Compendium Deluxe 1

home *** CD-ROM | disk | FTP | other *** search

/ Compendium Deluxe 1 / LSD Compendium Deluxe 1.iso / a / programming / misc / ada1110b.lha / AmigaSources / hdr.h < prev next >

Wrap

C/C++ Source or Header | 1993-07-10 | 65.2 KB | 2,206 lines

/* * Copyright (C) 1985-1992 New York University * * This file is part of the Ada/Ed-C system. See the Ada/Ed README file for * warranty (none) and distribution info and also the GNU General Public * License for more details. */ #ifndef _hdr_h #define _hdr_h /* Symbol Table Definitions for Ada/ED in C Change history: October 1989 gs Remove Sybmol Table stuff and put in separate file symbol.h. This is included from this file now. July 1989 gs Add field to node N_OPERAND to hold either the register or literal value associated with the Amiable generated code for the Node. 19-dec-86 gs Revise format for nodes storing left span (only) for terminal nodes, and overlaying these 2 fields with N_AST1 of nonterminal nodes. 18-sep-86 ds Revise format for nodes used for aggregates to permit overlaying of n_unq and n_ast3 fields. This requires that n_ast3 not be used for such nodes, so a new node kind - as_aggregate_list - has been introduced and placed in the n_ast1 position to hold the two nodes formerly kept in n_ast1 and n_ast2. The former n_ast3 field is now at n_ast2. The affected node kinds are as_array_aggregate, as_record_aggregate, as_array_ivalue and as_record_ivalue. 15-nov-85 gs add CONSTRAINT_ACCESS for subtypes of access types. 28-sep-85 ds defintions of ATTR_ codes moved to separate inclusion file attr.h 14-aug-85 ds Add new symbol table field type_kind, used only by code_generator. 8-jul-85 ds Add new member const_fixed to Const structure used for fixed-point values during code generation. 24-apr-85 ds Support only lower case na_ codes, drop NA_ codes. Only NA_ENUM was actually used in upper case and this has been changed to lower case where so used. Also, delete declarations for externs that return int's. 16-apr-65 ds Reduce the size of the FORDECLARED macro by using (new) procedures fordeclared_1() and fordeclared_2() to initiate and advance the iteration, respectively. 3-apr-85 add structure definitions for Symbolmap and Nodemap because node_map must be global (to all of chapter 12). 21-mar-85 Redefine associated_name to associated_symbols, a tuple of symbols. This is described in more detail in ghdr.c. 13-mar-85 The macro index_type was conditionally defined only for SEM since this name was used as variable name within code generator. Such usages within code generator have been changed to use name indx_type so that conditional definition no longer needed. This avoids need for SEM conditional option in this case. 12-mar-85 Redo declarations for nodes and symbols so that single struct used for both SEM and GEN with all fields defined. Fields used only by GEN are defined at end of symbol_s declaration so that (eventually) SEM can ignore these fields when building entries. For now will process full entry. I merged in corrections from sem hdr.c through version 1.46. 4-mar-85 Add codes CODE_SLOTS etc for string used as third arg to select entry (this is enumeration type). Add symbol table field associated_name. THis is used for example to give the _type variable associated with an array name, etc. In SETL version, these names are referenced by taking name and appending string... 21-jan-85 modify to permit use with GEN and SEM phases. Use conditional symbols GEN and SEM to indicate code needed only for that phase. 31-dec-84 add NEEDNAME field to indicate symbols for which identifier needed as name for SETL code generator. Add CONSTRAINT_ARRAY code for use when writing out signatures for constrained arrays. 26-dec-84 ds add N_COUNT field to hold NODE_COUNT values for node. 16-dec-84 ds Use overloads for literal map for enumeration types, accessed using macro literal_map. This was modified to used declared field due to original handling of misc_type_attributes. Now that type_attr field available, it is possible (and necessary to avoid bugs) to use overloads, as this avoids adding further code for special handling of declared map. 12-nov-84 ds make is_anonymous_task a procedure. 08-nov-84 ds add n_spans field to Node_s for spans information. NOTE: attempt to declare n_spans as short (which it should be) caused very mysterious failures in malloc for no apparent reason; hence n_spans is declared as int. add s_type_attr field to Symbol_s to hold 'miscellaneous_type_attributes' which cannot be kept in overloads field in C version. Change definition of misc_type_attributes macro so it now refers to new field and definition of private_dependents macro so it now refers to overloads. 31-oct-84 ds Node that numeric_constraint for 'delta' now has four nodes; new one is 'small' with new macro numeric_constraint_small. 15-oct-84 ds Merge const.h and pdecl.h files into this file. 6-oct-84 ds Revise to use seq_node tuple to record allocated nodes. Assigned values of N_SEQ start with 1, seq_node[i] is address of node with N_SEQ == i. seq_node_n is number of actual nodes so far allocated; only the first seq_node_n entries of seq_node correspond to allocated nodes, the remaining entries provide expansion space. Similarly for allocated symbols, using seq_symbol. SEQ_NODE_INC defines number of new entries allocated when have to expand seq_node, similarly for SEQ_SYMBOL_INC and seq_symbol. */ #define SEQ_NODE_INC 50 #define SEQ_SYMBOL_INC 50 /* ds 5-oct-84 Delete n_anon field for node, add n_seq field for node sequence number. Also add n_unit field for unit number of node. Change name of Symbol field sequence to s_seq and add s_unit field for a Symbol to hold unit number. ds 29-sep-84 I originally made numeric_constraint data structure (see na_subtype) because I thought lo, hi, etc were actual values. It turns out they are just nodes, and it is more convenient to represent contraints as tuples, as is done in SETL. I have modified code to do this, thbough macors numeric_constraint_low, etc, are still used. They can be edited out later. But structure numeric_constraint_s is now gone. ds 1 aug Modify symbol sequence handling as follows: unit_symbol is tuple of symbol table pointers for current unit - the i-th entry should have sequence number i. unit_symbols is number of symbols allocated. To avoid reallocating the tuple unit_symbol when each new symbol is allocated, unit_symbol_length will be allocated length, and must always be >= unit_symbols. Will also have unit_scope, a tuple with an entry for each new scope created within a compilation unit. ds 20 july Need to check macro definitions to see where need parentheses around argument names. ds 12 july add new symbol table field 'sequence' that contains sequence number assigned for symbols. The global SEQUENCE_NUMBER gives the number of the last symbol allocated; numbering starts with one. The global SEQUENCE_LIST is a tuple indexed by sequence numbers that gives the address of the corresponding symbol table entry. ---- This file contains the (proposed) definitions and declarations for symbol table access for Ada/Ed-C, the C version of Ada/Ed. Coding conventions: ------------------- In this section we outline a suggested set of coding conventions that are used in the remainder of this document. C does not have the 'end-of-line' comment using a single character, but requires that all comments begin with slash asterisk and end with asterisk slash (we are naming the characters here to avoid having a comment within a comment!). There are a variety of conventions used for writing comments in C, including the all too common convention of writing no comments at all (a practice we do not intend to follow). We follow the following convention: If a line begins with white space followed by the comment opener, then it is the first of possibly several comment lines. The comment continues up to and including the next line containing the comment closer. No code may follow the comment closer on the last line. If a line otherwise contains a comment opener, then it must also contain the comment closer, and no code must follow the comment closer. This permits each line to be classified as either code followed by a comment extending to the end of the line, code with no comment, or part of a comment, and makes it possible to write a filter to convert to other comment conventions. Significance of Character Case ------------------------------ In C, unlike SETL, the case of an identifier is significant, so that 'I' and 'i' are distinct identifiers. The SETL version uses such case-ignorant names at times; and these will have to be removed as part of the translation to C. In general, we will follow the following rules, with occasional exceptions, noted in comments below, which will simplify the translation of the SETL source to C; such exceptions will typically introduce mixed-case identifiers. 1. Names defined by #define are capitalized, i.e., defined constants and macros. 2. SETL strings used as discriminants will usually be changed to defined constants, with some appropriate prefix, and will thus be capitalized. 3. User-defined type names have their first letter capitalized. 4. Structure tags are to be defined with initial letter capitalized and are given the same name via a typdef. For example, struct Str { ... } Str; 5. Since much of the code will involve merely passing pointers to structures, we will often use type names that in fact just pointers. This requires two names, one for the structure and one for a pointer to the structure. Having two names permits us to write Str x,y,z; instead of Str *x,*y,*z; For such cases, our convention is to use Name for a pointer to the structure, and Name_s for the structure itself. This permits us to write code using the first approach, avoiding needless (since they are always implicit) asterisks in declarations and other references, such as casts, involving the type name. */ /* Include declarations for sets and tuples */ #include <stdlib.h> #include <stdio.h> #include <ctype.h> #include <string.h> #include "config.h" #include "arith.h" /* SYMBOL TABLE In SETL, the symbol table is represented as several maps: nature, type_of, scope_of, signature, overloads and alias. In C, a symbol table entry will be a structure, with fields corresponding to the SETL maps, in the order given above. Not all fields are defined for each entry. We provide a single declaration and leave it up to the allocator (which will be given the nature) to allocate the proper length entry. The alias field is used for derived subprograms and renamed subprograms, and also for types. The overloads field is defined for procedures, functions, literals and operators; for packages it is used to store private declarations. The SETL version has declared as a map; in the C version this will be a field in symbol table entries for which it is defined. The orig_name field gives the original name of the symbol. */ typedef struct Declaredmap_s *Declaredmap; #ifdef AMIABLE /* #include "gena.h" */ /* it includes symbol.h */ #else #include "symbol.h" #endif /* Each entry in the symbol type has a nature. The nature is represented as a string in the SETL source, and as one of the following constant values of the form na_ in C. */ /* AST declarations and definitions This file contains the declarations and definitions for the AST representation for the ADA/ED-C ADASEM phase. The following summary is taken from file [ada.dev]ast.doc (perhaps should merge in this whole file later): The following is a description of the nodes used in building the abstract syntax tree (AST) for an ADA source program (in the SETL version). The AST is represented as a collection of nodes on which certain maps are defined to form the tree. The nodes correspond to non-terminals such as "pragma", "object_declaration", etc... Nodes are represented by setl integers and are described as either equivalent to other nodes or by specifying values for the maps N_KIND, N_AST, N_VAL, and/or N_LIST that are defined on them. These maps are used as follows: N_KIND(node) is a string (for now) that names the node and serves as a discriminant for the other maps defined on it. N_AST(node) is a tuple of nodes, corresponding to the right-hand side of a production for the node. N_VAL(node) is defined for terminal nodes. For simple_names it is the source identifier for that node. The semantic actions use the N_VAL map to store miscellaneous information used by the EMIT routine. N_LIST(node) is a tuple of arbitrary length whose components are nodes of the same kind. N_LIST(node) is defined for all occurrences in the grammar of arbitrary repetitions. These are noted below as nodes of the form <node...>, and have the N_KIND 'list'. SEMANTIC PROCESSING. The semantic routines are driven by the N_KIND of each AST node. In most cases the AST is not directly modified by the semantic actions; rather, several maps, including the symbol table, are constructed and updated. The following maps are used by sem: N_UNQ(node) is the unique name corresponding to a name node. The N_KIND of such a node is always set to 'simple_name'. The N_UNQ is a symbol table pointer, through which all attributes of an entity are retrieved. N_NAMES(node) is the overloaded set of names coresponding to a name node before overload resolution is performed. It is used only during type resolution. The above two maps, N_UNQ and N_NAMES, are defined only on name nodes. N_TYPE(node) is the unique name of the type of an expression node. This field is set for all expressions and names that are not simple names. N_PTYPES(node) is the set of possible types of an overloaded construct, before resolution. It is also useless after type resolution. The maps N_TYPE and N_PTYPES are defined only on expression nodes. -- end excerpt from AST.DOC In C N_KIND will be represented as an integer constant of the form AS_. A node has N_UNQ and N_NAMES defined; or it has N_TYPE and N_PTYPES defined. Field n_sym is used for N_UNQ or N_TYPE, which are both pointers to symbol table entries. Field n_set is used for N_NAMES or N_PTYPES, which are both pointers to sets of symbol table pointers. The fields for N_LIST and N_VAL can overlap, and are hence declared as a union. The fields for N_AST and N_LIST can overlap since they are never both defined for a node Note that for the case where the SETL AST map id defined, the number of components is fixed for a particular kind, and at most five. We thus define fields n_ast1, n_ast2 ... n_ast5 and allocate only as many fields as needed for each particular node. It is not necessary to keep the number of AST entries, since the length can be deduced from N_KIND. */ typedef struct Span_s *Span; /* pointer to spans information */ typedef struct Span_s { short line; short col; } Span_s; typedef struct Node_s *Node; /* Node is pointer to Node_s */ #define NODE_SIZE sizeof(Node_s) typedef struct Node_s { #ifndef IBM_PC struct { unsigned n_kind:8; unsigned n_unit:8; unsigned n_side:1; unsigned n_overloaded:1; } n_flags; short n_seq; #else struct { unsigned n_kind:8; unsigned n_unit:8; } n_flag1; struct { unsigned n_side:1; unsigned n_seq:14; unsigned n_overloaded:1; } n_flag2; #endif union { Node n_ast1; Span_s n_span; } nu1; union { Node n_ast2; Tuple n_list; char *n_val; } nu2; union { Node n_ast3; Set n_names; Symbol n_unq; } nu3; union { Node n_ast4; Set n_ptypes; Symbol n_type; } nu4; #ifdef AMIABLE Operand_s n_operand; #endif } Node_s; struct unit { char *name; int isMain; char *libUnit; struct { char *fname; char *obsolete; char *currCodeSeg; char *localRefMap; char *compDate; } libInfo; struct { char *preComp; char *pragmaElab; char *compDate; char *symbols; int numberSymbols; char *unitDecl; } aisInfo; struct { char *tableAllocated; int nodeCount; int rootSeq; } treInfo; }; #define MAX_UNITS 100 #ifdef IBM_PC #define N_KIND(p) ((p)->n_flag1.n_kind) #define N_UNIT(p) ((p)->n_flag1.n_unit) #define N_OVERLOADED(p) ((p)->n_flag2.n_overloaded) #define N_SIDE(p) ((p)->n_flag2.n_side) #define N_SEQ(p) ((p)->n_flag2.n_seq) #else #define N_KIND(p) ((p)->n_flags.n_kind) #define N_UNIT(p) ((p)->n_flags.n_unit) #define N_OVERLOADED(p) ((p)->n_flags.n_overloaded) #define N_SIDE(p) ((p)->n_flags.n_side) #define N_SEQ(p) ((p)->n_seq) #endif #define N_SPAN0(p) ((p)->nu1.n_span.line) #define N_SPAN1(p) ((p)->nu1.n_span.col) #define N_AST1(p) ((p)->nu1.n_ast1) #define N_VAL(p) ((p)->nu2.n_val) #define N_LIST(p) ((p)->nu2.n_list) #define N_AST2(p) ((p)->nu2.n_ast2) #define N_AST3(p) ((p)->nu3.n_ast3) #define N_UNQ(p) ((p)->nu3.n_unq) #define N_NAMES(p) ((p)->nu3.n_names) #define N_AST4(p) ((p)->nu4.n_ast4) #define N_TYPE(p) ((p)->nu4.n_type) #define N_PTYPES(p) ((p)->nu4.n_ptypes) #ifdef AMIABLE #define N_OPERAND(p) ((p)->n_operand) #endif /* N_D_ macros define which fields defined for node according to kind */ #define N_D_AST1 1 #define N_D_AST2 2 #define N_D_AST3 4 #define N_D_AST4 8 #define N_D_LIST 16 #define N_D_VAL 32 #define N_D_UNQ 64 /* was N_CHAOS 128 ds 10-29-86 */ #define N_D_TYPE 256 #define N_AST1_DEFINED(p) (N_DEFINED[p]&N_D_AST1) #define N_AST2_DEFINED(p) (N_DEFINED[p]&N_D_AST2) #define N_AST3_DEFINED(p) (N_DEFINED[p]&N_D_AST3) #define N_AST4_DEFINED(p) (N_DEFINED[p]&N_D_AST4) #define N_VAL_DEFINED(p) (N_DEFINED[p]&N_D_VAL) /* N_UNQ_DEFINED and N_TYPE defined are set for nodes for which these fields * are to written to library files and for which the fields are to be set * during instantiation. */ #define N_UNQ_DEFINED(p) (N_DEFINED[p]&N_D_UNQ) #define N_TYPE_DEFINED(p) (N_DEFINED[p]&N_D_TYPE) #define N_LIST_DEFINED(p) (N_DEFINED[p]&N_D_LIST) /* From schonber Tue Feb 11 00:51:07 1986 Date: Tue, 11 Feb 86 00:51:03 est From: schonber Received: by nyu-acf2.arpa; Tue, 11 Feb 86 00:51:03 est To: shields Subject: data compression. Status: R Have you tried overlapping N_PTYPES and N_TYPE? This should be safe in all cases, as they have always disjoint usage. A further possibility is N_NAMES and N_UNQ, but this requires the use of some additional bit to indicate whether an entity is overloaded or not. By the way, N_NAMES is ONLY defined on simple names and ops, so maybe overlapped with unused N_AST field. THere is still some juice to be squeezed out. Franck says: N_NAMES and N_UNQ cannot be overlaid From chiabaut@nyu.arpa Thu Sep 11 10:27:20 1986 Received: from nyu.arpa by nyu-acf2.arpa; Thu, 11 Sep 86 10:27:17 edt Date: Thu, 11 Sep 86 10:28:48 edt From: chiabaut@nyu.arpa (Jerome Chiabaut) Message-Id: <8609111428.AA03384@nyu.arpa> Received: by nyu.arpa; Thu, 11 Sep 86 10:28:48 edt To: shields@nyu-acf2.arpa Subject: Re: overlaying n_unq and n_type Status: RO I'm not sure but i think that an aggregate or array-aggregate has more than two n_asts defined along with n_type and n_unq. In any case the idea is that for a few nodes it might require to change the shape of the tree. Jerome */ #define as_pragma 0 #define as_arg 1 #define as_obj_decl 2 #define as_const_decl 3 #define as_num_decl 4 #define as_type_decl 5 #define as_subtype_decl 6 #define as_subtype_indic 7 #define as_derived_type 8 #define as_range 9 #define as_range_attribute 10 #define as_constraint 11 #define as_enum 12 #define as_int_type 13 #define as_float_type 14 #define as_fixed_type 15 #define as_digits 16 #define as_delta 17 #define as_array_type 18 #define as_box 19 #define as_subtype 20 #define as_record 21 #define as_component_list 22 #define as_field 23 #define as_discr_spec 24 #define as_variant_decl 25 #define as_variant_choices 26 #define as_string 27 #define as_simple_choice 28 #define as_range_choice 29 #define as_choice_unresolved 30 #define as_others_choice 31 #define as_access_type 32 #define as_incomplete_decl 33 #define as_declarations 34 #define as_labels 35 #define as_character_literal 36 #define as_simple_name 37 #define as_call_unresolved 38 #define as_selector 39 #define as_all 40 #define as_attribute 41 #define as_aggregate 42 #define as_parenthesis 43 #define as_choice_list 44 #define as_op 45 #define as_in 46 #define as_notin 47 #define as_un_op 48 #define as_int_literal 49 #define as_real_literal 50 #define as_string_literal 51 #define as_null 52 #define as_name 53 #define as_qualify 54 #define as_new_init 55 #define as_new 56 #define as_statements 57 #define as_statement 58 #define as_null_s 59 #define as_assignment 60 #define as_if 61 #define as_cond_statements 62 #define as_condition 63 #define as_case 64 #define as_case_statements 65 #define as_loop 66 #define as_while 67 #define as_for 68 #define as_forrev 69 #define as_block 70 #define as_exit 71 #define as_return 72 #define as_goto 73 #define as_subprogram_decl 74 #define as_procedure 75 #define as_function 76 #define as_operator 77 #define as_formal 78 #define as_mode 79 #define as_subprogram 80 #define as_call 81 #define as_package_spec 82 #define as_package_body 83 #define as_private_decl 84 #define as_use 85 #define as_rename_obj 86 #define as_rename_ex 87 #define as_rename_pack 88 #define as_rename_sub 89 #define as_task_spec 90 #define as_task_type_spec 91 #define as_task 92 #define as_entry 93 #define as_entry_family 94 #define as_accept 95 #define as_delay 96 #define as_selective_wait 97 #define as_guard 98 #define as_accept_alt 99 #define as_delay_alt 100 #define as_terminate_alt 101 #define as_conditional_entry_call 102 #define as_timed_entry_call 103 #define as_abort 104 #define as_unit 105 #define as_with_use_list 106 #define as_with 107 #define as_subprogram_stub 108 #define as_package_stub 109 #define as_task_stub 110 #define as_separate 111 #define as_exception 112 #define as_except_decl 113 #define as_handler 114 #define as_others 115 #define as_raise 116 #define as_generic_function 117 #define as_generic_procedure 118 #define as_generic_package 119 #define as_generic_formals 120 #define as_generic_obj 121 #define as_generic_type 122 #define as_gen_priv_type 123 #define as_generic_subp 124 #define as_generic 125 #define as_package_instance 126 #define as_function_instance 127 #define as_procedure_instance 128 #define as_instance 129 #define as_length_clause 130 #define as_enum_rep_clause 131 #define as_rec_rep_clause 132 #define as_compon_clause 133 #define as_address_clause 134 #define as_any_op 135 #define as_opt 136 #define as_list 137 #define as_range_expression 138 #define as_arg_assoc_list 139 #define as_private 140 #define as_limited_private 141 #define as_code 142 #define as_line_no 143 #define as_index 144 #define as_slice 145 #define as_number 146 #define as_convert 147 #define as_entry_name 148 #define as_array_aggregate 149 #define as_record_aggregate 150 #define as_ecall 151 #define as_call_or_index 152 #define as_ivalue 153 #define as_qual_range 154 #define as_qual_index 155 #define as_qual_discr 156 #define as_qual_arange 157 #define as_qual_alength 158 #define as_qual_adiscr 159 #define as_qual_aindex 160 #define as_check_bounds 161 #define as_discr_ref 162 #define as_row 163 #define as_current_task 164 /* new node added aug 84 by banner */ #define as_check_discr 165 #define as_end 166 #define as_terminate 167 #define as_exception_accept 168 #define as_test_exception 169 #define as_create_task 170 #define as_predef 171 #define as_deleted 172 #define as_insert 173 #define as_arg_convert 174 #define as_end_activation 175 #define as_activate_spec 176 #define as_delayed_type 177 #define as_qual_sub 178 #define as_static_comp 179 #define as_array_ivalue 180 #define as_record_ivalue 181 #define as_expanded 182 #define as_choices 183 #define as_init_call 184 #define as_type_and_value 185 #define as_discard 186 /* as_unread used by C version to indicate node not yet read in. * The SETL test N_KIND(node)=om thus becomes * N_KIND(node)==as_unread in C */ #define as_unread 187 #define as_string_ivalue 188 #define as_instance_tuple 189 #define as_entry_family_name 190 /* The next two kinds are used internally only by astread */ #define as_astend 191 #define as_astnull 192 /* aggregate_list added 9-15-86 so can overlay n_ast3 and n_unq (DS) */ #define as_aggregate_list 193 #define as_interfaced 194 #define as_record_choice 195 #define as_subprogram_decl_tr 196 #define as_subprogram_tr 197 #define as_subprogram_stub_tr 198 #define as_rename_sub_tr 199 /* PARSE TREE ---------- This section describes changes and additions to the parse tree (nodes) in the C version: AS_ATTRIBUTE ------------ In the SETL version a node of this kind has as its first subnode (AST1) a name-type node with a string giving the attribute name. The SETL version alters this string by adding a prefix of "T_" or "O_" to indicate attributes for types and objects, respectively. In C this is done by storing a numeric attribute code in the N_VAL field for the as_attribute node. The attribute codes are chosen so that where "T_" and "O_" codes are needed, the codes are in the order base code, then "O_" code and then "T_" code. The procedure attribute_str() maps numeric codes to their string equivalents. The original subnode is left intact; all manipulation of codes is done with the numeric codes. The codes are defined in header file attr.h: */ /* SYMBOL TABLE ------------ Let us now continue out the description of the symbol table entries. Entries whose nature ends in '_SPEC' have the same form as the corresponding nature with '_SPEC' removed. The initial entry with '_SPEC' is made when the specification is seen; the other is made when the body is encountered. For a field, 'not used' means the field is defined but not used, so that its contents are not relevant (they will, and should, be NULL, the null pointer); */ /* * Define nature codes. To assist conversion of SETL source each * code is defined in both upper and lower case. */ #define na_op 1 #define na_un_op 2 #define na_attribute 3 #define na_obj 4 #define na_constant 5 #define na_type 6 #define na_subtype 7 #define na_array 8 #define na_record 9 #define na_enum 10 #define na_literal 11 #define na_access 12 #define na_aggregate 13 #define na_block 14 #define na_procedure_spec 15 #define na_function_spec 16 #define na_procedure 17 #define na_function 18 #define na_in 19 #define na_inout 20 #define na_out 21 #define na_package_spec 22 #define na_package 23 #define na_task_type 24 #define na_task_type_spec 25 #define na_task_obj 26 #define na_task_obj_spec 27 #define na_entry 28 #define na_entry_family 29 #define na_entry_former 30 #define na_generic_procedure_spec 31 #define na_generic_function_spec 32 #define na_generic_package_spec 33 #define na_generic_procedure 34 #define na_generic_function 35 #define na_generic_package 36 #define na_exception 37 #define na_private_part 38 #define na_void 39 #define na_null 40 #define na_discriminant 41 #define na_field 42 #define na_label 43 #define na_generic_part 44 /* na_subprog, na_body and na_task referenced in sem0.c, addedm15-may for debugging */ #define na_subprog 45 #define na_body 46 #define na_task 47 /* added for use with symbol_op (cf find_qual (8) and result_types(4))*/ #define na_task_body 48 /* NA_ACCESS: ---------- denotes an access type. type_of: map of an access type is itself. The macro designated_type(access_type) retrieves the unique name of the type of designated objects of the access type. this macro is also available on subtypes of access types. signature: designated_type (as noted above) overloads: not used type_attr: misc_type_attributes NA_AGGREGATE: ------------- Every array and record declaration introduces an aggre- gate into the environment, i.e. one possible meaning for any aggregate expression appearing in the code. In order to disambiguate such an expression, it is convenient to view aggregates as overloaded constructs. Each aggregate entry is uniquely characterized by its type (held in type_of). Note that constraining an array does not introduce a new aggregate. Properly speaking, it is only a new sequence type (an unbounded array) which introduces a new aggregate. type_of: (unique characterization of) type signature: not used overloads: set of other aggregate types (symbols) in same scope NA_ARRAY: --------- This denotes an entry for an array type. This entry is built when an array declaration is processed, and also when a task family or an entry family is elaborated. An array denotes an unconstrained type. type_of: unique name of parent sequence. This is invariably a generated name. For the predef defined STRING type, the parent sequence is STRING itself. all sequences are arrays whose type is themselves. signature: the pair : [ index-types, component-type] where index_types is a tuple of type names. overloads: type_attr: misc_type_attributes NA_ATTRIBUTE: ------------- An entry of this nature denotes a predefined attribute of the language. Many attributes are generic : they apply to classes of types. type_of: either a predefined type, (INTEGER or STRING) or the marker 'overloaded'. This indicates that the type returned by the attribute depends on the type of its arguments. signature,overloads: not used NA_BLOCK: --------- There are three kinds of blocks, distinguished by their overloads values: block, loop or handler. A loop has alias defined, so all entries need full length symbol table entry. In SETL, the type_of field is used to hold the block type. In C we use special codes stored in the overloads field. Blocks are not used for separate compilation, so it is an error if one occurs. signature: miscellaneous information. unused externally overloads: block type, one of the following */ #define BLOCK_BLOCK 0 #define BLOCK_LOOP 1 #define BLOCK_HANDLER 2 /* Note that the use of overloads for NA_BLOCK is internal to adasem. This information need not be written out externally. NA_CONSTANT: ------------ Denotes a symbol table slot for a user-defined constant or for the result of evaluating a static expression. type_of: specified type for constant. signature: Node for expression overloads: not used NA_ENTRY: --------- denotes an entry object. type_of: points to an entry former. signature: the formalslist, and is identical to that of a subprogram. overloads: identical to that of a subprogram NA_ENTRY_FAMILY: ---------------- denotes an entry family object. This is a non-over- loadable entity, constructed as an array of entries. type_of: anonymous array type whose component is an entry-type. signature: formal parameters overloads: not used NA_ENTRY_FORMER: ---------------- denotes the type of an entry object. Calls to entries (rendez-vous) have the same syntax as procedure calls, and named and default parameters are also present. type_of: task type in which it is declared (same as scope_of, in fact) * ?? As for procedures, the TYPE_OF an entry is the marker 'none'. signature: the list of formals for the entry, and has the same format as that of a procedure. An entry call does not return a value. overloads: not used NA_ENUM: -------- Denotes an enumeration type. type_of: parent type, if derived, or the typename itself for BOOLEAN, CHARACTER, and non-derived enumerations. signature: the range for the enumeration type in the format: ['range', low, high] where low and high are the integer values for the first and last enumeration literals of the type. overloads: a map, from the literals to their POS in the type. By default the range of this map is dense and 0-origined, but it can be modified by a representation clause. This map is accessed by the macro literal_map. It is represented in C as a tuple with element i being pointer to string for literal, element i+1 being value of literal map for the string. NA_EXCEPTION: ------------- NA_FUNCTION: ------------ identical to NA_PROCEDURE. type_of: the type which it returns. signature: same as for procedure. overloads: same as for procedure. NA_FUNCTION_SPEC: ----------------- An entry is given this nature after the corresponding subprogram specification is processed. When the subprogram body is encountered the specification is examined again, and if it coincides with an existing symbol table slot, the _spec tag is removed. signature: same as for procedure overloads: same as for procedure NA_GENERIC_FUNCTION_SPEC: NA_GENERIC_PROCEDURE_SPEC: NA_GENERIC_PACKAGE: --------------------- Generic objects have the same organization as their non-generic counterparts. The SIGNATURE of a generic entity has, in addition to its non-generic purpose, the further role of holding the list of generic parameters and the body of the generic object. type_of: unused. signature: The signature is a tuple with two elements: 1) gen_list, a tuple of pairs, each consisting of a symbol and a node. 2) form_list, a tuple of symbols overloads: not used NA_GENERIC_PACKAGE: --------------------- type_of: unused. signature: A tuple with three elements: 1) gen_list, a tuple of pairs, as for generic_procedure 2) decl_node, a node. 3) opt_priv_node, a node. overloads: not used NA_IN: ------ type_of: type of formal parameter signature: initial value (node) if there is one, else null node. overloads: not used. NA_INOUT: --------- type_of: type of formal parameter signature: initial value (node) if there is one, else null node. overloads: not used. NA_LABEL: --------- An entry for a label requires only a single bit, indicating whether the label has been defined. NA_LITERAL: ----------- an entry of this nature denotes an enumeration literal. type_of: entry holds the unique name of the enumeration type to which it belongs. signature: is the same as that of a parameterless function, i.e. []. overloads: same meaning as for operators NA_NULL ------- TBSL (almost certainly dead - does not occur in SETL) NA_OBJ: ------- Denotes a symbol table slot for a program variable. type_of: a unique name, namely that of a user-defined or internally generated type. signature: holds the initial value for fields of a record. Otherwise, it is not used. overloads: not used NA_OP: ------ An entry with this nature denotes a predefined operator in the language. type_of: for an operator is usually a generic marker which indicates the family of types to which this operator can apply. This marker obviates the need to introduce a separate operator for each type derived from the predefined types. These generic type marks are also given a symbol table slot. The following are used : 1) 'universal_integer' : for all arithmetic operators which apply to integers. 2) 'universal_real' : ditto for floating and fixed types. 3) 'array_type' : used for the (idiosyncratic) concatenation operator. 4) 'order_type' : for comparison operators. 5) 'boolean type' : for boolean operators. This simplifies the handling of arrays of booleans and derived(!) booleans. 6) 'string_type' : for string literals, which overload any array of characters. We represent these in C as values of following global variables: gtm_universal_integer gtm_universal_real gtm_array_type gtm_order_type gtm_boolean_type gtm_string_type */ /* signature: follows the same format as the one for a subprogram. The canonical parameter names LEFT and RIGHT are not stored explicitly but are known to an internal procedure order_arg_list in those rare cases when an operator is used in functional notation with named associatoins. Note that this need not appear when writing externally. overloads: has the same meaning as for subprograms. Given that operators are entered into the symbol table during initialization, at the outermost scope, most ope- rators in fact do not overload anything. For arithmetic ope- rators however, we give different unique names to the integer and floating point version of them. */ /* NA_OUT: ------- type_of: type of formal parameter signature: not used Note however that as a byproduct of initializing nodes for arguments, the signature may be a node, namely OPT_NODE. NA_PACKAGE: ----------- designates a package or a package specification. type_of: not used signature: not used overloads: pointer to private symbol table with private declarations of the package. */ /* * Private Declarations * * Access to entities declared in a package is through its declared map * (when inside the package) or through its visible map, which holds only the * declarations in the visible part of the package specification. * The private declarations for the package are stored as a map * in its overloads entry, and installed in the symbol table on * entry to the package body (i.e. when compiling the body). * The macro private_decls(pack) accesses the private decls, * which are installed when compiling the body of the package, * and removed after the body. * Access to entities in the specification is * via the mapping VISIBLE. The domain of visible(p) is a sub- * set of that of declared(p). * The private declarations of the package are stored in a * separate structure, private_decls(p). These private decla- * rations are installed in the global symbol table when the * body of the package is compiled, and removed afterwards, so * that only the visible attributes remain accessible outside * of the package. * * Note that in the C version, VISIBLE is not a separate map but * an attribute associated with each entry in a declared map. * alias: ?? * -- pdecl.doc * In the C version, private_decls is represented as a pointer to a tuple. * The pointer is needed so we can expand the tuple as necessary. * Each odd entry is a symbol table pointer and is followed by * a symbol table pointer to a symbol table information saved when * the declaration was installed (by private_decls_put). * * private_decls is accessed by the * following procedures: * * Private_declarations pdecl; * Symbol s1,s2; * * private_decls_new(n) - allocate new tuple for n symbols * * private_decls_put(pdecl,s1); * Install s1 in pdecl if not yet present. Copy current symbol * table field values for s1 into saved entry in pdecl. * * Symbol private_decls_get(pdecl,s1) * Return entry in pdecl for s1 if present, (Symbol)0 otherwise. * * private_decls_swap(s1,s2) * swaps symbol table blocks for s1 and s2. */ typedef struct Private_declarations_s { Tuple private_declarations_tuple; } Private_declarations_s; typedef Private_declarations_s *Private_declarations; typedef struct Forprivate_decls { Tuple forprivate_tup; int forprivate_i; int forprivate_n; } Forprivate_decls; #define FORPRIVATE_DECLS(s1,s2,pd,fp) \ fp.forprivate_tup = (Tuple) (pd)->private_declarations_tuple; \ fp.forprivate_n = tup_size(fp.forprivate_tup) ; \ for (fp.forprivate_i=1; fp.forprivate_i<=fp.forprivate_n;) { \ s1 = (Symbol) fp.forprivate_tup[fp.forprivate_i++]; \ s2 = (Symbol) fp.forprivate_tup[fp.forprivate_i++]; #define ENDFORPRIVATE_DECLS(fp) } /* NA_PACKAGE_SPEC: ---------------- overloads: not used signature: not used NA_PRIVATE_PART: ---------------- * 'private_part': for technical reasons, it is convenient to regard the * private part of a package specification as a scope in its own * right. This simplifies the segregation of private declarations * from visible ones, and the detection of sundry semantic errrors * overloads: not used signature: not used NA_PROCEDURE: ------------- Denotes a symbol table slot for a procedure subprogram. type_of: holds the marker NULL (ie., a null pointer). signature: describes the formal parameters of the procedure. It is a tuple of symbol table pointers to formal parameter entries (of type na_in, na_out, or na_inout). overloads: the set of procedures with the same name (source identifier, that is) which appear in the current lexical scope, and which are not hidden by the current entry. This set is built when the procedure specification is processed (See procedure : chain- overloads in the semantic actions) and is central to the overload resolution mechanism. overloads(id) always includes -id- itself. If the procedure is the only visible instance of the name, (i.e. it overloads no other identifier) then the overloads entry is the singleton set containing its own name. alias: ?? declared: ?? NA_PROCEDURE_SPEC: ------------------ An entry is given this nature after the corresponding subprogram specification is processed. When the subprogram body is encountered the specification is examined again, and if it coincides with an existing symbol table slot, the _spec tag is removed. Apart from generic objects, all remaining types of entries denote type entities. signature: same as for procedure. overloads: same as for procedure. Other fields same as procedure spec. NA_TYPE: -------- This denotes an entry for a new, or derived type from one of the numeric types 'INTEGER', 'FLOAT', or '$FIXED', or from another numeric type, or the error type 'any'. All such entries have all symbol fields, including alias, defined. type_of: unique name of the type from which the current one is derived. signature: the constraint (see 'subtype' below). Private and incomplete type declarations produce type entries whose type_of field is 'private','limited private' or 'incomplete' respectively. overloads: The map 'misc_type_attributes' collects various useful predi- cates on types, in particular composite and generic types. This information is kept in the field type_attr. In C, misc_type_attributes can be represented as a byte, assigning a bit for each of the necessary attributes: flag TA_ISPRIVATE is used only when writing out files in C format to flag the case of a private or limited private type. This is determined in SETL by a symbol having a TYPE_OF corresponding to 'private' or 'limited private', and in C by TYPE_OF being symbol_private or symbol_limited private. For such symbols we set TA_ISPRIVATE when writing the symbol to indicate that the signature has the same format as for a record (na_record), even though the nature is not necessarily na_record. */ #define TA_ISPRIVATE 1 #define TA_INCOMPLETE 2 #define TA_LIMITED 4 #define TA_LIMITED_PRIVATE 8 #define TA_PRIVATE 16 #define TA_OUT 32 /* GENERIC corresponds to '$generic' in SETL */ #define TA_GENERIC 64 /* CONSTRAIN corresponds to '$constrain' in SETL */ #define TA_CONSTRAIN 128 /* NA_RECORD: ---------- Denotes a record type. type_of: the type name itself. signature: a tuple formatted as follows (in SETL): [ [invariant_part, variant_part], discriminant_list , declared_map, discr_decl_tree ] the discriminant list is [] for a record without discrimants. In the above, invariant_part and variant_part refer to AST nodes. In C version, signature is quintuple as follows: [ invariant_part, variant_part, discriminant_list, declared_map, discr_decl_tree] invariant_part, variant_part and discr_decl_tree are nodes. discriminant list is tuple of symbols. The declared_map is the (declared) map from selector names to the unique names which have been assigned to them. (NOTE: This is redundant if we keep declared for record) Note that this is initialized in process_discr in ch 3. overloads: not used (was misc_type_attributes) type_attr: misc_type_attributes NA_SUBTYPE: ----------- This denotes an entry for a declared subtype, or for a range descriptor which has been promoted to a subtype. type_of: unique name of the type of which the current one is a subtype, i.e. its base type. signature: the constraint which defines the subtype. For numeric types, the constraint has one of the following formats : ['range' low, high] for integer subtypes. ['digits', low, high, digits] for floating types. ['delta', low, high, delta,small] for fixed types. In each case, -low- and -high- are the bounds of the subytpe. digits is a SETL integer, delta is a SETL floating point. Both of them have their standard Ada meaning . Components other than the first are nodes. Add CONSTRAINT_DISCR (ds 14 aug) to correspond to 'discr' case in SETL. For this type numeric_constraint_value[0] will be tuple with list of discriminant values. This will be referenced as numeric_constraint_discr. Add CONSTRAINT_ARRRY (ds 1 jan 84) to correspond to constrained aray case. For this type the SETL signature has the same form as form an array type - a pair whose first component is a tuple of symbols and whose second component is a symbol. In C we keep the signature as a tuple, using CONSTRAINT_ARRAY only when writing and reading the ais file in c format. Add CONSTRAINT_ACCESS for subtypes of access types. In SETL, this was just the designated type. Here, for uniformity of subtypes, we keep it as a constraint (a tuple whose first element is CONSTRAINT_ACCESS, and second is the designated type). The macro designated_type has been changed to reflect this. In C we represent the constraint as the following structure: */ #define CONSTRAINT_RANGE 0 #define CONSTRAINT_DIGITS 1 #define CONSTRAINT_DELTA 2 #define CONSTRAINT_DISCR 3 #define CONSTRAINT_ARRAY 4 #define CONSTRAINT_ACCESS 6 #define numeric_constraint_kind(p) p[1] #define numeric_constraint_low(p) p[2] #define numeric_constraint_high(p) p[3] #define numeric_constraint_digits(p) p[4] #define numeric_constraint_delta(p) p[4] #define numeric_constraint_small(p) p[5] #define numeric_constraint_discr(p) p[2] /* new instances of above allocated by new_constraint(type) */ /* for record subtypes, the constraint as the format : ['discr', list of discriminant values] For records we also need the discriminant map. In SETL this is a map from symbols to nodes. The only operations used are initialization, assignment, retrieval and interation, so in C we keep this map as a tuple with successive pairs of elements representing the domain (Symbol) and range (Node) values. The following procedures are used to access the map: Node discr_map_get(dmap, sym); Symbol sym; retrieve value of map for symbol sym. Tuple discr_map_put(dmap,sym,nod) Symbol sym; Node nod; set value of map dmap for symbol sym to node nod. Note that symbols corresponding to private types have a signature of the same form as for a record; see discussion of TA_ISPRIVATE above. overloads: field present, but not used. declared: not used alias: yes, for all types, the root type */ /* NA_TASK_OBJ ----------- Denotes a task type. A task object is either an object of task type, or (in the case of a task family) an array of objects of task type. A task contains declarations for entries. Access is obtained by means of the -declared- map for the task. type_of: task type signature: not used overloads: not used declared: defined NA_TASK_OBJ_SPEC ---------------- NA_TASK_TYPE: ------------- Same as NA_TASK. NA_TASK_TYPE_SPEC ----------------- Same as NA_TASK_TYPE. NA_VOID ------- TBSL A symbol of this type is generated for the symbol "$used". It has as signature a tuple of symbols: signature: tuple of symbols corresponding to the macro use_declarations in SETL. */ /* DECLARED definitions and declarations In the SETL version, program identifiers are related to unique names using a global map DECLARED. Every lexical scope corresponds to an entry in DECLARED. In turn, this entry is a map from source identifiers to their unique names in that scope. DECLARED : scope-name -> (map : identifier -> unique-name ) The scope manipulation routines and the name resolution routines revolve around DECLARED. In the SETL code, access is typically via DECLARED(SCOPE_OF(name))(identifier) In C, each symbol table entry has a field scope_of that gives the symbol table entry of the scope for the entry. The declared field of the entry for the scope is a pointer to the declared map from identifiers to symbol table entries. Since access to the declared map involves a hashed lookup, we will access the map using procedures (or macros that ultimately invoke procedures). For example, SETL code DECLARED(s)(id) will become dcl_get(s,id) for retrievial, or dcl_put(s,id,sym) or dcl_put_vis(s,id,sym,vis) to explicitly set visibility: dcl_put(s,id,sym) is same as dcl_put_vis(s,id,sym,TRUE); for assignment,where s will often have one of the forms scope_name or scope_of(name) Our intent is that, as much as possible, the existing SETL code can be translated to C using dcl_get and dcl_put, using a 'paging' algorithm to bring in names from other compilation units, possibly on library files, in a manner transparent to most of the code. This paging process will involve marking names as one of the following: installed symbol table entry available not installed name and location known, but symbol table entry not yet installed. An identifer that is not installed will have an indication of the compilation unit and 'file address' of the information needed to install the name. Installation consists of converting the external 'file address' to an internal pointer. Note that map VISIBLE maintained in the SETL version is represented in the C version by the attribute is_visible associated with an entry in a declared map. */ typedef struct Declaredmap_s { short dmap_curlen; /* current number of elements */ short dmap_maxlen; /* maximum number of elements */ struct Dment *dmap_table; /* pointer to entry list */ } Declaredmap_s; /* Each entry in the hash table is a pointer to a declared map entry */ typedef struct Dment { struct { #ifdef IBM_PC /* avoid overlaying on PC, due to bad generated code ds 4-26-86 */ unsigned dment_idnum ; /* source identifier number */ unsigned dment_visible ; /* non-zero if visible */ #else unsigned dment_idnum : 15; /* source identifier number */ unsigned dment_visible : 1; /* non-zero if visible */ #endif } dment_i; Symbol dment_symbol; /* symbol table pointer */ } Dment; /* dment_id is the character string for the source identifier. dment_hash is the hash code of the string. This is kept so that the hash table size can be changed as the map grows without avoiding the need to rehash all the strings. dment_next is a pointer to the next entry on the hash chain, or NULL for the end of a chain. dment_symbol is a symbol table pointer for an installed name, or NULL if the name not yet installed. dment_install is a boolean, initially FALSE, that is set when the identifier has been installed in the symbol table. Note that it may be possible to avoid the use of this bit and just say a symbol is not installed if dment_symbol is NULL. dment_file and dment_fileoff are the file identifier and file offset; the details of these have to be worked out. */ /* A tentative list of the procedures that are used to access a declared map is as follows: Declaredmap dcl_new(nh) int nh; initializes declared map with nh hash headers (zero may be given, in which case default values are taken). dcl_free(dmap) Declaredmap dmap; free space occupied by declared map dmap. Symbol dcl_get(dmap,s) Declaredmap dmap; char * s; return current value of declared map dmap for string s, or NULL if map not defined for s. Symbol dcl_get_vis(dmap,s) Declaredmap dmap; char * s; return current value of declared map dmap for string s, or NULL if map not defined for s. This returns current value only if it is visible. dcl_put_vis(dmap,str,sym,vis) Declaredmap dmap; char *str; Symbol sym; int vis; define value of declared map dmap for string s to be symbol s. This sets visible attribute for entered sym to vis. dcl_put(dmap,str,sym) ... same as dcl_put_vis(dmap,str,sym,TRUE). dcl_undef(dmap,str) Declaredmap dmap, char *str; undefine declared map dmap for string s; i.e., remove entry from map if present. The form of an iteration over a declared map is: FORDECLARED(str,sym, dmap, dmapiv) which iterates over declared map dmap using dmapiv to control iteration. For each pass, str is set to identifier, sym to symbol table pointer. Fordeclared dmapiv; Declaredmap dmap; char *str; Symbol sym; FORDECLARED(str,sym,dmap,dmapiv) ENDFORDECLARED(dmapiv) Within a FORDECLARED iteration, the macro IS_VISIBLE may used to get visibility status of the current variable, by IS_VISIBLE(dmapiv). ISVISIBLE may be used on the left hand side to change visibility, IS_VISIBLE(dmapiv) = 0 . */ typedef struct Fordeclared { Declaredmap fordeclared_map; unsigned short fordeclared_i; unsigned short fordeclared_n; struct Dment *fordeclared_dment; } Fordeclared; extern char *dstrings; #define FORDECLARED(str,sym,dmap,iv) \ iv.fordeclared_map = dmap; \ iv.fordeclared_n = iv.fordeclared_map->dmap_curlen;\ iv.fordeclared_dment = iv.fordeclared_map->dmap_table; \ for (iv.fordeclared_i=0; iv.fordeclared_i<iv.fordeclared_n; \ (iv.fordeclared_i++,iv.fordeclared_dment++)) { \ str = dstrings + iv.fordeclared_dment->dment_i.dment_idnum;\ sym = iv.fordeclared_dment->dment_symbol;\ #define ENDFORDECLARED(iv) } #define IS_VISIBLE(iv) iv.fordeclared_dment->dment_i.dment_visible #define SETDECLAREDVISIBLE(iv,n) \ IS_VISIBLE(iv) = n; /* The form of an iteration over a declared map, visiting only the visible entries, is: FORVISIBLE(str,sym, dmap, dmapiv) which iterates over declared map dmap using dmapiv to control iteration, processing only visible entries. For each pass, str is set to identifier, sym to symbol table pointer. Fordeclared dmapiv; Declaredmap dmap; char *str; Symbol sym; FORVISIBLE(str,sym,dmap,dmapiv) ENDFORVISIBLE(dmapiv) */ #define FORVISIBLE(str,sym,dmap,iv) \ FORDECLARED(str,sym,dmap,iv) \ if (IS_VISIBLE(iv)==0) continue; #define ENDFORVISIBLE(iv) } /* Declarations for use by evalstat. * Values returned by evalstat are as follows: */ typedef struct Const_s *Const; typedef struct Const_s { int const_kind; union { int const_int; int *const_uint; double const_real; char *const_str; Rational const_rat; long const_fixed; } const_value; } Const_s; #ifdef IVALUE /* Const should be ivalue. Define both until Sem changed */ typedef struct Ivalue_s *Ivalue; typedef struct Ivalue_s { int ivalue_kind; union { int ivalue_int; int *ivalue_uint; double ivalue_real; char *ivalue_str; Rational ivalue_rat; long ivalue_fixed; } ivalue_value; } Ivalue_s; #endif /* kinds for const_kind */ #define CONST_OM 0 #define CONST_INT 1 #define CONST_REAL 2 #define CONST_STR 3 #define CONST_RAT 4 #define CONST_CONSTRAINT_ERROR 5 #define CONST_UINT 6 #define CONST_FIXED 7 /* predicates for const_kind */ #define is_const_om(c) ((c)->const_kind == CONST_OM) #define is_const_int(c) ((c)->const_kind == CONST_INT) #define is_const_real(c) ((c)->const_kind == CONST_REAL) #define is_const_str(c) ((c)->const_kind == CONST_STR) #define is_const_rat(c) ((c)->const_kind == CONST_RAT) #define is_const_constraint_error(c) ((c)->const_kind == CONST_CONSTRAINT_ERROR) #define is_const_uint(c) ((c)->const_kind == CONST_UINT) #define is_const_fixed(c) ((c)->const_kind == CONST_FIXED) /* kinds for const_kind */ #ifdef IVALUE /* CONST_ should be IVALUE_. Define both until convert Sem */ #define IVALUE_OM 0 #define IVALUE_INT 1 #define IVALUE_REAL 2 #define IVALUE_STR 3 #define IVALUE_RAT 4 #define IVALUE_CONSTRAINT_ERROR 5 #define IVALUE_UINT 6 #define IVALUE_FIXED 7 #endif /* Macros to access values of a Const */ #define INTV(op) (op)->const_value.const_int #define UINTV(op) (op)->const_value.const_uint #define REALV(op) (op)->const_value.const_real #define RATV(op) (op)->const_value.const_rat #define FIXEDV(op) (op)->const_value.const_fixed /* The SETL maps rename_map and type_map are maps from symbols * to symbols. In the C version these are kept as type Symbolmap, * currently represented using a tuple map, i.e., a tuple with * successive pairs of domain values followed by the range value. */ typedef struct Symbolmap_s { Tuple symbolmap_tuple; } Symbolmap_s; typedef Symbolmap_s *Symbolmap; /* and is accessed using the following procedures (defined in 12.c): */ /* Iteration over symbolmaps is done as follows: */ typedef struct Forsymbol { Tuple forsymbolmap_tuple; int forsymbolmap_i; int forsymbolmap_n; } Forsymbol; #define FORSYMBOL(s1,s2,pd,fp) \ fp.forsymbolmap_tuple = (Tuple) (pd)->symbolmap_tuple; \ fp.forsymbolmap_n = tup_size(fp.forsymbolmap_tuple) ; \ for (fp.forsymbolmap_i=1; fp.forsymbolmap_i<=fp.forsymbolmap_n;) { \ s1 = (Symbol) fp.forsymbolmap_tuple[fp.forsymbolmap_i++]; \ s2 = (Symbol) fp.forsymbolmap_tuple[fp.forsymbolmap_i++]; #define ENDFORSYMBOL(fp) } typedef struct Nodemap_s { Tuple nodemap_tuple; } Nodemap_s; typedef Nodemap_s *Nodemap; /* *T+ UTILITY MACROS FOR SEMANTIC MODULES. *S+ Global macros for the use of all modules: */ /* We define the utility macros in lower case since that is the * case used in the SETL source. */ /* macro find; assert exists endm; */ #define find ??? /* macro top(x); x(#x) endm; */ /* *S+ Macros for manipulation of the AST. */ /*macro is_empty(node); (node=[]) endm ; */ #define is_empty(node) ((int)node[0] == 0) /* macro attribute_name(node) ; ;N_VAL(N_AST(node)(1)) endm; */ /* In C this returns integer attribute code stored in ast1 node of node */ /* this returns integer not string */ #define attribute_name(node) N_VAL(N_AST1(node)) /* In C we define attribute_kind to return integer attribute code */ #define attribute_kind(node) N_VAL(N_AST1((node))) /* macro SYMBTAB(name); [NATURE(name),TYPE_OF(name),SIGNATURE(name),OVERLOADS(name)] endm; */ /* * For library and stub manipulation, the scope_of map must also * be saved and restored. The following macro is usedfor that purpose. */ /* macro SYMBTABF(name) ; [NATURE(name), TYPE_OF(name), SIGNATURE(name), OVERLOADS(name), SCOPE_OF(name), ALIAS(name) ] endm ; */ /* * The following is a lexical macro which converts a double-quoted * character which appears as a designator, into a one-character * string. This distinction is imposed by the need to differentiate * the literal a from the character "a" when they appear in an * enumeration literal, but nowhere else. */ /*macro char_to_name(desig) ; (if # desig = 3 and desig(1) = '"' and desig(3) = '"' then desig(2) else desig end) endm ; */ /* *S+ Macros for accessing type information. */ /* macro root_type(type_mark) ; ALIAS(type_mark) endm ; $$ES68 */ #define root_type(type_mark) ((type_mark)->alias) /* macro index_types(array_type) ; signature(array_type)(1) endm ; */ #define index_types(array_type) ((Tuple) ((array_type)->signature)[1]) /* * macro component_type(array_type) ; * signature(array_type)(2) * endm ; */ #define component_type(array_type) ((Symbol) ((array_type)->signature)[2]) /* macro no_dimensions(array_type) ; (#index_types(array_type)) endm ; */ /* macro index_type(array_type) ; (index_types(array_type)(1)) endm ; */ #define index_type(array_type) ((Tuple) (index_types(array_type))[1]) /* macro literal_map(enumeration) ; overloads(enumeration) endm ; */ #define literal_map(enumeration) ((enumeration)->overloads) /* macro record_declarations(record); signature(record) endm ; */ #define record_declarations(record) ((record)->signature) /* In C, will have the following reference a single quadruple ds 14 aug */ /* macro all_components(record) ; signature(record)(1) endm ; */ #ifdef TBSN -- all_components should be dead ds 14 aug #define all_components(record) (((record)->signature)[1]) #endif /* the macros for records that access components of the signature * cannot be written on the left hand side due to use of (Tuple) * cast to get indexing right. The left hand case is done by * explicit assignment; see record_decl(3b.c) and process_discr (3b.c). * ds 31-dec-84 */ /* macro discriminant_list(record) ; signature(record)(2) endm ; */ #define discriminant_list(record) ((Tuple) ((record)->signature)[3]) /* macro invariant_part(record) ; all_components(record)(1) endm ;*/ #define invariant_part(record) ((Tuple) ((record)->signature)[1]) /* macro variant_part(record) ; all_components(record)(2) endm ; */ #define variant_part(record) ((Tuple) ((record)->signature)[2]) /* macro declared_components(record); signature(record)(3) endm ; */ /* The value returned by declared_components is a declared map */ #define declared_components(record) ((Tuple) ((record)->signature)[4]) /* TBSLshould this be four or five ds 14 aug */ /* macro discr_decl_tree(record); signature(record)(4) endm ; */ #define discr_decl_tree(record) ((Tuple) ((record)->signature)[5]) /* macro has_discriminants(record) ; (discriminant_list(root_type(record)) /= []) endm ; */ #define has_discriminants(record) \ (discriminant_list(root_type(record)) != (Tuple) 0 && \ tup_size(discriminant_list(root_type(record))) != 0) /* macro designated_type(access_type) ; (signature(access_type)) endm ; */ /* In the C version, the signature of a subtype of an access type is a * tuple whose first element is CONSTRAINT_ACCESS, and whose second * is the designated type symbol */ #define designated_type(access_type) \ (((NATURE(access_type)==na_subtype) \ ? (Symbol)((access_type)->signature)[2] \ : (Symbol)((access_type)->signature) )) /* macro is_range_attribute(e) ; (is_tuple(e) and e(1) = '''' and e(2) = 'RANGE') endm ; */ /* macro label_status(lab) ; signature(lab) endm ; */ #define label_status(lab) ((lab)->signature) /* default_expr has a node as its value (ds 15 may) */ /* macro default_expr(nam) ; signature(nam) endm ;*/ #define default_expr(nam) ((nam)->signature) #define formal_decl_tree(proc_name) ((proc_name)->init_proc) /* formal_decl_tree stores the formals subtree for subsequent conformance * checks. It is defined for a subprogram_spec only, and therefore * overlays the init_proc fields used during code generation . */ /* *S+ Macros for scope manipulation. */ /* *$BB5 */ /* macro init_scope_info ; ['STANDARD#0', '', ['STANDARD#0', 'UNMENTIONABLE#0'], [], ['ASCII'], '' ] endm; */ /* macro is_comp_unit ; (#scope_st = 0) endm ; macro scope_info ; [scope_name, prefix, open_scopes, used_mods, vis_mods, suffix] endm ; */ /* make c macro name upper case since no arguments*/ #define IS_COMP_UNIT (tup_size(scope_st)==0) /*S+ Miscellaneous type predicates.*/ /*macro is_access(name) ; (NATURE(root_type(name)) = na_access) endm ; */ /* is_access now a procedure in the C version (smisc.c) */ /*macro is_identifier(name) ; is_string(name) endm ; */ /* Make this a procedure int is_identifier();*/ /*macro is_literal(name) ; nature(name) = 'literal' endm ; */ #define is_literal(name) (NATURE((name)) == na_literal) /*macro is_overloaded ; (is_set(name)) endm ; */ /* this is field N_OVERLOADED in C version */ /* macro is_constant(name) ; (is_identifier(name) and nature(name) = 'constant') endm ; */ /* In C is_identifier for symbol is just test that not null */ #define is_constant(name) ((name)!=(Symbol) 0 && NATURE((name))== na_constant) /*macro is_proc(proc_name) ; nature(proc_name) in {'procedure', 'function','generic_procedure', 'generic_function'} endm; */ /* #define is_proc(proc_name) ( (TMP := nature(proc_name) == na_procedure ? TRUE : (TMP == na_function || TMP == na_generic_procedure \ || TMP == na_generic function) ? TRUE : FALSE ) ) */ /*macro is_anonymous(typ) ; (original_name(typ) =='') endm ;*/ /* in C anonymous types are given an original name beginnning with & */ #define is_anonymous(typ) (*(ORIG_NAME(typ)) == '&') /*macro is_first_named_subtype(typ); (is_anonymous(base_type(typ))) endm;*/ /* macro is_array(typ) ; (nature(base_type(typ)) = 'array') endm ; */ #define is_array(typ) (NATURE(base_type(typ)) == na_array) /**$PK3 ES118 */ /* macro is_formal(id) ; (nature(id) in ['in', 'out', 'inout']) endm ; */ #define is_formal(id) (NATURE(id)==na_in || NATURE(id)==na_out \ || NATURE(id)==na_inout) /* macro is_generic(scope ; nat) ; ((nat:=nature(scope)) /= om and match(nat,'generic_') /= om) endm ; */ /* macro misc_type_attributes(type_mark) ; overloads(type_mark) endm; */ /* In C misc_type_attributes are kept in separate field */ #define misc_type_attributes(type_mark) TYPE_ATTR(type_mark) /* macro private_dependents(type_mark) ; $$ES132 misc_type_attributes(type_mark) endm; */ /* In C, private_dependents is kept in overloads field */ #define private_dependents(type_mark) OVERLOADS(type_mark) /* macro private_decls(package); overloads(package) endm; */ /* this yields a tuple ds 25 jul */ #define private_decls(package) OVERLOADS(package) /* *$ES48 macro use_declarations(package) ; signature(declared(package)('$used')) endm; */ /* Since use_declarations appears on left hand side - we avoid * defining a macro for it in C (at least for now) ds 4 aug */ #ifdef TBSN #define use_declarations(package) \ SIGNATURE(dcl_get(DECLARED(package),"$used")) #endif /* macro is_anonymous_task(name) ; (is_task_type(name) and 'task_type:' in original_name(name)) endm ; */ /* make a procedure in C (by not defining as macro!) */ /* macro is_integer_type(x) ; (x in {'INTEGER', 'SHORT_INTEGER', 'LONG_INTEGER', 'universal_integer'}) endm ; */ /* * The following macro is used by cross-reference making statements. * spans(node) is either [l_span] or [l_span, r_span] , * l_span := [line_no, col1, col2]. macro TO_XREF(name) ; if spans /= om and spans(current_node) /= om then XREF(name) := ( XREF(name) ? {} ) + { spans(current_node)(1)(1) } ; end if endm; */ #define TO_XREF(name) #define TO_ERRFILE(text) to_errfile(text) /* streq() in misc.c tests strings for equality */ #define errmsg_l(msg1,msg2,lrm,node) \ errmsg(strjoin(msg1,msg2),lrm,node) #define errmsg_l_id(msg1,msg2,name,lrm,node) \ errmsg_id(strjoin(msg1,msg2),name,lrm,node) #define errmsg_l_str(msg1,msg2,str,lrm,node) \ errmsg_str(strjoin(msg1,msg2),str,lrm,node) extern int N_DEFINED[]; /* power_of_2. In SETL this procedure returns a tuple. In C it sets * global variables whose names begin with power_of_2_. Also needed * are two constants corresponding to the SETL strings 'exact'and * 'approximate'. */ #define POWER_OF_2_EXACT 0 #define POWER_OF_2_APPROXIMATE 1 #ifdef GEN #include "ghdr.h" #endif #define TAG_RECORD 0 #define TAG_TASK 1 #define TAG_ACCESS 2 #define TAG_ARRAY 3 #define TAG_ENUM 4 #define TAG_FIXED 5 #define TAG_INT 6 #define TAG_FLOAT 7 #endif