t/0, [m2 - "f2", m3 - "f3", m1 - "f1"]).
%
% the second argument will contain the (nonempty) list "f1", "f2", "f3".
%
% On the other hand, if the mapping in the foreign enum definition is
% *not* a bijection, then we will not generate a checked_foreign_enum
% structure for it.
%
:- type checked_foreign_enum
---> checked_foreign_enum(item_foreign_enum_info, one_or_more(string)).
%---------------------------------------------------------------------------%
%
% The representation of a checked-to-be-consistent set of inst definitions
% for every inst constructor defined in a module.
%
:- type inst_ctor_checked_map == map(inst_ctor, inst_ctor_checked_defn).
:- type inst_ctor_checked_defn
---> checked_defn_inst(std_inst_defn, src_defns_inst).
:- type std_inst_defn
---> std_inst_defn(std_inst_status, item_inst_defn_info).
:- type std_inst_status
---> std_inst_exported
% The inst definition is exported.
; std_inst_abstract_exported
% Only the inst name is exported. Its definition is private.
; std_inst_all_private.
% Everything about the inst is private.
:- type src_defns_inst
---> src_defns_inst(
% The inst definition (if any) in the interface.
maybe(item_inst_defn_info),
% The inst definition (if any) in the implementation.
maybe(item_inst_defn_info)
).
%---------------------------------------------------------------------------%
%
% The representation of a checked-to-be-consistent set of mode definitions
% for every mode constructor defined in a module.
%
:- type mode_ctor_checked_map == map(mode_ctor, mode_ctor_checked_defn).
:- type mode_ctor_checked_defn
---> checked_defn_mode(std_mode_defn, src_defns_mode).
:- type std_mode_defn
---> std_mode_defn(std_mode_status, item_mode_defn_info).
:- type std_mode_status
---> std_mode_exported
% The mode definition is exported.
; std_mode_abstract_exported
% Only the mode name is exported. Its definition is private.
; std_mode_all_private.
% Everything about the mode is private.
:- type src_defns_mode
---> src_defns_mode(
% The mode definition (if any) in the interface.
maybe(item_mode_defn_info),
% The mode definition (if any) in the implementation.
maybe(item_mode_defn_info)
).
%---------------------------------------------------------------------------%
%
% Information about the representations of types defined in other modules.
%
% This type and type_ctor_checked_defn are closely related.
% The principal differences are the following.
%
% - type_ctor_checked_defn deals with solver types. Since solver types
% have no representation information themselves (they are represented
% by values of another type), this type does not deal with them.
%
% - One of the purposes of type_ctor_checked_defn is to decide
% what items to include in interface files, for use by code using
% the compiler's ancient approach to deciding type representation,
% where each compiler invocation that generated code decided for itself
% how every type it had access to was represented, including the types
% imported from other modules. This means that it needs to contain
% either whole items (of particular kinds), or information from which
% whole items can be reconstructed.
%
% - The above consideration also requires a type_ctor_checked_defn
% to specify the status of the type. On the other hand, values of
% this type have no use for status information. Status information
% is used only for checking whether an access to a type should be
% allowed or not; the only use of values of this type is to help
% compute type representations.
%
% - Only this type needs to contain representation information.
% A value of the type_ctor_checked_defn type needs to contain *part*
% of the information from which this representation information is
% computed for its type, but not *all* of it; some of that information
% comes from information about the representation of *other* types.
%
% One sort-of difference is while both contain information that has been
% checked by a compiler invocation, values of this type that have been
% read in from an interface file, while checked by another compiler
% invocation before being written out, may be corrupted in the filesystem.
% However, while this danger is always present, we need not take any
% special steps to guard against it, precisely because no perfect defense
% is possible.
%
% XXX TYPE_REPN Consider whether we can split this type into two,
% one for the tcrepns that can occur in .int3 files, and one for the
% tcrepns that can occur in .int/.int2 files.
%
:- type type_ctor_repn_info
---> tcrepn_is_word_aligned_ptr
; tcrepn_is_eqv_to(mer_type)
; tcrepn_is_subtype_of(type_ctor)
; tcrepn_du(du_repn)
; tcrepn_foreign(c_j_cs_repn).
% A type that has a discriminated union definition in Mercury
% may also have a definition in each of our foreign languages,
% If it is an direct_dummy or enum type, that definition may be
% either a foreign type definition or a foreign enum definition;
% otherwise, it can only be a foreign type definition.
:- type du_repn
---> dur_direct_dummy(direct_dummy_repn)
; dur_enum(enum_repn)
; dur_notag(notag_repn)
; dur_gen_only_functor(gen_du_only_functor_repn)
; dur_gen_more_functors(gen_du_more_functors_repn).
% When targeting C, many argument packing decisions depend on
% three properties of the target platform, i.e. on the combination
% of the target hardware and the target grade:
%
% - whether the target is 64 or 32 bit;
% - whether the grade is an spf (single-precision float) grade; and
% - whether the grade allows the direct arg optimization.
%
% These have eight combinations, but the spf grade component has
% no effect on argument packing on 64 bit targets (a float is one word
% either way), so only six are meaningful.
%
% If the decision represented by the T parameter happens to be the same
% on all six platforms, that decision can be represented by c_repns_same.
%
% If they are different on 64 vs 32 bit platforms, but are consistent
% for each word size, then they can be represented by c_repns_64_32.
%
% If neither is the case, we can record all six decisions using
% c_repns_all.
%
% XXX We should look for other partitions of the set of six platforms
% which often have identical decision results; one could be da vs noda.
%
% The name of this type is c_repns because argument packing applies
% only to the low level data representation, which is applicable only
% when targeting C.
:- type c_repns(T)
---> c_repns_same(
c_repn_same :: T
)
; c_repns_64_32(
c_repn_all_64 :: T,
c_repn_all_32 :: T
)
; c_repns_all(
c_repn_64_nospf_noda :: T,
c_repn_64_nospf_da :: T,
% c_repn_64_spf_noda :: T, % not needed; see above
% c_repn_64_spf_da :: T, % not needed; see above
c_repn_32_nospf_noda :: T,
c_repn_32_nospf_da :: T,
c_repn_32_spf_noda :: T,
c_repn_32_spf_da :: T
).
%---------------------%
:- type direct_dummy_repn
---> direct_dummy_repn(
% The type is a direct dummy type that satisfies the
% requirements of du_type_is_dummy.
% The name of the one functor in the type, which must be
% arity 0. Its representation will be dummy_tag.
dummy_functor_name :: string,
% Any foreign type or foreign enum definitions for the type.
dummy_foreign :: c_j_cs_enum_repn
).
%---------------------%
:- type enum_repn
---> enum_repn(
% The type is an enum type that satisfies the requirements
% of non_sub_du_type_is_enum.
% The list of the functor names (all arity 0). We store
% the first two separately to enforce the structural invariant
% that an enum must have at least two functors.
%
% The representation of functor #N in Mercury will be
% int_tag(int_tag_int(N)), with counting starting at 0.
%
% We do not care about the 32 vs 64 bit distinction here,
% because the definition of an enum type with more than 2^32
% function symbols will cause a compiler to run out of memory
% for a *very* long time to come.
enum_functor1 :: string,
enum_functor2 :: string,
enum_functors3plus :: list(string),
% Any foreign type or foreign enum definitions for the type.
enum_foreign :: c_j_cs_enum_repn
).
%---------------------%
:- type notag_repn
---> notag_repn(
% The name of the one functor in the type, which must be
% arity 1. Its representation will be no_tag.
% The representation of the argument be *recorded*
% as a full word at offset 0, but this should never be
% looked up, since the argument will actually be stored
% wherever the whole term is stored.
notag_functor_name :: string,
% The type of the one functor's one argument.
% We record this because without this information,
% we cannot recognize that a notag type whose argument size
% is less than one word can itself be stored in less than
% one word.
notag_functor_arg_type :: mer_type,
% The foreign language definitions for this type, if any.
notag_foreign :: c_j_cs_repn
).
%---------------------%
:- type gen_du_only_functor_repn
---> gen_du_only_functor_repn(
% The name of the data constructor. The arity is given by
% the length of list of argument types. The lists of argument
% representations in all of the nonconstant_repns inside
% the c_repns must also ave this length.
only_functor :: string,
% The types of the constructor's arguments, after
% the expansion of both equivalence types and notag types.
only_deref_arg_types :: list(mer_type),
% The representation of this functor for each possible
% target platform with the low level data representation.
% The nonconstant_repn cannot be ncr_direct_arg.
% XXX TYPE_REPN could we encode that invariant in the type?
only_arg_repns :: c_repns(only_nonconstant_repn),
% The foreign language definitions for this type, if any.
only_foreign :: c_j_cs_repn
).
:- type gen_du_more_functors_repn
---> gen_du_more_functors_repn(
% The first, second and any later functors in the type,
% in declaration order, i.e. ordered on the functors'
% original ordinal numbers.
more_functor1 :: gen_du_functor_repn,
more_functor2 :: gen_du_functor_repn,
more_functors3plus :: list(gen_du_functor_repn),
% The foreign language definitions for this type, if any.
more_foreign :: c_j_cs_repn
).
%---------------------%
:- type gen_du_functor_repn
---> gen_du_constant_functor_repn(
% The name of the data constructor. The arity is zero.
gducf_functor :: string,
% The representation of this functor for each possible
% target platform with the low level data representation.
gducf_functor_repn :: c_repns(constant_repn)
)
; gen_du_nonconstant_functor_repn(
% The name of the data constructor. The arity is given by
% the length of list of argument types. The lists of argument
% representations in all of the nonconstant_repns inside
% the c_repns must also ave this length.
gduncf_functor :: string,
% The types of the constructor's arguments, after
% the expansion of both equivalence types and notag types.
%
% Logically, the type of each argument belongs with
% the representation of that argument, but we have to store
% up to six versions of the representation, and we don't want
% a duplicate copy of the type next to each version.
gduncf_deref_arg_types :: list(mer_type),
% The representation of this functor for each possible
% target platform with the low level data representation.
gduncf_functor_repn :: c_repns(more_nonconstant_repn)
).
:- type constant_repn
---> constant_repn(
% The ptag is 0. The next two fields specify the value
% and the size of the local secondary tag.
cr_sectag :: uint,
cr_sectag_size :: lsectag_word_or_size
).
:- type only_nonconstant_repn
---> oncr_local_cell(only_nonconstant_local_cell_repn)
; oncr_remote_cell(only_nonconstant_remote_cell_repn).
:- type more_nonconstant_repn
---> mncr_local_cell(more_nonconstant_local_cell_repn)
; mncr_remote_cell(more_nonconstant_remote_cell_repn)
; mncr_direct_arg(ptag).
:- type only_nonconstant_local_cell_repn
---> only_nonconstant_local_cell_repn(
% The ptag and local sectag are both implicitly 0u.
onclcr_arg_repns :: one_or_more(local_arg_repn)
).
:- type more_nonconstant_local_cell_repn
---> more_nonconstant_local_cell_repn(
% The ptag is implicitly 0u.
mnclcr_sectag :: cell_local_sectag,
mnclcr_arg_repns :: one_or_more(local_arg_repn)
).
:- type only_nonconstant_remote_cell_repn
---> only_nonconstant_remote_cell_repn(
% The ptag is both implicitly 0u, and there is
% no remote sectag.
ncrcr_arg_repns :: one_or_more(remote_arg_repn)
).
:- type more_nonconstant_remote_cell_repn
---> more_nonconstant_remote_cell_repn(
ncrcr_ptag :: ptag,
ncrcr_sectag :: cell_remote_sectag,
ncrcr_arg_repns :: one_or_more(remote_arg_repn)
).
:- type cell_local_sectag
---> cell_local_sectag(
clss_sectag :: uint,
clss_sectag_size :: uint8
).
:- type cell_remote_sectag
---> cell_remote_no_sectag
; cell_remote_sectag(
crss_sectag :: uint,
crss_sectag_size :: rsectag_word_or_size
).
:- type lsectag_word_or_size
---> lsectag_rest_of_word(uint8)
; lsectag_part_of_word(uint8).
:- type rsectag_word_or_size
---> rsectag_full_word
; rsectag_part_of_word(uint8).
:- type local_arg_repn
---> local_partial(
lp_shift :: uint,
lp_fill :: fill_kind_size
)
; local_none.
:- type remote_arg_repn
---> remote_full(
rf_arg_only_offset :: arg_only_offset,
rf_cell_offset :: cell_offset
)
; remote_double(
rd_arg_only_offset :: arg_only_offset,
rd_cell_offset :: cell_offset,
rd_kind :: double_word_kind
)
; remote_partial_first(
rpf_arg_only_offset :: arg_only_offset,
rpf_cell_offset :: cell_offset,
rpf_shift :: uint8,
rpf_fill :: fill_kind_size
)
; remote_partial_shifted(
rps_arg_only_offset :: arg_only_offset,
rps_cell_offset :: cell_offset,
rps_shift :: uint8,
rps_fill :: fill_kind_size
)
; remote_none_shifted(
rns_arg_only_offset :: arg_only_offset,
rns_cell_offset :: cell_offset
)
; remote_none_nowhere.
:- type fill_kind_size
---> fk_enum(uint) % XXX TYPE_REPN should be uint8
; fk_int8
; fk_int16
; fk_int32
; fk_uint8
; fk_uint16
; fk_uint32
; fk_char21.
% XXX TYPE_REPN should return uint8
:- func fill_kind_size_num_bits(fill_kind_size) = uint.
%---------------------%
:- type foreign_type_lang_repn
---> foreign_type_lang_repn(
ftlr_lang :: foreign_language,
ftlr_foreign_type :: foreign_type_repn
).
:- type foreign_type_repn
---> foreign_type_repn(
% The name of the foreign type that represents values
% of this Mercury type.
ftr_foreign_type :: string,
% The assertions about this foreign type.
ftr_assertions :: foreign_type_assertions
).
:- type enum_foreign_repn
---> enum_foreign_type(foreign_type_repn)
; enum_foreign_enum(one_or_more(string)).
%---------------------------------------------------------------------------%
%
% Pragmas.
%
:- type decl_pragma
---> decl_pragma_obsolete_pred(pragma_info_obsolete_pred)
; decl_pragma_obsolete_proc(pragma_info_obsolete_proc)
; decl_pragma_format_call(pragma_info_format_call)
; decl_pragma_type_spec(pragma_info_type_spec)
; decl_pragma_oisu(pragma_info_oisu)
; decl_pragma_terminates(pred_pfu_name_arity)
; decl_pragma_does_not_terminate(pred_pfu_name_arity)
; decl_pragma_check_termination(pred_pfu_name_arity)
; decl_pragma_termination_info(pragma_info_termination_info)
; decl_pragma_termination2_info(pragma_info_termination2_info)
; decl_pragma_structure_sharing(pragma_info_structure_sharing)
; decl_pragma_structure_reuse(pragma_info_structure_reuse).
:- type impl_pragma
---> impl_pragma_foreign_decl(pragma_info_foreign_decl)
; impl_pragma_foreign_code(pragma_info_foreign_code)
; impl_pragma_foreign_proc(pragma_info_foreign_proc)
; impl_pragma_foreign_proc_export(pragma_info_foreign_proc_export)
; impl_pragma_external_proc(pragma_info_external_proc)
; impl_pragma_fact_table(pragma_info_fact_table)
; impl_pragma_tabled(pragma_info_tabled)
; impl_pragma_inline(pred_pfu_name_arity)
; impl_pragma_no_inline(pred_pfu_name_arity)
; impl_pragma_consider_used(pred_pfu_name_arity)
; impl_pragma_mode_check_clauses(pred_pfu_name_arity)
; impl_pragma_no_detism_warning(pred_pfu_name_arity)
; impl_pragma_require_tail_rec(pragma_info_require_tail_rec)
; impl_pragma_promise_pure(pred_pfu_name_arity)
; impl_pragma_promise_semipure(pred_pfu_name_arity)
; impl_pragma_promise_eqv_clauses(pred_pfu_name_arity)
; impl_pragma_require_feature_set(pragma_info_require_feature_set).
:- type generated_pragma
---> gen_pragma_unused_args(pragma_info_unused_args)
; gen_pragma_exceptions(pragma_info_exceptions)
; gen_pragma_trailing_info(pragma_info_trailing_info)
; gen_pragma_mm_tabling_info(pragma_info_mm_tabling_info).
:- type pred_marker_pragma_kind
---> pmpk_inline
; pmpk_noinline
; pmpk_promise_pure
; pmpk_promise_semipure
; pmpk_promise_eqv_clauses
; pmpk_terminates
; pmpk_does_not_terminate
; pmpk_mode_check_clauses.
:- type pragma_info_pred_marker
---> pragma_info_pred_marker(
pred_pf_name_arity,
pred_marker_pragma_kind
).
:- type item_pred_marker == item_pragma_info(pragma_info_pred_marker).
:- type item_type_spec == item_pragma_info(pragma_info_type_spec).
:- type item_termination == item_pragma_info(pragma_info_termination_info).
:- type item_termination2 == item_pragma_info(pragma_info_termination2_info).
:- type item_struct_sharing == item_pragma_info(pragma_info_structure_sharing).
:- type item_struct_reuse == item_pragma_info(pragma_info_structure_reuse).
:- type item_foreign_proc == item_pragma_info(pragma_info_foreign_proc).
:- type item_tabled == item_pragma_info(pragma_info_tabled).
:- type item_unused_args == item_pragma_info(pragma_info_unused_args).
:- type item_exceptions == item_pragma_info(pragma_info_exceptions).
:- type item_trailing == item_pragma_info(pragma_info_trailing_info).
:- type item_mm_tabling == item_pragma_info(pragma_info_mm_tabling_info).
% Foreign language interfacing pragmas.
:- type pragma_info_foreign_decl
---> pragma_info_foreign_decl(
% A foreign language declaration, such as C header code.
decl_lang :: foreign_language,
decl_is_local :: foreign_decl_is_local,
decl_decl :: foreign_literal_or_include
).
:- type pragma_info_foreign_code
---> pragma_info_foreign_code(
code_lang :: foreign_language,
code_code :: foreign_literal_or_include
).
:- type pragma_info_foreign_proc
---> pragma_info_foreign_proc(
% Set of foreign proc attributes, such as:
% what language this code is in
% whether or not the code may call Mercury,
% whether or not the code is thread-safe
% PredName, Predicate or Function, Vars/Mode,
% VarNames, Foreign Code Implementation Info
proc_attrs :: pragma_foreign_proc_attributes,
proc_name :: sym_name,
proc_p_or_f :: pred_or_func,
proc_vars :: list(pragma_var),
proc_varset :: prog_varset,
proc_instvarset :: inst_varset,
proc_impl :: pragma_foreign_proc_impl
).
:- type pragma_info_foreign_proc_export
---> pragma_info_foreign_proc_export(
exp_maybe_attrs :: item_maybe_attrs,
exp_language :: foreign_language,
% Predname, Predicate/function, Modes, foreign function name.
exp_pred_id :: proc_pf_name_modes,
exp_foreign_name :: string,
% Specified the names of any variables in the modes above.
% Used for generating error messages about foreign_export
% pragmas for undeclared modes.
exp_varaset :: prog_varset
).
:- type pragma_info_external_proc
---> pragma_info_external_proc(
% The specified procedure(s) is/are implemented outside
% of Mercury code, for the named backend if there is one,
% or if there isn't a named backend, then for all backends.
external_name :: pred_pf_name_arity,
external_maybe_backend :: maybe(backend)
).
% Optimization pragmas.
:- type pragma_info_type_spec
---> pragma_info_type_spec(
tspec_pfumm :: pred_func_or_unknown_maybe_modes,
% The existing predicate name.
tspec_pred_name :: sym_name,
% The name of the module from whose (source or interface) file
% we read the type_spec pragma. This will always name
% the module that contain the pragma, because we never put
% a type_spec pragma into any interface file other than
% an interface file of the module containing the pragma.
tspec_module_name :: module_name,
% The type substitution (using the variable names
% from the pred declaration).
tspec_tsubst :: type_subst,
% The varset of the term containing the pragma, coerced
% to being a tvarset (since no part of the pragma except
% the type substitution may contain variables).
%
% All variables in this tvarset have to have explicit names.
% If the original pragma contains anonymous variables, the
% code constructing this pragma_info_type_spec will give
% those variable names.
%
% The reason for this requirement is that the process
% of writing out an anonymous variable and reading it back in
% will produce a non-anonymous variable. Since the names
% (if any) of the variables in tspec_tsubst are an input
% to the code that constructs the name of the type-specialized
% predicate, we would get a discrepancy between the predicate
% name constructed by compiler invocations that know the
% variable as unnamed (this will be the invocation that
% compiles the module containing the type_spec pragma,
% which constructs the code of the type specialized predicate),
% and compiler invocations that know that variable as named
% (this will be all the invocations that read the original
% module's .int file, which will be constructing many of
% the *calls* to the type specialized predicate). The result
% will be calls to the type specialized predicate that refer
% to it by the wrong name, leading to link errors.
%
% By giving all anonymous variables in the type_spec pragma
% in the original source file as soon as we have parsed it,
% and then always using the resulting names, we avoid this
% problem.
tspec_tvarset :: tvarset,
% The equivalence types used.
tspec_items :: set(recomp_item_id)
).
:- type pragma_info_unused_args
---> pragma_info_unused_args(
% This pragma Should only appear in .opt files.
unused_proc_id :: proc_pf_name_arity_mn,
% The argument positions of the unused arguments.
% Used for intermodule unused argument removal.
unused_args :: list(int)
).
:- type pragma_info_exceptions
---> pragma_info_exceptions(
% This pragma should only appear in `.opt' and
% `.trans_opt' files.
exceptions_proc_id :: proc_pf_name_arity_mn,
exceptions_status :: exception_status
).
:- type pragma_info_trailing_info
---> pragma_info_trailing_info(
% This pragma should only appear in `.trans_opt' files.
trailing_info_proc_id :: proc_pf_name_arity_mn,
trailing_info_status :: trailing_status
).
:- type pragma_info_mm_tabling_info
---> pragma_info_mm_tabling_info(
% This pragma should only appear in `.opt' and
% `.trans_opt' files.
mm_tabling_info_proc_id :: proc_pf_name_arity_mn,
mm_tabling_info_status :: mm_tabling_status
).
:- type pragma_info_require_tail_rec
---> pragma_info_require_tail_rec(
rtr_proc_id :: pred_or_proc_pfumm_name,
rtr_require_tailrec :: require_tail_recursion
% This parameter only makes sense when options contains
% either rtro_mutual_rec_only or rtro_all_recursion.
% TODO, currently unused, may be used later to implement one
% of Zoltan's suggestions here:
% http://www.mercurylang.org/list-archives/developers/
% 2015-November/016482.html
% rtr_maybe_scc :: maybe(list(
% pred_or_proc_pfumm_name))
).
% Evaluation method pragmas.
:- type pragma_info_tabled
---> pragma_info_tabled(
% Tabling type, Predname, Arity, PredOrFunc?, Mode?
tabled_method :: tabled_eval_method,
tabled_name :: pred_or_proc_pfumm_name,
tabled_attributes :: maybe(table_attributes)
).
:- type pragma_info_fact_table
---> pragma_info_fact_table(
% Predname and Arity, Fact file name.
fact_table_pred :: pred_pfu_name_arity,
fact_table_filename :: string
).
:- type pragma_info_oisu
---> pragma_info_oisu(
oisu_type_ctor :: type_ctor,
oisu_creator_preds :: list(pred_pf_name_arity),
oisu_transformer_preds :: list(pred_pf_name_arity),
oisu_destroyer_preds :: list(pred_pf_name_arity)
).
% Termination analysis pragmas.
:- type pragma_info_termination_info
---> pragma_info_termination_info(
% The list(mer_mode) is the declared argmodes of the
% procedure, unless there are no declared argmodes, in which
% case the inferred argmodes are used. This pragma is used to
% define information about a predicates termination
% properties. It is most useful where the compiler has
% insufficient information to be able to analyse the
% predicate. This includes c_code, and imported predicates.
% termination_info pragmas are used in opt and trans_opt
% files.
terminfo_pred_id :: proc_pf_name_modes,
terminfo_args :: maybe(pragma_arg_size_info),
terminfo_term :: maybe(pragma_termination_info)
).
:- type pragma_info_termination2_info
---> pragma_info_termination2_info(
terminfo2_pred_id :: proc_pf_name_modes,
terminfo2_args :: maybe(pragma_constr_arg_size_info),
terminfo2_args2 :: maybe(pragma_constr_arg_size_info),
terminfo2_term :: maybe(pragma_termination_info)
).
% CTGC pragmas: structure sharing / structure reuse analysis.
:- type pragma_info_structure_sharing
---> pragma_info_structure_sharing(
% After structure sharing analysis, the compiler generates
% structure sharing pragmas to be stored in and read from
% optimization interface files.
%
% The list of modes consists of the declared argmodes
% (or inferred argmodes if there are no declared ones).
sharing_pred_id :: proc_pf_name_modes,
sharing_headvars :: list(prog_var),
sharing_headvar_types :: list(mer_type),
% The prog_varset and tvarset are meaningful only when
% writing out this pragma; add_pragma.m ignores both varsets.
sharing_varset :: prog_varset,
sharing_tvarset :: tvarset,
% As of 2019 10 29, and probably long before then,
% the compiler *always* fills this slot with `yes(...)'.
% A `no' would mean that the relevant information is not
% available, but in that case, we simply do not write out
% this pragma.
sharing_description :: maybe(structure_sharing_domain)
).
:- type pragma_info_structure_reuse
---> pragma_info_structure_reuse(
% After reuse analysis, the compiler generates structure reuse
% pragmas to be stored in and read from optimization interface
% files.
%
% The list of modes consists of the declared argmodes
% (or inferred argmodes if there are no declared ones).
%
% The last sym_name (reuse_optimised_name) stores the name
% of the optimised version of the exported predicate.
% XXX As of 2019 10 29, the word "reuse_optimised_name"
% appears nowhere in the compiler apart from this comment.
reuse_pred_id :: proc_pf_name_modes,
reuse_headvars :: list(prog_var),
reuse_headvar_types :: list(mer_type),
% The prog_varset and tvarset are meaningful only when
% writing out this pragma; add_pragma.m ignores both varsets.
reuse_varset :: prog_varset,
reuse_tvarset :: tvarset,
% As of 2019 10 29, and probably long before then,
% the compiler *always* fills this slot with `yes(...)'.
% A `no' would mean that the relevant information is not
% available, but in that case, we simply do not write out
% this pragma.
reuse_description :: maybe(structure_reuse_domain)
).
% Misc pragmas.
:- type pragma_info_obsolete_pred
---> pragma_info_obsolete_pred(
pred_pfu_name_arity,
list(sym_name_arity)
).
:- type pragma_info_obsolete_proc
---> pragma_info_obsolete_proc(
proc_pf_name_modes,
list(sym_name_arity)
).
:- type pragma_info_format_call
---> pragma_info_format_call(
pred_pf_name_arity,
one_or_more(format_string_values)
).
:- type pragma_info_require_feature_set
---> pragma_info_require_feature_set(
rfs_feature_set :: set(required_feature)
).
% These types identify predicates, functions and/or procedures in pragmas.
:- type pred_pfu_name_arity
---> pred_pfu_name_arity(
ppfuna_pfu :: pred_func_or_unknown,
ppfuna_pred_name :: sym_name,
ppfuna_arity :: user_arity
).
:- type proc_pf_name_arity_mn
---> proc_pf_name_arity_mn(
ppfnamn_pf :: pred_or_func,
ppfnamn_pred_name :: sym_name,
ppfnamn_arity :: user_arity,
ppfnamn_mode_num :: mode_num
).
:- type proc_pf_name_modes
---> proc_pf_name_modes(
ppfnm_pf :: pred_or_func,
ppfnm_pred_name :: sym_name,
ppfnm_arity :: list(mer_mode)
).
:- type pred_or_proc_pfumm_name
---> pred_or_proc_pfumm_name(
ppfummn_pfumm :: pred_func_or_unknown_maybe_modes,
ppfummn_pred_name :: sym_name
).
:- type pred_func_or_unknown
---> pfu_predicate
; pfu_function
; pfu_unknown.
:- type pred_func_or_unknown_maybe_modes
---> pfumm_predicate(modes_or_arity)
; pfumm_function(modes_or_arity)
; pfumm_unknown(user_arity).
:- type modes_or_arity
---> moa_modes(list(mer_mode))
; moa_arity(user_arity).
:- func pfu_to_maybe_pred_or_func(pred_func_or_unknown) = maybe(pred_or_func).
:- func maybe_pred_or_func_to_pfu(maybe(pred_or_func)) = pred_func_or_unknown.
:- pred pfumm_to_maybe_pf_arity_maybe_modes(
pred_func_or_unknown_maybe_modes::in, maybe(pred_or_func)::out,
user_arity::out, maybe(list(mer_mode))::out) is det.
%---------------------------------------------------------------------------%
%
% Goals.
%
% Here is how goals are represented in the parse tree.
% The three most frequent kinds of goals are first, to give them
% their own primary tags on 32 bit machines, and
% the seven most frequent kinds of goals are first, to give them
% their own primary tags on 64 bit machines.
%
% During a bootcheck in august 2015, the frequencies of occurrence
% of the various goal kinds were these:
%
% goal_unify 1360701
% goal_conj 1316066 when we had a conj_expr for each ","
% goal_call 1263403
%
% goal_true 135352
% goal_if_then_else 128052
% goal_disj 116547 when we had a disj_expr for each ";"
% goal_not 7080
%
% goal_fail 5219
% goal_pro_purity 1492
% goal_trace 1356
% goal_pro_eqv_solns 913
% goal_some_state_vars 620 now goal_quant/some/state
% goal_some 192 now goal_quant/some/ordinary
% goal_req_compl_switch 172
% goal_par_conj 132 when we had a par_conj_expr for each "&"
% goal_implies 129
% goal_all 78 now goal_quant/all/ordinary
% goal_req_detism 49
% goal_try 35
% goal_equivalent 18
% goal_event 17
% goal_req_arm_detism 14
% goal_pro_arbitrary 12
% goal_pro_eqv_soln_sets 8
% goal_atomic 2
% goal_all_state_vars 0 now goal_quant/all/state
:- type quant_type
---> quant_some
; quant_all.
:- type quant_vars_kind
---> quant_ordinary_vars
; quant_state_vars.
:- type plain_or_dot_var
---> podv_plain(prog_var)
% V: a plain variable.
; podv_dot(prog_var).
% !.SV: the current state of this state variable.
:- type goal
% The most frequent kinds of goals.
---> unify_expr(prog_context, prog_term, prog_term, purity)
; call_expr(prog_context, sym_name, list(prog_term), purity)
; conj_expr(prog_context, goal, list(goal))
% nonempty plain conjunction
% NOTE: We could replace this with
% conj_expr(prog_context, goal, goal, list(goal))
% to encode the invariant that
% - a conjunction has at least one conjunction operator, and
% - that operator has two argument goals.
% However, no part of the current compiler can exploit
% this extra information.
% NOTE: On the other hand, we could also replace this with
% conj_expr(prog_context, list(goal))
% letting a conj_expr with an empty list of goals take over
% the role of true_expr. However, that would make the parse tree
% representation of plain conjunctions differ from the
% representation of parallel conjunctions. And the most
% frequent goal that does not now have its own primary tag
% on 64 bit machines, fail_expr, is infrequent enough that
% giving it its own primary tag would not materially improve
% performance, and even if it were frequent enough, it could be
% folded into disj_exprs in a similar way.
; true_expr(prog_context)
% empty conjunction
; if_then_else_expr(
prog_context,
list(prog_var), % SomeVars
list(prog_var), % StateVars
goal, % Cond
goal, % Then
goal % Else
)
; disj_expr(prog_context, goal, goal, list(goal))
% nonempty disjunction; will contain at least two goals.
; not_expr(prog_context, goal)
% The other kinds of goals.
; fail_expr(prog_context)
% empty disjunction
; par_conj_expr(prog_context, goal, list(goal))
% nonempty parallel conjunction
; quant_expr(
% Existential or universal quantification?
quant_type,
% Are the variables ordinary variables or state variables?
quant_vars_kind,
prog_context,
list(prog_var),
goal
)
; promise_purity_expr(prog_context, purity, goal)
; promise_equivalent_solutions_expr(
prog_context,
list(prog_var), % OrdinaryVars
list(prog_var), % StateVars (!V)
list(prog_var), % DotStateVars (!.V)
list(prog_var), % ColonStateVars (!:V)
goal
)
; promise_equivalent_solution_sets_expr(
prog_context,
list(prog_var), % OrdinaryVars
list(prog_var), % StateVars (!V)
list(prog_var), % DotStateVars (!.V)
list(prog_var), % ColonStateVars (!:V)
goal
)
; promise_equivalent_solution_arbitrary_expr(
prog_context,
list(prog_var), % OrdinaryVars
list(prog_var), % StateVars (!V)
list(prog_var), % DotStateVars (!.V)
list(prog_var), % ColonStateVars (!:V)
goal
)
; require_detism_expr(
prog_context,
determinism,
goal
)
; require_complete_switch_expr(
prog_context,
plain_or_dot_var,
goal
)
; require_switch_arms_detism_expr(
prog_context,
plain_or_dot_var,
determinism,
goal
)
; disable_warnings_expr(
% Disable the given one or more warnings
% in the goal inside the scope.
prog_context,
goal_warning,
list(goal_warning),
goal
)
; trace_expr(
texpr_context :: prog_context,
texpr_compiletime :: maybe(trace_expr(trace_compiletime)),
texpr_runtime :: maybe(trace_expr(trace_runtime)),
texpr_maybe_io :: maybe(prog_var),
texpr_mutable_vars :: list(trace_mutable_var),
texpr_goal :: goal
)
; atomic_expr(
% Subgoals of the atomic goal are parsed into the following
% datatype. During the creation of the parse tree, all
% subterms of the "orelse" operator are flattened and placed
% into a list. If this is the case, the first "orelse"
% alternative is stored in "main_goal" whilst the other
% alternatives are stored in "orelse_alternatives". If there
% are no "or_else" operators within the atomic subgoal,
% the subgoal is stored in "main_goal" whilst the
% "orelse_alternatives" list remains empty.
aexpr_context :: prog_context,
aexpr_outer :: atomic_component_state,
aexpr_inner :: atomic_component_state,
aexpr_output_vars :: maybe(list(prog_var)),
aexpr_main_goal :: goal,
aexpr_orelse_goals :: list(goal)
)
; try_expr(
tryexpr_context :: prog_context,
tryexpr_maybe_io :: maybe(prog_var),
tryexpr_goal :: goal,
tryexpr_then :: goal,
tryexpr_maybe_else :: maybe(goal),
tryexpr_catches :: list(catch_expr),
tryexpr_maybe_catch_any :: maybe(catch_any_expr)
)
; implies_expr(prog_context, goal, goal)
% implies_expr(_, A, B) represents either A => B or B <= A.
; equivalent_expr(prog_context, goal, goal)
% equivalent_expr(_, A, B) represents A <=> B.
; event_expr(prog_context, string, list(prog_term)).
:- type catch_expr
---> catch_expr(
catch_pattern :: prog_term,
catch_goal :: goal
).
:- type catch_any_expr
---> catch_any_expr(
catch_any_var :: prog_var,
catch_any_goal :: goal
).
%---------------------------------------------------------------------------%
:- func get_item_context(item) = prog_context.
:- func get_goal_context(goal) = prog_context.
%---------------------------------------------------------------------------%
:- type contains_foreign_code
---> foreign_code_langs_known(set(foreign_language))
; foreign_code_langs_unknown.
:- type contains_foreign_export
---> contains_foreign_export
; contains_no_foreign_export.
%---------------------------------------------------------------------------%
%---------------------------------------------------------------------------%
:- implementation.
:- import_module parse_tree.prog_util.
:- import_module term.
:- import_module varset.
%---------------------------------------------------------------------------%
init_aug_compilation_unit(ParseTreeModuleSrc, AugCompUnit) :-
map.init(AncestorIntSpecs),
map.init(DirectIntSpecs),
map.init(IndirectIntSpecs),
map.init(PlainOpts),
map.init(TransOpts),
map.init(IntForOptSpecs),
map.init(TypeRepnSpecs),
map.init(VersionNumbers),
AugCompUnit = aug_compilation_unit(ParseTreeModuleSrc,
AncestorIntSpecs, DirectIntSpecs, IndirectIntSpecs,
PlainOpts, TransOpts, IntForOptSpecs, TypeRepnSpecs, VersionNumbers).
%---------------------------------------------------------------------------%
%
% Mutable variables.
%
mutable_var_thread_local(Const) = Local :-
( if
Const = mutable_is_not_constant(_AttachToIO, IsLocal),
% Const = mutable_is_constant would imply mutable_not_thread_local
IsLocal = mutable_is_thread_local
then
Local = mutable_thread_local
else
Local = mutable_not_thread_local
).
mutable_thread_local_trailed(Local) = Trail :-
(
Local = mutable_is_not_thread_local(Trail)
;
Local = mutable_is_thread_local,
Trail = mutable_untrailed
).
%---------------------------------------------------------------------------%
fill_kind_size_num_bits(FillKindSize) = NumBits :-
(
FillKindSize = fk_enum(NumBits)
;
( FillKindSize = fk_int8
; FillKindSize = fk_uint8
),
NumBits = 8u
;
( FillKindSize = fk_int16
; FillKindSize = fk_uint16
),
NumBits = 16u
;
( FillKindSize = fk_int32
; FillKindSize = fk_uint32
),
NumBits = 32u
;
FillKindSize = fk_char21,
NumBits = 21u
).
%---------------------------------------------------------------------------%
pfu_to_maybe_pred_or_func(pfu_predicate) = yes(pf_predicate).
pfu_to_maybe_pred_or_func(pfu_function) = yes(pf_function).
pfu_to_maybe_pred_or_func(pfu_unknown) = no.
maybe_pred_or_func_to_pfu(yes(pf_predicate)) = pfu_predicate.
maybe_pred_or_func_to_pfu(yes(pf_function)) = pfu_function.
maybe_pred_or_func_to_pfu(no) = pfu_unknown.
pfumm_to_maybe_pf_arity_maybe_modes(PFUMM, MaybePredOrFunc, UserArity,
MaybeModes) :-
(
(
PFUMM = pfumm_predicate(ModesOrArity),
PredOrFunc = pf_predicate
;
PFUMM = pfumm_function(ModesOrArity),
PredOrFunc = pf_function
),
MaybePredOrFunc = yes(PredOrFunc),
(
ModesOrArity = moa_modes(Modes),
list.length(Modes, NumModes),
PredFormArity = pred_form_arity(NumModes),
user_arity_pred_form_arity(PredOrFunc, UserArity, PredFormArity),
MaybeModes = yes(Modes)
;
ModesOrArity = moa_arity(UserArity),
MaybeModes = no
)
;
PFUMM = pfumm_unknown(UserArity),
MaybePredOrFunc = no,
MaybeModes = no
).
%---------------------------------------------------------------------------%
get_item_context(Item) = Context :-
(
Item = item_clause(ItemClause),
Context = ItemClause ^ cl_context
;
Item = item_type_defn(ItemTypeDefn),
Context = ItemTypeDefn ^ td_context
;
Item = item_inst_defn(ItemInstDefn),
Context = ItemInstDefn ^ id_context
;
Item = item_mode_defn(ItemModeDefn),
Context = ItemModeDefn ^ md_context
;
Item = item_pred_decl(ItemPredDecl),
Context = ItemPredDecl ^ pf_context
;
Item = item_mode_decl(ItemModeDecl),
Context = ItemModeDecl ^ pfm_context
;
Item = item_foreign_enum(ItemForeignEnum),
Context = ItemForeignEnum ^ fe_context
;
Item = item_foreign_export_enum(ItemForeignExportEnum),
Context = ItemForeignExportEnum ^ fee_context
;
Item = item_decl_pragma(ItemDeclPragma),
Context = ItemDeclPragma ^ prag_context
;
Item = item_impl_pragma(ItemImplPragma),
Context = ItemImplPragma ^ prag_context
;
Item = item_generated_pragma(ItemGenPragma),
Context = ItemGenPragma ^ prag_context
;
Item = item_promise(ItemPromise),
Context = ItemPromise ^ prom_context
;
Item = item_typeclass(ItemTypeClass),
Context = ItemTypeClass ^ tc_context
;
Item = item_instance(ItemInstance),
Context = ItemInstance ^ ci_context
;
Item = item_initialise(ItemInitialise),
Context = ItemInitialise ^ init_context
;
Item = item_finalise(ItemFinalise),
Context = ItemFinalise ^ final_context
;
Item = item_mutable(ItemMutable),
Context = ItemMutable ^ mut_context
;
Item = item_type_repn(ItemTypeRepn),
Context = ItemTypeRepn ^ tr_context
).
get_goal_context(Goal) = Context :-
( Goal = conj_expr(Context, _, _)
; Goal = par_conj_expr(Context, _, _)
; Goal = true_expr(Context)
; Goal = disj_expr(Context, _, _, _)
; Goal = fail_expr(Context)
; Goal = quant_expr(_, _, Context, _, _)
; Goal = promise_purity_expr(Context, _, _)
; Goal = promise_equivalent_solutions_expr(Context, _, _, _, _, _)
; Goal = promise_equivalent_solution_sets_expr(Context, _, _, _, _, _)
; Goal = promise_equivalent_solution_arbitrary_expr(Context, _, _, _, _, _)
; Goal = require_detism_expr(Context, _, _)
; Goal = require_complete_switch_expr(Context, _, _)
; Goal = require_switch_arms_detism_expr(Context, _, _, _)
; Goal = disable_warnings_expr(Context, _, _, _)
; Goal = trace_expr(Context, _, _, _, _, _)
; Goal = atomic_expr(Context, _, _, _, _, _)
; Goal = try_expr(Context, _, _, _, _, _, _)
; Goal = implies_expr(Context, _, _)
; Goal = equivalent_expr(Context, _, _)
; Goal = not_expr(Context, _)
; Goal = if_then_else_expr(Context, _, _, _, _, _)
; Goal = event_expr(Context, _, _)
; Goal = call_expr(Context, _, _, _)
; Goal = unify_expr(Context, _, _, _)
).
%---------------------------------------------------------------------------%
:- end_module parse_tree.prog_item.
%---------------------------------------------------------------------------%