1 # Ocean Interpreter - Jamison Creek version
3 Ocean is intended to be a compiled language, so this interpreter is
4 not targeted at being the final product. It is, rather, an intermediate
5 stage and fills that role in two distinct ways.
7 Firstly, it exists as a platform to experiment with the early language
8 design. An interpreter is easy to write and easy to get working, so
9 the barrier for entry is lower if I aim to start with an interpreter.
11 Secondly, the plan for the Ocean compiler is to write it in the
12 [Ocean language](http://ocean-lang.org). To achieve this we naturally
13 need some sort of boot-strap process and this interpreter - written in
14 portable C - will fill that role. It will be used to bootstrap the
17 Two features that are not needed to fill either of these roles are
18 performance and completeness. The interpreter only needs to be fast
19 enough to run small test programs and occasionally to run the compiler
20 on itself. It only needs to be complete enough to test aspects of the
21 design which are developed before the compiler is working, and to run
22 the compiler on itself. Any features not used by the compiler when
23 compiling itself are superfluous. They may be included anyway, but
26 Nonetheless, the interpreter should end up being reasonably complete,
27 and any performance bottlenecks which appear and are easily fixed, will
32 This third version of the interpreter exists to test out some initial
33 ideas relating to types. Particularly it adds arrays (indexed from
34 zero) and simple structures. Basic control flow and variable scoping
35 are already fairly well established, as are basic numerical and
38 Some operators that have only recently been added, and so have not
39 generated all that much experience yet are "and then" and "or else" as
40 short-circuit Boolean operators, and the "if ... else" trinary
41 operator which can select between two expressions based on a third
42 (which appears syntactically in the middle).
44 The "func" clause currently only allows a "main" function to be
45 declared. That will be extended when proper function support is added.
47 An element that is present purely to make a usable language, and
48 without any expectation that they will remain, is the "print" statement
49 which performs simple output.
51 The current scalar types are "number", "Boolean", and "string".
52 Boolean will likely stay in its current form, the other two might, but
53 could just as easily be changed.
57 Versions of the interpreter which obviously do not support a complete
58 language will be named after creeks and streams. This one is Jamison
61 Once we have something reasonably resembling a complete language, the
62 names of rivers will be used.
63 Early versions of the compiler will be named after seas. Major
64 releases of the compiler will be named after oceans. Hopefully I will
65 be finished once I get to the Pacific Ocean release.
69 As well as parsing and executing a program, the interpreter can print
70 out the program from the parsed internal structure. This is useful
71 for validating the parsing.
72 So the main requirements of the interpreter are:
74 - Parse the program, possibly with tracing,
75 - Analyse the parsed program to ensure consistency,
77 - Execute the "main" function in the program, if no parsing or
78 consistency errors were found.
80 This is all performed by a single C program extracted with
83 There will be two formats for printing the program: a default and one
84 that uses bracketing. So a `--bracket` command line option is needed
85 for that. Normally the first code section found is used, however an
86 alternate section can be requested so that a file (such as this one)
87 can contain multiple programs. This is effected with the `--section`
90 This code must be compiled with `-fplan9-extensions` so that anonymous
91 structures can be used.
93 ###### File: oceani.mk
95 myCFLAGS := -Wall -g -fplan9-extensions
96 CFLAGS := $(filter-out $(myCFLAGS),$(CFLAGS)) $(myCFLAGS)
97 myLDLIBS:= libparser.o libscanner.o libmdcode.o -licuuc
98 LDLIBS := $(filter-out $(myLDLIBS),$(LDLIBS)) $(myLDLIBS)
100 all :: $(LDLIBS) oceani
101 oceani.c oceani.h : oceani.mdc parsergen
102 ./parsergen -o oceani --LALR --tag Parser oceani.mdc
103 oceani.mk: oceani.mdc md2c
106 oceani: oceani.o $(LDLIBS)
107 $(CC) $(CFLAGS) -o oceani oceani.o $(LDLIBS)
109 ###### Parser: header
111 struct parse_context;
113 struct parse_context {
114 struct token_config config;
123 #define container_of(ptr, type, member) ({ \
124 const typeof( ((type *)0)->member ) *__mptr = (ptr); \
125 (type *)( (char *)__mptr - offsetof(type,member) );})
127 #define config2context(_conf) container_of(_conf, struct parse_context, \
130 ###### Parser: reduce
131 struct parse_context *c = config2context(config);
139 #include <sys/mman.h>
158 static char Usage[] =
159 "Usage: oceani --trace --print --noexec --brackets --section=SectionName prog.ocn\n";
160 static const struct option long_options[] = {
161 {"trace", 0, NULL, 't'},
162 {"print", 0, NULL, 'p'},
163 {"noexec", 0, NULL, 'n'},
164 {"brackets", 0, NULL, 'b'},
165 {"section", 1, NULL, 's'},
168 const char *options = "tpnbs";
170 static void pr_err(char *msg) // NOTEST
172 fprintf(stderr, "%s\n", msg); // NOTEST
175 int main(int argc, char *argv[])
180 struct section *s, *ss;
181 char *section = NULL;
182 struct parse_context context = {
184 .ignored = (1 << TK_mark),
185 .number_chars = ".,_+- ",
190 int doprint=0, dotrace=0, doexec=1, brackets=0;
192 while ((opt = getopt_long(argc, argv, options, long_options, NULL))
195 case 't': dotrace=1; break;
196 case 'p': doprint=1; break;
197 case 'n': doexec=0; break;
198 case 'b': brackets=1; break;
199 case 's': section = optarg; break;
200 default: fprintf(stderr, Usage);
204 if (optind >= argc) {
205 fprintf(stderr, "oceani: no input file given\n");
208 fd = open(argv[optind], O_RDONLY);
210 fprintf(stderr, "oceani: cannot open %s\n", argv[optind]);
213 context.file_name = argv[optind];
214 len = lseek(fd, 0, 2);
215 file = mmap(NULL, len, PROT_READ, MAP_SHARED, fd, 0);
216 s = code_extract(file, file+len, pr_err);
218 fprintf(stderr, "oceani: could not find any code in %s\n",
223 ## context initialization
226 for (ss = s; ss; ss = ss->next) {
227 struct text sec = ss->section;
228 if (sec.len == strlen(section) &&
229 strncmp(sec.txt, section, sec.len) == 0)
233 fprintf(stderr, "oceani: cannot find section %s\n",
239 parse_oceani(ss->code, &context.config, dotrace ? stderr : NULL);
242 fprintf(stderr, "oceani: no main function found.\n");
243 context.parse_error = 1;
245 if (context.prog && doprint) {
248 print_exec(context.prog, 0, brackets);
250 if (context.prog && doexec && !context.parse_error) {
251 if (!analyse_prog(context.prog, &context)) {
252 fprintf(stderr, "oceani: type error in program - not running.\n");
255 interp_prog(&context, context.prog, argc - optind, argv+optind);
257 free_exec(context.prog);
260 struct section *t = s->next;
266 ## free context types
267 ## free context storage
268 exit(context.parse_error ? 1 : 0);
273 The four requirements of parse, analyse, print, interpret apply to
274 each language element individually so that is how most of the code
277 Three of the four are fairly self explanatory. The one that requires
278 a little explanation is the analysis step.
280 The current language design does not require the types of variables to
281 be declared, but they must still have a single type. Different
282 operations impose different requirements on the variables, for example
283 addition requires both arguments to be numeric, and assignment
284 requires the variable on the left to have the same type as the
285 expression on the right.
287 Analysis involves propagating these type requirements around and
288 consequently setting the type of each variable. If any requirements
289 are violated (e.g. a string is compared with a number) or if a
290 variable needs to have two different types, then an error is raised
291 and the program will not run.
293 If the same variable is declared in both branchs of an 'if/else', or
294 in all cases of a 'switch' then the multiple instances may be merged
295 into just one variable if the variable is referenced after the
296 conditional statement. When this happens, the types must naturally be
297 consistent across all the branches. When the variable is not used
298 outside the if, the variables in the different branches are distinct
299 and can be of different types.
301 Undeclared names may only appear in "use" statements and "case" expressions.
302 These names are given a type of "label" and a unique value.
303 This allows them to fill the role of a name in an enumerated type, which
304 is useful for testing the `switch` statement.
306 As we will see, the condition part of a `while` statement can return
307 either a Boolean or some other type. This requires that the expected
308 type that gets passed around comprises a type and a flag to indicate
309 that `Tbool` is also permitted.
311 As there are, as yet, no distinct types that are compatible, there
312 isn't much subtlety in the analysis. When we have distinct number
313 types, this will become more interesting.
317 When analysis discovers an inconsistency it needs to report an error;
318 just refusing to run the code ensures that the error doesn't cascade,
319 but by itself it isn't very useful. A clear understanding of the sort
320 of error message that are useful will help guide the process of
323 At a simplistic level, the only sort of error that type analysis can
324 report is that the type of some construct doesn't match a contextual
325 requirement. For example, in `4 + "hello"` the addition provides a
326 contextual requirement for numbers, but `"hello"` is not a number. In
327 this particular example no further information is needed as the types
328 are obvious from local information. When a variable is involved that
329 isn't the case. It may be helpful to explain why the variable has a
330 particular type, by indicating the location where the type was set,
331 whether by declaration or usage.
333 Using a recursive-descent analysis we can easily detect a problem at
334 multiple locations. In "`hello:= "there"; 4 + hello`" the addition
335 will detect that one argument is not a number and the usage of `hello`
336 will detect that a number was wanted, but not provided. In this
337 (early) version of the language, we will generate error reports at
338 multiple locations, so the use of `hello` will report an error and
339 explain were the value was set, and the addition will report an error
340 and say why numbers are needed. To be able to report locations for
341 errors, each language element will need to record a file location
342 (line and column) and each variable will need to record the language
343 element where its type was set. For now we will assume that each line
344 of an error message indicates one location in the file, and up to 2
345 types. So we provide a `printf`-like function which takes a format, a
346 location (a `struct exec` which has not yet been introduced), and 2
347 types. "`%1`" reports the first type, "`%2`" reports the second. We
348 will need a function to print the location, once we know how that is
349 stored. e As will be explained later, there are sometimes extra rules for
350 type matching and they might affect error messages, we need to pass those
353 As well as type errors, we sometimes need to report problems with
354 tokens, which might be unexpected or might name a type that has not
355 been defined. For these we have `tok_err()` which reports an error
356 with a given token. Each of the error functions sets the flag in the
357 context so indicate that parsing failed.
361 static void fput_loc(struct exec *loc, FILE *f);
363 ###### core functions
365 static void type_err(struct parse_context *c,
366 char *fmt, struct exec *loc,
367 struct type *t1, int rules, struct type *t2)
369 fprintf(stderr, "%s:", c->file_name);
370 fput_loc(loc, stderr);
371 for (; *fmt ; fmt++) {
378 case '%': fputc(*fmt, stderr); break; // NOTEST
379 default: fputc('?', stderr); break; // NOTEST
381 type_print(t1, stderr);
384 type_print(t2, stderr);
393 static void tok_err(struct parse_context *c, char *fmt, struct token *t)
395 fprintf(stderr, "%s:%d:%d: %s: %.*s\n", c->file_name, t->line, t->col, fmt,
396 t->txt.len, t->txt.txt);
400 ## Entities: declared and predeclared.
402 There are various "things" that the language and/or the interpreter
403 needs to know about to parse and execute a program. These include
404 types, variables, values, and executable code. These are all lumped
405 together under the term "entities" (calling them "objects" would be
406 confusing) and introduced here. The following section will present the
407 different specific code elements which comprise or manipulate these
412 Values come in a wide range of types, with more likely to be added.
413 Each type needs to be able to print its own values (for convenience at
414 least) as well as to compare two values, at least for equality and
415 possibly for order. For now, values might need to be duplicated and
416 freed, though eventually such manipulations will be better integrated
419 Rather than requiring every numeric type to support all numeric
420 operations (add, multiple, etc), we allow types to be able to present
421 as one of a few standard types: integer, float, and fraction. The
422 existence of these conversion functions eventually enable types to
423 determine if they are compatible with other types, though such types
424 have not yet been implemented.
426 Named type are stored in a simple linked list. Objects of each type are
427 "values" which are often passed around by value.
434 ## value union fields
442 void (*init)(struct type *type, struct value *val);
443 void (*prepare_type)(struct parse_context *c, struct type *type, int parse_time);
444 void (*print)(struct type *type, struct value *val);
445 void (*print_type)(struct type *type, FILE *f);
446 int (*cmp_order)(struct type *t1, struct type *t2,
447 struct value *v1, struct value *v2);
448 int (*cmp_eq)(struct type *t1, struct type *t2,
449 struct value *v1, struct value *v2);
450 void (*dup)(struct type *type, struct value *vold, struct value *vnew);
451 void (*free)(struct type *type, struct value *val);
452 void (*free_type)(struct type *t);
453 long long (*to_int)(struct value *v);
454 double (*to_float)(struct value *v);
455 int (*to_mpq)(mpq_t *q, struct value *v);
464 struct type *typelist;
468 static struct type *find_type(struct parse_context *c, struct text s)
470 struct type *l = c->typelist;
473 text_cmp(l->name, s) != 0)
478 static struct type *add_type(struct parse_context *c, struct text s,
483 n = calloc(1, sizeof(*n));
486 n->next = c->typelist;
491 static void free_type(struct type *t)
493 /* The type is always a reference to something in the
494 * context, so we don't need to free anything.
498 static void free_value(struct type *type, struct value *v)
504 static void type_print(struct type *type, FILE *f)
507 fputs("*unknown*type*", f); // NOTEST
508 else if (type->name.len)
509 fprintf(f, "%.*s", type->name.len, type->name.txt);
510 else if (type->print_type)
511 type->print_type(type, f);
513 fputs("*invalid*type*", f); // NOTEST
516 static void val_init(struct type *type, struct value *val)
518 if (type && type->init)
519 type->init(type, val);
522 static void dup_value(struct type *type,
523 struct value *vold, struct value *vnew)
525 if (type && type->dup)
526 type->dup(type, vold, vnew);
529 static int value_cmp(struct type *tl, struct type *tr,
530 struct value *left, struct value *right)
532 if (tl && tl->cmp_order)
533 return tl->cmp_order(tl, tr, left, right);
534 if (tl && tl->cmp_eq) // NOTEST
535 return tl->cmp_eq(tl, tr, left, right); // NOTEST
539 static void print_value(struct type *type, struct value *v)
541 if (type && type->print)
542 type->print(type, v);
544 printf("*Unknown*"); // NOTEST
549 static void free_value(struct type *type, struct value *v);
550 static int type_compat(struct type *require, struct type *have, int rules);
551 static void type_print(struct type *type, FILE *f);
552 static void val_init(struct type *type, struct value *v);
553 static void dup_value(struct type *type,
554 struct value *vold, struct value *vnew);
555 static int value_cmp(struct type *tl, struct type *tr,
556 struct value *left, struct value *right);
557 static void print_value(struct type *type, struct value *v);
559 ###### free context types
561 while (context.typelist) {
562 struct type *t = context.typelist;
564 context.typelist = t->next;
570 Type can be specified for local variables, for fields in a structure,
571 for formal parameters to functions, and possibly elsewhere. Different
572 rules may apply in different contexts. As a minimum, a named type may
573 always be used. Currently the type of a formal parameter can be
574 different from types in other contexts, so we have a separate grammar
580 Type -> IDENTIFIER ${
581 $0 = find_type(c, $1.txt);
584 "error: undefined type", &$1);
591 FormalType -> Type ${ $0 = $<1; }$
592 ## formal type grammar
596 Values of the base types can be numbers, which we represent as
597 multi-precision fractions, strings, Booleans and labels. When
598 analysing the program we also need to allow for places where no value
599 is meaningful (type `Tnone`) and where we don't know what type to
600 expect yet (type is `NULL`).
602 Values are never shared, they are always copied when used, and freed
603 when no longer needed.
605 When propagating type information around the program, we need to
606 determine if two types are compatible, where type `NULL` is compatible
607 with anything. There are two special cases with type compatibility,
608 both related to the Conditional Statement which will be described
609 later. In some cases a Boolean can be accepted as well as some other
610 primary type, and in others any type is acceptable except a label (`Vlabel`).
611 A separate function encoding these cases will simplify some code later.
613 ###### type functions
615 int (*compat)(struct type *this, struct type *other);
619 static int type_compat(struct type *require, struct type *have, int rules)
621 if ((rules & Rboolok) && have == Tbool)
623 if ((rules & Rnolabel) && have == Tlabel)
625 if (!require || !have)
629 return require->compat(require, have);
631 return require == have;
636 #include "parse_string.h"
637 #include "parse_number.h"
640 myLDLIBS := libnumber.o libstring.o -lgmp
641 LDLIBS := $(filter-out $(myLDLIBS),$(LDLIBS)) $(myLDLIBS)
643 ###### type union fields
644 enum vtype {Vnone, Vstr, Vnum, Vbool, Vlabel} vtype;
646 ###### value union fields
653 static void _free_value(struct type *type, struct value *v)
657 switch (type->vtype) {
659 case Vstr: free(v->str.txt); break;
660 case Vnum: mpq_clear(v->num); break;
666 ###### value functions
668 static void _val_init(struct type *type, struct value *val)
670 switch(type->vtype) {
671 case Vnone: // NOTEST
674 mpq_init(val->num); break;
676 val->str.txt = malloc(1);
688 static void _dup_value(struct type *type,
689 struct value *vold, struct value *vnew)
691 switch (type->vtype) {
692 case Vnone: // NOTEST
695 vnew->label = vold->label;
698 vnew->bool = vold->bool;
702 mpq_set(vnew->num, vold->num);
705 vnew->str.len = vold->str.len;
706 vnew->str.txt = malloc(vnew->str.len);
707 memcpy(vnew->str.txt, vold->str.txt, vnew->str.len);
712 static int _value_cmp(struct type *tl, struct type *tr,
713 struct value *left, struct value *right)
717 return tl - tr; // NOTEST
719 case Vlabel: cmp = left->label == right->label ? 0 : 1; break;
720 case Vnum: cmp = mpq_cmp(left->num, right->num); break;
721 case Vstr: cmp = text_cmp(left->str, right->str); break;
722 case Vbool: cmp = left->bool - right->bool; break;
723 case Vnone: cmp = 0; // NOTEST
728 static void _print_value(struct type *type, struct value *v)
730 switch (type->vtype) {
731 case Vnone: // NOTEST
732 printf("*no-value*"); break; // NOTEST
733 case Vlabel: // NOTEST
734 printf("*label-%p*", v->label); break; // NOTEST
736 printf("%.*s", v->str.len, v->str.txt); break;
738 printf("%s", v->bool ? "True":"False"); break;
743 mpf_set_q(fl, v->num);
744 gmp_printf("%Fg", fl);
751 static void _free_value(struct type *type, struct value *v);
753 static struct type base_prototype = {
755 .print = _print_value,
756 .cmp_order = _value_cmp,
757 .cmp_eq = _value_cmp,
762 static struct type *Tbool, *Tstr, *Tnum, *Tnone, *Tlabel;
765 static struct type *add_base_type(struct parse_context *c, char *n,
766 enum vtype vt, int size)
768 struct text txt = { n, strlen(n) };
771 t = add_type(c, txt, &base_prototype);
774 t->align = size > sizeof(void*) ? sizeof(void*) : size;
775 if (t->size & (t->align - 1))
776 t->size = (t->size | (t->align - 1)) + 1; // NOTEST
780 ###### context initialization
782 Tbool = add_base_type(&context, "Boolean", Vbool, sizeof(char));
783 Tstr = add_base_type(&context, "string", Vstr, sizeof(struct text));
784 Tnum = add_base_type(&context, "number", Vnum, sizeof(mpq_t));
785 Tnone = add_base_type(&context, "none", Vnone, 0);
786 Tlabel = add_base_type(&context, "label", Vlabel, sizeof(void*));
790 Variables are scoped named values. We store the names in a linked list
791 of "bindings" sorted in lexical order, and use sequential search and
798 struct binding *next; // in lexical order
802 This linked list is stored in the parse context so that "reduce"
803 functions can find or add variables, and so the analysis phase can
804 ensure that every variable gets a type.
808 struct binding *varlist; // In lexical order
812 static struct binding *find_binding(struct parse_context *c, struct text s)
814 struct binding **l = &c->varlist;
819 (cmp = text_cmp((*l)->name, s)) < 0)
823 n = calloc(1, sizeof(*n));
830 Each name can be linked to multiple variables defined in different
831 scopes. Each scope starts where the name is declared and continues
832 until the end of the containing code block. Scopes of a given name
833 cannot nest, so a declaration while a name is in-scope is an error.
835 ###### binding fields
836 struct variable *var;
840 struct variable *previous;
842 struct binding *name;
843 struct exec *where_decl;// where name was declared
844 struct exec *where_set; // where type was set
848 While the naming seems strange, we include local constants in the
849 definition of variables. A name declared `var := value` can
850 subsequently be changed, but a name declared `var ::= value` cannot -
853 ###### variable fields
856 Scopes in parallel branches can be partially merged. More
857 specifically, if a given name is declared in both branches of an
858 if/else then its scope is a candidate for merging. Similarly if
859 every branch of an exhaustive switch (e.g. has an "else" clause)
860 declares a given name, then the scopes from the branches are
861 candidates for merging.
863 Note that names declared inside a loop (which is only parallel to
864 itself) are never visible after the loop. Similarly names defined in
865 scopes which are not parallel, such as those started by `for` and
866 `switch`, are never visible after the scope. Only variables defined in
867 both `then` and `else` (including the implicit then after an `if`, and
868 excluding `then` used with `for`) and in all `case`s and `else` of a
869 `switch` or `while` can be visible beyond the `if`/`switch`/`while`.
871 Labels, which are a bit like variables, follow different rules.
872 Labels are not explicitly declared, but if an undeclared name appears
873 in a context where a label is legal, that effectively declares the
874 name as a label. The declaration remains in force (or in scope) at
875 least to the end of the immediately containing block and conditionally
876 in any larger containing block which does not declare the name in some
877 other way. Importantly, the conditional scope extension happens even
878 if the label is only used in one parallel branch of a conditional --
879 when used in one branch it is treated as having been declared in all
882 Merge candidates are tentatively visible beyond the end of the
883 branching statement which creates them. If the name is used, the
884 merge is affirmed and they become a single variable visible at the
885 outer layer. If not - if it is redeclared first - the merge lapses.
887 To track scopes we have an extra stack, implemented as a linked list,
888 which roughly parallels the parse stack and which is used exclusively
889 for scoping. When a new scope is opened, a new frame is pushed and
890 the child-count of the parent frame is incremented. This child-count
891 is used to distinguish between the first of a set of parallel scopes,
892 in which declared variables must not be in scope, and subsequent
893 branches, whether they may already be conditionally scoped.
895 To push a new frame *before* any code in the frame is parsed, we need a
896 grammar reduction. This is most easily achieved with a grammar
897 element which derives the empty string, and creates the new scope when
898 it is recognised. This can be placed, for example, between a keyword
899 like "if" and the code following it.
903 struct scope *parent;
909 struct scope *scope_stack;
912 static void scope_pop(struct parse_context *c)
914 struct scope *s = c->scope_stack;
916 c->scope_stack = s->parent;
921 static void scope_push(struct parse_context *c)
923 struct scope *s = calloc(1, sizeof(*s));
925 c->scope_stack->child_count += 1;
926 s->parent = c->scope_stack;
934 OpenScope -> ${ scope_push(c); }$
935 ClosePara -> ${ var_block_close(c, CloseParallel); }$
937 Each variable records a scope depth and is in one of four states:
939 - "in scope". This is the case between the declaration of the
940 variable and the end of the containing block, and also between
941 the usage with affirms a merge and the end of that block.
943 The scope depth is not greater than the current parse context scope
944 nest depth. When the block of that depth closes, the state will
945 change. To achieve this, all "in scope" variables are linked
946 together as a stack in nesting order.
948 - "pending". The "in scope" block has closed, but other parallel
949 scopes are still being processed. So far, every parallel block at
950 the same level that has closed has declared the name.
952 The scope depth is the depth of the last parallel block that
953 enclosed the declaration, and that has closed.
955 - "conditionally in scope". The "in scope" block and all parallel
956 scopes have closed, and no further mention of the name has been
957 seen. This state includes a secondary nest depth which records the
958 outermost scope seen since the variable became conditionally in
959 scope. If a use of the name is found, the variable becomes "in
960 scope" and that secondary depth becomes the recorded scope depth.
961 If the name is declared as a new variable, the old variable becomes
962 "out of scope" and the recorded scope depth stays unchanged.
964 - "out of scope". The variable is neither in scope nor conditionally
965 in scope. It is permanently out of scope now and can be removed from
966 the "in scope" stack.
968 ###### variable fields
969 int depth, min_depth;
970 enum { OutScope, PendingScope, CondScope, InScope } scope;
971 struct variable *in_scope;
975 struct variable *in_scope;
977 All variables with the same name are linked together using the
978 'previous' link. Those variable that have been affirmatively merged all
979 have a 'merged' pointer that points to one primary variable - the most
980 recently declared instance. When merging variables, we need to also
981 adjust the 'merged' pointer on any other variables that had previously
982 been merged with the one that will no longer be primary.
984 A variable that is no longer the most recent instance of a name may
985 still have "pending" scope, if it might still be merged with most
986 recent instance. These variables don't really belong in the
987 "in_scope" list, but are not immediately removed when a new instance
988 is found. Instead, they are detected and ignored when considering the
989 list of in_scope names.
991 The storage of the value of a variable will be described later. For now
992 we just need to know that when a variable goes out of scope, it might
993 need to be freed. For this we need to be able to find it, so assume that
994 `var_value()` will provide that.
996 ###### variable fields
997 struct variable *merged;
1001 static void variable_merge(struct variable *primary, struct variable *secondary)
1005 if (primary->merged)
1007 primary = primary->merged; // NOTEST
1009 for (v = primary->previous; v; v=v->previous)
1010 if (v == secondary || v == secondary->merged ||
1011 v->merged == secondary ||
1012 (v->merged && v->merged == secondary->merged)) {
1013 v->scope = OutScope;
1014 v->merged = primary;
1018 ###### forward decls
1019 static struct value *var_value(struct parse_context *c, struct variable *v);
1021 ###### free context vars
1023 while (context.varlist) {
1024 struct binding *b = context.varlist;
1025 struct variable *v = b->var;
1026 context.varlist = b->next;
1029 struct variable *t = v;
1032 free_value(t->type, var_value(&context, t));
1034 // This is a global constant
1035 free_exec(t->where_decl);
1040 #### Manipulating Bindings
1042 When a name is conditionally visible, a new declaration discards the
1043 old binding - the condition lapses. Conversely a usage of the name
1044 affirms the visibility and extends it to the end of the containing
1045 block - i.e. the block that contains both the original declaration and
1046 the latest usage. This is determined from `min_depth`. When a
1047 conditionally visible variable gets affirmed like this, it is also
1048 merged with other conditionally visible variables with the same name.
1050 When we parse a variable declaration we either report an error if the
1051 name is currently bound, or create a new variable at the current nest
1052 depth if the name is unbound or bound to a conditionally scoped or
1053 pending-scope variable. If the previous variable was conditionally
1054 scoped, it and its homonyms becomes out-of-scope.
1056 When we parse a variable reference (including non-declarative assignment
1057 "foo = bar") we report an error if the name is not bound or is bound to
1058 a pending-scope variable; update the scope if the name is bound to a
1059 conditionally scoped variable; or just proceed normally if the named
1060 variable is in scope.
1062 When we exit a scope, any variables bound at this level are either
1063 marked out of scope or pending-scoped, depending on whether the scope
1064 was sequential or parallel. Here a "parallel" scope means the "then"
1065 or "else" part of a conditional, or any "case" or "else" branch of a
1066 switch. Other scopes are "sequential".
1068 When exiting a parallel scope we check if there are any variables that
1069 were previously pending and are still visible. If there are, then
1070 there weren't redeclared in the most recent scope, so they cannot be
1071 merged and must become out-of-scope. If it is not the first of
1072 parallel scopes (based on `child_count`), we check that there was a
1073 previous binding that is still pending-scope. If there isn't, the new
1074 variable must now be out-of-scope.
1076 When exiting a sequential scope that immediately enclosed parallel
1077 scopes, we need to resolve any pending-scope variables. If there was
1078 no `else` clause, and we cannot determine that the `switch` was exhaustive,
1079 we need to mark all pending-scope variable as out-of-scope. Otherwise
1080 all pending-scope variables become conditionally scoped.
1083 enum closetype { CloseSequential, CloseParallel, CloseElse };
1085 ###### ast functions
1087 static struct variable *var_decl(struct parse_context *c, struct text s)
1089 struct binding *b = find_binding(c, s);
1090 struct variable *v = b->var;
1092 switch (v ? v->scope : OutScope) {
1094 /* Caller will report the error */
1098 v && v->scope == CondScope;
1100 v->scope = OutScope;
1104 v = calloc(1, sizeof(*v));
1105 v->previous = b->var;
1108 v->min_depth = v->depth = c->scope_depth;
1110 v->in_scope = c->in_scope;
1115 static struct variable *var_ref(struct parse_context *c, struct text s)
1117 struct binding *b = find_binding(c, s);
1118 struct variable *v = b->var;
1119 struct variable *v2;
1121 switch (v ? v->scope : OutScope) {
1124 /* Caller will report the error */
1127 /* All CondScope variables of this name need to be merged
1128 * and become InScope
1130 v->depth = v->min_depth;
1132 for (v2 = v->previous;
1133 v2 && v2->scope == CondScope;
1135 variable_merge(v, v2);
1143 static void var_block_close(struct parse_context *c, enum closetype ct)
1145 /* Close off all variables that are in_scope */
1146 struct variable *v, **vp, *v2;
1149 for (vp = &c->in_scope;
1150 v = *vp, v && v->depth > c->scope_depth && v->min_depth > c->scope_depth;
1152 if (v->name->var == v) switch (ct) {
1154 case CloseParallel: /* handle PendingScope */
1158 if (c->scope_stack->child_count == 1)
1159 v->scope = PendingScope;
1160 else if (v->previous &&
1161 v->previous->scope == PendingScope)
1162 v->scope = PendingScope;
1163 else if (v->type == Tlabel) // UNTESTED
1164 v->scope = PendingScope; // UNTESTED
1165 else if (v->name->var == v) // UNTESTED
1166 v->scope = OutScope; // UNTESTED
1167 if (ct == CloseElse) {
1168 /* All Pending variables with this name
1169 * are now Conditional */
1171 v2 && v2->scope == PendingScope;
1173 v2->scope = CondScope;
1178 v2 && v2->scope == PendingScope;
1180 if (v2->type != Tlabel)
1181 v2->scope = OutScope;
1183 case OutScope: break; // UNTESTED
1186 case CloseSequential:
1187 if (v->type == Tlabel)
1188 v->scope = PendingScope;
1191 v->scope = OutScope;
1194 /* There was no 'else', so we can only become
1195 * conditional if we know the cases were exhaustive,
1196 * and that doesn't mean anything yet.
1197 * So only labels become conditional..
1200 v2 && v2->scope == PendingScope;
1202 if (v2->type == Tlabel) {
1203 v2->scope = CondScope;
1204 v2->min_depth = c->scope_depth;
1206 v2->scope = OutScope;
1209 case OutScope: break;
1213 if (v->scope == OutScope || v->name->var != v)
1222 The value of a variable is store separately from the variable, on an
1223 analogue of a stack frame. There are (currently) two frames that can be
1224 active. A global frame which currently only stores constants, and a
1225 stacked frame which stores local variables. Each variable knows if it
1226 is global or not, and what its index into the frame is.
1228 Values in the global frame are known immediately they are relevant, so
1229 the frame needs to be reallocated as it grows so it can store those
1230 values. The local frame doesn't get values until the interpreted phase
1231 is started, so there is no need to allocate until the size is known.
1233 ###### variable fields
1237 ###### parse context
1239 short global_size, global_alloc;
1241 void *global, *local;
1243 ###### ast functions
1245 static struct value *var_value(struct parse_context *c, struct variable *v)
1248 if (!c->local || !v->type)
1250 if (v->frame_pos + v->type->size > c->local_size) {
1251 printf("INVALID frame_pos\n"); // NOTEST
1254 return c->local + v->frame_pos;
1256 if (c->global_size > c->global_alloc) {
1257 int old = c->global_alloc;
1258 c->global_alloc = (c->global_size | 1023) + 1024;
1259 c->global = realloc(c->global, c->global_alloc);
1260 memset(c->global + old, 0, c->global_alloc - old);
1262 return c->global + v->frame_pos;
1265 static struct value *global_alloc(struct parse_context *c, struct type *t,
1266 struct variable *v, struct value *init)
1269 struct variable scratch;
1271 if (t->prepare_type)
1272 t->prepare_type(c, t, 1); // NOTEST
1274 if (c->global_size & (t->align - 1))
1275 c->global_size = (c->global_size + t->align) & ~(t->align-1); // UNTESTED
1280 v->frame_pos = c->global_size;
1282 c->global_size += v->type->size;
1283 ret = var_value(c, v);
1285 memcpy(ret, init, t->size);
1291 As global values are found -- struct field initializers, labels etc --
1292 `global_alloc()` is called to record the value in the global frame.
1294 When the program is fully parsed, we need to walk the list of variables
1295 to find any that weren't merged away and that aren't global, and to
1296 calculate the frame size and assign a frame position for each variable.
1297 For this we have `scope_finalize()`.
1299 ###### ast functions
1301 static void scope_finalize(struct parse_context *c)
1305 for (b = c->varlist; b; b = b->next) {
1307 for (v = b->var; v; v = v->previous) {
1308 struct type *t = v->type;
1309 if (v->merged && v->merged != v)
1313 if (c->local_size & (t->align - 1))
1314 c->local_size = (c->local_size + t->align) & ~(t->align-1);
1315 v->frame_pos = c->local_size;
1316 c->local_size += v->type->size;
1319 c->local = calloc(1, c->local_size);
1322 ###### free context storage
1323 free(context.global);
1324 free(context.local);
1328 Executables can be lots of different things. In many cases an
1329 executable is just an operation combined with one or two other
1330 executables. This allows for expressions and lists etc. Other times an
1331 executable is something quite specific like a constant or variable name.
1332 So we define a `struct exec` to be a general executable with a type, and
1333 a `struct binode` which is a subclass of `exec`, forms a node in a
1334 binary tree, and holds an operation. There will be other subclasses,
1335 and to access these we need to be able to `cast` the `exec` into the
1336 various other types. The first field in any `struct exec` is the type
1337 from the `exec_types` enum.
1340 #define cast(structname, pointer) ({ \
1341 const typeof( ((struct structname *)0)->type) *__mptr = &(pointer)->type; \
1342 if (__mptr && *__mptr != X##structname) abort(); \
1343 (struct structname *)( (char *)__mptr);})
1345 #define new(structname) ({ \
1346 struct structname *__ptr = ((struct structname *)calloc(1,sizeof(struct structname))); \
1347 __ptr->type = X##structname; \
1348 __ptr->line = -1; __ptr->column = -1; \
1351 #define new_pos(structname, token) ({ \
1352 struct structname *__ptr = ((struct structname *)calloc(1,sizeof(struct structname))); \
1353 __ptr->type = X##structname; \
1354 __ptr->line = token.line; __ptr->column = token.col; \
1363 enum exec_types type;
1371 struct exec *left, *right;
1374 ###### ast functions
1376 static int __fput_loc(struct exec *loc, FILE *f)
1380 if (loc->line >= 0) {
1381 fprintf(f, "%d:%d: ", loc->line, loc->column);
1384 if (loc->type == Xbinode)
1385 return __fput_loc(cast(binode,loc)->left, f) ||
1386 __fput_loc(cast(binode,loc)->right, f); // NOTEST
1389 static void fput_loc(struct exec *loc, FILE *f)
1391 if (!__fput_loc(loc, f))
1392 fprintf(f, "??:??: "); // NOTEST
1395 Each different type of `exec` node needs a number of functions defined,
1396 a bit like methods. We must be able to free it, print it, analyse it
1397 and execute it. Once we have specific `exec` types we will need to
1398 parse them too. Let's take this a bit more slowly.
1402 The parser generator requires a `free_foo` function for each struct
1403 that stores attributes and they will often be `exec`s and subtypes
1404 there-of. So we need `free_exec` which can handle all the subtypes,
1405 and we need `free_binode`.
1407 ###### ast functions
1409 static void free_binode(struct binode *b)
1414 free_exec(b->right);
1418 ###### core functions
1419 static void free_exec(struct exec *e)
1428 ###### forward decls
1430 static void free_exec(struct exec *e);
1432 ###### free exec cases
1433 case Xbinode: free_binode(cast(binode, e)); break;
1437 Printing an `exec` requires that we know the current indent level for
1438 printing line-oriented components. As will become clear later, we
1439 also want to know what sort of bracketing to use.
1441 ###### ast functions
1443 static void do_indent(int i, char *str)
1450 ###### core functions
1451 static void print_binode(struct binode *b, int indent, int bracket)
1455 ## print binode cases
1459 static void print_exec(struct exec *e, int indent, int bracket)
1465 print_binode(cast(binode, e), indent, bracket); break;
1470 ###### forward decls
1472 static void print_exec(struct exec *e, int indent, int bracket);
1476 As discussed, analysis involves propagating type requirements around the
1477 program and looking for errors.
1479 So `propagate_types` is passed an expected type (being a `struct type`
1480 pointer together with some `val_rules` flags) that the `exec` is
1481 expected to return, and returns the type that it does return, either
1482 of which can be `NULL` signifying "unknown". An `ok` flag is passed
1483 by reference. It is set to `0` when an error is found, and `2` when
1484 any change is made. If it remains unchanged at `1`, then no more
1485 propagation is needed.
1489 enum val_rules {Rnolabel = 1<<0, Rboolok = 1<<1, Rnoconstant = 2<<1};
1493 if (rules & Rnolabel)
1494 fputs(" (labels not permitted)", stderr);
1497 ###### core functions
1499 static struct type *propagate_types(struct exec *prog, struct parse_context *c, int *ok,
1500 struct type *type, int rules);
1501 static struct type *__propagate_types(struct exec *prog, struct parse_context *c, int *ok,
1502 struct type *type, int rules)
1509 switch (prog->type) {
1512 struct binode *b = cast(binode, prog);
1514 ## propagate binode cases
1518 ## propagate exec cases
1523 static struct type *propagate_types(struct exec *prog, struct parse_context *c, int *ok,
1524 struct type *type, int rules)
1526 struct type *ret = __propagate_types(prog, c, ok, type, rules);
1535 Interpreting an `exec` doesn't require anything but the `exec`. State
1536 is stored in variables and each variable will be directly linked from
1537 within the `exec` tree. The exception to this is the `main` function
1538 which needs to look at command line arguments. This function will be
1539 interpreted separately.
1541 Each `exec` can return a value combined with a type in `struct lrval`.
1542 The type may be `Tnone` but must be non-NULL. Some `exec`s will return
1543 the location of a value, which can be updated, in `lval`. Others will
1544 set `lval` to NULL indicating that there is a value of appropriate type
1547 ###### core functions
1551 struct value rval, *lval;
1554 static struct lrval _interp_exec(struct parse_context *c, struct exec *e);
1556 static struct value interp_exec(struct parse_context *c, struct exec *e,
1557 struct type **typeret)
1559 struct lrval ret = _interp_exec(c, e);
1561 if (!ret.type) abort();
1563 *typeret = ret.type;
1565 dup_value(ret.type, ret.lval, &ret.rval);
1569 static struct value *linterp_exec(struct parse_context *c, struct exec *e,
1570 struct type **typeret)
1572 struct lrval ret = _interp_exec(c, e);
1575 *typeret = ret.type;
1577 free_value(ret.type, &ret.rval);
1581 static struct lrval _interp_exec(struct parse_context *c, struct exec *e)
1584 struct value rv = {}, *lrv = NULL;
1585 struct type *rvtype;
1587 rvtype = ret.type = Tnone;
1589 ret.lval = lrv; // UNTESTED
1590 ret.rval = rv; // UNTESTED
1591 return ret; // UNTESTED
1597 struct binode *b = cast(binode, e);
1598 struct value left, right, *lleft;
1599 struct type *ltype, *rtype;
1600 ltype = rtype = Tnone;
1602 ## interp binode cases
1604 free_value(ltype, &left);
1605 free_value(rtype, &right);
1608 ## interp exec cases
1618 Now that we have the shape of the interpreter in place we can add some
1619 complex types and connected them in to the data structures and the
1620 different phases of parse, analyse, print, interpret.
1622 Thus far we have arrays and structs.
1626 Arrays can be declared by giving a size and a type, as `[size]type' so
1627 `freq:[26]number` declares `freq` to be an array of 26 numbers. The
1628 size can be either a literal number, or a named constant. Some day an
1629 arbitrary expression will be supported.
1631 As a formal parameter to a function, the array can be declared with a
1632 new variable as the size: `name:[size::number]string`. The `size`
1633 variable is set to the size of the array and must be a constant. As
1634 `number` is the only supported type, it can be left out:
1635 `name:[size::]string`.
1637 Arrays cannot be assigned. When pointers are introduced we will also
1638 introduce array slices which can refer to part or all of an array -
1639 the assignment syntax will create a slice. For now, an array can only
1640 ever be referenced by the name it is declared with. It is likely that
1641 a "`copy`" primitive will eventually be define which can be used to
1642 make a copy of an array with controllable recursive depth.
1644 For now we have two sorts of array, those with fixed size either because
1645 it is given as a literal number or because it is a struct member (which
1646 cannot have a runtime-changing size), and those with a size that is
1647 determined at runtime - local variables with a const size. The former
1648 have their size calculated at parse time, the latter at run time.
1650 For the latter type, the `size` field of the type is the size of a
1651 pointer, and the array is reallocated every time it comes into scope.
1653 We differentiate struct fields with a const size from local variables
1654 with a const size by whether they are prepared at parse time or not.
1656 ###### type union fields
1659 int unspec; // size is unspecified - vsize must be set.
1662 struct variable *vsize;
1663 struct type *member;
1666 ###### value union fields
1667 void *array; // used if not static_size
1669 ###### value functions
1671 static void array_prepare_type(struct parse_context *c, struct type *type,
1674 struct value *vsize;
1676 if (!type->array.vsize || type->array.static_size)
1679 vsize = var_value(c, type->array.vsize);
1681 mpz_tdiv_q(q, mpq_numref(vsize->num), mpq_denref(vsize->num));
1682 type->array.size = mpz_get_si(q);
1686 type->array.static_size = 1;
1687 type->size = type->array.size * type->array.member->size;
1688 type->align = type->array.member->align;
1692 static void array_init(struct type *type, struct value *val)
1695 void *ptr = val->ptr;
1699 if (!type->array.static_size) {
1700 val->array = calloc(type->array.size,
1701 type->array.member->size);
1704 for (i = 0; i < type->array.size; i++) {
1706 v = (void*)ptr + i * type->array.member->size;
1707 val_init(type->array.member, v);
1711 static void array_free(struct type *type, struct value *val)
1714 void *ptr = val->ptr;
1716 if (!type->array.static_size)
1718 for (i = 0; i < type->array.size; i++) {
1720 v = (void*)ptr + i * type->array.member->size;
1721 free_value(type->array.member, v);
1723 if (!type->array.static_size)
1727 static int array_compat(struct type *require, struct type *have)
1729 if (have->compat != require->compat)
1730 return 0; // UNTESTED
1731 /* Both are arrays, so we can look at details */
1732 if (!type_compat(require->array.member, have->array.member, 0))
1734 if (have->array.unspec && require->array.unspec) {
1735 if (have->array.vsize && require->array.vsize &&
1736 have->array.vsize != require->array.vsize) // UNTESTED
1737 /* sizes might not be the same */
1738 return 0; // UNTESTED
1741 if (have->array.unspec || require->array.unspec)
1742 return 1; // UNTESTED
1743 if (require->array.vsize == NULL && have->array.vsize == NULL)
1744 return require->array.size == have->array.size;
1746 return require->array.vsize == have->array.vsize; // UNTESTED
1749 static void array_print_type(struct type *type, FILE *f)
1752 if (type->array.vsize) {
1753 struct binding *b = type->array.vsize->name;
1754 fprintf(f, "%.*s%s]", b->name.len, b->name.txt,
1755 type->array.unspec ? "::" : "");
1757 fprintf(f, "%d]", type->array.size);
1758 type_print(type->array.member, f);
1761 static struct type array_prototype = {
1763 .prepare_type = array_prepare_type,
1764 .print_type = array_print_type,
1765 .compat = array_compat,
1767 .size = sizeof(void*),
1768 .align = sizeof(void*),
1771 ###### declare terminals
1776 | [ NUMBER ] Type ${ {
1779 struct text noname = { "", 0 };
1782 $0 = t = add_type(c, noname, &array_prototype);
1783 t->array.member = $<4;
1784 t->array.vsize = NULL;
1785 if (number_parse(num, tail, $2.txt) == 0)
1786 tok_err(c, "error: unrecognised number", &$2);
1788 tok_err(c, "error: unsupported number suffix", &$2);
1790 t->array.size = mpz_get_ui(mpq_numref(num));
1791 if (mpz_cmp_ui(mpq_denref(num), 1) != 0) {
1792 tok_err(c, "error: array size must be an integer",
1794 } else if (mpz_cmp_ui(mpq_numref(num), 1UL << 30) >= 0)
1795 tok_err(c, "error: array size is too large",
1799 t->array.static_size = 1;
1800 t->size = t->array.size * t->array.member->size;
1801 t->align = t->array.member->align;
1804 | [ IDENTIFIER ] Type ${ {
1805 struct variable *v = var_ref(c, $2.txt);
1806 struct text noname = { "", 0 };
1809 tok_err(c, "error: name undeclared", &$2);
1810 else if (!v->constant)
1811 tok_err(c, "error: array size must be a constant", &$2);
1813 $0 = add_type(c, noname, &array_prototype);
1814 $0->array.member = $<4;
1816 $0->array.vsize = v;
1821 OptType -> Type ${ $0 = $<1; }$
1824 ###### formal type grammar
1826 | [ IDENTIFIER :: OptType ] Type ${ {
1827 struct variable *v = var_decl(c, $ID.txt);
1828 struct text noname = { "", 0 };
1834 $0 = add_type(c, noname, &array_prototype);
1835 $0->array.member = $<6;
1837 $0->array.unspec = 1;
1838 $0->array.vsize = v;
1844 ###### variable grammar
1846 | Variable [ Expression ] ${ {
1847 struct binode *b = new(binode);
1854 ###### print binode cases
1856 print_exec(b->left, -1, bracket);
1858 print_exec(b->right, -1, bracket);
1862 ###### propagate binode cases
1864 /* left must be an array, right must be a number,
1865 * result is the member type of the array
1867 propagate_types(b->right, c, ok, Tnum, 0);
1868 t = propagate_types(b->left, c, ok, NULL, rules & Rnoconstant);
1869 if (!t || t->compat != array_compat) {
1870 type_err(c, "error: %1 cannot be indexed", prog, t, 0, NULL);
1873 if (!type_compat(type, t->array.member, rules)) {
1874 type_err(c, "error: have %1 but need %2", prog,
1875 t->array.member, rules, type);
1877 return t->array.member;
1881 ###### interp binode cases
1887 lleft = linterp_exec(c, b->left, <ype);
1888 right = interp_exec(c, b->right, &rtype);
1890 mpz_tdiv_q(q, mpq_numref(right.num), mpq_denref(right.num));
1894 if (ltype->array.static_size)
1897 ptr = *(void**)lleft;
1898 rvtype = ltype->array.member;
1899 if (i >= 0 && i < ltype->array.size)
1900 lrv = ptr + i * rvtype->size;
1902 val_init(ltype->array.member, &rv);
1909 A `struct` is a data-type that contains one or more other data-types.
1910 It differs from an array in that each member can be of a different
1911 type, and they are accessed by name rather than by number. Thus you
1912 cannot choose an element by calculation, you need to know what you
1915 The language makes no promises about how a given structure will be
1916 stored in memory - it is free to rearrange fields to suit whatever
1917 criteria seems important.
1919 Structs are declared separately from program code - they cannot be
1920 declared in-line in a variable declaration like arrays can. A struct
1921 is given a name and this name is used to identify the type - the name
1922 is not prefixed by the word `struct` as it would be in C.
1924 Structs are only treated as the same if they have the same name.
1925 Simply having the same fields in the same order is not enough. This
1926 might change once we can create structure initializers from a list of
1929 Each component datum is identified much like a variable is declared,
1930 with a name, one or two colons, and a type. The type cannot be omitted
1931 as there is no opportunity to deduce the type from usage. An initial
1932 value can be given following an equals sign, so
1934 ##### Example: a struct type
1940 would declare a type called "complex" which has two number fields,
1941 each initialised to zero.
1943 Struct will need to be declared separately from the code that uses
1944 them, so we will need to be able to print out the declaration of a
1945 struct when reprinting the whole program. So a `print_type_decl` type
1946 function will be needed.
1948 ###### type union fields
1960 ###### type functions
1961 void (*print_type_decl)(struct type *type, FILE *f);
1963 ###### value functions
1965 static void structure_init(struct type *type, struct value *val)
1969 for (i = 0; i < type->structure.nfields; i++) {
1971 v = (void*) val->ptr + type->structure.fields[i].offset;
1972 if (type->structure.fields[i].init)
1973 dup_value(type->structure.fields[i].type,
1974 type->structure.fields[i].init,
1977 val_init(type->structure.fields[i].type, v);
1981 static void structure_free(struct type *type, struct value *val)
1985 for (i = 0; i < type->structure.nfields; i++) {
1987 v = (void*)val->ptr + type->structure.fields[i].offset;
1988 free_value(type->structure.fields[i].type, v);
1992 static void structure_free_type(struct type *t)
1995 for (i = 0; i < t->structure.nfields; i++)
1996 if (t->structure.fields[i].init) {
1997 free_value(t->structure.fields[i].type,
1998 t->structure.fields[i].init);
2000 free(t->structure.fields);
2003 static struct type structure_prototype = {
2004 .init = structure_init,
2005 .free = structure_free,
2006 .free_type = structure_free_type,
2007 .print_type_decl = structure_print_type,
2021 ###### free exec cases
2023 free_exec(cast(fieldref, e)->left);
2027 ###### declare terminals
2030 ###### variable grammar
2032 | Variable . IDENTIFIER ${ {
2033 struct fieldref *fr = new_pos(fieldref, $2);
2040 ###### print exec cases
2044 struct fieldref *f = cast(fieldref, e);
2045 print_exec(f->left, -1, bracket);
2046 printf(".%.*s", f->name.len, f->name.txt);
2050 ###### ast functions
2051 static int find_struct_index(struct type *type, struct text field)
2054 for (i = 0; i < type->structure.nfields; i++)
2055 if (text_cmp(type->structure.fields[i].name, field) == 0)
2060 ###### propagate exec cases
2064 struct fieldref *f = cast(fieldref, prog);
2065 struct type *st = propagate_types(f->left, c, ok, NULL, 0);
2068 type_err(c, "error: unknown type for field access", f->left, // UNTESTED
2070 else if (st->init != structure_init)
2071 type_err(c, "error: field reference attempted on %1, not a struct",
2072 f->left, st, 0, NULL);
2073 else if (f->index == -2) {
2074 f->index = find_struct_index(st, f->name);
2076 type_err(c, "error: cannot find requested field in %1",
2077 f->left, st, 0, NULL);
2079 if (f->index >= 0) {
2080 struct type *ft = st->structure.fields[f->index].type;
2081 if (!type_compat(type, ft, rules))
2082 type_err(c, "error: have %1 but need %2", prog,
2089 ###### interp exec cases
2092 struct fieldref *f = cast(fieldref, e);
2094 struct value *lleft = linterp_exec(c, f->left, <ype);
2095 lrv = (void*)lleft->ptr + ltype->structure.fields[f->index].offset;
2096 rvtype = ltype->structure.fields[f->index].type;
2102 struct fieldlist *prev;
2106 ###### ast functions
2107 static void free_fieldlist(struct fieldlist *f)
2111 free_fieldlist(f->prev);
2113 free_value(f->f.type, f->f.init); // UNTESTED
2114 free(f->f.init); // UNTESTED
2119 ###### top level grammar
2120 DeclareStruct -> struct IDENTIFIER FieldBlock Newlines ${ {
2122 add_type(c, $2.txt, &structure_prototype);
2124 struct fieldlist *f;
2126 for (f = $3; f; f=f->prev)
2129 t->structure.nfields = cnt;
2130 t->structure.fields = calloc(cnt, sizeof(struct field));
2133 int a = f->f.type->align;
2135 t->structure.fields[cnt] = f->f;
2136 if (t->size & (a-1))
2137 t->size = (t->size | (a-1)) + 1;
2138 t->structure.fields[cnt].offset = t->size;
2139 t->size += ((f->f.type->size - 1) | (a-1)) + 1;
2148 FieldBlock -> { IN OptNL FieldLines OUT OptNL } ${ $0 = $<FL; }$
2149 | { SimpleFieldList } ${ $0 = $<SFL; }$
2150 | IN OptNL FieldLines OUT ${ $0 = $<FL; }$
2151 | SimpleFieldList EOL ${ $0 = $<SFL; }$
2153 FieldLines -> SimpleFieldList Newlines ${ $0 = $<SFL; }$
2154 | FieldLines SimpleFieldList Newlines ${
2159 SimpleFieldList -> Field ${ $0 = $<F; }$
2160 | SimpleFieldList ; Field ${
2164 | SimpleFieldList ; ${
2167 | ERROR ${ tok_err(c, "Syntax error in struct field", &$1); }$
2169 Field -> IDENTIFIER : Type = Expression ${ {
2172 $0 = calloc(1, sizeof(struct fieldlist));
2173 $0->f.name = $1.txt;
2178 propagate_types($<5, c, &ok, $3, 0);
2181 c->parse_error = 1; // UNTESTED
2183 struct value vl = interp_exec(c, $5, NULL);
2184 $0->f.init = global_alloc(c, $0->f.type, NULL, &vl);
2187 | IDENTIFIER : Type ${
2188 $0 = calloc(1, sizeof(struct fieldlist));
2189 $0->f.name = $1.txt;
2191 if ($0->f.type->prepare_type)
2192 $0->f.type->prepare_type(c, $0->f.type, 1);
2195 ###### forward decls
2196 static void structure_print_type(struct type *t, FILE *f);
2198 ###### value functions
2199 static void structure_print_type(struct type *t, FILE *f) // UNTESTED
2203 fprintf(f, "struct %.*s\n", t->name.len, t->name.txt);
2205 for (i = 0; i < t->structure.nfields; i++) {
2206 struct field *fl = t->structure.fields + i;
2207 fprintf(f, " %.*s : ", fl->name.len, fl->name.txt);
2208 type_print(fl->type, f);
2209 if (fl->type->print && fl->init) {
2211 if (fl->type == Tstr)
2212 fprintf(f, "\""); // UNTESTED
2213 print_value(fl->type, fl->init);
2214 if (fl->type == Tstr)
2215 fprintf(f, "\""); // UNTESTED
2221 ###### print type decls
2223 struct type *t; // UNTESTED
2226 while (target != 0) {
2228 for (t = context.typelist; t ; t=t->next)
2229 if (t->print_type_decl) {
2238 t->print_type_decl(t, stdout);
2246 A function is a named chunk of code which can be passed parameters and
2247 can return results. Each function has an implicit type which includes
2248 the set of parameters and the return value. As yet these types cannot
2249 be declared separate from the function itself.
2251 In fact, only one function is currently possible - `main`. `main` is
2252 passed an array of strings together with the size of the array, and
2253 doesn't return anything. The strings are command line arguments.
2255 The parameters can be specified either in parentheses as a list, such as
2257 ##### Example: function 1
2259 func main(av:[ac::number]string)
2262 or as an indented list of one parameter per line
2264 ##### Example: function 2
2267 argv:[argc::number]string
2279 MainFunction -> func main ( OpenScope Args ) Block Newlines ${
2282 $0->left = reorder_bilist($<Ar);
2284 var_block_close(c, CloseSequential);
2285 if (c->scope_stack && !c->parse_error) abort();
2287 | func main IN OpenScope OptNL Args OUT OptNL do Block Newlines ${
2290 $0->left = reorder_bilist($<Ar);
2292 var_block_close(c, CloseSequential);
2293 if (c->scope_stack && !c->parse_error) abort();
2295 | func main NEWLINE OpenScope OptNL do Block Newlines ${
2300 var_block_close(c, CloseSequential);
2301 if (c->scope_stack && !c->parse_error) abort();
2304 Args -> ${ $0 = NULL; }$
2305 | Varlist ${ $0 = $<1; }$
2306 | Varlist ; ${ $0 = $<1; }$
2307 | Varlist NEWLINE ${ $0 = $<1; }$
2309 Varlist -> Varlist ; ArgDecl ${ // UNTESTED
2323 ArgDecl -> IDENTIFIER : FormalType ${ {
2324 struct variable *v = var_decl(c, $1.txt);
2330 ## Executables: the elements of code
2332 Each code element needs to be parsed, printed, analysed,
2333 interpreted, and freed. There are several, so let's just start with
2334 the easy ones and work our way up.
2338 We have already met values as separate objects. When manifest
2339 constants appear in the program text, that must result in an executable
2340 which has a constant value. So the `val` structure embeds a value in
2353 ###### ast functions
2354 struct val *new_val(struct type *T, struct token tk)
2356 struct val *v = new_pos(val, tk);
2367 $0 = new_val(Tbool, $1);
2371 $0 = new_val(Tbool, $1);
2375 $0 = new_val(Tnum, $1);
2378 if (number_parse($0->val.num, tail, $1.txt) == 0)
2379 mpq_init($0->val.num); // UNTESTED
2381 tok_err(c, "error: unsupported number suffix",
2386 $0 = new_val(Tstr, $1);
2389 string_parse(&$1, '\\', &$0->val.str, tail);
2391 tok_err(c, "error: unsupported string suffix",
2396 $0 = new_val(Tstr, $1);
2399 string_parse(&$1, '\\', &$0->val.str, tail);
2401 tok_err(c, "error: unsupported string suffix",
2406 ###### print exec cases
2409 struct val *v = cast(val, e);
2410 if (v->vtype == Tstr)
2412 print_value(v->vtype, &v->val);
2413 if (v->vtype == Tstr)
2418 ###### propagate exec cases
2421 struct val *val = cast(val, prog);
2422 if (!type_compat(type, val->vtype, rules))
2423 type_err(c, "error: expected %1%r found %2",
2424 prog, type, rules, val->vtype);
2428 ###### interp exec cases
2430 rvtype = cast(val, e)->vtype;
2431 dup_value(rvtype, &cast(val, e)->val, &rv);
2434 ###### ast functions
2435 static void free_val(struct val *v)
2438 free_value(v->vtype, &v->val);
2442 ###### free exec cases
2443 case Xval: free_val(cast(val, e)); break;
2445 ###### ast functions
2446 // Move all nodes from 'b' to 'rv', reversing their order.
2447 // In 'b' 'left' is a list, and 'right' is the last node.
2448 // In 'rv', left' is the first node and 'right' is a list.
2449 static struct binode *reorder_bilist(struct binode *b)
2451 struct binode *rv = NULL;
2454 struct exec *t = b->right;
2458 b = cast(binode, b->left);
2468 Just as we used a `val` to wrap a value into an `exec`, we similarly
2469 need a `var` to wrap a `variable` into an exec. While each `val`
2470 contained a copy of the value, each `var` holds a link to the variable
2471 because it really is the same variable no matter where it appears.
2472 When a variable is used, we need to remember to follow the `->merged`
2473 link to find the primary instance.
2481 struct variable *var;
2489 VariableDecl -> IDENTIFIER : ${ {
2490 struct variable *v = var_decl(c, $1.txt);
2491 $0 = new_pos(var, $1);
2496 v = var_ref(c, $1.txt);
2498 type_err(c, "error: variable '%v' redeclared",
2500 type_err(c, "info: this is where '%v' was first declared",
2501 v->where_decl, NULL, 0, NULL);
2504 | IDENTIFIER :: ${ {
2505 struct variable *v = var_decl(c, $1.txt);
2506 $0 = new_pos(var, $1);
2512 v = var_ref(c, $1.txt);
2514 type_err(c, "error: variable '%v' redeclared",
2516 type_err(c, "info: this is where '%v' was first declared",
2517 v->where_decl, NULL, 0, NULL);
2520 | IDENTIFIER : Type ${ {
2521 struct variable *v = var_decl(c, $1.txt);
2522 $0 = new_pos(var, $1);
2529 v = var_ref(c, $1.txt);
2531 type_err(c, "error: variable '%v' redeclared",
2533 type_err(c, "info: this is where '%v' was first declared",
2534 v->where_decl, NULL, 0, NULL);
2537 | IDENTIFIER :: Type ${ {
2538 struct variable *v = var_decl(c, $1.txt);
2539 $0 = new_pos(var, $1);
2547 v = var_ref(c, $1.txt);
2549 type_err(c, "error: variable '%v' redeclared",
2551 type_err(c, "info: this is where '%v' was first declared",
2552 v->where_decl, NULL, 0, NULL);
2557 Variable -> IDENTIFIER ${ {
2558 struct variable *v = var_ref(c, $1.txt);
2559 $0 = new_pos(var, $1);
2561 /* This might be a label - allocate a var just in case */
2562 v = var_decl(c, $1.txt);
2569 cast(var, $0)->var = v;
2573 ###### print exec cases
2576 struct var *v = cast(var, e);
2578 struct binding *b = v->var->name;
2579 printf("%.*s", b->name.len, b->name.txt);
2586 if (loc && loc->type == Xvar) {
2587 struct var *v = cast(var, loc);
2589 struct binding *b = v->var->name;
2590 fprintf(stderr, "%.*s", b->name.len, b->name.txt);
2592 fputs("???", stderr); // NOTEST
2594 fputs("NOTVAR", stderr); // NOTEST
2597 ###### propagate exec cases
2601 struct var *var = cast(var, prog);
2602 struct variable *v = var->var;
2604 type_err(c, "%d:BUG: no variable!!", prog, NULL, 0, NULL); // NOTEST
2605 return Tnone; // NOTEST
2609 if (v->constant && (rules & Rnoconstant)) {
2610 type_err(c, "error: Cannot assign to a constant: %v",
2611 prog, NULL, 0, NULL);
2612 type_err(c, "info: name was defined as a constant here",
2613 v->where_decl, NULL, 0, NULL);
2616 if (v->type == Tnone && v->where_decl == prog)
2617 type_err(c, "error: variable used but not declared: %v",
2618 prog, NULL, 0, NULL);
2619 if (v->type == NULL) {
2620 if (type && *ok != 0) {
2622 v->where_set = prog;
2627 if (!type_compat(type, v->type, rules)) {
2628 type_err(c, "error: expected %1%r but variable '%v' is %2", prog,
2629 type, rules, v->type);
2630 type_err(c, "info: this is where '%v' was set to %1", v->where_set,
2631 v->type, rules, NULL);
2638 ###### interp exec cases
2641 struct var *var = cast(var, e);
2642 struct variable *v = var->var;
2645 v = v->merged; // UNTESTED
2646 lrv = var_value(c, v);
2651 ###### ast functions
2653 static void free_var(struct var *v)
2658 ###### free exec cases
2659 case Xvar: free_var(cast(var, e)); break;
2661 ### Expressions: Conditional
2663 Our first user of the `binode` will be conditional expressions, which
2664 is a bit odd as they actually have three components. That will be
2665 handled by having 2 binodes for each expression. The conditional
2666 expression is the lowest precedence operator which is why we define it
2667 first - to start the precedence list.
2669 Conditional expressions are of the form "value `if` condition `else`
2670 other_value". They associate to the right, so everything to the right
2671 of `else` is part of an else value, while only a higher-precedence to
2672 the left of `if` is the if values. Between `if` and `else` there is no
2673 room for ambiguity, so a full conditional expression is allowed in
2685 Expression -> Expression if Expression else Expression $$ifelse ${ {
2686 struct binode *b1 = new(binode);
2687 struct binode *b2 = new(binode);
2696 ## expression grammar
2698 ###### print binode cases
2701 b2 = cast(binode, b->right);
2702 if (bracket) printf("(");
2703 print_exec(b2->left, -1, bracket);
2705 print_exec(b->left, -1, bracket);
2707 print_exec(b2->right, -1, bracket);
2708 if (bracket) printf(")");
2711 ###### propagate binode cases
2714 /* cond must be Tbool, others must match */
2715 struct binode *b2 = cast(binode, b->right);
2718 propagate_types(b->left, c, ok, Tbool, 0);
2719 t = propagate_types(b2->left, c, ok, type, Rnolabel);
2720 t2 = propagate_types(b2->right, c, ok, type ?: t, Rnolabel);
2724 ###### interp binode cases
2727 struct binode *b2 = cast(binode, b->right);
2728 left = interp_exec(c, b->left, <ype);
2730 rv = interp_exec(c, b2->left, &rvtype); // UNTESTED
2732 rv = interp_exec(c, b2->right, &rvtype);
2736 ### Expressions: Boolean
2738 The next class of expressions to use the `binode` will be Boolean
2739 expressions. "`and then`" and "`or else`" are similar to `and` and `or`
2740 have same corresponding precendence. The difference is that they don't
2741 evaluate the second expression if not necessary.
2750 ###### expr precedence
2755 ###### expression grammar
2756 | Expression or Expression ${ {
2757 struct binode *b = new(binode);
2763 | Expression or else Expression ${ {
2764 struct binode *b = new(binode);
2771 | Expression and Expression ${ {
2772 struct binode *b = new(binode);
2778 | Expression and then Expression ${ {
2779 struct binode *b = new(binode);
2786 | not Expression ${ {
2787 struct binode *b = new(binode);
2793 ###### print binode cases
2795 if (bracket) printf("(");
2796 print_exec(b->left, -1, bracket);
2798 print_exec(b->right, -1, bracket);
2799 if (bracket) printf(")");
2802 if (bracket) printf("(");
2803 print_exec(b->left, -1, bracket);
2804 printf(" and then ");
2805 print_exec(b->right, -1, bracket);
2806 if (bracket) printf(")");
2809 if (bracket) printf("(");
2810 print_exec(b->left, -1, bracket);
2812 print_exec(b->right, -1, bracket);
2813 if (bracket) printf(")");
2816 if (bracket) printf("(");
2817 print_exec(b->left, -1, bracket);
2818 printf(" or else ");
2819 print_exec(b->right, -1, bracket);
2820 if (bracket) printf(")");
2823 if (bracket) printf("(");
2825 print_exec(b->right, -1, bracket);
2826 if (bracket) printf(")");
2829 ###### propagate binode cases
2835 /* both must be Tbool, result is Tbool */
2836 propagate_types(b->left, c, ok, Tbool, 0);
2837 propagate_types(b->right, c, ok, Tbool, 0);
2838 if (type && type != Tbool)
2839 type_err(c, "error: %1 operation found where %2 expected", prog,
2843 ###### interp binode cases
2845 rv = interp_exec(c, b->left, &rvtype);
2846 right = interp_exec(c, b->right, &rtype);
2847 rv.bool = rv.bool && right.bool;
2850 rv = interp_exec(c, b->left, &rvtype);
2852 rv = interp_exec(c, b->right, NULL);
2855 rv = interp_exec(c, b->left, &rvtype);
2856 right = interp_exec(c, b->right, &rtype);
2857 rv.bool = rv.bool || right.bool;
2860 rv = interp_exec(c, b->left, &rvtype);
2862 rv = interp_exec(c, b->right, NULL);
2865 rv = interp_exec(c, b->right, &rvtype);
2869 ### Expressions: Comparison
2871 Of slightly higher precedence that Boolean expressions are Comparisons.
2872 A comparison takes arguments of any comparable type, but the two types
2875 To simplify the parsing we introduce an `eop` which can record an
2876 expression operator, and the `CMPop` non-terminal will match one of them.
2883 ###### ast functions
2884 static void free_eop(struct eop *e)
2898 ###### expr precedence
2899 $LEFT < > <= >= == != CMPop
2901 ###### expression grammar
2902 | Expression CMPop Expression ${ {
2903 struct binode *b = new(binode);
2913 CMPop -> < ${ $0.op = Less; }$
2914 | > ${ $0.op = Gtr; }$
2915 | <= ${ $0.op = LessEq; }$
2916 | >= ${ $0.op = GtrEq; }$
2917 | == ${ $0.op = Eql; }$
2918 | != ${ $0.op = NEql; }$
2920 ###### print binode cases
2928 if (bracket) printf("(");
2929 print_exec(b->left, -1, bracket);
2931 case Less: printf(" < "); break;
2932 case LessEq: printf(" <= "); break;
2933 case Gtr: printf(" > "); break;
2934 case GtrEq: printf(" >= "); break;
2935 case Eql: printf(" == "); break;
2936 case NEql: printf(" != "); break;
2937 default: abort(); // NOTEST
2939 print_exec(b->right, -1, bracket);
2940 if (bracket) printf(")");
2943 ###### propagate binode cases
2950 /* Both must match but not be labels, result is Tbool */
2951 t = propagate_types(b->left, c, ok, NULL, Rnolabel);
2953 propagate_types(b->right, c, ok, t, 0);
2955 t = propagate_types(b->right, c, ok, NULL, Rnolabel); // UNTESTED
2957 t = propagate_types(b->left, c, ok, t, 0); // UNTESTED
2959 if (!type_compat(type, Tbool, 0))
2960 type_err(c, "error: Comparison returns %1 but %2 expected", prog,
2961 Tbool, rules, type);
2964 ###### interp binode cases
2973 left = interp_exec(c, b->left, <ype);
2974 right = interp_exec(c, b->right, &rtype);
2975 cmp = value_cmp(ltype, rtype, &left, &right);
2978 case Less: rv.bool = cmp < 0; break;
2979 case LessEq: rv.bool = cmp <= 0; break;
2980 case Gtr: rv.bool = cmp > 0; break;
2981 case GtrEq: rv.bool = cmp >= 0; break;
2982 case Eql: rv.bool = cmp == 0; break;
2983 case NEql: rv.bool = cmp != 0; break;
2984 default: rv.bool = 0; break; // NOTEST
2989 ### Expressions: The rest
2991 The remaining expressions with the highest precedence are arithmetic,
2992 string concatenation, and string conversion. String concatenation
2993 (`++`) has the same precedence as multiplication and division, but lower
2996 String conversion is a temporary feature until I get a better type
2997 system. `$` is a prefix operator which expects a string and returns
3000 `+` and `-` are both infix and prefix operations (where they are
3001 absolute value and negation). These have different operator names.
3003 We also have a 'Bracket' operator which records where parentheses were
3004 found. This makes it easy to reproduce these when printing. Possibly I
3005 should only insert brackets were needed for precedence.
3015 ###### expr precedence
3021 ###### expression grammar
3022 | Expression Eop Expression ${ {
3023 struct binode *b = new(binode);
3030 | Expression Top Expression ${ {
3031 struct binode *b = new(binode);
3038 | ( Expression ) ${ {
3039 struct binode *b = new_pos(binode, $1);
3044 | Uop Expression ${ {
3045 struct binode *b = new(binode);
3050 | Value ${ $0 = $<1; }$
3051 | Variable ${ $0 = $<1; }$
3054 Eop -> + ${ $0.op = Plus; }$
3055 | - ${ $0.op = Minus; }$
3057 Uop -> + ${ $0.op = Absolute; }$
3058 | - ${ $0.op = Negate; }$
3059 | $ ${ $0.op = StringConv; }$
3061 Top -> * ${ $0.op = Times; }$
3062 | / ${ $0.op = Divide; }$
3063 | % ${ $0.op = Rem; }$
3064 | ++ ${ $0.op = Concat; }$
3066 ###### print binode cases
3073 if (bracket) printf("(");
3074 print_exec(b->left, indent, bracket);
3076 case Plus: fputs(" + ", stdout); break;
3077 case Minus: fputs(" - ", stdout); break;
3078 case Times: fputs(" * ", stdout); break;
3079 case Divide: fputs(" / ", stdout); break;
3080 case Rem: fputs(" % ", stdout); break;
3081 case Concat: fputs(" ++ ", stdout); break;
3082 default: abort(); // NOTEST
3084 print_exec(b->right, indent, bracket);
3085 if (bracket) printf(")");
3090 if (bracket) printf("(");
3092 case Absolute: fputs("+", stdout); break;
3093 case Negate: fputs("-", stdout); break;
3094 case StringConv: fputs("$", stdout); break;
3095 default: abort(); // NOTEST
3097 print_exec(b->right, indent, bracket);
3098 if (bracket) printf(")");
3102 print_exec(b->right, indent, bracket);
3106 ###### propagate binode cases
3112 /* both must be numbers, result is Tnum */
3115 /* as propagate_types ignores a NULL,
3116 * unary ops fit here too */
3117 propagate_types(b->left, c, ok, Tnum, 0);
3118 propagate_types(b->right, c, ok, Tnum, 0);
3119 if (!type_compat(type, Tnum, 0))
3120 type_err(c, "error: Arithmetic returns %1 but %2 expected", prog,
3125 /* both must be Tstr, result is Tstr */
3126 propagate_types(b->left, c, ok, Tstr, 0);
3127 propagate_types(b->right, c, ok, Tstr, 0);
3128 if (!type_compat(type, Tstr, 0))
3129 type_err(c, "error: Concat returns %1 but %2 expected", prog,
3134 /* op must be string, result is number */
3135 propagate_types(b->left, c, ok, Tstr, 0);
3136 if (!type_compat(type, Tnum, 0))
3137 type_err(c, // UNTESTED
3138 "error: Can only convert string to number, not %1",
3139 prog, type, 0, NULL);
3143 return propagate_types(b->right, c, ok, type, 0);
3145 ###### interp binode cases
3148 rv = interp_exec(c, b->left, &rvtype);
3149 right = interp_exec(c, b->right, &rtype);
3150 mpq_add(rv.num, rv.num, right.num);
3153 rv = interp_exec(c, b->left, &rvtype);
3154 right = interp_exec(c, b->right, &rtype);
3155 mpq_sub(rv.num, rv.num, right.num);
3158 rv = interp_exec(c, b->left, &rvtype);
3159 right = interp_exec(c, b->right, &rtype);
3160 mpq_mul(rv.num, rv.num, right.num);
3163 rv = interp_exec(c, b->left, &rvtype);
3164 right = interp_exec(c, b->right, &rtype);
3165 mpq_div(rv.num, rv.num, right.num);
3170 left = interp_exec(c, b->left, <ype);
3171 right = interp_exec(c, b->right, &rtype);
3172 mpz_init(l); mpz_init(r); mpz_init(rem);
3173 mpz_tdiv_q(l, mpq_numref(left.num), mpq_denref(left.num));
3174 mpz_tdiv_q(r, mpq_numref(right.num), mpq_denref(right.num));
3175 mpz_tdiv_r(rem, l, r);
3176 val_init(Tnum, &rv);
3177 mpq_set_z(rv.num, rem);
3178 mpz_clear(r); mpz_clear(l); mpz_clear(rem);
3183 rv = interp_exec(c, b->right, &rvtype);
3184 mpq_neg(rv.num, rv.num);
3187 rv = interp_exec(c, b->right, &rvtype);
3188 mpq_abs(rv.num, rv.num);
3191 rv = interp_exec(c, b->right, &rvtype);
3194 left = interp_exec(c, b->left, <ype);
3195 right = interp_exec(c, b->right, &rtype);
3197 rv.str = text_join(left.str, right.str);
3200 right = interp_exec(c, b->right, &rvtype);
3204 struct text tx = right.str;
3207 if (tx.txt[0] == '-') {
3208 neg = 1; // UNTESTED
3209 tx.txt++; // UNTESTED
3210 tx.len--; // UNTESTED
3212 if (number_parse(rv.num, tail, tx) == 0)
3213 mpq_init(rv.num); // UNTESTED
3215 mpq_neg(rv.num, rv.num); // UNTESTED
3217 printf("Unsupported suffix: %.*s\n", tx.len, tx.txt); // UNTESTED
3221 ###### value functions
3223 static struct text text_join(struct text a, struct text b)
3226 rv.len = a.len + b.len;
3227 rv.txt = malloc(rv.len);
3228 memcpy(rv.txt, a.txt, a.len);
3229 memcpy(rv.txt+a.len, b.txt, b.len);
3233 ### Blocks, Statements, and Statement lists.
3235 Now that we have expressions out of the way we need to turn to
3236 statements. There are simple statements and more complex statements.
3237 Simple statements do not contain (syntactic) newlines, complex statements do.
3239 Statements often come in sequences and we have corresponding simple
3240 statement lists and complex statement lists.
3241 The former comprise only simple statements separated by semicolons.
3242 The later comprise complex statements and simple statement lists. They are
3243 separated by newlines. Thus the semicolon is only used to separate
3244 simple statements on the one line. This may be overly restrictive,
3245 but I'm not sure I ever want a complex statement to share a line with
3248 Note that a simple statement list can still use multiple lines if
3249 subsequent lines are indented, so
3251 ###### Example: wrapped simple statement list
3256 is a single simple statement list. This might allow room for
3257 confusion, so I'm not set on it yet.
3259 A simple statement list needs no extra syntax. A complex statement
3260 list has two syntactic forms. It can be enclosed in braces (much like
3261 C blocks), or it can be introduced by an indent and continue until an
3262 unindented newline (much like Python blocks). With this extra syntax
3263 it is referred to as a block.
3265 Note that a block does not have to include any newlines if it only
3266 contains simple statements. So both of:
3268 if condition: a=b; d=f
3270 if condition { a=b; print f }
3274 In either case the list is constructed from a `binode` list with
3275 `Block` as the operator. When parsing the list it is most convenient
3276 to append to the end, so a list is a list and a statement. When using
3277 the list it is more convenient to consider a list to be a statement
3278 and a list. So we need a function to re-order a list.
3279 `reorder_bilist` serves this purpose.
3281 The only stand-alone statement we introduce at this stage is `pass`
3282 which does nothing and is represented as a `NULL` pointer in a `Block`
3283 list. Other stand-alone statements will follow once the infrastructure
3294 Block -> { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3295 | { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3296 | SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
3297 | SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
3298 | IN OptNL Statementlist OUT ${ $0 = $<Sl; }$
3300 OpenBlock -> OpenScope { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3301 | OpenScope { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3302 | OpenScope SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
3303 | OpenScope SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
3304 | IN OpenScope OptNL Statementlist OUT ${ $0 = $<Sl; }$
3306 UseBlock -> { OpenScope IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3307 | { OpenScope SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3308 | IN OpenScope OptNL Statementlist OUT ${ $0 = $<Sl; }$
3310 ColonBlock -> { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3311 | { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3312 | : SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
3313 | : SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
3314 | : IN OptNL Statementlist OUT ${ $0 = $<Sl; }$
3316 Statementlist -> ComplexStatements ${ $0 = reorder_bilist($<CS); }$
3318 ComplexStatements -> ComplexStatements ComplexStatement ${
3328 | ComplexStatement ${
3340 ComplexStatement -> SimpleStatements Newlines ${
3341 $0 = reorder_bilist($<SS);
3343 | SimpleStatements ; Newlines ${
3344 $0 = reorder_bilist($<SS);
3346 ## ComplexStatement Grammar
3349 SimpleStatements -> SimpleStatements ; SimpleStatement ${
3355 | SimpleStatement ${
3363 SimpleStatement -> pass ${ $0 = NULL; }$
3364 | ERROR ${ tok_err(c, "Syntax error in statement", &$1); }$
3365 ## SimpleStatement Grammar
3367 ###### print binode cases
3371 if (b->left == NULL) // UNTESTED
3372 printf("pass"); // UNTESTED
3374 print_exec(b->left, indent, bracket); // UNTESTED
3375 if (b->right) { // UNTESTED
3376 printf("; "); // UNTESTED
3377 print_exec(b->right, indent, bracket); // UNTESTED
3380 // block, one per line
3381 if (b->left == NULL)
3382 do_indent(indent, "pass\n");
3384 print_exec(b->left, indent, bracket);
3386 print_exec(b->right, indent, bracket);
3390 ###### propagate binode cases
3393 /* If any statement returns something other than Tnone
3394 * or Tbool then all such must return same type.
3395 * As each statement may be Tnone or something else,
3396 * we must always pass NULL (unknown) down, otherwise an incorrect
3397 * error might occur. We never return Tnone unless it is
3402 for (e = b; e; e = cast(binode, e->right)) {
3403 t = propagate_types(e->left, c, ok, NULL, rules);
3404 if ((rules & Rboolok) && t == Tbool)
3406 if (t && t != Tnone && t != Tbool) {
3410 type_err(c, "error: expected %1%r, found %2",
3411 e->left, type, rules, t);
3417 ###### interp binode cases
3419 while (rvtype == Tnone &&
3422 rv = interp_exec(c, b->left, &rvtype);
3423 b = cast(binode, b->right);
3427 ### The Print statement
3429 `print` is a simple statement that takes a comma-separated list of
3430 expressions and prints the values separated by spaces and terminated
3431 by a newline. No control of formatting is possible.
3433 `print` faces the same list-ordering issue as blocks, and uses the
3439 ##### expr precedence
3442 ###### SimpleStatement Grammar
3444 | print ExpressionList ${
3445 $0 = reorder_bilist($<2);
3447 | print ExpressionList , ${
3452 $0 = reorder_bilist($0);
3463 ExpressionList -> ExpressionList , Expression ${
3476 ###### print binode cases
3479 do_indent(indent, "print");
3483 print_exec(b->left, -1, bracket);
3487 b = cast(binode, b->right);
3493 ###### propagate binode cases
3496 /* don't care but all must be consistent */
3497 propagate_types(b->left, c, ok, NULL, Rnolabel);
3498 propagate_types(b->right, c, ok, NULL, Rnolabel);
3501 ###### interp binode cases
3507 for ( ; b; b = cast(binode, b->right))
3511 left = interp_exec(c, b->left, <ype);
3512 print_value(ltype, &left);
3513 free_value(ltype, &left);
3524 ###### Assignment statement
3526 An assignment will assign a value to a variable, providing it hasn't
3527 been declared as a constant. The analysis phase ensures that the type
3528 will be correct so the interpreter just needs to perform the
3529 calculation. There is a form of assignment which declares a new
3530 variable as well as assigning a value. If a name is assigned before
3531 it is declared, and error will be raised as the name is created as
3532 `Tlabel` and it is illegal to assign to such names.
3538 ###### declare terminals
3541 ###### SimpleStatement Grammar
3542 | Variable = Expression ${
3548 | VariableDecl = Expression ${
3556 if ($1->var->where_set == NULL) {
3558 "Variable declared with no type or value: %v",
3568 ###### print binode cases
3571 do_indent(indent, "");
3572 print_exec(b->left, indent, bracket);
3574 print_exec(b->right, indent, bracket);
3581 struct variable *v = cast(var, b->left)->var;
3582 do_indent(indent, "");
3583 print_exec(b->left, indent, bracket);
3584 if (cast(var, b->left)->var->constant) {
3585 if (v->where_decl == v->where_set) {
3587 type_print(v->type, stdout);
3592 if (v->where_decl == v->where_set) {
3594 type_print(v->type, stdout);
3601 print_exec(b->right, indent, bracket);
3608 ###### propagate binode cases
3612 /* Both must match and not be labels,
3613 * Type must support 'dup',
3614 * For Assign, left must not be constant.
3617 t = propagate_types(b->left, c, ok, NULL,
3618 Rnolabel | (b->op == Assign ? Rnoconstant : 0));
3623 if (propagate_types(b->right, c, ok, t, 0) != t)
3624 if (b->left->type == Xvar)
3625 type_err(c, "info: variable '%v' was set as %1 here.",
3626 cast(var, b->left)->var->where_set, t, rules, NULL);
3628 t = propagate_types(b->right, c, ok, NULL, Rnolabel);
3630 propagate_types(b->left, c, ok, t,
3631 (b->op == Assign ? Rnoconstant : 0));
3633 if (t && t->dup == NULL)
3634 type_err(c, "error: cannot assign value of type %1", b, t, 0, NULL);
3639 ###### interp binode cases
3642 lleft = linterp_exec(c, b->left, <ype);
3643 right = interp_exec(c, b->right, &rtype);
3645 free_value(ltype, lleft);
3646 dup_value(ltype, &right, lleft);
3653 struct variable *v = cast(var, b->left)->var;
3657 val = var_value(c, v);
3658 free_value(v->type, val);
3659 if (v->type->prepare_type)
3660 v->type->prepare_type(c, v->type, 0);
3662 right = interp_exec(c, b->right, &rtype);
3663 memcpy(val, &right, rtype->size);
3666 val_init(v->type, val);
3671 ### The `use` statement
3673 The `use` statement is the last "simple" statement. It is needed when
3674 the condition in a conditional statement is a block. `use` works much
3675 like `return` in C, but only completes the `condition`, not the whole
3681 ###### expr precedence
3684 ###### SimpleStatement Grammar
3686 $0 = new_pos(binode, $1);
3689 if ($0->right->type == Xvar) {
3690 struct var *v = cast(var, $0->right);
3691 if (v->var->type == Tnone) {
3692 /* Convert this to a label */
3695 v->var->type = Tlabel;
3696 val = global_alloc(c, Tlabel, v->var, NULL);
3702 ###### print binode cases
3705 do_indent(indent, "use ");
3706 print_exec(b->right, -1, bracket);
3711 ###### propagate binode cases
3714 /* result matches value */
3715 return propagate_types(b->right, c, ok, type, 0);
3717 ###### interp binode cases
3720 rv = interp_exec(c, b->right, &rvtype);
3723 ### The Conditional Statement
3725 This is the biggy and currently the only complex statement. This
3726 subsumes `if`, `while`, `do/while`, `switch`, and some parts of `for`.
3727 It is comprised of a number of parts, all of which are optional though
3728 set combinations apply. Each part is (usually) a key word (`then` is
3729 sometimes optional) followed by either an expression or a code block,
3730 except the `casepart` which is a "key word and an expression" followed
3731 by a code block. The code-block option is valid for all parts and,
3732 where an expression is also allowed, the code block can use the `use`
3733 statement to report a value. If the code block does not report a value
3734 the effect is similar to reporting `True`.
3736 The `else` and `case` parts, as well as `then` when combined with
3737 `if`, can contain a `use` statement which will apply to some
3738 containing conditional statement. `for` parts, `do` parts and `then`
3739 parts used with `for` can never contain a `use`, except in some
3740 subordinate conditional statement.
3742 If there is a `forpart`, it is executed first, only once.
3743 If there is a `dopart`, then it is executed repeatedly providing
3744 always that the `condpart` or `cond`, if present, does not return a non-True
3745 value. `condpart` can fail to return any value if it simply executes
3746 to completion. This is treated the same as returning `True`.
3748 If there is a `thenpart` it will be executed whenever the `condpart`
3749 or `cond` returns True (or does not return any value), but this will happen
3750 *after* `dopart` (when present).
3752 If `elsepart` is present it will be executed at most once when the
3753 condition returns `False` or some value that isn't `True` and isn't
3754 matched by any `casepart`. If there are any `casepart`s, they will be
3755 executed when the condition returns a matching value.
3757 The particular sorts of values allowed in case parts has not yet been
3758 determined in the language design, so nothing is prohibited.
3760 The various blocks in this complex statement potentially provide scope
3761 for variables as described earlier. Each such block must include the
3762 "OpenScope" nonterminal before parsing the block, and must call
3763 `var_block_close()` when closing the block.
3765 The code following "`if`", "`switch`" and "`for`" does not get its own
3766 scope, but is in a scope covering the whole statement, so names
3767 declared there cannot be redeclared elsewhere. Similarly the
3768 condition following "`while`" is in a scope the covers the body
3769 ("`do`" part) of the loop, and which does not allow conditional scope
3770 extension. Code following "`then`" (both looping and non-looping),
3771 "`else`" and "`case`" each get their own local scope.
3773 The type requirements on the code block in a `whilepart` are quite
3774 unusal. It is allowed to return a value of some identifiable type, in
3775 which case the loop aborts and an appropriate `casepart` is run, or it
3776 can return a Boolean, in which case the loop either continues to the
3777 `dopart` (on `True`) or aborts and runs the `elsepart` (on `False`).
3778 This is different both from the `ifpart` code block which is expected to
3779 return a Boolean, or the `switchpart` code block which is expected to
3780 return the same type as the casepart values. The correct analysis of
3781 the type of the `whilepart` code block is the reason for the
3782 `Rboolok` flag which is passed to `propagate_types()`.
3784 The `cond_statement` cannot fit into a `binode` so a new `exec` is
3793 struct exec *action;
3794 struct casepart *next;
3796 struct cond_statement {
3798 struct exec *forpart, *condpart, *dopart, *thenpart, *elsepart;
3799 struct casepart *casepart;
3802 ###### ast functions
3804 static void free_casepart(struct casepart *cp)
3808 free_exec(cp->value);
3809 free_exec(cp->action);
3816 static void free_cond_statement(struct cond_statement *s)
3820 free_exec(s->forpart);
3821 free_exec(s->condpart);
3822 free_exec(s->dopart);
3823 free_exec(s->thenpart);
3824 free_exec(s->elsepart);
3825 free_casepart(s->casepart);
3829 ###### free exec cases
3830 case Xcond_statement: free_cond_statement(cast(cond_statement, e)); break;
3832 ###### ComplexStatement Grammar
3833 | CondStatement ${ $0 = $<1; }$
3835 ###### expr precedence
3836 $TERM for then while do
3843 // A CondStatement must end with EOL, as does CondSuffix and
3845 // ForPart, ThenPart, SwitchPart, CasePart are non-empty and
3846 // may or may not end with EOL
3847 // WhilePart and IfPart include an appropriate Suffix
3849 // Both ForPart and Whilepart open scopes, and CondSuffix only
3850 // closes one - so in the first branch here we have another to close.
3851 CondStatement -> ForPart OptNL ThenPart OptNL WhilePart CondSuffix ${
3854 $0->thenpart = $<TP;
3855 $0->condpart = $WP.condpart; $WP.condpart = NULL;
3856 $0->dopart = $WP.dopart; $WP.dopart = NULL;
3857 var_block_close(c, CloseSequential);
3859 | ForPart OptNL WhilePart CondSuffix ${
3862 $0->condpart = $WP.condpart; $WP.condpart = NULL;
3863 $0->dopart = $WP.dopart; $WP.dopart = NULL;
3864 var_block_close(c, CloseSequential);
3866 | WhilePart CondSuffix ${
3868 $0->condpart = $WP.condpart; $WP.condpart = NULL;
3869 $0->dopart = $WP.dopart; $WP.dopart = NULL;
3871 | SwitchPart OptNL CasePart CondSuffix ${
3873 $0->condpart = $<SP;
3874 $CP->next = $0->casepart;
3875 $0->casepart = $<CP;
3877 | SwitchPart : IN OptNL CasePart CondSuffix OUT Newlines ${
3879 $0->condpart = $<SP;
3880 $CP->next = $0->casepart;
3881 $0->casepart = $<CP;
3883 | IfPart IfSuffix ${
3885 $0->condpart = $IP.condpart; $IP.condpart = NULL;
3886 $0->thenpart = $IP.thenpart; $IP.thenpart = NULL;
3887 // This is where we close an "if" statement
3888 var_block_close(c, CloseSequential);
3891 CondSuffix -> IfSuffix ${
3893 // This is where we close scope of the whole
3894 // "for" or "while" statement
3895 var_block_close(c, CloseSequential);
3897 | Newlines CasePart CondSuffix ${
3899 $CP->next = $0->casepart;
3900 $0->casepart = $<CP;
3902 | CasePart CondSuffix ${
3904 $CP->next = $0->casepart;
3905 $0->casepart = $<CP;
3908 IfSuffix -> Newlines ${ $0 = new(cond_statement); }$
3909 | Newlines ElsePart ${ $0 = $<EP; }$
3910 | ElsePart ${$0 = $<EP; }$
3912 ElsePart -> else OpenBlock Newlines ${
3913 $0 = new(cond_statement);
3914 $0->elsepart = $<OB;
3915 var_block_close(c, CloseElse);
3917 | else OpenScope CondStatement ${
3918 $0 = new(cond_statement);
3919 $0->elsepart = $<CS;
3920 var_block_close(c, CloseElse);
3924 CasePart -> case Expression OpenScope ColonBlock ${
3925 $0 = calloc(1,sizeof(struct casepart));
3928 var_block_close(c, CloseParallel);
3932 // These scopes are closed in CondSuffix
3933 ForPart -> for OpenBlock ${
3937 ThenPart -> then OpenBlock ${
3939 var_block_close(c, CloseSequential);
3943 // This scope is closed in CondSuffix
3944 WhilePart -> while UseBlock OptNL do Block ${
3948 | while OpenScope Expression ColonBlock ${
3949 $0.condpart = $<Exp;
3953 IfPart -> if UseBlock OptNL then OpenBlock ClosePara ${
3957 | if OpenScope Expression OpenScope ColonBlock ClosePara ${
3961 | if OpenScope Expression OpenScope OptNL then Block ClosePara ${
3967 // This scope is closed in CondSuffix
3968 SwitchPart -> switch OpenScope Expression ${
3971 | switch UseBlock ${
3975 ###### print exec cases
3977 case Xcond_statement:
3979 struct cond_statement *cs = cast(cond_statement, e);
3980 struct casepart *cp;
3982 do_indent(indent, "for");
3983 if (bracket) printf(" {\n"); else printf("\n");
3984 print_exec(cs->forpart, indent+1, bracket);
3987 do_indent(indent, "} then {\n");
3989 do_indent(indent, "then\n");
3990 print_exec(cs->thenpart, indent+1, bracket);
3992 if (bracket) do_indent(indent, "}\n");
3996 if (cs->condpart && cs->condpart->type == Xbinode &&
3997 cast(binode, cs->condpart)->op == Block) {
3999 do_indent(indent, "while {\n");
4001 do_indent(indent, "while\n");
4002 print_exec(cs->condpart, indent+1, bracket);
4004 do_indent(indent, "} do {\n");
4006 do_indent(indent, "do\n");
4007 print_exec(cs->dopart, indent+1, bracket);
4009 do_indent(indent, "}\n");
4011 do_indent(indent, "while ");
4012 print_exec(cs->condpart, 0, bracket);
4017 print_exec(cs->dopart, indent+1, bracket);
4019 do_indent(indent, "}\n");
4024 do_indent(indent, "switch");
4026 do_indent(indent, "if");
4027 if (cs->condpart && cs->condpart->type == Xbinode &&
4028 cast(binode, cs->condpart)->op == Block) {
4029 if (bracket) // UNTESTED
4030 printf(" {\n"); // UNTESTED
4032 printf(":\n"); // UNTESTED
4033 print_exec(cs->condpart, indent+1, bracket); // UNTESTED
4034 if (bracket) // UNTESTED
4035 do_indent(indent, "}\n"); // UNTESTED
4036 if (cs->thenpart) { // UNTESTED
4037 do_indent(indent, "then:\n"); // UNTESTED
4038 print_exec(cs->thenpart, indent+1, bracket); // UNTESTED
4042 print_exec(cs->condpart, 0, bracket);
4048 print_exec(cs->thenpart, indent+1, bracket);
4050 do_indent(indent, "}\n");
4055 for (cp = cs->casepart; cp; cp = cp->next) {
4056 do_indent(indent, "case ");
4057 print_exec(cp->value, -1, 0);
4062 print_exec(cp->action, indent+1, bracket);
4064 do_indent(indent, "}\n");
4067 do_indent(indent, "else");
4072 print_exec(cs->elsepart, indent+1, bracket);
4074 do_indent(indent, "}\n");
4079 ###### propagate exec cases
4080 case Xcond_statement:
4082 // forpart and dopart must return Tnone
4083 // thenpart must return Tnone if there is a dopart,
4084 // otherwise it is like elsepart.
4086 // be bool if there is no casepart
4087 // match casepart->values if there is a switchpart
4088 // either be bool or match casepart->value if there
4090 // elsepart and casepart->action must match the return type
4091 // expected of this statement.
4092 struct cond_statement *cs = cast(cond_statement, prog);
4093 struct casepart *cp;
4095 t = propagate_types(cs->forpart, c, ok, Tnone, 0);
4096 if (!type_compat(Tnone, t, 0))
4097 *ok = 0; // UNTESTED
4098 t = propagate_types(cs->dopart, c, ok, Tnone, 0);
4099 if (!type_compat(Tnone, t, 0))
4100 *ok = 0; // UNTESTED
4102 t = propagate_types(cs->thenpart, c, ok, Tnone, 0);
4103 if (!type_compat(Tnone, t, 0))
4104 *ok = 0; // UNTESTED
4106 if (cs->casepart == NULL)
4107 propagate_types(cs->condpart, c, ok, Tbool, 0);
4109 /* Condpart must match case values, with bool permitted */
4111 for (cp = cs->casepart;
4112 cp && !t; cp = cp->next)
4113 t = propagate_types(cp->value, c, ok, NULL, 0);
4114 if (!t && cs->condpart)
4115 t = propagate_types(cs->condpart, c, ok, NULL, Rboolok); // UNTESTED
4116 // Now we have a type (I hope) push it down
4118 for (cp = cs->casepart; cp; cp = cp->next)
4119 propagate_types(cp->value, c, ok, t, 0);
4120 propagate_types(cs->condpart, c, ok, t, Rboolok);
4123 // (if)then, else, and case parts must return expected type.
4124 if (!cs->dopart && !type)
4125 type = propagate_types(cs->thenpart, c, ok, NULL, rules);
4127 type = propagate_types(cs->elsepart, c, ok, NULL, rules);
4128 for (cp = cs->casepart;
4130 cp = cp->next) // UNTESTED
4131 type = propagate_types(cp->action, c, ok, NULL, rules); // UNTESTED
4134 propagate_types(cs->thenpart, c, ok, type, rules);
4135 propagate_types(cs->elsepart, c, ok, type, rules);
4136 for (cp = cs->casepart; cp ; cp = cp->next)
4137 propagate_types(cp->action, c, ok, type, rules);
4143 ###### interp exec cases
4144 case Xcond_statement:
4146 struct value v, cnd;
4147 struct type *vtype, *cndtype;
4148 struct casepart *cp;
4149 struct cond_statement *cs = cast(cond_statement, e);
4152 interp_exec(c, cs->forpart, NULL);
4155 cnd = interp_exec(c, cs->condpart, &cndtype);
4157 cndtype = Tnone; // UNTESTED
4158 if (!(cndtype == Tnone ||
4159 (cndtype == Tbool && cnd.bool != 0)))
4161 // cnd is Tnone or Tbool, doesn't need to be freed
4163 interp_exec(c, cs->dopart, NULL);
4166 rv = interp_exec(c, cs->thenpart, &rvtype);
4167 if (rvtype != Tnone || !cs->dopart)
4169 free_value(rvtype, &rv);
4172 } while (cs->dopart);
4174 for (cp = cs->casepart; cp; cp = cp->next) {
4175 v = interp_exec(c, cp->value, &vtype);
4176 if (value_cmp(cndtype, vtype, &v, &cnd) == 0) {
4177 free_value(vtype, &v);
4178 free_value(cndtype, &cnd);
4179 rv = interp_exec(c, cp->action, &rvtype);
4182 free_value(vtype, &v);
4184 free_value(cndtype, &cnd);
4186 rv = interp_exec(c, cs->elsepart, &rvtype);
4193 ### Top level structure
4195 All the language elements so far can be used in various places. Now
4196 it is time to clarify what those places are.
4198 At the top level of a file there will be a number of declarations.
4199 Many of the things that can be declared haven't been described yet,
4200 such as functions, procedures, imports, and probably more.
4201 For now there are two sorts of things that can appear at the top
4202 level. They are predefined constants, `struct` types, and the `main`
4203 function. While the syntax will allow the `main` function to appear
4204 multiple times, that will trigger an error if it is actually attempted.
4206 The various declarations do not return anything. They store the
4207 various declarations in the parse context.
4209 ###### Parser: grammar
4212 Ocean -> OptNL DeclarationList
4214 ## declare terminals
4221 DeclarationList -> Declaration
4222 | DeclarationList Declaration
4224 Declaration -> ERROR Newlines ${
4225 tok_err(c, // UNTESTED
4226 "error: unhandled parse error", &$1);
4232 ## top level grammar
4236 ### The `const` section
4238 As well as being defined in with the code that uses them, constants
4239 can be declared at the top level. These have full-file scope, so they
4240 are always `InScope`. The value of a top level constant can be given
4241 as an expression, and this is evaluated immediately rather than in the
4242 later interpretation stage. Once we add functions to the language, we
4243 will need rules concern which, if any, can be used to define a top
4246 Constants are defined in a section that starts with the reserved word
4247 `const` and then has a block with a list of assignment statements.
4248 For syntactic consistency, these must use the double-colon syntax to
4249 make it clear that they are constants. Type can also be given: if
4250 not, the type will be determined during analysis, as with other
4253 As the types constants are inserted at the head of a list, printing
4254 them in the same order that they were read is not straight forward.
4255 We take a quadratic approach here and count the number of constants
4256 (variables of depth 0), then count down from there, each time
4257 searching through for the Nth constant for decreasing N.
4259 ###### top level grammar
4263 DeclareConstant -> const { IN OptNL ConstList OUT OptNL } Newlines
4264 | const { SimpleConstList } Newlines
4265 | const IN OptNL ConstList OUT Newlines
4266 | const SimpleConstList Newlines
4268 ConstList -> ConstList SimpleConstLine
4270 SimpleConstList -> SimpleConstList ; Const
4273 SimpleConstLine -> SimpleConstList Newlines
4274 | ERROR Newlines ${ tok_err(c, "Syntax error in constant", &$1); }$
4277 CType -> Type ${ $0 = $<1; }$
4280 Const -> IDENTIFIER :: CType = Expression ${ {
4284 v = var_decl(c, $1.txt);
4286 struct var *var = new_pos(var, $1);
4287 v->where_decl = var;
4292 v = var_ref(c, $1.txt);
4293 tok_err(c, "error: name already declared", &$1);
4294 type_err(c, "info: this is where '%v' was first declared",
4295 v->where_decl, NULL, 0, NULL);
4299 propagate_types($5, c, &ok, $3, 0);
4304 struct value res = interp_exec(c, $5, &v->type);
4305 global_alloc(c, v->type, v, &res);
4309 ###### print const decls
4314 while (target != 0) {
4316 for (v = context.in_scope; v; v=v->in_scope)
4317 if (v->depth == 0) {
4328 struct value *val = var_value(&context, v);
4329 printf(" %.*s :: ", v->name->name.len, v->name->name.txt);
4330 type_print(v->type, stdout);
4332 if (v->type == Tstr)
4334 print_value(v->type, val);
4335 if (v->type == Tstr)
4343 ### Finally the whole `main` function.
4345 An Ocean program can currently have only one function - `main` - and
4346 that must exist. It expects an array of strings with a provided size.
4347 Following this is a `block` which is the code to execute.
4349 As this is the top level, several things are handled a bit
4351 The function is not interpreted by `interp_exec` as that isn't
4352 passed the argument list which the program requires. Similarly type
4353 analysis is a bit more interesting at this level.
4355 ###### top level grammar
4357 DeclareFunction -> MainFunction ${ {
4359 type_err(c, "\"main\" defined a second time",
4365 ###### print binode cases
4368 do_indent(indent, "func main(");
4369 for (b2 = cast(binode, b->left); b2; b2 = cast(binode, b2->right)) {
4370 struct variable *v = cast(var, b2->left)->var;
4372 print_exec(b2->left, 0, 0);
4374 type_print(v->type, stdout);
4380 print_exec(b->right, indent+1, bracket);
4382 do_indent(indent, "}\n");
4385 ###### propagate binode cases
4387 case Func: abort(); // NOTEST
4389 ###### core functions
4391 static int analyse_prog(struct exec *prog, struct parse_context *c)
4393 struct binode *bp = cast(binode, prog);
4397 struct type *argv_type;
4398 struct text argv_type_name = { " argv", 5 };
4403 argv_type = add_type(c, argv_type_name, &array_prototype);
4404 argv_type->array.member = Tstr;
4405 argv_type->array.unspec = 1;
4407 for (b = cast(binode, bp->left); b; b = cast(binode, b->right)) {
4411 propagate_types(b->left, c, &ok, argv_type, 0);
4413 default: /* invalid */ // NOTEST
4414 propagate_types(b->left, c, &ok, Tnone, 0); // NOTEST
4420 propagate_types(bp->right, c, &ok, Tnone, 0);
4425 /* Make sure everything is still consistent */
4426 propagate_types(bp->right, c, &ok, Tnone, 0);
4428 return 0; // UNTESTED
4433 static void interp_prog(struct parse_context *c, struct exec *prog,
4434 int argc, char **argv)
4436 struct binode *p = cast(binode, prog);
4444 al = cast(binode, p->left);
4446 struct var *v = cast(var, al->left);
4447 struct value *vl = var_value(c, v->var);
4457 mpq_set_ui(argcq, argc, 1);
4458 memcpy(var_value(c, t->array.vsize), &argcq, sizeof(argcq));
4459 t->prepare_type(c, t, 0);
4460 array_init(v->var->type, vl);
4461 for (i = 0; i < argc; i++) {
4462 struct value *vl2 = vl->array + i * v->var->type->array.member->size;
4465 arg.str.txt = argv[i];
4466 arg.str.len = strlen(argv[i]);
4467 free_value(Tstr, vl2);
4468 dup_value(Tstr, &arg, vl2);
4472 al = cast(binode, al->right);
4474 v = interp_exec(c, p->right, &vtype);
4475 free_value(vtype, &v);
4478 ###### interp binode cases
4480 case Func: abort(); // NOTEST
4482 ## And now to test it out.
4484 Having a language requires having a "hello world" program. I'll
4485 provide a little more than that: a program that prints "Hello world"
4486 finds the GCD of two numbers, prints the first few elements of
4487 Fibonacci, performs a binary search for a number, and a few other
4488 things which will likely grow as the languages grows.
4490 ###### File: oceani.mk
4493 @echo "===== DEMO ====="
4494 ./oceani --section "demo: hello" oceani.mdc 55 33
4500 four ::= 2 + 2 ; five ::= 10/2
4501 const pie ::= "I like Pie";
4502 cake ::= "The cake is"
4513 print "Hello World, what lovely oceans you have!"
4514 print "Are there", five, "?"
4515 print pi, pie, "but", cake
4517 A := $argv[1]; B := $argv[2]
4519 /* When a variable is defined in both branches of an 'if',
4520 * and used afterwards, the variables are merged.
4526 print "Is", A, "bigger than", B,"? ", bigger
4527 /* If a variable is not used after the 'if', no
4528 * merge happens, so types can be different
4531 double:string = "yes"
4532 print A, "is more than twice", B, "?", double
4535 print "double", B, "is", double
4540 if a > 0 and then b > 0:
4546 print "GCD of", A, "and", B,"is", a
4548 print a, "is not positive, cannot calculate GCD"
4550 print b, "is not positive, cannot calculate GCD"
4555 print "Fibonacci:", f1,f2,
4556 then togo = togo - 1
4564 /* Binary search... */
4569 mid := (lo + hi) / 2
4581 print "Yay, I found", target
4583 print "Closest I found was", mid
4588 // "middle square" PRNG. Not particularly good, but one my
4589 // Dad taught me - the first one I ever heard of.
4590 for i:=1; then i = i + 1; while i < size:
4591 n := list[i-1] * list[i-1]
4592 list[i] = (n / 100) % 10 000
4594 print "Before sort:",
4595 for i:=0; then i = i + 1; while i < size:
4599 for i := 1; then i=i+1; while i < size:
4600 for j:=i-1; then j=j-1; while j >= 0:
4601 if list[j] > list[j+1]:
4605 print " After sort:",
4606 for i:=0; then i = i + 1; while i < size:
4610 if 1 == 2 then print "yes"; else print "no"
4614 bob.alive = (bob.name == "Hello")
4615 print "bob", "is" if bob.alive else "isn't", "alive"