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;
122 #define container_of(ptr, type, member) ({ \
123 const typeof( ((type *)0)->member ) *__mptr = (ptr); \
124 (type *)( (char *)__mptr - offsetof(type,member) );})
126 #define config2context(_conf) container_of(_conf, struct parse_context, \
129 ###### Parser: reduce
130 struct parse_context *c = config2context(config);
138 #include <sys/mman.h>
157 static char Usage[] =
158 "Usage: oceani --trace --print --noexec --brackets --section=SectionName prog.ocn\n";
159 static const struct option long_options[] = {
160 {"trace", 0, NULL, 't'},
161 {"print", 0, NULL, 'p'},
162 {"noexec", 0, NULL, 'n'},
163 {"brackets", 0, NULL, 'b'},
164 {"section", 1, NULL, 's'},
167 const char *options = "tpnbs";
169 static void pr_err(char *msg) // NOTEST
171 fprintf(stderr, "%s\n", msg); // NOTEST
174 int main(int argc, char *argv[])
179 struct section *s = NULL, *ss;
180 char *section = NULL;
181 struct parse_context context = {
183 .ignored = (1 << TK_mark),
184 .number_chars = ".,_+- ",
189 int doprint=0, dotrace=0, doexec=1, brackets=0;
191 while ((opt = getopt_long(argc, argv, options, long_options, NULL))
194 case 't': dotrace=1; break;
195 case 'p': doprint=1; break;
196 case 'n': doexec=0; break;
197 case 'b': brackets=1; break;
198 case 's': section = optarg; break;
199 default: fprintf(stderr, Usage);
203 if (optind >= argc) {
204 fprintf(stderr, "oceani: no input file given\n");
207 fd = open(argv[optind], O_RDONLY);
209 fprintf(stderr, "oceani: cannot open %s\n", argv[optind]);
212 context.file_name = argv[optind];
213 len = lseek(fd, 0, 2);
214 file = mmap(NULL, len, PROT_READ, MAP_SHARED, fd, 0);
215 s = code_extract(file, file+len, pr_err);
217 fprintf(stderr, "oceani: could not find any code in %s\n",
222 ## context initialization
225 for (ss = s; ss; ss = ss->next) {
226 struct text sec = ss->section;
227 if (sec.len == strlen(section) &&
228 strncmp(sec.txt, section, sec.len) == 0)
232 fprintf(stderr, "oceani: cannot find section %s\n",
239 fprintf(stderr, "oceani: no code found in requested section\n"); // NOTEST
240 goto cleanup; // NOTEST
243 parse_oceani(ss->code, &context.config, dotrace ? stderr : NULL);
245 resolve_consts(&context);
246 prepare_types(&context);
247 if (!context.parse_error && !analyse_funcs(&context)) {
248 fprintf(stderr, "oceani: type error in program - not running.\n");
249 context.parse_error += 1;
257 if (doexec && !context.parse_error)
258 interp_main(&context, argc - optind, argv + optind);
261 struct section *t = s->next;
266 // FIXME parser should pop scope even on error
267 while (context.scope_depth > 0)
271 ## free context types
272 ## free context storage
273 exit(context.parse_error ? 1 : 0);
278 The four requirements of parse, analyse, print, interpret apply to
279 each language element individually so that is how most of the code
282 Three of the four are fairly self explanatory. The one that requires
283 a little explanation is the analysis step.
285 The current language design does not require the types of variables to
286 be declared, but they must still have a single type. Different
287 operations impose different requirements on the variables, for example
288 addition requires both arguments to be numeric, and assignment
289 requires the variable on the left to have the same type as the
290 expression on the right.
292 Analysis involves propagating these type requirements around and
293 consequently setting the type of each variable. If any requirements
294 are violated (e.g. a string is compared with a number) or if a
295 variable needs to have two different types, then an error is raised
296 and the program will not run.
298 If the same variable is declared in both branchs of an 'if/else', or
299 in all cases of a 'switch' then the multiple instances may be merged
300 into just one variable if the variable is referenced after the
301 conditional statement. When this happens, the types must naturally be
302 consistent across all the branches. When the variable is not used
303 outside the if, the variables in the different branches are distinct
304 and can be of different types.
306 Undeclared names may only appear in "use" statements and "case" expressions.
307 These names are given a type of "label" and a unique value.
308 This allows them to fill the role of a name in an enumerated type, which
309 is useful for testing the `switch` statement.
311 As we will see, the condition part of a `while` statement can return
312 either a Boolean or some other type. This requires that the expected
313 type that gets passed around comprises a type and a flag to indicate
314 that `Tbool` is also permitted.
316 As there are, as yet, no distinct types that are compatible, there
317 isn't much subtlety in the analysis. When we have distinct number
318 types, this will become more interesting.
322 When analysis discovers an inconsistency it needs to report an error;
323 just refusing to run the code ensures that the error doesn't cascade,
324 but by itself it isn't very useful. A clear understanding of the sort
325 of error message that are useful will help guide the process of
328 At a simplistic level, the only sort of error that type analysis can
329 report is that the type of some construct doesn't match a contextual
330 requirement. For example, in `4 + "hello"` the addition provides a
331 contextual requirement for numbers, but `"hello"` is not a number. In
332 this particular example no further information is needed as the types
333 are obvious from local information. When a variable is involved that
334 isn't the case. It may be helpful to explain why the variable has a
335 particular type, by indicating the location where the type was set,
336 whether by declaration or usage.
338 Using a recursive-descent analysis we can easily detect a problem at
339 multiple locations. In "`hello:= "there"; 4 + hello`" the addition
340 will detect that one argument is not a number and the usage of `hello`
341 will detect that a number was wanted, but not provided. In this
342 (early) version of the language, we will generate error reports at
343 multiple locations, so the use of `hello` will report an error and
344 explain were the value was set, and the addition will report an error
345 and say why numbers are needed. To be able to report locations for
346 errors, each language element will need to record a file location
347 (line and column) and each variable will need to record the language
348 element where its type was set. For now we will assume that each line
349 of an error message indicates one location in the file, and up to 2
350 types. So we provide a `printf`-like function which takes a format, a
351 location (a `struct exec` which has not yet been introduced), and 2
352 types. "`%1`" reports the first type, "`%2`" reports the second. We
353 will need a function to print the location, once we know how that is
354 stored. e As will be explained later, there are sometimes extra rules for
355 type matching and they might affect error messages, we need to pass those
358 As well as type errors, we sometimes need to report problems with
359 tokens, which might be unexpected or might name a type that has not
360 been defined. For these we have `tok_err()` which reports an error
361 with a given token. Each of the error functions sets the flag in the
362 context so indicate that parsing failed.
366 static void fput_loc(struct exec *loc, FILE *f);
367 static void type_err(struct parse_context *c,
368 char *fmt, struct exec *loc,
369 struct type *t1, int rules, struct type *t2);
371 ###### core functions
373 static void type_err(struct parse_context *c,
374 char *fmt, struct exec *loc,
375 struct type *t1, int rules, struct type *t2)
377 fprintf(stderr, "%s:", c->file_name);
378 fput_loc(loc, stderr);
379 for (; *fmt ; fmt++) {
386 case '%': fputc(*fmt, stderr); break; // NOTEST
387 default: fputc('?', stderr); break; // NOTEST
389 type_print(t1, stderr);
392 type_print(t2, stderr);
401 static void tok_err(struct parse_context *c, char *fmt, struct token *t)
403 fprintf(stderr, "%s:%d:%d: %s: %.*s\n", c->file_name, t->line, t->col, fmt,
404 t->txt.len, t->txt.txt);
408 ## Entities: declared and predeclared.
410 There are various "things" that the language and/or the interpreter
411 needs to know about to parse and execute a program. These include
412 types, variables, values, and executable code. These are all lumped
413 together under the term "entities" (calling them "objects" would be
414 confusing) and introduced here. The following section will present the
415 different specific code elements which comprise or manipulate these
420 Executables can be lots of different things. In many cases an
421 executable is just an operation combined with one or two other
422 executables. This allows for expressions and lists etc. Other times an
423 executable is something quite specific like a constant or variable name.
424 So we define a `struct exec` to be a general executable with a type, and
425 a `struct binode` which is a subclass of `exec`, forms a node in a
426 binary tree, and holds an operation. There will be other subclasses,
427 and to access these we need to be able to `cast` the `exec` into the
428 various other types. The first field in any `struct exec` is the type
429 from the `exec_types` enum.
432 #define cast(structname, pointer) ({ \
433 const typeof( ((struct structname *)0)->type) *__mptr = &(pointer)->type; \
434 if (__mptr && *__mptr != X##structname) abort(); \
435 (struct structname *)( (char *)__mptr);})
437 #define new(structname) ({ \
438 struct structname *__ptr = ((struct structname *)calloc(1,sizeof(struct structname))); \
439 __ptr->type = X##structname; \
440 __ptr->line = -1; __ptr->column = -1; \
443 #define new_pos(structname, token) ({ \
444 struct structname *__ptr = ((struct structname *)calloc(1,sizeof(struct structname))); \
445 __ptr->type = X##structname; \
446 __ptr->line = token.line; __ptr->column = token.col; \
455 enum exec_types type;
464 struct exec *left, *right;
469 static int __fput_loc(struct exec *loc, FILE *f)
473 if (loc->line >= 0) {
474 fprintf(f, "%d:%d: ", loc->line, loc->column);
477 if (loc->type == Xbinode)
478 return __fput_loc(cast(binode,loc)->left, f) ||
479 __fput_loc(cast(binode,loc)->right, f); // NOTEST
482 static void fput_loc(struct exec *loc, FILE *f)
484 if (!__fput_loc(loc, f))
485 fprintf(f, "??:??: ");
488 Each different type of `exec` node needs a number of functions defined,
489 a bit like methods. We must be able to free it, print it, analyse it
490 and execute it. Once we have specific `exec` types we will need to
491 parse them too. Let's take this a bit more slowly.
495 The parser generator requires a `free_foo` function for each struct
496 that stores attributes and they will often be `exec`s and subtypes
497 there-of. So we need `free_exec` which can handle all the subtypes,
498 and we need `free_binode`.
502 static void free_binode(struct binode *b)
511 ###### core functions
512 static void free_exec(struct exec *e)
523 static void free_exec(struct exec *e);
525 ###### free exec cases
526 case Xbinode: free_binode(cast(binode, e)); break;
530 Printing an `exec` requires that we know the current indent level for
531 printing line-oriented components. As will become clear later, we
532 also want to know what sort of bracketing to use.
536 static void do_indent(int i, char *str)
543 ###### core functions
544 static void print_binode(struct binode *b, int indent, int bracket)
548 ## print binode cases
552 static void print_exec(struct exec *e, int indent, int bracket)
558 print_binode(cast(binode, e), indent, bracket); break;
563 do_indent(indent, "/* FREE");
564 for (v = e->to_free; v; v = v->next_free) {
565 printf(" %.*s", v->name->name.len, v->name->name.txt);
566 printf("[%d,%d]", v->scope_start, v->scope_end);
567 if (v->frame_pos >= 0)
568 printf("(%d+%d)", v->frame_pos,
569 v->type ? v->type->size:0);
577 static void print_exec(struct exec *e, int indent, int bracket);
581 As discussed, analysis involves propagating type requirements around the
582 program and looking for errors.
584 So `propagate_types` is passed an expected type (being a `struct type`
585 pointer together with some `val_rules` flags) that the `exec` is
586 expected to return, and returns the type that it does return, either of
587 which can be `NULL` signifying "unknown". A `prop_err` flag set is
588 passed by reference. It has `Efail` set when an error is found, and
589 `Eretry` when the type for some element is set via propagation. If
590 any expression cannot be evaluated immediately, `Enoconst` is set.
592 If it remains unchanged at `0`, then no more propagation is needed.
596 enum val_rules {Rnolabel = 1<<0, Rboolok = 1<<1, Rnoconstant = 1<<2};
597 enum prop_err {Efail = 1<<0, Eretry = 1<<1, Enoconst = 1<<2};
601 if (rules & Rnolabel)
602 fputs(" (labels not permitted)", stderr);
606 static struct type *propagate_types(struct exec *prog, struct parse_context *c, enum prop_err *perr,
607 struct type *type, int rules);
608 ###### core functions
610 static struct type *__propagate_types(struct exec *prog, struct parse_context *c, enum prop_err *perr,
611 struct type *type, int rules)
618 switch (prog->type) {
621 struct binode *b = cast(binode, prog);
623 ## propagate binode cases
627 ## propagate exec cases
632 static struct type *propagate_types(struct exec *prog, struct parse_context *c, enum prop_err *perr,
633 struct type *type, int rules)
635 int pre_err = c->parse_error;
636 struct type *ret = __propagate_types(prog, c, perr, type, rules);
638 if (c->parse_error > pre_err)
645 Interpreting an `exec` doesn't require anything but the `exec`. State
646 is stored in variables and each variable will be directly linked from
647 within the `exec` tree. The exception to this is the `main` function
648 which needs to look at command line arguments. This function will be
649 interpreted separately.
651 Each `exec` can return a value combined with a type in `struct lrval`.
652 The type may be `Tnone` but must be non-NULL. Some `exec`s will return
653 the location of a value, which can be updated, in `lval`. Others will
654 set `lval` to NULL indicating that there is a value of appropriate type
658 static struct value interp_exec(struct parse_context *c, struct exec *e,
659 struct type **typeret);
660 ###### core functions
664 struct value rval, *lval;
667 /* If dest is passed, dtype must give the expected type, and
668 * result can go there, in which case type is returned as NULL.
670 static struct lrval _interp_exec(struct parse_context *c, struct exec *e,
671 struct value *dest, struct type *dtype);
673 static struct value interp_exec(struct parse_context *c, struct exec *e,
674 struct type **typeret)
676 struct lrval ret = _interp_exec(c, e, NULL, NULL);
678 if (!ret.type) abort();
682 dup_value(ret.type, ret.lval, &ret.rval);
686 static struct value *linterp_exec(struct parse_context *c, struct exec *e,
687 struct type **typeret)
689 struct lrval ret = _interp_exec(c, e, NULL, NULL);
691 if (!ret.type) abort();
695 free_value(ret.type, &ret.rval);
699 /* dinterp_exec is used when the destination type is certain and
700 * the value has a place to go.
702 static void dinterp_exec(struct parse_context *c, struct exec *e,
703 struct value *dest, struct type *dtype,
706 struct lrval ret = _interp_exec(c, e, dest, dtype);
710 free_value(dtype, dest);
712 dup_value(dtype, ret.lval, dest);
714 memcpy(dest, &ret.rval, dtype->size);
717 static struct lrval _interp_exec(struct parse_context *c, struct exec *e,
718 struct value *dest, struct type *dtype)
720 /* If the result is copied to dest, ret.type is set to NULL */
722 struct value rv = {}, *lrv = NULL;
725 rvtype = ret.type = Tnone;
735 struct binode *b = cast(binode, e);
736 struct value left, right, *lleft;
737 struct type *ltype, *rtype;
738 ltype = rtype = Tnone;
740 ## interp binode cases
742 free_value(ltype, &left);
743 free_value(rtype, &right);
753 ## interp exec cleanup
759 Values come in a wide range of types, with more likely to be added.
760 Each type needs to be able to print its own values (for convenience at
761 least) as well as to compare two values, at least for equality and
762 possibly for order. For now, values might need to be duplicated and
763 freed, though eventually such manipulations will be better integrated
766 Rather than requiring every numeric type to support all numeric
767 operations (add, multiply, etc), we allow types to be able to present
768 as one of a few standard types: integer, float, and fraction. The
769 existence of these conversion functions eventually enable types to
770 determine if they are compatible with other types, though such types
771 have not yet been implemented.
773 Named type are stored in a simple linked list. Objects of each type are
774 "values" which are often passed around by value.
776 There are both explicitly named types, and anonymous types. Anonymous
777 cannot be accessed by name, but are used internally and have a name
778 which might be reported in error messages.
785 ## value union fields
794 void (*init)(struct type *type, struct value *val);
795 void (*prepare_type)(struct parse_context *c, struct type *type, int parse_time);
796 void (*print)(struct type *type, struct value *val, FILE *f);
797 void (*print_type)(struct type *type, FILE *f);
798 int (*cmp_order)(struct type *t1, struct type *t2,
799 struct value *v1, struct value *v2);
800 int (*cmp_eq)(struct type *t1, struct type *t2,
801 struct value *v1, struct value *v2);
802 void (*dup)(struct type *type, struct value *vold, struct value *vnew);
803 void (*free)(struct type *type, struct value *val);
804 void (*free_type)(struct type *t);
805 long long (*to_int)(struct value *v);
806 double (*to_float)(struct value *v);
807 int (*to_mpq)(mpq_t *q, struct value *v);
816 struct type *typelist;
823 static struct type *find_type(struct parse_context *c, struct text s)
825 struct type *t = c->typelist;
827 while (t && (t->anon ||
828 text_cmp(t->name, s) != 0))
833 static struct type *_add_type(struct parse_context *c, struct text s,
834 struct type *proto, int anon)
838 n = calloc(1, sizeof(*n));
842 n->next = c->typelist;
847 static struct type *add_type(struct parse_context *c, struct text s,
850 return _add_type(c, s, proto, 0);
853 static struct type *add_anon_type(struct parse_context *c,
854 struct type *proto, char *name, ...)
860 vasprintf(&t.txt, name, ap);
862 t.len = strlen(name);
863 return _add_type(c, t, proto, 1);
866 static void free_type(struct type *t)
868 /* The type is always a reference to something in the
869 * context, so we don't need to free anything.
873 static void free_value(struct type *type, struct value *v)
877 memset(v, 0x5a, type->size);
881 static void type_print(struct type *type, FILE *f)
884 fputs("*unknown*type*", f); // NOTEST
885 else if (type->name.len && !type->anon)
886 fprintf(f, "%.*s", type->name.len, type->name.txt);
887 else if (type->print_type)
888 type->print_type(type, f);
890 fputs("*invalid*type*", f);
893 static void val_init(struct type *type, struct value *val)
895 if (type && type->init)
896 type->init(type, val);
899 static void dup_value(struct type *type,
900 struct value *vold, struct value *vnew)
902 if (type && type->dup)
903 type->dup(type, vold, vnew);
906 static int value_cmp(struct type *tl, struct type *tr,
907 struct value *left, struct value *right)
909 if (tl && tl->cmp_order)
910 return tl->cmp_order(tl, tr, left, right);
911 if (tl && tl->cmp_eq) // NOTEST
912 return tl->cmp_eq(tl, tr, left, right); // NOTEST
916 static void print_value(struct type *type, struct value *v, FILE *f)
918 if (type && type->print)
919 type->print(type, v, f);
921 fprintf(f, "*Unknown*"); // NOTEST
924 static void prepare_types(struct parse_context *c)
928 for (t = c->typelist; t; t = t->next)
930 t->prepare_type(c, t, 1);
935 static void free_value(struct type *type, struct value *v);
936 static int type_compat(struct type *require, struct type *have, int rules);
937 static void type_print(struct type *type, FILE *f);
938 static void val_init(struct type *type, struct value *v);
939 static void dup_value(struct type *type,
940 struct value *vold, struct value *vnew);
941 static int value_cmp(struct type *tl, struct type *tr,
942 struct value *left, struct value *right);
943 static void print_value(struct type *type, struct value *v, FILE *f);
945 ###### free context types
947 while (context.typelist) {
948 struct type *t = context.typelist;
950 context.typelist = t->next;
958 Type can be specified for local variables, for fields in a structure,
959 for formal parameters to functions, and possibly elsewhere. Different
960 rules may apply in different contexts. As a minimum, a named type may
961 always be used. Currently the type of a formal parameter can be
962 different from types in other contexts, so we have a separate grammar
968 Type -> IDENTIFIER ${
969 $0 = find_type(c, $1.txt);
972 "error: undefined type", &$1);
979 FormalType -> Type ${ $0 = $<1; }$
980 ## formal type grammar
984 Values of the base types can be numbers, which we represent as
985 multi-precision fractions, strings, Booleans and labels. When
986 analysing the program we also need to allow for places where no value
987 is meaningful (type `Tnone`) and where we don't know what type to
988 expect yet (type is `NULL`).
990 Values are never shared, they are always copied when used, and freed
991 when no longer needed.
993 When propagating type information around the program, we need to
994 determine if two types are compatible, where type `NULL` is compatible
995 with anything. There are two special cases with type compatibility,
996 both related to the Conditional Statement which will be described
997 later. In some cases a Boolean can be accepted as well as some other
998 primary type, and in others any type is acceptable except a label (`Vlabel`).
999 A separate function encoding these cases will simplify some code later.
1001 ###### type functions
1003 int (*compat)(struct type *this, struct type *other);
1005 ###### ast functions
1007 static int type_compat(struct type *require, struct type *have, int rules)
1009 if ((rules & Rboolok) && have == Tbool)
1011 if ((rules & Rnolabel) && have == Tlabel)
1013 if (!require || !have)
1016 if (require->compat)
1017 return require->compat(require, have);
1019 return require == have;
1024 #include "parse_string.h"
1025 #include "parse_number.h"
1028 myLDLIBS := libnumber.o libstring.o -lgmp
1029 LDLIBS := $(filter-out $(myLDLIBS),$(LDLIBS)) $(myLDLIBS)
1031 ###### type union fields
1032 enum vtype {Vnone, Vstr, Vnum, Vbool, Vlabel} vtype;
1034 ###### value union fields
1040 ###### ast functions
1041 static void _free_value(struct type *type, struct value *v)
1045 switch (type->vtype) {
1047 case Vstr: free(v->str.txt); break;
1048 case Vnum: mpq_clear(v->num); break;
1054 ###### value functions
1056 static void _val_init(struct type *type, struct value *val)
1058 switch(type->vtype) {
1059 case Vnone: // NOTEST
1062 mpq_init(val->num); break;
1064 val->str.txt = malloc(1);
1076 static void _dup_value(struct type *type,
1077 struct value *vold, struct value *vnew)
1079 switch (type->vtype) {
1080 case Vnone: // NOTEST
1083 vnew->label = vold->label;
1086 vnew->bool = vold->bool;
1089 mpq_init(vnew->num);
1090 mpq_set(vnew->num, vold->num);
1093 vnew->str.len = vold->str.len;
1094 vnew->str.txt = malloc(vnew->str.len);
1095 memcpy(vnew->str.txt, vold->str.txt, vnew->str.len);
1100 static int _value_cmp(struct type *tl, struct type *tr,
1101 struct value *left, struct value *right)
1105 return tl - tr; // NOTEST
1106 switch (tl->vtype) {
1107 case Vlabel: cmp = left->label == right->label ? 0 : 1; break;
1108 case Vnum: cmp = mpq_cmp(left->num, right->num); break;
1109 case Vstr: cmp = text_cmp(left->str, right->str); break;
1110 case Vbool: cmp = left->bool - right->bool; break;
1111 case Vnone: cmp = 0; // NOTEST
1116 static void _print_value(struct type *type, struct value *v, FILE *f)
1118 switch (type->vtype) {
1119 case Vnone: // NOTEST
1120 fprintf(f, "*no-value*"); break; // NOTEST
1121 case Vlabel: // NOTEST
1122 fprintf(f, "*label-%p*", v->label); break; // NOTEST
1124 fprintf(f, "%.*s", v->str.len, v->str.txt); break;
1126 fprintf(f, "%s", v->bool ? "True":"False"); break;
1131 mpf_set_q(fl, v->num);
1132 gmp_fprintf(f, "%.10Fg", fl);
1139 static void _free_value(struct type *type, struct value *v);
1141 static struct type base_prototype = {
1143 .print = _print_value,
1144 .cmp_order = _value_cmp,
1145 .cmp_eq = _value_cmp,
1147 .free = _free_value,
1150 static struct type *Tbool, *Tstr, *Tnum, *Tnone, *Tlabel;
1152 ###### ast functions
1153 static struct type *add_base_type(struct parse_context *c, char *n,
1154 enum vtype vt, int size)
1156 struct text txt = { n, strlen(n) };
1159 t = add_type(c, txt, &base_prototype);
1162 t->align = size > sizeof(void*) ? sizeof(void*) : size;
1163 if (t->size & (t->align - 1))
1164 t->size = (t->size | (t->align - 1)) + 1; // NOTEST
1168 ###### context initialization
1170 Tbool = add_base_type(&context, "Boolean", Vbool, sizeof(char));
1171 Tstr = add_base_type(&context, "string", Vstr, sizeof(struct text));
1172 Tnum = add_base_type(&context, "number", Vnum, sizeof(mpq_t));
1173 Tnone = add_base_type(&context, "none", Vnone, 0);
1174 Tlabel = add_base_type(&context, "label", Vlabel, sizeof(void*));
1178 We have already met values as separate objects. When manifest constants
1179 appear in the program text, that must result in an executable which has
1180 a constant value. So the `val` structure embeds a value in an
1193 ###### ast functions
1194 struct val *new_val(struct type *T, struct token tk)
1196 struct val *v = new_pos(val, tk);
1207 $0 = new_val(Tbool, $1);
1211 $0 = new_val(Tbool, $1);
1216 $0 = new_val(Tnum, $1);
1217 if (number_parse($0->val.num, tail, $1.txt) == 0)
1218 mpq_init($0->val.num); // UNTESTED
1220 tok_err(c, "error: unsupported number suffix",
1225 $0 = new_val(Tstr, $1);
1226 string_parse(&$1, '\\', &$0->val.str, tail);
1228 tok_err(c, "error: unsupported string suffix",
1233 $0 = new_val(Tstr, $1);
1234 string_parse(&$1, '\\', &$0->val.str, tail);
1236 tok_err(c, "error: unsupported string suffix",
1240 ###### print exec cases
1243 struct val *v = cast(val, e);
1244 if (v->vtype == Tstr)
1246 // FIXME how to ensure numbers have same precision.
1247 print_value(v->vtype, &v->val, stdout);
1248 if (v->vtype == Tstr)
1253 ###### propagate exec cases
1256 struct val *val = cast(val, prog);
1257 if (!type_compat(type, val->vtype, rules))
1258 type_err(c, "error: expected %1%r found %2",
1259 prog, type, rules, val->vtype);
1263 ###### interp exec cases
1265 rvtype = cast(val, e)->vtype;
1266 dup_value(rvtype, &cast(val, e)->val, &rv);
1269 ###### ast functions
1270 static void free_val(struct val *v)
1273 free_value(v->vtype, &v->val);
1277 ###### free exec cases
1278 case Xval: free_val(cast(val, e)); break;
1280 ###### ast functions
1281 // Move all nodes from 'b' to 'rv', reversing their order.
1282 // In 'b' 'left' is a list, and 'right' is the last node.
1283 // In 'rv', left' is the first node and 'right' is a list.
1284 static struct binode *reorder_bilist(struct binode *b)
1286 struct binode *rv = NULL;
1289 struct exec *t = b->right;
1293 b = cast(binode, b->left);
1303 Variables are scoped named values. We store the names in a linked list
1304 of "bindings" sorted in lexical order, and use sequential search and
1311 struct binding *next; // in lexical order
1315 This linked list is stored in the parse context so that "reduce"
1316 functions can find or add variables, and so the analysis phase can
1317 ensure that every variable gets a type.
1319 ###### parse context
1321 struct binding *varlist; // In lexical order
1323 ###### ast functions
1325 static struct binding *find_binding(struct parse_context *c, struct text s)
1327 struct binding **l = &c->varlist;
1332 (cmp = text_cmp((*l)->name, s)) < 0)
1336 n = calloc(1, sizeof(*n));
1343 Each name can be linked to multiple variables defined in different
1344 scopes. Each scope starts where the name is declared and continues
1345 until the end of the containing code block. Scopes of a given name
1346 cannot nest, so a declaration while a name is in-scope is an error.
1348 ###### binding fields
1349 struct variable *var;
1353 struct variable *previous;
1355 struct binding *name;
1356 struct exec *where_decl;// where name was declared
1357 struct exec *where_set; // where type was set
1361 When a scope closes, the values of the variables might need to be freed.
1362 This happens in the context of some `struct exec` and each `exec` will
1363 need to know which variables need to be freed when it completes.
1366 struct variable *to_free;
1368 ####### variable fields
1369 struct exec *cleanup_exec;
1370 struct variable *next_free;
1372 ####### interp exec cleanup
1375 for (v = e->to_free; v; v = v->next_free) {
1376 struct value *val = var_value(c, v);
1377 free_value(v->type, val);
1381 ###### ast functions
1382 static void variable_unlink_exec(struct variable *v)
1384 struct variable **vp;
1385 if (!v->cleanup_exec)
1387 for (vp = &v->cleanup_exec->to_free;
1388 *vp; vp = &(*vp)->next_free) {
1392 v->cleanup_exec = NULL;
1397 While the naming seems strange, we include local constants in the
1398 definition of variables. A name declared `var := value` can
1399 subsequently be changed, but a name declared `var ::= value` cannot -
1402 ###### variable fields
1405 Scopes in parallel branches can be partially merged. More
1406 specifically, if a given name is declared in both branches of an
1407 if/else then its scope is a candidate for merging. Similarly if
1408 every branch of an exhaustive switch (e.g. has an "else" clause)
1409 declares a given name, then the scopes from the branches are
1410 candidates for merging.
1412 Note that names declared inside a loop (which is only parallel to
1413 itself) are never visible after the loop. Similarly names defined in
1414 scopes which are not parallel, such as those started by `for` and
1415 `switch`, are never visible after the scope. Only variables defined in
1416 both `then` and `else` (including the implicit then after an `if`, and
1417 excluding `then` used with `for`) and in all `case`s and `else` of a
1418 `switch` or `while` can be visible beyond the `if`/`switch`/`while`.
1420 Labels, which are a bit like variables, follow different rules.
1421 Labels are not explicitly declared, but if an undeclared name appears
1422 in a context where a label is legal, that effectively declares the
1423 name as a label. The declaration remains in force (or in scope) at
1424 least to the end of the immediately containing block and conditionally
1425 in any larger containing block which does not declare the name in some
1426 other way. Importantly, the conditional scope extension happens even
1427 if the label is only used in one parallel branch of a conditional --
1428 when used in one branch it is treated as having been declared in all
1431 Merge candidates are tentatively visible beyond the end of the
1432 branching statement which creates them. If the name is used, the
1433 merge is affirmed and they become a single variable visible at the
1434 outer layer. If not - if it is redeclared first - the merge lapses.
1436 To track scopes we have an extra stack, implemented as a linked list,
1437 which roughly parallels the parse stack and which is used exclusively
1438 for scoping. When a new scope is opened, a new frame is pushed and
1439 the child-count of the parent frame is incremented. This child-count
1440 is used to distinguish between the first of a set of parallel scopes,
1441 in which declared variables must not be in scope, and subsequent
1442 branches, whether they may already be conditionally scoped.
1444 We need a total ordering of scopes so we can easily compare to variables
1445 to see if they are concurrently in scope. To achieve this we record a
1446 `scope_count` which is actually a count of both beginnings and endings
1447 of scopes. Then each variable has a record of the scope count where it
1448 enters scope, and where it leaves.
1450 To push a new frame *before* any code in the frame is parsed, we need a
1451 grammar reduction. This is most easily achieved with a grammar
1452 element which derives the empty string, and creates the new scope when
1453 it is recognised. This can be placed, for example, between a keyword
1454 like "if" and the code following it.
1458 struct scope *parent;
1462 ###### parse context
1465 struct scope *scope_stack;
1467 ###### variable fields
1468 int scope_start, scope_end;
1470 ###### ast functions
1471 static void scope_pop(struct parse_context *c)
1473 struct scope *s = c->scope_stack;
1475 c->scope_stack = s->parent;
1477 c->scope_depth -= 1;
1478 c->scope_count += 1;
1481 static void scope_push(struct parse_context *c)
1483 struct scope *s = calloc(1, sizeof(*s));
1485 c->scope_stack->child_count += 1;
1486 s->parent = c->scope_stack;
1488 c->scope_depth += 1;
1489 c->scope_count += 1;
1495 OpenScope -> ${ scope_push(c); }$
1497 Each variable records a scope depth and is in one of four states:
1499 - "in scope". This is the case between the declaration of the
1500 variable and the end of the containing block, and also between
1501 the usage with affirms a merge and the end of that block.
1503 The scope depth is not greater than the current parse context scope
1504 nest depth. When the block of that depth closes, the state will
1505 change. To achieve this, all "in scope" variables are linked
1506 together as a stack in nesting order.
1508 - "pending". The "in scope" block has closed, but other parallel
1509 scopes are still being processed. So far, every parallel block at
1510 the same level that has closed has declared the name.
1512 The scope depth is the depth of the last parallel block that
1513 enclosed the declaration, and that has closed.
1515 - "conditionally in scope". The "in scope" block and all parallel
1516 scopes have closed, and no further mention of the name has been seen.
1517 This state includes a secondary nest depth (`min_depth`) which records
1518 the outermost scope seen since the variable became conditionally in
1519 scope. If a use of the name is found, the variable becomes "in scope"
1520 and that secondary depth becomes the recorded scope depth. If the
1521 name is declared as a new variable, the old variable becomes "out of
1522 scope" and the recorded scope depth stays unchanged.
1524 - "out of scope". The variable is neither in scope nor conditionally
1525 in scope. It is permanently out of scope now and can be removed from
1526 the "in scope" stack. When a variable becomes out-of-scope it is
1527 moved to a separate list (`out_scope`) of variables which have fully
1528 known scope. This will be used at the end of each function to assign
1529 each variable a place in the stack frame.
1531 ###### variable fields
1532 int depth, min_depth;
1533 enum { OutScope, PendingScope, CondScope, InScope } scope;
1534 struct variable *in_scope;
1536 ###### parse context
1538 struct variable *in_scope;
1539 struct variable *out_scope;
1541 All variables with the same name are linked together using the
1542 'previous' link. Those variable that have been affirmatively merged all
1543 have a 'merged' pointer that points to one primary variable - the most
1544 recently declared instance. When merging variables, we need to also
1545 adjust the 'merged' pointer on any other variables that had previously
1546 been merged with the one that will no longer be primary.
1548 A variable that is no longer the most recent instance of a name may
1549 still have "pending" scope, if it might still be merged with most
1550 recent instance. These variables don't really belong in the
1551 "in_scope" list, but are not immediately removed when a new instance
1552 is found. Instead, they are detected and ignored when considering the
1553 list of in_scope names.
1555 The storage of the value of a variable will be described later. For now
1556 we just need to know that when a variable goes out of scope, it might
1557 need to be freed. For this we need to be able to find it, so assume that
1558 `var_value()` will provide that.
1560 ###### variable fields
1561 struct variable *merged;
1563 ###### ast functions
1565 static void variable_merge(struct variable *primary, struct variable *secondary)
1569 primary = primary->merged;
1571 for (v = primary->previous; v; v=v->previous)
1572 if (v == secondary || v == secondary->merged ||
1573 v->merged == secondary ||
1574 v->merged == secondary->merged) {
1575 v->scope = OutScope;
1576 v->merged = primary;
1577 if (v->scope_start < primary->scope_start)
1578 primary->scope_start = v->scope_start;
1579 if (v->scope_end > primary->scope_end)
1580 primary->scope_end = v->scope_end; // NOTEST
1581 variable_unlink_exec(v);
1585 ###### forward decls
1586 static struct value *var_value(struct parse_context *c, struct variable *v);
1588 ###### free global vars
1590 while (context.varlist) {
1591 struct binding *b = context.varlist;
1592 struct variable *v = b->var;
1593 context.varlist = b->next;
1596 struct variable *next = v->previous;
1598 if (v->global && v->frame_pos >= 0) {
1599 free_value(v->type, var_value(&context, v));
1600 if (v->depth == 0 && v->type->free == function_free)
1601 // This is a function constant
1602 free_exec(v->where_decl);
1609 #### Manipulating Bindings
1611 When a name is conditionally visible, a new declaration discards the old
1612 binding - the condition lapses. Similarly when we reach the end of a
1613 function (outermost non-global scope) any conditional scope must lapse.
1614 Conversely a usage of the name affirms the visibility and extends it to
1615 the end of the containing block - i.e. the block that contains both the
1616 original declaration and the latest usage. This is determined from
1617 `min_depth`. When a conditionally visible variable gets affirmed like
1618 this, it is also merged with other conditionally visible variables with
1621 When we parse a variable declaration we either report an error if the
1622 name is currently bound, or create a new variable at the current nest
1623 depth if the name is unbound or bound to a conditionally scoped or
1624 pending-scope variable. If the previous variable was conditionally
1625 scoped, it and its homonyms becomes out-of-scope.
1627 When we parse a variable reference (including non-declarative assignment
1628 "foo = bar") we report an error if the name is not bound or is bound to
1629 a pending-scope variable; update the scope if the name is bound to a
1630 conditionally scoped variable; or just proceed normally if the named
1631 variable is in scope.
1633 When we exit a scope, any variables bound at this level are either
1634 marked out of scope or pending-scoped, depending on whether the scope
1635 was sequential or parallel. Here a "parallel" scope means the "then"
1636 or "else" part of a conditional, or any "case" or "else" branch of a
1637 switch. Other scopes are "sequential".
1639 When exiting a parallel scope we check if there are any variables that
1640 were previously pending and are still visible. If there are, then
1641 they weren't redeclared in the most recent scope, so they cannot be
1642 merged and must become out-of-scope. If it is not the first of
1643 parallel scopes (based on `child_count`), we check that there was a
1644 previous binding that is still pending-scope. If there isn't, the new
1645 variable must now be out-of-scope.
1647 When exiting a sequential scope that immediately enclosed parallel
1648 scopes, we need to resolve any pending-scope variables. If there was
1649 no `else` clause, and we cannot determine that the `switch` was exhaustive,
1650 we need to mark all pending-scope variable as out-of-scope. Otherwise
1651 all pending-scope variables become conditionally scoped.
1654 enum closetype { CloseSequential, CloseFunction, CloseParallel, CloseElse };
1656 ###### ast functions
1658 static struct variable *var_decl(struct parse_context *c, struct text s)
1660 struct binding *b = find_binding(c, s);
1661 struct variable *v = b->var;
1663 switch (v ? v->scope : OutScope) {
1665 /* Caller will report the error */
1669 v && v->scope == CondScope;
1671 v->scope = OutScope;
1675 v = calloc(1, sizeof(*v));
1676 v->previous = b->var;
1680 v->min_depth = v->depth = c->scope_depth;
1682 v->in_scope = c->in_scope;
1683 v->scope_start = c->scope_count;
1689 static struct variable *var_ref(struct parse_context *c, struct text s)
1691 struct binding *b = find_binding(c, s);
1692 struct variable *v = b->var;
1693 struct variable *v2;
1695 switch (v ? v->scope : OutScope) {
1698 /* Caller will report the error */
1701 /* All CondScope variables of this name need to be merged
1702 * and become InScope
1704 v->depth = v->min_depth;
1706 for (v2 = v->previous;
1707 v2 && v2->scope == CondScope;
1709 variable_merge(v, v2);
1717 static int var_refile(struct parse_context *c, struct variable *v)
1719 /* Variable just went out of scope. Add it to the out_scope
1720 * list, sorted by ->scope_start
1722 struct variable **vp = &c->out_scope;
1723 while ((*vp) && (*vp)->scope_start < v->scope_start)
1724 vp = &(*vp)->in_scope;
1730 static void var_block_close(struct parse_context *c, enum closetype ct,
1733 /* Close off all variables that are in_scope.
1734 * Some variables in c->scope may already be not-in-scope,
1735 * such as when a PendingScope variable is hidden by a new
1736 * variable with the same name.
1737 * So we check for v->name->var != v and drop them.
1738 * If we choose to make a variable OutScope, we drop it
1741 struct variable *v, **vp, *v2;
1744 for (vp = &c->in_scope;
1745 (v = *vp) && v->min_depth > c->scope_depth;
1746 (v->scope == OutScope || v->name->var != v)
1747 ? (*vp = v->in_scope, var_refile(c, v))
1748 : ( vp = &v->in_scope, 0)) {
1749 v->min_depth = c->scope_depth;
1750 if (v->name->var != v)
1751 /* This is still in scope, but we haven't just
1755 v->min_depth = c->scope_depth;
1756 if (v->scope == InScope)
1757 v->scope_end = c->scope_count;
1758 if (v->scope == InScope && e && !v->global) {
1759 /* This variable gets cleaned up when 'e' finishes */
1760 variable_unlink_exec(v);
1761 v->cleanup_exec = e;
1762 v->next_free = e->to_free;
1767 case CloseParallel: /* handle PendingScope */
1771 if (c->scope_stack->child_count == 1)
1772 /* first among parallel branches */
1773 v->scope = PendingScope;
1774 else if (v->previous &&
1775 v->previous->scope == PendingScope)
1776 /* all previous branches used name */
1777 v->scope = PendingScope;
1778 else if (v->type == Tlabel)
1779 /* Labels remain pending even when not used */
1780 v->scope = PendingScope; // UNTESTED
1782 v->scope = OutScope;
1783 if (ct == CloseElse) {
1784 /* All Pending variables with this name
1785 * are now Conditional */
1787 v2 && v2->scope == PendingScope;
1789 v2->scope = CondScope;
1793 /* Not possible as it would require
1794 * parallel scope to be nested immediately
1795 * in a parallel scope, and that never
1799 /* Not possible as we already tested for
1806 if (v->scope == CondScope)
1807 /* Condition cannot continue past end of function */
1810 case CloseSequential:
1811 if (v->type == Tlabel)
1812 v->scope = PendingScope;
1815 v->scope = OutScope;
1818 /* There was no 'else', so we can only become
1819 * conditional if we know the cases were exhaustive,
1820 * and that doesn't mean anything yet.
1821 * So only labels become conditional..
1824 v2 && v2->scope == PendingScope;
1826 if (v2->type == Tlabel)
1827 v2->scope = CondScope;
1829 v2->scope = OutScope;
1832 case OutScope: break;
1841 The value of a variable is store separately from the variable, on an
1842 analogue of a stack frame. There are (currently) two frames that can be
1843 active. A global frame which currently only stores constants, and a
1844 stacked frame which stores local variables. Each variable knows if it
1845 is global or not, and what its index into the frame is.
1847 Values in the global frame are known immediately they are relevant, so
1848 the frame needs to be reallocated as it grows so it can store those
1849 values. The local frame doesn't get values until the interpreted phase
1850 is started, so there is no need to allocate until the size is known.
1852 We initialize the `frame_pos` to an impossible value, so that we can
1853 tell if it was set or not later.
1855 ###### variable fields
1859 ###### variable init
1862 ###### parse context
1864 short global_size, global_alloc;
1866 void *global, *local;
1868 ###### forward decls
1869 static struct value *global_alloc(struct parse_context *c, struct type *t,
1870 struct variable *v, struct value *init);
1872 ###### ast functions
1874 static struct value *var_value(struct parse_context *c, struct variable *v)
1877 if (!c->local || !v->type)
1878 return NULL; // UNTESTED
1879 if (v->frame_pos + v->type->size > c->local_size) {
1880 printf("INVALID frame_pos\n"); // NOTEST
1883 return c->local + v->frame_pos;
1885 if (c->global_size > c->global_alloc) {
1886 int old = c->global_alloc;
1887 c->global_alloc = (c->global_size | 1023) + 1024;
1888 c->global = realloc(c->global, c->global_alloc);
1889 memset(c->global + old, 0, c->global_alloc - old);
1891 return c->global + v->frame_pos;
1894 static struct value *global_alloc(struct parse_context *c, struct type *t,
1895 struct variable *v, struct value *init)
1898 struct variable scratch;
1900 if (t->prepare_type)
1901 t->prepare_type(c, t, 1); // NOTEST
1903 if (c->global_size & (t->align - 1))
1904 c->global_size = (c->global_size + t->align) & ~(t->align-1); // NOTEST
1909 v->frame_pos = c->global_size;
1911 c->global_size += v->type->size;
1912 ret = var_value(c, v);
1914 memcpy(ret, init, t->size);
1920 As global values are found -- struct field initializers, labels etc --
1921 `global_alloc()` is called to record the value in the global frame.
1923 When the program is fully parsed, each function is analysed, we need to
1924 walk the list of variables local to that function and assign them an
1925 offset in the stack frame. For this we have `scope_finalize()`.
1927 We keep the stack from dense by re-using space for between variables
1928 that are not in scope at the same time. The `out_scope` list is sorted
1929 by `scope_start` and as we process a varible, we move it to an FIFO
1930 stack. For each variable we consider, we first discard any from the
1931 stack anything that went out of scope before the new variable came in.
1932 Then we place the new variable just after the one at the top of the
1935 ###### ast functions
1937 static void scope_finalize(struct parse_context *c, struct type *ft)
1939 int size = ft->function.local_size;
1940 struct variable *next = ft->function.scope;
1941 struct variable *done = NULL;
1944 struct variable *v = next;
1945 struct type *t = v->type;
1952 if (v->frame_pos >= 0)
1954 while (done && done->scope_end < v->scope_start)
1955 done = done->in_scope;
1957 pos = done->frame_pos + done->type->size;
1959 pos = ft->function.local_size;
1960 if (pos & (t->align - 1))
1961 pos = (pos + t->align) & ~(t->align-1);
1963 if (size < pos + v->type->size)
1964 size = pos + v->type->size;
1968 c->out_scope = NULL;
1969 ft->function.local_size = size;
1972 ###### free context storage
1973 free(context.global);
1975 #### Variables as executables
1977 Just as we used a `val` to wrap a value into an `exec`, we similarly
1978 need a `var` to wrap a `variable` into an exec. While each `val`
1979 contained a copy of the value, each `var` holds a link to the variable
1980 because it really is the same variable no matter where it appears.
1981 When a variable is used, we need to remember to follow the `->merged`
1982 link to find the primary instance.
1984 When a variable is declared, it may or may not be given an explicit
1985 type. We need to record which so that we can report the parsed code
1994 struct variable *var;
1997 ###### variable fields
2005 VariableDecl -> IDENTIFIER : ${ {
2006 struct variable *v = var_decl(c, $1.txt);
2007 $0 = new_pos(var, $1);
2012 v = var_ref(c, $1.txt);
2014 type_err(c, "error: variable '%v' redeclared",
2016 type_err(c, "info: this is where '%v' was first declared",
2017 v->where_decl, NULL, 0, NULL);
2020 | IDENTIFIER :: ${ {
2021 struct variable *v = var_decl(c, $1.txt);
2022 $0 = new_pos(var, $1);
2028 v = var_ref(c, $1.txt);
2030 type_err(c, "error: variable '%v' redeclared",
2032 type_err(c, "info: this is where '%v' was first declared",
2033 v->where_decl, NULL, 0, NULL);
2036 | IDENTIFIER : Type ${ {
2037 struct variable *v = var_decl(c, $1.txt);
2038 $0 = new_pos(var, $1);
2044 v->explicit_type = 1;
2046 v = var_ref(c, $1.txt);
2048 type_err(c, "error: variable '%v' redeclared",
2050 type_err(c, "info: this is where '%v' was first declared",
2051 v->where_decl, NULL, 0, NULL);
2054 | IDENTIFIER :: Type ${ {
2055 struct variable *v = var_decl(c, $1.txt);
2056 $0 = new_pos(var, $1);
2063 v->explicit_type = 1;
2065 v = var_ref(c, $1.txt);
2067 type_err(c, "error: variable '%v' redeclared",
2069 type_err(c, "info: this is where '%v' was first declared",
2070 v->where_decl, NULL, 0, NULL);
2075 Variable -> IDENTIFIER ${ {
2076 struct variable *v = var_ref(c, $1.txt);
2077 $0 = new_pos(var, $1);
2079 /* This might be a global const or a label
2080 * Allocate a var with impossible type Tnone,
2081 * which will be adjusted when we find out what it is,
2082 * or will trigger an error.
2084 v = var_decl(c, $1.txt);
2091 cast(var, $0)->var = v;
2094 ###### print exec cases
2097 struct var *v = cast(var, e);
2099 struct binding *b = v->var->name;
2100 printf("%.*s", b->name.len, b->name.txt);
2107 if (loc && loc->type == Xvar) {
2108 struct var *v = cast(var, loc);
2110 struct binding *b = v->var->name;
2111 fprintf(stderr, "%.*s", b->name.len, b->name.txt);
2113 fputs("???", stderr); // NOTEST
2115 fputs("NOTVAR", stderr);
2118 ###### propagate exec cases
2122 struct var *var = cast(var, prog);
2123 struct variable *v = var->var;
2125 type_err(c, "%d:BUG: no variable!!", prog, NULL, 0, NULL); // NOTEST
2126 return Tnone; // NOTEST
2129 if (v->constant && (rules & Rnoconstant)) {
2130 type_err(c, "error: Cannot assign to a constant: %v",
2131 prog, NULL, 0, NULL);
2132 type_err(c, "info: name was defined as a constant here",
2133 v->where_decl, NULL, 0, NULL);
2136 if (v->type == Tnone && v->where_decl == prog)
2137 type_err(c, "error: variable used but not declared: %v",
2138 prog, NULL, 0, NULL);
2139 if (v->type == NULL) {
2140 if (type && !(*perr & Efail)) {
2142 v->where_set = prog;
2145 } else if (!type_compat(type, v->type, rules)) {
2146 type_err(c, "error: expected %1%r but variable '%v' is %2", prog,
2147 type, rules, v->type);
2148 type_err(c, "info: this is where '%v' was set to %1", v->where_set,
2149 v->type, rules, NULL);
2151 if (!v->global || v->frame_pos < 0)
2158 ###### interp exec cases
2161 struct var *var = cast(var, e);
2162 struct variable *v = var->var;
2165 lrv = var_value(c, v);
2170 ###### ast functions
2172 static void free_var(struct var *v)
2177 ###### free exec cases
2178 case Xvar: free_var(cast(var, e)); break;
2183 Now that we have the shape of the interpreter in place we can add some
2184 complex types and connected them in to the data structures and the
2185 different phases of parse, analyse, print, interpret.
2187 Being "complex" the language will naturally have syntax to access
2188 specifics of objects of these types. These will fit into the grammar as
2189 "Terms" which are the things that are combined with various operators to
2190 form "Expression". Where a Term is formed by some operation on another
2191 Term, the subordinate Term will always come first, so for example a
2192 member of an array will be expressed as the Term for the array followed
2193 by an index in square brackets. The strict rule of using postfix
2194 operations makes precedence irrelevant within terms. To provide a place
2195 to put the grammar for each terms of each type, we will start out by
2196 introducing the "Term" grammar production, with contains at least a
2197 simple "Value" (to be explained later).
2201 Term -> Value ${ $0 = $<1; }$
2202 | Variable ${ $0 = $<1; }$
2205 Thus far the complex types we have are arrays and structs.
2209 Arrays can be declared by giving a size and a type, as `[size]type' so
2210 `freq:[26]number` declares `freq` to be an array of 26 numbers. The
2211 size can be either a literal number, or a named constant. Some day an
2212 arbitrary expression will be supported.
2214 As a formal parameter to a function, the array can be declared with a
2215 new variable as the size: `name:[size::number]string`. The `size`
2216 variable is set to the size of the array and must be a constant. As
2217 `number` is the only supported type, it can be left out:
2218 `name:[size::]string`.
2220 Arrays cannot be assigned. When pointers are introduced we will also
2221 introduce array slices which can refer to part or all of an array -
2222 the assignment syntax will create a slice. For now, an array can only
2223 ever be referenced by the name it is declared with. It is likely that
2224 a "`copy`" primitive will eventually be define which can be used to
2225 make a copy of an array with controllable recursive depth.
2227 For now we have two sorts of array, those with fixed size either because
2228 it is given as a literal number or because it is a struct member (which
2229 cannot have a runtime-changing size), and those with a size that is
2230 determined at runtime - local variables with a const size. The former
2231 have their size calculated at parse time, the latter at run time.
2233 For the latter type, the `size` field of the type is the size of a
2234 pointer, and the array is reallocated every time it comes into scope.
2236 We differentiate struct fields with a const size from local variables
2237 with a const size by whether they are prepared at parse time or not.
2239 ###### type union fields
2242 int unspec; // size is unspecified - vsize must be set.
2245 struct variable *vsize;
2246 struct type *member;
2249 ###### value union fields
2250 void *array; // used if not static_size
2252 ###### value functions
2254 static void array_prepare_type(struct parse_context *c, struct type *type,
2257 struct value *vsize;
2259 if (type->array.static_size)
2261 if (type->array.unspec && parse_time)
2263 if (parse_time && type->array.vsize && !type->array.vsize->global)
2266 if (type->array.vsize) {
2267 vsize = var_value(c, type->array.vsize);
2271 mpz_tdiv_q(q, mpq_numref(vsize->num), mpq_denref(vsize->num));
2272 type->array.size = mpz_get_si(q);
2276 if (parse_time && type->array.member->size) {
2277 type->array.static_size = 1;
2278 type->size = type->array.size * type->array.member->size;
2279 type->align = type->array.member->align;
2283 static void array_init(struct type *type, struct value *val)
2286 void *ptr = val->ptr;
2290 if (!type->array.static_size) {
2291 val->array = calloc(type->array.size,
2292 type->array.member->size);
2295 for (i = 0; i < type->array.size; i++) {
2297 v = (void*)ptr + i * type->array.member->size;
2298 val_init(type->array.member, v);
2302 static void array_free(struct type *type, struct value *val)
2305 void *ptr = val->ptr;
2307 if (!type->array.static_size)
2309 for (i = 0; i < type->array.size; i++) {
2311 v = (void*)ptr + i * type->array.member->size;
2312 free_value(type->array.member, v);
2314 if (!type->array.static_size)
2318 static int array_compat(struct type *require, struct type *have)
2320 if (have->compat != require->compat)
2322 /* Both are arrays, so we can look at details */
2323 if (!type_compat(require->array.member, have->array.member, 0))
2325 if (have->array.unspec && require->array.unspec) {
2326 if (have->array.vsize && require->array.vsize &&
2327 have->array.vsize != require->array.vsize) // UNTESTED
2328 /* sizes might not be the same */
2329 return 0; // UNTESTED
2332 if (have->array.unspec || require->array.unspec)
2333 return 1; // UNTESTED
2334 if (require->array.vsize == NULL && have->array.vsize == NULL)
2335 return require->array.size == have->array.size;
2337 return require->array.vsize == have->array.vsize; // UNTESTED
2340 static void array_print_type(struct type *type, FILE *f)
2343 if (type->array.vsize) {
2344 struct binding *b = type->array.vsize->name;
2345 fprintf(f, "%.*s%s]", b->name.len, b->name.txt,
2346 type->array.unspec ? "::" : "");
2347 } else if (type->array.size)
2348 fprintf(f, "%d]", type->array.size);
2351 type_print(type->array.member, f);
2354 static struct type array_prototype = {
2356 .prepare_type = array_prepare_type,
2357 .print_type = array_print_type,
2358 .compat = array_compat,
2360 .size = sizeof(void*),
2361 .align = sizeof(void*),
2364 ###### declare terminals
2369 | [ NUMBER ] Type ${ {
2375 if (number_parse(num, tail, $2.txt) == 0)
2376 tok_err(c, "error: unrecognised number", &$2);
2378 tok_err(c, "error: unsupported number suffix", &$2);
2381 elements = mpz_get_ui(mpq_numref(num));
2382 if (mpz_cmp_ui(mpq_denref(num), 1) != 0) {
2383 tok_err(c, "error: array size must be an integer",
2385 } else if (mpz_cmp_ui(mpq_numref(num), 1UL << 30) >= 0)
2386 tok_err(c, "error: array size is too large",
2391 $0 = t = add_anon_type(c, &array_prototype, "array[%d]", elements );
2392 t->array.size = elements;
2393 t->array.member = $<4;
2394 t->array.vsize = NULL;
2397 | [ IDENTIFIER ] Type ${ {
2398 struct variable *v = var_ref(c, $2.txt);
2401 tok_err(c, "error: name undeclared", &$2);
2402 else if (!v->constant)
2403 tok_err(c, "error: array size must be a constant", &$2);
2405 $0 = add_anon_type(c, &array_prototype, "array[%.*s]", $2.txt.len, $2.txt.txt);
2406 $0->array.member = $<4;
2408 $0->array.vsize = v;
2413 OptType -> Type ${ $0 = $<1; }$
2416 ###### formal type grammar
2418 | [ IDENTIFIER :: OptType ] Type ${ {
2419 struct variable *v = var_decl(c, $ID.txt);
2425 $0 = add_anon_type(c, &array_prototype, "array[var]");
2426 $0->array.member = $<6;
2428 $0->array.unspec = 1;
2429 $0->array.vsize = v;
2437 | Term [ Expression ] ${ {
2438 struct binode *b = new(binode);
2445 ###### print binode cases
2447 print_exec(b->left, -1, bracket);
2449 print_exec(b->right, -1, bracket);
2453 ###### propagate binode cases
2455 /* left must be an array, right must be a number,
2456 * result is the member type of the array
2458 propagate_types(b->right, c, perr, Tnum, 0);
2459 t = propagate_types(b->left, c, perr, NULL, rules & Rnoconstant);
2460 if (!t || t->compat != array_compat) {
2461 type_err(c, "error: %1 cannot be indexed", prog, t, 0, NULL);
2464 if (!type_compat(type, t->array.member, rules)) {
2465 type_err(c, "error: have %1 but need %2", prog,
2466 t->array.member, rules, type);
2468 return t->array.member;
2472 ###### interp binode cases
2478 lleft = linterp_exec(c, b->left, <ype);
2479 right = interp_exec(c, b->right, &rtype);
2481 mpz_tdiv_q(q, mpq_numref(right.num), mpq_denref(right.num));
2485 if (ltype->array.static_size)
2488 ptr = *(void**)lleft;
2489 rvtype = ltype->array.member;
2490 if (i >= 0 && i < ltype->array.size)
2491 lrv = ptr + i * rvtype->size;
2493 val_init(ltype->array.member, &rv); // UNSAFE
2500 A `struct` is a data-type that contains one or more other data-types.
2501 It differs from an array in that each member can be of a different
2502 type, and they are accessed by name rather than by number. Thus you
2503 cannot choose an element by calculation, you need to know what you
2506 The language makes no promises about how a given structure will be
2507 stored in memory - it is free to rearrange fields to suit whatever
2508 criteria seems important.
2510 Structs are declared separately from program code - they cannot be
2511 declared in-line in a variable declaration like arrays can. A struct
2512 is given a name and this name is used to identify the type - the name
2513 is not prefixed by the word `struct` as it would be in C.
2515 Structs are only treated as the same if they have the same name.
2516 Simply having the same fields in the same order is not enough. This
2517 might change once we can create structure initializers from a list of
2520 Each component datum is identified much like a variable is declared,
2521 with a name, one or two colons, and a type. The type cannot be omitted
2522 as there is no opportunity to deduce the type from usage. An initial
2523 value can be given following an equals sign, so
2525 ##### Example: a struct type
2531 would declare a type called "complex" which has two number fields,
2532 each initialised to zero.
2534 Struct will need to be declared separately from the code that uses
2535 them, so we will need to be able to print out the declaration of a
2536 struct when reprinting the whole program. So a `print_type_decl` type
2537 function will be needed.
2539 ###### type union fields
2548 } *fields; // This is created when field_list is analysed.
2550 struct fieldlist *prev;
2553 } *field_list; // This is created during parsing
2556 ###### type functions
2557 void (*print_type_decl)(struct type *type, FILE *f);
2559 ###### value functions
2561 static void structure_init(struct type *type, struct value *val)
2565 for (i = 0; i < type->structure.nfields; i++) {
2567 v = (void*) val->ptr + type->structure.fields[i].offset;
2568 if (type->structure.fields[i].init)
2569 dup_value(type->structure.fields[i].type,
2570 type->structure.fields[i].init,
2573 val_init(type->structure.fields[i].type, v);
2577 static void structure_free(struct type *type, struct value *val)
2581 for (i = 0; i < type->structure.nfields; i++) {
2583 v = (void*)val->ptr + type->structure.fields[i].offset;
2584 free_value(type->structure.fields[i].type, v);
2588 static void free_fieldlist(struct fieldlist *f)
2592 free_fieldlist(f->prev);
2597 static void structure_free_type(struct type *t)
2600 for (i = 0; i < t->structure.nfields; i++)
2601 if (t->structure.fields[i].init) {
2602 free_value(t->structure.fields[i].type,
2603 t->structure.fields[i].init);
2605 free(t->structure.fields);
2606 free_fieldlist(t->structure.field_list);
2609 static void structure_prepare_type(struct parse_context *c,
2610 struct type *t, int parse_time)
2613 struct fieldlist *f;
2615 if (!parse_time || t->structure.fields)
2618 for (f = t->structure.field_list; f; f=f->prev) {
2622 if (f->f.type->prepare_type)
2623 f->f.type->prepare_type(c, f->f.type, 1);
2624 if (f->init == NULL)
2628 propagate_types(f->init, c, &perr, f->f.type, 0);
2629 } while (perr & Eretry);
2631 c->parse_error += 1; // NOTEST
2634 t->structure.nfields = cnt;
2635 t->structure.fields = calloc(cnt, sizeof(struct field));
2636 f = t->structure.field_list;
2638 int a = f->f.type->align;
2640 t->structure.fields[cnt] = f->f;
2641 if (t->size & (a-1))
2642 t->size = (t->size | (a-1)) + 1;
2643 t->structure.fields[cnt].offset = t->size;
2644 t->size += ((f->f.type->size - 1) | (a-1)) + 1;
2648 if (f->init && !c->parse_error) {
2649 struct value vl = interp_exec(c, f->init, NULL);
2650 t->structure.fields[cnt].init =
2651 global_alloc(c, f->f.type, NULL, &vl);
2658 static struct type structure_prototype = {
2659 .init = structure_init,
2660 .free = structure_free,
2661 .free_type = structure_free_type,
2662 .print_type_decl = structure_print_type,
2663 .prepare_type = structure_prepare_type,
2677 ###### free exec cases
2679 free_exec(cast(fieldref, e)->left);
2683 ###### declare terminals
2688 | Term . IDENTIFIER ${ {
2689 struct fieldref *fr = new_pos(fieldref, $2);
2696 ###### print exec cases
2700 struct fieldref *f = cast(fieldref, e);
2701 print_exec(f->left, -1, bracket);
2702 printf(".%.*s", f->name.len, f->name.txt);
2706 ###### ast functions
2707 static int find_struct_index(struct type *type, struct text field)
2710 for (i = 0; i < type->structure.nfields; i++)
2711 if (text_cmp(type->structure.fields[i].name, field) == 0)
2716 ###### propagate exec cases
2720 struct fieldref *f = cast(fieldref, prog);
2721 struct type *st = propagate_types(f->left, c, perr, NULL, 0);
2724 type_err(c, "error: unknown type for field access", f->left, // UNTESTED
2726 else if (st->init != structure_init)
2727 type_err(c, "error: field reference attempted on %1, not a struct",
2728 f->left, st, 0, NULL);
2729 else if (f->index == -2) {
2730 f->index = find_struct_index(st, f->name);
2732 type_err(c, "error: cannot find requested field in %1",
2733 f->left, st, 0, NULL);
2735 if (f->index >= 0) {
2736 struct type *ft = st->structure.fields[f->index].type;
2737 if (!type_compat(type, ft, rules))
2738 type_err(c, "error: have %1 but need %2", prog,
2745 ###### interp exec cases
2748 struct fieldref *f = cast(fieldref, e);
2750 struct value *lleft = linterp_exec(c, f->left, <ype);
2751 lrv = (void*)lleft->ptr + ltype->structure.fields[f->index].offset;
2752 rvtype = ltype->structure.fields[f->index].type;
2756 ###### top level grammar
2757 DeclareStruct -> struct IDENTIFIER FieldBlock Newlines ${ {
2759 add_type(c, $2.txt, &structure_prototype);
2760 t->structure.field_list = $<FB;
2764 FieldBlock -> { IN OptNL FieldLines OUT OptNL } ${ $0 = $<FL; }$
2765 | { SimpleFieldList } ${ $0 = $<SFL; }$
2766 | IN OptNL FieldLines OUT ${ $0 = $<FL; }$
2767 | SimpleFieldList EOL ${ $0 = $<SFL; }$
2769 FieldLines -> SimpleFieldList Newlines ${ $0 = $<SFL; }$
2770 | FieldLines SimpleFieldList Newlines ${
2775 SimpleFieldList -> Field ${ $0 = $<F; }$
2776 | SimpleFieldList ; Field ${
2780 | SimpleFieldList ; ${
2783 | ERROR ${ tok_err(c, "Syntax error in struct field", &$1); }$
2785 Field -> IDENTIFIER : Type = Expression ${ {
2786 $0 = calloc(1, sizeof(struct fieldlist));
2787 $0->f.name = $ID.txt;
2788 $0->f.type = $<Type;
2792 | IDENTIFIER : Type ${
2793 $0 = calloc(1, sizeof(struct fieldlist));
2794 $0->f.name = $ID.txt;
2795 $0->f.type = $<Type;
2798 ###### forward decls
2799 static void structure_print_type(struct type *t, FILE *f);
2801 ###### value functions
2802 static void structure_print_type(struct type *t, FILE *f)
2806 fprintf(f, "struct %.*s\n", t->name.len, t->name.txt);
2808 for (i = 0; i < t->structure.nfields; i++) {
2809 struct field *fl = t->structure.fields + i;
2810 fprintf(f, " %.*s : ", fl->name.len, fl->name.txt);
2811 type_print(fl->type, f);
2812 if (fl->type->print && fl->init) {
2814 if (fl->type == Tstr)
2815 fprintf(f, "\""); // UNTESTED
2816 print_value(fl->type, fl->init, f);
2817 if (fl->type == Tstr)
2818 fprintf(f, "\""); // UNTESTED
2824 ###### print type decls
2829 while (target != 0) {
2831 for (t = context.typelist; t ; t=t->next)
2832 if (!t->anon && t->print_type_decl &&
2842 t->print_type_decl(t, stdout);
2850 A function is a chunk of code which can be passed parameters and can
2851 return results. Each function has a type which includes the set of
2852 parameters and the return value. As yet these types cannot be declared
2853 separately from the function itself.
2855 The parameters can be specified either in parentheses as a ';' separated
2858 ##### Example: function 1
2860 func main(av:[ac::number]string; env:[envc::number]string)
2863 or as an indented list of one parameter per line (though each line can
2864 be a ';' separated list)
2866 ##### Example: function 2
2869 argv:[argc::number]string
2870 env:[envc::number]string
2874 In the first case a return type can follow the parentheses after a colon,
2875 in the second it is given on a line starting with the word `return`.
2877 ##### Example: functions that return
2879 func add(a:number; b:number): number
2889 Rather than returning a type, the function can specify a set of local
2890 variables to return as a struct. The values of these variables when the
2891 function exits will be provided to the caller. For this the return type
2892 is replaced with a block of result declarations, either in parentheses
2893 or bracketed by `return` and `do`.
2895 ##### Example: functions returning multiple variables
2897 func to_cartesian(rho:number; theta:number):(x:number; y:number)
2910 For constructing the lists we use a `List` binode, which will be
2911 further detailed when Expression Lists are introduced.
2913 ###### type union fields
2916 struct binode *params;
2917 struct type *return_type;
2918 struct variable *scope;
2919 int inline_result; // return value is at start of 'local'
2923 ###### value union fields
2924 struct exec *function;
2926 ###### type functions
2927 void (*check_args)(struct parse_context *c, enum prop_err *perr,
2928 struct type *require, struct exec *args);
2930 ###### value functions
2932 static void function_free(struct type *type, struct value *val)
2934 free_exec(val->function);
2935 val->function = NULL;
2938 static int function_compat(struct type *require, struct type *have)
2940 // FIXME can I do anything here yet?
2944 static void function_check_args(struct parse_context *c, enum prop_err *perr,
2945 struct type *require, struct exec *args)
2947 /* This should be 'compat', but we don't have a 'tuple' type to
2948 * hold the type of 'args'
2950 struct binode *arg = cast(binode, args);
2951 struct binode *param = require->function.params;
2954 struct var *pv = cast(var, param->left);
2956 type_err(c, "error: insufficient arguments to function.",
2957 args, NULL, 0, NULL);
2961 propagate_types(arg->left, c, perr, pv->var->type, 0);
2962 param = cast(binode, param->right);
2963 arg = cast(binode, arg->right);
2966 type_err(c, "error: too many arguments to function.",
2967 args, NULL, 0, NULL);
2970 static void function_print(struct type *type, struct value *val, FILE *f)
2972 print_exec(val->function, 1, 0);
2975 static void function_print_type_decl(struct type *type, FILE *f)
2979 for (b = type->function.params; b; b = cast(binode, b->right)) {
2980 struct variable *v = cast(var, b->left)->var;
2981 fprintf(f, "%.*s%s", v->name->name.len, v->name->name.txt,
2982 v->constant ? "::" : ":");
2983 type_print(v->type, f);
2988 if (type->function.return_type != Tnone) {
2990 if (type->function.inline_result) {
2992 struct type *t = type->function.return_type;
2994 for (i = 0; i < t->structure.nfields; i++) {
2995 struct field *fl = t->structure.fields + i;
2998 fprintf(f, "%.*s:", fl->name.len, fl->name.txt);
2999 type_print(fl->type, f);
3003 type_print(type->function.return_type, f);
3008 static void function_free_type(struct type *t)
3010 free_exec(t->function.params);
3013 static struct type function_prototype = {
3014 .size = sizeof(void*),
3015 .align = sizeof(void*),
3016 .free = function_free,
3017 .compat = function_compat,
3018 .check_args = function_check_args,
3019 .print = function_print,
3020 .print_type_decl = function_print_type_decl,
3021 .free_type = function_free_type,
3024 ###### declare terminals
3034 FuncName -> IDENTIFIER ${ {
3035 struct variable *v = var_decl(c, $1.txt);
3036 struct var *e = new_pos(var, $1);
3042 v = var_ref(c, $1.txt);
3044 type_err(c, "error: function '%v' redeclared",
3046 type_err(c, "info: this is where '%v' was first declared",
3047 v->where_decl, NULL, 0, NULL);
3053 Args -> ArgsLine NEWLINE ${ $0 = $<AL; }$
3054 | Args ArgsLine NEWLINE ${ {
3055 struct binode *b = $<AL;
3056 struct binode **bp = &b;
3058 bp = (struct binode **)&(*bp)->left;
3063 ArgsLine -> ${ $0 = NULL; }$
3064 | Varlist ${ $0 = $<1; }$
3065 | Varlist ; ${ $0 = $<1; }$
3067 Varlist -> Varlist ; ArgDecl ${
3081 ArgDecl -> IDENTIFIER : FormalType ${ {
3082 struct variable *v = var_decl(c, $1.txt);
3088 ##### Function calls
3090 A function call can appear either as an expression or as a statement.
3091 We use a new 'Funcall' binode type to link the function with a list of
3092 arguments, form with the 'List' nodes.
3094 We have already seen the "Term" which is how a function call can appear
3095 in an expression. To parse a function call into a statement we include
3096 it in the "SimpleStatement Grammar" which will be described later.
3102 | Term ( ExpressionList ) ${ {
3103 struct binode *b = new(binode);
3106 b->right = reorder_bilist($<EL);
3110 struct binode *b = new(binode);
3117 ###### SimpleStatement Grammar
3119 | Term ( ExpressionList ) ${ {
3120 struct binode *b = new(binode);
3123 b->right = reorder_bilist($<EL);
3127 ###### print binode cases
3130 do_indent(indent, "");
3131 print_exec(b->left, -1, bracket);
3133 for (b = cast(binode, b->right); b; b = cast(binode, b->right)) {
3136 print_exec(b->left, -1, bracket);
3146 ###### propagate binode cases
3149 /* Every arg must match formal parameter, and result
3150 * is return type of function
3152 struct binode *args = cast(binode, b->right);
3153 struct var *v = cast(var, b->left);
3155 if (!v->var->type || v->var->type->check_args == NULL) {
3156 type_err(c, "error: attempt to call a non-function.",
3157 prog, NULL, 0, NULL);
3161 v->var->type->check_args(c, perr, v->var->type, args);
3162 return v->var->type->function.return_type;
3165 ###### interp binode cases
3168 struct var *v = cast(var, b->left);
3169 struct type *t = v->var->type;
3170 void *oldlocal = c->local;
3171 int old_size = c->local_size;
3172 void *local = calloc(1, t->function.local_size);
3173 struct value *fbody = var_value(c, v->var);
3174 struct binode *arg = cast(binode, b->right);
3175 struct binode *param = t->function.params;
3178 struct var *pv = cast(var, param->left);
3179 struct type *vtype = NULL;
3180 struct value val = interp_exec(c, arg->left, &vtype);
3182 c->local = local; c->local_size = t->function.local_size;
3183 lval = var_value(c, pv->var);
3184 c->local = oldlocal; c->local_size = old_size;
3185 memcpy(lval, &val, vtype->size);
3186 param = cast(binode, param->right);
3187 arg = cast(binode, arg->right);
3189 c->local = local; c->local_size = t->function.local_size;
3190 if (t->function.inline_result && dtype) {
3191 _interp_exec(c, fbody->function, NULL, NULL);
3192 memcpy(dest, local, dtype->size);
3193 rvtype = ret.type = NULL;
3195 rv = interp_exec(c, fbody->function, &rvtype);
3196 c->local = oldlocal; c->local_size = old_size;
3201 ## Complex executables: statements and expressions
3203 Now that we have types and values and variables and most of the basic
3204 Terms which provide access to these, we can explore the more complex
3205 code that combine all of these to get useful work done. Specifically
3206 statements and expressions.
3208 Expressions are various combinations of Terms. We will use operator
3209 precedence to ensure correct parsing. The simplest Expression is just a
3210 Term - others will follow.
3215 Expression -> Term ${ $0 = $<Term; }$
3216 ## expression grammar
3218 ### Expressions: Conditional
3220 Our first user of the `binode` will be conditional expressions, which
3221 is a bit odd as they actually have three components. That will be
3222 handled by having 2 binodes for each expression. The conditional
3223 expression is the lowest precedence operator which is why we define it
3224 first - to start the precedence list.
3226 Conditional expressions are of the form "value `if` condition `else`
3227 other_value". They associate to the right, so everything to the right
3228 of `else` is part of an else value, while only a higher-precedence to
3229 the left of `if` is the if values. Between `if` and `else` there is no
3230 room for ambiguity, so a full conditional expression is allowed in
3236 ###### declare terminals
3240 ###### expression grammar
3242 | Expression if Expression else Expression $$ifelse ${ {
3243 struct binode *b1 = new(binode);
3244 struct binode *b2 = new(binode);
3254 ###### print binode cases
3257 b2 = cast(binode, b->right);
3258 if (bracket) printf("(");
3259 print_exec(b2->left, -1, bracket);
3261 print_exec(b->left, -1, bracket);
3263 print_exec(b2->right, -1, bracket);
3264 if (bracket) printf(")");
3267 ###### propagate binode cases
3270 /* cond must be Tbool, others must match */
3271 struct binode *b2 = cast(binode, b->right);
3274 propagate_types(b->left, c, perr, Tbool, 0);
3275 t = propagate_types(b2->left, c, perr, type, Rnolabel);
3276 t2 = propagate_types(b2->right, c, perr, type ?: t, Rnolabel);
3280 ###### interp binode cases
3283 struct binode *b2 = cast(binode, b->right);
3284 left = interp_exec(c, b->left, <ype);
3286 rv = interp_exec(c, b2->left, &rvtype); // UNTESTED
3288 rv = interp_exec(c, b2->right, &rvtype);
3294 We take a brief detour, now that we have expressions, to describe lists
3295 of expressions. These will be needed for function parameters and
3296 possibly other situations. They seem generic enough to introduce here
3297 to be used elsewhere.
3299 And ExpressionList will use the `List` type of `binode`, building up at
3300 the end. And place where they are used will probably call
3301 `reorder_bilist()` to get a more normal first/next arrangement.
3303 ###### declare terminals
3306 `List` execs have no implicit semantics, so they are never propagated or
3307 interpreted. The can be printed as a comma separate list, which is how
3308 they are parsed. Note they are also used for function formal parameter
3309 lists. In that case a separate function is used to print them.
3311 ###### print binode cases
3315 print_exec(b->left, -1, bracket);
3318 b = cast(binode, b->right);
3322 ###### propagate binode cases
3323 case List: abort(); // NOTEST
3324 ###### interp binode cases
3325 case List: abort(); // NOTEST
3330 ExpressionList -> ExpressionList , Expression ${
3343 ### Expressions: Boolean
3345 The next class of expressions to use the `binode` will be Boolean
3346 expressions. "`and then`" and "`or else`" are similar to `and` and `or`
3347 have same corresponding precendence. The difference is that they don't
3348 evaluate the second expression if not necessary.
3357 ###### declare terminals
3362 ###### expression grammar
3363 | Expression or Expression ${ {
3364 struct binode *b = new(binode);
3370 | Expression or else Expression ${ {
3371 struct binode *b = new(binode);
3378 | Expression and Expression ${ {
3379 struct binode *b = new(binode);
3385 | Expression and then Expression ${ {
3386 struct binode *b = new(binode);
3393 | not Expression ${ {
3394 struct binode *b = new(binode);
3400 ###### print binode cases
3402 if (bracket) printf("(");
3403 print_exec(b->left, -1, bracket);
3405 print_exec(b->right, -1, bracket);
3406 if (bracket) printf(")");
3409 if (bracket) printf("(");
3410 print_exec(b->left, -1, bracket);
3411 printf(" and then ");
3412 print_exec(b->right, -1, bracket);
3413 if (bracket) printf(")");
3416 if (bracket) printf("(");
3417 print_exec(b->left, -1, bracket);
3419 print_exec(b->right, -1, bracket);
3420 if (bracket) printf(")");
3423 if (bracket) printf("(");
3424 print_exec(b->left, -1, bracket);
3425 printf(" or else ");
3426 print_exec(b->right, -1, bracket);
3427 if (bracket) printf(")");
3430 if (bracket) printf("(");
3432 print_exec(b->right, -1, bracket);
3433 if (bracket) printf(")");
3436 ###### propagate binode cases
3442 /* both must be Tbool, result is Tbool */
3443 propagate_types(b->left, c, perr, Tbool, 0);
3444 propagate_types(b->right, c, perr, Tbool, 0);
3445 if (type && type != Tbool)
3446 type_err(c, "error: %1 operation found where %2 expected", prog,
3450 ###### interp binode cases
3452 rv = interp_exec(c, b->left, &rvtype);
3453 right = interp_exec(c, b->right, &rtype);
3454 rv.bool = rv.bool && right.bool;
3457 rv = interp_exec(c, b->left, &rvtype);
3459 rv = interp_exec(c, b->right, NULL);
3462 rv = interp_exec(c, b->left, &rvtype);
3463 right = interp_exec(c, b->right, &rtype);
3464 rv.bool = rv.bool || right.bool;
3467 rv = interp_exec(c, b->left, &rvtype);
3469 rv = interp_exec(c, b->right, NULL);
3472 rv = interp_exec(c, b->right, &rvtype);
3476 ### Expressions: Comparison
3478 Of slightly higher precedence that Boolean expressions are Comparisons.
3479 A comparison takes arguments of any comparable type, but the two types
3482 To simplify the parsing we introduce an `eop` which can record an
3483 expression operator, and the `CMPop` non-terminal will match one of them.
3490 ###### ast functions
3491 static void free_eop(struct eop *e)
3505 ###### declare terminals
3506 $LEFT < > <= >= == != CMPop
3508 ###### expression grammar
3509 | Expression CMPop Expression ${ {
3510 struct binode *b = new(binode);
3520 CMPop -> < ${ $0.op = Less; }$
3521 | > ${ $0.op = Gtr; }$
3522 | <= ${ $0.op = LessEq; }$
3523 | >= ${ $0.op = GtrEq; }$
3524 | == ${ $0.op = Eql; }$
3525 | != ${ $0.op = NEql; }$
3527 ###### print binode cases
3535 if (bracket) printf("(");
3536 print_exec(b->left, -1, bracket);
3538 case Less: printf(" < "); break;
3539 case LessEq: printf(" <= "); break;
3540 case Gtr: printf(" > "); break;
3541 case GtrEq: printf(" >= "); break;
3542 case Eql: printf(" == "); break;
3543 case NEql: printf(" != "); break;
3544 default: abort(); // NOTEST
3546 print_exec(b->right, -1, bracket);
3547 if (bracket) printf(")");
3550 ###### propagate binode cases
3557 /* Both must match but not be labels, result is Tbool */
3558 t = propagate_types(b->left, c, perr, NULL, Rnolabel);
3560 propagate_types(b->right, c, perr, t, 0);
3562 t = propagate_types(b->right, c, perr, NULL, Rnolabel); // UNTESTED
3564 t = propagate_types(b->left, c, perr, t, 0); // UNTESTED
3566 if (!type_compat(type, Tbool, 0))
3567 type_err(c, "error: Comparison returns %1 but %2 expected", prog,
3568 Tbool, rules, type);
3571 ###### interp binode cases
3580 left = interp_exec(c, b->left, <ype);
3581 right = interp_exec(c, b->right, &rtype);
3582 cmp = value_cmp(ltype, rtype, &left, &right);
3585 case Less: rv.bool = cmp < 0; break;
3586 case LessEq: rv.bool = cmp <= 0; break;
3587 case Gtr: rv.bool = cmp > 0; break;
3588 case GtrEq: rv.bool = cmp >= 0; break;
3589 case Eql: rv.bool = cmp == 0; break;
3590 case NEql: rv.bool = cmp != 0; break;
3591 default: rv.bool = 0; break; // NOTEST
3596 ### Expressions: Arithmetic etc.
3598 The remaining expressions with the highest precedence are arithmetic,
3599 string concatenation, and string conversion. String concatenation
3600 (`++`) has the same precedence as multiplication and division, but lower
3603 String conversion is a temporary feature until I get a better type
3604 system. `$` is a prefix operator which expects a string and returns
3607 `+` and `-` are both infix and prefix operations (where they are
3608 absolute value and negation). These have different operator names.
3610 We also have a 'Bracket' operator which records where parentheses were
3611 found. This makes it easy to reproduce these when printing. Possibly I
3612 should only insert brackets were needed for precedence. Putting
3613 parentheses around an expression converts it into a Term,
3623 ###### declare terminals
3629 ###### expression grammar
3630 | Expression Eop Expression ${ {
3631 struct binode *b = new(binode);
3638 | Expression Top Expression ${ {
3639 struct binode *b = new(binode);
3646 | Uop Expression ${ {
3647 struct binode *b = new(binode);
3655 | ( Expression ) ${ {
3656 struct binode *b = new_pos(binode, $1);
3665 Eop -> + ${ $0.op = Plus; }$
3666 | - ${ $0.op = Minus; }$
3668 Uop -> + ${ $0.op = Absolute; }$
3669 | - ${ $0.op = Negate; }$
3670 | $ ${ $0.op = StringConv; }$
3672 Top -> * ${ $0.op = Times; }$
3673 | / ${ $0.op = Divide; }$
3674 | % ${ $0.op = Rem; }$
3675 | ++ ${ $0.op = Concat; }$
3677 ###### print binode cases
3684 if (bracket) printf("(");
3685 print_exec(b->left, indent, bracket);
3687 case Plus: fputs(" + ", stdout); break;
3688 case Minus: fputs(" - ", stdout); break;
3689 case Times: fputs(" * ", stdout); break;
3690 case Divide: fputs(" / ", stdout); break;
3691 case Rem: fputs(" % ", stdout); break;
3692 case Concat: fputs(" ++ ", stdout); break;
3693 default: abort(); // NOTEST
3695 print_exec(b->right, indent, bracket);
3696 if (bracket) printf(")");
3701 if (bracket) printf("(");
3703 case Absolute: fputs("+", stdout); break;
3704 case Negate: fputs("-", stdout); break;
3705 case StringConv: fputs("$", stdout); break;
3706 default: abort(); // NOTEST
3708 print_exec(b->right, indent, bracket);
3709 if (bracket) printf(")");
3713 print_exec(b->right, indent, bracket);
3717 ###### propagate binode cases
3723 /* both must be numbers, result is Tnum */
3726 /* as propagate_types ignores a NULL,
3727 * unary ops fit here too */
3728 propagate_types(b->left, c, perr, Tnum, 0);
3729 propagate_types(b->right, c, perr, Tnum, 0);
3730 if (!type_compat(type, Tnum, 0))
3731 type_err(c, "error: Arithmetic returns %1 but %2 expected", prog,
3736 /* both must be Tstr, result is Tstr */
3737 propagate_types(b->left, c, perr, Tstr, 0);
3738 propagate_types(b->right, c, perr, Tstr, 0);
3739 if (!type_compat(type, Tstr, 0))
3740 type_err(c, "error: Concat returns %1 but %2 expected", prog,
3745 /* op must be string, result is number */
3746 propagate_types(b->left, c, perr, Tstr, 0);
3747 if (!type_compat(type, Tnum, 0))
3748 type_err(c, // UNTESTED
3749 "error: Can only convert string to number, not %1",
3750 prog, type, 0, NULL);
3754 return propagate_types(b->right, c, perr, type, 0);
3756 ###### interp binode cases
3759 rv = interp_exec(c, b->left, &rvtype);
3760 right = interp_exec(c, b->right, &rtype);
3761 mpq_add(rv.num, rv.num, right.num);
3764 rv = interp_exec(c, b->left, &rvtype);
3765 right = interp_exec(c, b->right, &rtype);
3766 mpq_sub(rv.num, rv.num, right.num);
3769 rv = interp_exec(c, b->left, &rvtype);
3770 right = interp_exec(c, b->right, &rtype);
3771 mpq_mul(rv.num, rv.num, right.num);
3774 rv = interp_exec(c, b->left, &rvtype);
3775 right = interp_exec(c, b->right, &rtype);
3776 mpq_div(rv.num, rv.num, right.num);
3781 left = interp_exec(c, b->left, <ype);
3782 right = interp_exec(c, b->right, &rtype);
3783 mpz_init(l); mpz_init(r); mpz_init(rem);
3784 mpz_tdiv_q(l, mpq_numref(left.num), mpq_denref(left.num));
3785 mpz_tdiv_q(r, mpq_numref(right.num), mpq_denref(right.num));
3786 mpz_tdiv_r(rem, l, r);
3787 val_init(Tnum, &rv);
3788 mpq_set_z(rv.num, rem);
3789 mpz_clear(r); mpz_clear(l); mpz_clear(rem);
3794 rv = interp_exec(c, b->right, &rvtype);
3795 mpq_neg(rv.num, rv.num);
3798 rv = interp_exec(c, b->right, &rvtype);
3799 mpq_abs(rv.num, rv.num);
3802 rv = interp_exec(c, b->right, &rvtype);
3805 left = interp_exec(c, b->left, <ype);
3806 right = interp_exec(c, b->right, &rtype);
3808 rv.str = text_join(left.str, right.str);
3811 right = interp_exec(c, b->right, &rvtype);
3815 struct text tx = right.str;
3818 if (tx.txt[0] == '-') {
3819 neg = 1; // UNTESTED
3820 tx.txt++; // UNTESTED
3821 tx.len--; // UNTESTED
3823 if (number_parse(rv.num, tail, tx) == 0)
3824 mpq_init(rv.num); // UNTESTED
3826 mpq_neg(rv.num, rv.num); // UNTESTED
3828 printf("Unsupported suffix: %.*s\n", tx.len, tx.txt); // UNTESTED
3832 ###### value functions
3834 static struct text text_join(struct text a, struct text b)
3837 rv.len = a.len + b.len;
3838 rv.txt = malloc(rv.len);
3839 memcpy(rv.txt, a.txt, a.len);
3840 memcpy(rv.txt+a.len, b.txt, b.len);
3844 ### Blocks, Statements, and Statement lists.
3846 Now that we have expressions out of the way we need to turn to
3847 statements. There are simple statements and more complex statements.
3848 Simple statements do not contain (syntactic) newlines, complex statements do.
3850 Statements often come in sequences and we have corresponding simple
3851 statement lists and complex statement lists.
3852 The former comprise only simple statements separated by semicolons.
3853 The later comprise complex statements and simple statement lists. They are
3854 separated by newlines. Thus the semicolon is only used to separate
3855 simple statements on the one line. This may be overly restrictive,
3856 but I'm not sure I ever want a complex statement to share a line with
3859 Note that a simple statement list can still use multiple lines if
3860 subsequent lines are indented, so
3862 ###### Example: wrapped simple statement list
3867 is a single simple statement list. This might allow room for
3868 confusion, so I'm not set on it yet.
3870 A simple statement list needs no extra syntax. A complex statement
3871 list has two syntactic forms. It can be enclosed in braces (much like
3872 C blocks), or it can be introduced by an indent and continue until an
3873 unindented newline (much like Python blocks). With this extra syntax
3874 it is referred to as a block.
3876 Note that a block does not have to include any newlines if it only
3877 contains simple statements. So both of:
3879 if condition: a=b; d=f
3881 if condition { a=b; print f }
3885 In either case the list is constructed from a `binode` list with
3886 `Block` as the operator. When parsing the list it is most convenient
3887 to append to the end, so a list is a list and a statement. When using
3888 the list it is more convenient to consider a list to be a statement
3889 and a list. So we need a function to re-order a list.
3890 `reorder_bilist` serves this purpose.
3892 The only stand-alone statement we introduce at this stage is `pass`
3893 which does nothing and is represented as a `NULL` pointer in a `Block`
3894 list. Other stand-alone statements will follow once the infrastructure
3897 As many statements will use binodes, we declare a binode pointer 'b' in
3898 the common header for all reductions to use.
3900 ###### Parser: reduce
3911 Block -> { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3912 | { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3913 | SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
3914 | SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
3915 | IN OptNL Statementlist OUT ${ $0 = $<Sl; }$
3917 OpenBlock -> OpenScope { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3918 | OpenScope { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3919 | OpenScope SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
3920 | OpenScope SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
3921 | IN OpenScope OptNL Statementlist OUT ${ $0 = $<Sl; }$
3923 UseBlock -> { OpenScope IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3924 | { OpenScope SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3925 | IN OpenScope OptNL Statementlist OUT ${ $0 = $<Sl; }$
3927 ColonBlock -> { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3928 | { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3929 | : SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
3930 | : SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
3931 | : IN OptNL Statementlist OUT ${ $0 = $<Sl; }$
3933 Statementlist -> ComplexStatements ${ $0 = reorder_bilist($<CS); }$
3935 ComplexStatements -> ComplexStatements ComplexStatement ${
3945 | ComplexStatement ${
3957 ComplexStatement -> SimpleStatements Newlines ${
3958 $0 = reorder_bilist($<SS);
3960 | SimpleStatements ; Newlines ${
3961 $0 = reorder_bilist($<SS);
3963 ## ComplexStatement Grammar
3966 SimpleStatements -> SimpleStatements ; SimpleStatement ${
3972 | SimpleStatement ${
3981 SimpleStatement -> pass ${ $0 = NULL; }$
3982 | ERROR ${ tok_err(c, "Syntax error in statement", &$1); }$
3983 ## SimpleStatement Grammar
3985 ###### print binode cases
3989 if (b->left == NULL) // UNTESTED
3990 printf("pass"); // UNTESTED
3992 print_exec(b->left, indent, bracket); // UNTESTED
3993 if (b->right) { // UNTESTED
3994 printf("; "); // UNTESTED
3995 print_exec(b->right, indent, bracket); // UNTESTED
3998 // block, one per line
3999 if (b->left == NULL)
4000 do_indent(indent, "pass\n");
4002 print_exec(b->left, indent, bracket);
4004 print_exec(b->right, indent, bracket);
4008 ###### propagate binode cases
4011 /* If any statement returns something other than Tnone
4012 * or Tbool then all such must return same type.
4013 * As each statement may be Tnone or something else,
4014 * we must always pass NULL (unknown) down, otherwise an incorrect
4015 * error might occur. We never return Tnone unless it is
4020 for (e = b; e; e = cast(binode, e->right)) {
4021 t = propagate_types(e->left, c, perr, NULL, rules);
4022 if ((rules & Rboolok) && (t == Tbool || t == Tnone))
4024 if (t == Tnone && e->right)
4025 /* Only the final statement *must* return a value
4033 type_err(c, "error: expected %1%r, found %2",
4034 e->left, type, rules, t);
4040 ###### interp binode cases
4042 while (rvtype == Tnone &&
4045 rv = interp_exec(c, b->left, &rvtype);
4046 b = cast(binode, b->right);
4050 ### The Print statement
4052 `print` is a simple statement that takes a comma-separated list of
4053 expressions and prints the values separated by spaces and terminated
4054 by a newline. No control of formatting is possible.
4056 `print` uses `ExpressionList` to collect the expressions and stores them
4057 on the left side of a `Print` binode unlessthere is a trailing comma
4058 when the list is stored on the `right` side and no trailing newline is
4064 ##### declare terminals
4067 ###### SimpleStatement Grammar
4069 | print ExpressionList ${
4070 $0 = b = new(binode);
4073 b->left = reorder_bilist($<EL);
4075 | print ExpressionList , ${ {
4076 $0 = b = new(binode);
4078 b->right = reorder_bilist($<EL);
4082 $0 = b = new(binode);
4088 ###### print binode cases
4091 do_indent(indent, "print");
4093 print_exec(b->right, -1, bracket);
4096 print_exec(b->left, -1, bracket);
4101 ###### propagate binode cases
4104 /* don't care but all must be consistent */
4106 b = cast(binode, b->left);
4108 b = cast(binode, b->right);
4110 propagate_types(b->left, c, perr, NULL, Rnolabel);
4111 b = cast(binode, b->right);
4115 ###### interp binode cases
4119 struct binode *b2 = cast(binode, b->left);
4121 b2 = cast(binode, b->right);
4122 for (; b2; b2 = cast(binode, b2->right)) {
4123 left = interp_exec(c, b2->left, <ype);
4124 print_value(ltype, &left, stdout);
4125 free_value(ltype, &left);
4129 if (b->right == NULL)
4135 ###### Assignment statement
4137 An assignment will assign a value to a variable, providing it hasn't
4138 been declared as a constant. The analysis phase ensures that the type
4139 will be correct so the interpreter just needs to perform the
4140 calculation. There is a form of assignment which declares a new
4141 variable as well as assigning a value. If a name is assigned before
4142 it is declared, and error will be raised as the name is created as
4143 `Tlabel` and it is illegal to assign to such names.
4149 ###### declare terminals
4152 ###### SimpleStatement Grammar
4153 | Term = Expression ${
4154 $0 = b= new(binode);
4159 | VariableDecl = Expression ${
4160 $0 = b= new(binode);
4167 if ($1->var->where_set == NULL) {
4169 "Variable declared with no type or value: %v",
4173 $0 = b = new(binode);
4180 ###### print binode cases
4183 do_indent(indent, "");
4184 print_exec(b->left, indent, bracket);
4186 print_exec(b->right, indent, bracket);
4193 struct variable *v = cast(var, b->left)->var;
4194 do_indent(indent, "");
4195 print_exec(b->left, indent, bracket);
4196 if (cast(var, b->left)->var->constant) {
4198 if (v->explicit_type) {
4199 type_print(v->type, stdout);
4204 if (v->explicit_type) {
4205 type_print(v->type, stdout);
4211 print_exec(b->right, indent, bracket);
4218 ###### propagate binode cases
4222 /* Both must match and not be labels,
4223 * Type must support 'dup',
4224 * For Assign, left must not be constant.
4227 t = propagate_types(b->left, c, perr, NULL,
4228 Rnolabel | (b->op == Assign ? Rnoconstant : 0));
4233 if (propagate_types(b->right, c, perr, t, 0) != t)
4234 if (b->left->type == Xvar)
4235 type_err(c, "info: variable '%v' was set as %1 here.",
4236 cast(var, b->left)->var->where_set, t, rules, NULL);
4238 t = propagate_types(b->right, c, perr, NULL, Rnolabel);
4240 propagate_types(b->left, c, perr, t,
4241 (b->op == Assign ? Rnoconstant : 0));
4243 if (t && t->dup == NULL && t->name.txt[0] != ' ') // HACK
4244 type_err(c, "error: cannot assign value of type %1", b, t, 0, NULL);
4249 ###### interp binode cases
4252 lleft = linterp_exec(c, b->left, <ype);
4254 dinterp_exec(c, b->right, lleft, ltype, 1);
4260 struct variable *v = cast(var, b->left)->var;
4263 val = var_value(c, v);
4264 if (v->type->prepare_type)
4265 v->type->prepare_type(c, v->type, 0);
4267 dinterp_exec(c, b->right, val, v->type, 0);
4269 val_init(v->type, val);
4273 ### The `use` statement
4275 The `use` statement is the last "simple" statement. It is needed when a
4276 statement block can return a value. This includes the body of a
4277 function which has a return type, and the "condition" code blocks in
4278 `if`, `while`, and `switch` statements.
4283 ###### declare terminals
4286 ###### SimpleStatement Grammar
4288 $0 = b = new_pos(binode, $1);
4291 if (b->right->type == Xvar) {
4292 struct var *v = cast(var, b->right);
4293 if (v->var->type == Tnone) {
4294 /* Convert this to a label */
4297 v->var->type = Tlabel;
4298 val = global_alloc(c, Tlabel, v->var, NULL);
4304 ###### print binode cases
4307 do_indent(indent, "use ");
4308 print_exec(b->right, -1, bracket);
4313 ###### propagate binode cases
4316 /* result matches value */
4317 return propagate_types(b->right, c, perr, type, 0);
4319 ###### interp binode cases
4322 rv = interp_exec(c, b->right, &rvtype);
4325 ### The Conditional Statement
4327 This is the biggy and currently the only complex statement. This
4328 subsumes `if`, `while`, `do/while`, `switch`, and some parts of `for`.
4329 It is comprised of a number of parts, all of which are optional though
4330 set combinations apply. Each part is (usually) a key word (`then` is
4331 sometimes optional) followed by either an expression or a code block,
4332 except the `casepart` which is a "key word and an expression" followed
4333 by a code block. The code-block option is valid for all parts and,
4334 where an expression is also allowed, the code block can use the `use`
4335 statement to report a value. If the code block does not report a value
4336 the effect is similar to reporting `True`.
4338 The `else` and `case` parts, as well as `then` when combined with
4339 `if`, can contain a `use` statement which will apply to some
4340 containing conditional statement. `for` parts, `do` parts and `then`
4341 parts used with `for` can never contain a `use`, except in some
4342 subordinate conditional statement.
4344 If there is a `forpart`, it is executed first, only once.
4345 If there is a `dopart`, then it is executed repeatedly providing
4346 always that the `condpart` or `cond`, if present, does not return a non-True
4347 value. `condpart` can fail to return any value if it simply executes
4348 to completion. This is treated the same as returning `True`.
4350 If there is a `thenpart` it will be executed whenever the `condpart`
4351 or `cond` returns True (or does not return any value), but this will happen
4352 *after* `dopart` (when present).
4354 If `elsepart` is present it will be executed at most once when the
4355 condition returns `False` or some value that isn't `True` and isn't
4356 matched by any `casepart`. If there are any `casepart`s, they will be
4357 executed when the condition returns a matching value.
4359 The particular sorts of values allowed in case parts has not yet been
4360 determined in the language design, so nothing is prohibited.
4362 The various blocks in this complex statement potentially provide scope
4363 for variables as described earlier. Each such block must include the
4364 "OpenScope" nonterminal before parsing the block, and must call
4365 `var_block_close()` when closing the block.
4367 The code following "`if`", "`switch`" and "`for`" does not get its own
4368 scope, but is in a scope covering the whole statement, so names
4369 declared there cannot be redeclared elsewhere. Similarly the
4370 condition following "`while`" is in a scope the covers the body
4371 ("`do`" part) of the loop, and which does not allow conditional scope
4372 extension. Code following "`then`" (both looping and non-looping),
4373 "`else`" and "`case`" each get their own local scope.
4375 The type requirements on the code block in a `whilepart` are quite
4376 unusal. It is allowed to return a value of some identifiable type, in
4377 which case the loop aborts and an appropriate `casepart` is run, or it
4378 can return a Boolean, in which case the loop either continues to the
4379 `dopart` (on `True`) or aborts and runs the `elsepart` (on `False`).
4380 This is different both from the `ifpart` code block which is expected to
4381 return a Boolean, or the `switchpart` code block which is expected to
4382 return the same type as the casepart values. The correct analysis of
4383 the type of the `whilepart` code block is the reason for the
4384 `Rboolok` flag which is passed to `propagate_types()`.
4386 The `cond_statement` cannot fit into a `binode` so a new `exec` is
4387 defined. As there are two scopes which cover multiple parts - one for
4388 the whole statement and one for "while" and "do" - and as we will use
4389 the 'struct exec' to track scopes, we actually need two new types of
4390 exec. One is a `binode` for the looping part, the rest is the
4391 `cond_statement`. The `cond_statement` will use an auxilliary `struct
4392 casepart` to track a list of case parts.
4403 struct exec *action;
4404 struct casepart *next;
4406 struct cond_statement {
4408 struct exec *forpart, *condpart, *thenpart, *elsepart;
4409 struct binode *looppart;
4410 struct casepart *casepart;
4413 ###### ast functions
4415 static void free_casepart(struct casepart *cp)
4419 free_exec(cp->value);
4420 free_exec(cp->action);
4427 static void free_cond_statement(struct cond_statement *s)
4431 free_exec(s->forpart);
4432 free_exec(s->condpart);
4433 free_exec(s->looppart);
4434 free_exec(s->thenpart);
4435 free_exec(s->elsepart);
4436 free_casepart(s->casepart);
4440 ###### free exec cases
4441 case Xcond_statement: free_cond_statement(cast(cond_statement, e)); break;
4443 ###### ComplexStatement Grammar
4444 | CondStatement ${ $0 = $<1; }$
4446 ###### declare terminals
4447 $TERM for then while do
4454 // A CondStatement must end with EOL, as does CondSuffix and
4456 // ForPart, ThenPart, SwitchPart, CasePart are non-empty and
4457 // may or may not end with EOL
4458 // WhilePart and IfPart include an appropriate Suffix
4460 // ForPart, SwitchPart, and IfPart open scopes, o we have to close
4461 // them. WhilePart opens and closes its own scope.
4462 CondStatement -> ForPart OptNL ThenPart OptNL WhilePart CondSuffix ${
4465 $0->thenpart = $<TP;
4466 $0->looppart = $<WP;
4467 var_block_close(c, CloseSequential, $0);
4469 | ForPart OptNL WhilePart CondSuffix ${
4472 $0->looppart = $<WP;
4473 var_block_close(c, CloseSequential, $0);
4475 | WhilePart CondSuffix ${
4477 $0->looppart = $<WP;
4479 | SwitchPart OptNL CasePart CondSuffix ${
4481 $0->condpart = $<SP;
4482 $CP->next = $0->casepart;
4483 $0->casepart = $<CP;
4484 var_block_close(c, CloseSequential, $0);
4486 | SwitchPart : IN OptNL CasePart CondSuffix OUT Newlines ${
4488 $0->condpart = $<SP;
4489 $CP->next = $0->casepart;
4490 $0->casepart = $<CP;
4491 var_block_close(c, CloseSequential, $0);
4493 | IfPart IfSuffix ${
4495 $0->condpart = $IP.condpart; $IP.condpart = NULL;
4496 $0->thenpart = $IP.thenpart; $IP.thenpart = NULL;
4497 // This is where we close an "if" statement
4498 var_block_close(c, CloseSequential, $0);
4501 CondSuffix -> IfSuffix ${
4504 | Newlines CasePart CondSuffix ${
4506 $CP->next = $0->casepart;
4507 $0->casepart = $<CP;
4509 | CasePart CondSuffix ${
4511 $CP->next = $0->casepart;
4512 $0->casepart = $<CP;
4515 IfSuffix -> Newlines ${ $0 = new(cond_statement); }$
4516 | Newlines ElsePart ${ $0 = $<EP; }$
4517 | ElsePart ${$0 = $<EP; }$
4519 ElsePart -> else OpenBlock Newlines ${
4520 $0 = new(cond_statement);
4521 $0->elsepart = $<OB;
4522 var_block_close(c, CloseElse, $0->elsepart);
4524 | else OpenScope CondStatement ${
4525 $0 = new(cond_statement);
4526 $0->elsepart = $<CS;
4527 var_block_close(c, CloseElse, $0->elsepart);
4531 CasePart -> case Expression OpenScope ColonBlock ${
4532 $0 = calloc(1,sizeof(struct casepart));
4535 var_block_close(c, CloseParallel, $0->action);
4539 // These scopes are closed in CondStatement
4540 ForPart -> for OpenBlock ${
4544 ThenPart -> then OpenBlock ${
4546 var_block_close(c, CloseSequential, $0);
4550 // This scope is closed in CondStatement
4551 WhilePart -> while UseBlock OptNL do OpenBlock ${
4556 var_block_close(c, CloseSequential, $0->right);
4557 var_block_close(c, CloseSequential, $0);
4559 | while OpenScope Expression OpenScope ColonBlock ${
4564 var_block_close(c, CloseSequential, $0->right);
4565 var_block_close(c, CloseSequential, $0);
4569 IfPart -> if UseBlock OptNL then OpenBlock ${
4572 var_block_close(c, CloseParallel, $0.thenpart);
4574 | if OpenScope Expression OpenScope ColonBlock ${
4577 var_block_close(c, CloseParallel, $0.thenpart);
4579 | if OpenScope Expression OpenScope OptNL then Block ${
4582 var_block_close(c, CloseParallel, $0.thenpart);
4586 // This scope is closed in CondStatement
4587 SwitchPart -> switch OpenScope Expression ${
4590 | switch UseBlock ${
4594 ###### print binode cases
4596 if (b->left && b->left->type == Xbinode &&
4597 cast(binode, b->left)->op == Block) {
4599 do_indent(indent, "while {\n");
4601 do_indent(indent, "while\n");
4602 print_exec(b->left, indent+1, bracket);
4604 do_indent(indent, "} do {\n");
4606 do_indent(indent, "do\n");
4607 print_exec(b->right, indent+1, bracket);
4609 do_indent(indent, "}\n");
4611 do_indent(indent, "while ");
4612 print_exec(b->left, 0, bracket);
4617 print_exec(b->right, indent+1, bracket);
4619 do_indent(indent, "}\n");
4623 ###### print exec cases
4625 case Xcond_statement:
4627 struct cond_statement *cs = cast(cond_statement, e);
4628 struct casepart *cp;
4630 do_indent(indent, "for");
4631 if (bracket) printf(" {\n"); else printf("\n");
4632 print_exec(cs->forpart, indent+1, bracket);
4635 do_indent(indent, "} then {\n");
4637 do_indent(indent, "then\n");
4638 print_exec(cs->thenpart, indent+1, bracket);
4640 if (bracket) do_indent(indent, "}\n");
4643 print_exec(cs->looppart, indent, bracket);
4647 do_indent(indent, "switch");
4649 do_indent(indent, "if");
4650 if (cs->condpart && cs->condpart->type == Xbinode &&
4651 cast(binode, cs->condpart)->op == Block) {
4656 print_exec(cs->condpart, indent+1, bracket);
4658 do_indent(indent, "}\n");
4660 do_indent(indent, "then\n");
4661 print_exec(cs->thenpart, indent+1, bracket);
4665 print_exec(cs->condpart, 0, bracket);
4671 print_exec(cs->thenpart, indent+1, bracket);
4673 do_indent(indent, "}\n");
4678 for (cp = cs->casepart; cp; cp = cp->next) {
4679 do_indent(indent, "case ");
4680 print_exec(cp->value, -1, 0);
4685 print_exec(cp->action, indent+1, bracket);
4687 do_indent(indent, "}\n");
4690 do_indent(indent, "else");
4695 print_exec(cs->elsepart, indent+1, bracket);
4697 do_indent(indent, "}\n");
4702 ###### propagate binode cases
4704 t = propagate_types(b->right, c, perr, Tnone, 0);
4705 if (!type_compat(Tnone, t, 0))
4706 *perr |= Efail; // UNTESTED
4707 return propagate_types(b->left, c, perr, type, rules);
4709 ###### propagate exec cases
4710 case Xcond_statement:
4712 // forpart and looppart->right must return Tnone
4713 // thenpart must return Tnone if there is a loopart,
4714 // otherwise it is like elsepart.
4716 // be bool if there is no casepart
4717 // match casepart->values if there is a switchpart
4718 // either be bool or match casepart->value if there
4720 // elsepart and casepart->action must match the return type
4721 // expected of this statement.
4722 struct cond_statement *cs = cast(cond_statement, prog);
4723 struct casepart *cp;
4725 t = propagate_types(cs->forpart, c, perr, Tnone, 0);
4726 if (!type_compat(Tnone, t, 0))
4727 *perr |= Efail; // UNTESTED
4730 t = propagate_types(cs->thenpart, c, perr, Tnone, 0);
4731 if (!type_compat(Tnone, t, 0))
4732 *perr |= Efail; // UNTESTED
4734 if (cs->casepart == NULL) {
4735 propagate_types(cs->condpart, c, perr, Tbool, 0);
4736 propagate_types(cs->looppart, c, perr, Tbool, 0);
4738 /* Condpart must match case values, with bool permitted */
4740 for (cp = cs->casepart;
4741 cp && !t; cp = cp->next)
4742 t = propagate_types(cp->value, c, perr, NULL, 0);
4743 if (!t && cs->condpart)
4744 t = propagate_types(cs->condpart, c, perr, NULL, Rboolok); // UNTESTED
4745 if (!t && cs->looppart)
4746 t = propagate_types(cs->looppart, c, perr, NULL, Rboolok); // UNTESTED
4747 // Now we have a type (I hope) push it down
4749 for (cp = cs->casepart; cp; cp = cp->next)
4750 propagate_types(cp->value, c, perr, t, 0);
4751 propagate_types(cs->condpart, c, perr, t, Rboolok);
4752 propagate_types(cs->looppart, c, perr, t, Rboolok);
4755 // (if)then, else, and case parts must return expected type.
4756 if (!cs->looppart && !type)
4757 type = propagate_types(cs->thenpart, c, perr, NULL, rules);
4759 type = propagate_types(cs->elsepart, c, perr, NULL, rules);
4760 for (cp = cs->casepart;
4762 cp = cp->next) // UNTESTED
4763 type = propagate_types(cp->action, c, perr, NULL, rules); // UNTESTED
4766 propagate_types(cs->thenpart, c, perr, type, rules);
4767 propagate_types(cs->elsepart, c, perr, type, rules);
4768 for (cp = cs->casepart; cp ; cp = cp->next)
4769 propagate_types(cp->action, c, perr, type, rules);
4775 ###### interp binode cases
4777 // This just performs one iterration of the loop
4778 rv = interp_exec(c, b->left, &rvtype);
4779 if (rvtype == Tnone ||
4780 (rvtype == Tbool && rv.bool != 0))
4781 // rvtype is Tnone or Tbool, doesn't need to be freed
4782 interp_exec(c, b->right, NULL);
4785 ###### interp exec cases
4786 case Xcond_statement:
4788 struct value v, cnd;
4789 struct type *vtype, *cndtype;
4790 struct casepart *cp;
4791 struct cond_statement *cs = cast(cond_statement, e);
4794 interp_exec(c, cs->forpart, NULL);
4796 while ((cnd = interp_exec(c, cs->looppart, &cndtype)),
4797 cndtype == Tnone || (cndtype == Tbool && cnd.bool != 0))
4798 interp_exec(c, cs->thenpart, NULL);
4800 cnd = interp_exec(c, cs->condpart, &cndtype);
4801 if ((cndtype == Tnone ||
4802 (cndtype == Tbool && cnd.bool != 0))) {
4803 // cnd is Tnone or Tbool, doesn't need to be freed
4804 rv = interp_exec(c, cs->thenpart, &rvtype);
4805 // skip else (and cases)
4809 for (cp = cs->casepart; cp; cp = cp->next) {
4810 v = interp_exec(c, cp->value, &vtype);
4811 if (value_cmp(cndtype, vtype, &v, &cnd) == 0) {
4812 free_value(vtype, &v);
4813 free_value(cndtype, &cnd);
4814 rv = interp_exec(c, cp->action, &rvtype);
4817 free_value(vtype, &v);
4819 free_value(cndtype, &cnd);
4821 rv = interp_exec(c, cs->elsepart, &rvtype);
4828 ### Top level structure
4830 All the language elements so far can be used in various places. Now
4831 it is time to clarify what those places are.
4833 At the top level of a file there will be a number of declarations.
4834 Many of the things that can be declared haven't been described yet,
4835 such as functions, procedures, imports, and probably more.
4836 For now there are two sorts of things that can appear at the top
4837 level. They are predefined constants, `struct` types, and the `main`
4838 function. While the syntax will allow the `main` function to appear
4839 multiple times, that will trigger an error if it is actually attempted.
4841 The various declarations do not return anything. They store the
4842 various declarations in the parse context.
4844 ###### Parser: grammar
4847 Ocean -> OptNL DeclarationList
4849 ## declare terminals
4857 DeclarationList -> Declaration
4858 | DeclarationList Declaration
4860 Declaration -> ERROR Newlines ${
4861 tok_err(c, // UNTESTED
4862 "error: unhandled parse error", &$1);
4868 ## top level grammar
4872 ### The `const` section
4874 As well as being defined in with the code that uses them, constants can
4875 be declared at the top level. These have full-file scope, so they are
4876 always `InScope`, even before(!) they have been declared. The value of
4877 a top level constant can be given as an expression, and this is
4878 evaluated after parsing and before execution.
4880 A function call can be used to evaluate a constant, but it will not have
4881 access to any program state, once such statement becomes meaningful.
4882 e.g. arguments and filesystem will not be visible.
4884 Constants are defined in a section that starts with the reserved word
4885 `const` and then has a block with a list of assignment statements.
4886 For syntactic consistency, these must use the double-colon syntax to
4887 make it clear that they are constants. Type can also be given: if
4888 not, the type will be determined during analysis, as with other
4891 ###### parse context
4892 struct binode *constlist;
4894 ###### top level grammar
4898 DeclareConstant -> const { IN OptNL ConstList OUT OptNL } Newlines
4899 | const { SimpleConstList } Newlines
4900 | const IN OptNL ConstList OUT Newlines
4901 | const SimpleConstList Newlines
4903 ConstList -> ConstList SimpleConstLine
4906 SimpleConstList -> SimpleConstList ; Const
4910 SimpleConstLine -> SimpleConstList Newlines
4911 | ERROR Newlines ${ tok_err(c, "Syntax error in constant", &$1); }$
4914 CType -> Type ${ $0 = $<1; }$
4918 Const -> IDENTIFIER :: CType = Expression ${ {
4920 struct binode *bl, *bv;
4921 struct var *var = new_pos(var, $ID);
4923 v = var_decl(c, $ID.txt);
4925 v->where_decl = var;
4931 v = var_ref(c, $1.txt);
4932 if (v->type == Tnone) {
4933 v->where_decl = var;
4939 tok_err(c, "error: name already declared", &$1);
4940 type_err(c, "info: this is where '%v' was first declared",
4941 v->where_decl, NULL, 0, NULL);
4953 bl->left = c->constlist;
4958 ###### core functions
4959 static void resolve_consts(struct parse_context *c)
4963 enum { none, some, cannot } progress = none;
4965 c->constlist = reorder_bilist(c->constlist);
4968 for (b = cast(binode, c->constlist); b;
4969 b = cast(binode, b->right)) {
4971 struct binode *vb = cast(binode, b->left);
4972 struct var *v = cast(var, vb->left);
4973 if (v->var->frame_pos >= 0)
4977 propagate_types(vb->right, c, &perr,
4979 } while (perr & Eretry);
4981 c->parse_error += 1;
4982 else if (!(perr & Enoconst)) {
4984 struct value res = interp_exec(
4985 c, vb->right, &v->var->type);
4986 global_alloc(c, v->var->type, v->var, &res);
4988 if (progress == cannot)
4989 type_err(c, "error: const %v cannot be resolved.",
4999 progress = cannot; break;
5001 progress = none; break;
5006 ###### print const decls
5011 for (b = cast(binode, context.constlist); b;
5012 b = cast(binode, b->right)) {
5013 struct binode *vb = cast(binode, b->left);
5014 struct var *vr = cast(var, vb->left);
5015 struct variable *v = vr->var;
5021 printf(" %.*s :: ", v->name->name.len, v->name->name.txt);
5022 type_print(v->type, stdout);
5024 print_exec(vb->right, -1, 0);
5029 ###### free const decls
5030 free_binode(context.constlist);
5032 ### Function declarations
5034 The code in an Ocean program is all stored in function declarations.
5035 One of the functions must be named `main` and it must accept an array of
5036 strings as a parameter - the command line arguments.
5038 As this is the top level, several things are handled a bit differently.
5039 The function is not interpreted by `interp_exec` as that isn't passed
5040 the argument list which the program requires. Similarly type analysis
5041 is a bit more interesting at this level.
5043 ###### ast functions
5045 static struct type *handle_results(struct parse_context *c,
5046 struct binode *results)
5048 /* Create a 'struct' type from the results list, which
5049 * is a list for 'struct var'
5051 struct type *t = add_anon_type(c, &structure_prototype,
5052 " function result");
5056 for (b = results; b; b = cast(binode, b->right))
5058 t->structure.nfields = cnt;
5059 t->structure.fields = calloc(cnt, sizeof(struct field));
5061 for (b = results; b; b = cast(binode, b->right)) {
5062 struct var *v = cast(var, b->left);
5063 struct field *f = &t->structure.fields[cnt++];
5064 int a = v->var->type->align;
5065 f->name = v->var->name->name;
5066 f->type = v->var->type;
5068 f->offset = t->size;
5069 v->var->frame_pos = f->offset;
5070 t->size += ((f->type->size - 1) | (a-1)) + 1;
5073 variable_unlink_exec(v->var);
5075 free_binode(results);
5079 static struct variable *declare_function(struct parse_context *c,
5080 struct variable *name,
5081 struct binode *args,
5083 struct binode *results,
5087 struct value fn = {.function = code};
5089 var_block_close(c, CloseFunction, code);
5090 t = add_anon_type(c, &function_prototype,
5091 "func %.*s", name->name->name.len,
5092 name->name->name.txt);
5094 t->function.params = reorder_bilist(args);
5096 ret = handle_results(c, reorder_bilist(results));
5097 t->function.inline_result = 1;
5098 t->function.local_size = ret->size;
5100 t->function.return_type = ret;
5101 global_alloc(c, t, name, &fn);
5102 name->type->function.scope = c->out_scope;
5107 var_block_close(c, CloseFunction, NULL);
5109 c->out_scope = NULL;
5113 ###### declare terminals
5116 ###### top level grammar
5119 DeclareFunction -> func FuncName ( OpenScope ArgsLine ) Block Newlines ${
5120 $0 = declare_function(c, $<FN, $<Ar, Tnone, NULL, $<Bl);
5122 | func FuncName IN OpenScope Args OUT OptNL do Block Newlines ${
5123 $0 = declare_function(c, $<FN, $<Ar, Tnone, NULL, $<Bl);
5125 | func FuncName NEWLINE OpenScope OptNL do Block Newlines ${
5126 $0 = declare_function(c, $<FN, NULL, Tnone, NULL, $<Bl);
5128 | func FuncName ( OpenScope ArgsLine ) : Type Block Newlines ${
5129 $0 = declare_function(c, $<FN, $<Ar, $<Ty, NULL, $<Bl);
5131 | func FuncName ( OpenScope ArgsLine ) : ( ArgsLine ) Block Newlines ${
5132 $0 = declare_function(c, $<FN, $<AL, NULL, $<AL2, $<Bl);
5134 | func FuncName IN OpenScope Args OUT OptNL return Type Newlines do Block Newlines ${
5135 $0 = declare_function(c, $<FN, $<Ar, $<Ty, NULL, $<Bl);
5137 | func FuncName NEWLINE OpenScope return Type Newlines do Block Newlines ${
5138 $0 = declare_function(c, $<FN, NULL, $<Ty, NULL, $<Bl);
5140 | func FuncName IN OpenScope Args OUT OptNL return IN Args OUT OptNL do Block Newlines ${
5141 $0 = declare_function(c, $<FN, $<Ar, NULL, $<Ar2, $<Bl);
5143 | func FuncName NEWLINE OpenScope return IN Args OUT OptNL do Block Newlines ${
5144 $0 = declare_function(c, $<FN, NULL, NULL, $<Ar, $<Bl);
5147 ###### print func decls
5152 while (target != 0) {
5154 for (v = context.in_scope; v; v=v->in_scope)
5155 if (v->depth == 0 && v->type && v->type->check_args) {
5164 struct value *val = var_value(&context, v);
5165 printf("func %.*s", v->name->name.len, v->name->name.txt);
5166 v->type->print_type_decl(v->type, stdout);
5168 print_exec(val->function, 0, brackets);
5170 print_value(v->type, val, stdout);
5171 printf("/* frame size %d */\n", v->type->function.local_size);
5177 ###### core functions
5179 static int analyse_funcs(struct parse_context *c)
5183 for (v = c->in_scope; v; v = v->in_scope) {
5187 if (v->depth != 0 || !v->type || !v->type->check_args)
5189 ret = v->type->function.inline_result ?
5190 Tnone : v->type->function.return_type;
5191 val = var_value(c, v);
5194 propagate_types(val->function, c, &perr, ret, 0);
5195 } while (!(perr & Efail) && (perr & Eretry));
5196 if (!(perr & Efail))
5197 /* Make sure everything is still consistent */
5198 propagate_types(val->function, c, &perr, ret, 0);
5201 if (!v->type->function.inline_result &&
5202 !v->type->function.return_type->dup) {
5203 type_err(c, "error: function cannot return value of type %1",
5204 v->where_decl, v->type->function.return_type, 0, NULL);
5207 scope_finalize(c, v->type);
5212 static int analyse_main(struct type *type, struct parse_context *c)
5214 struct binode *bp = type->function.params;
5218 struct type *argv_type;
5220 argv_type = add_anon_type(c, &array_prototype, "argv");
5221 argv_type->array.member = Tstr;
5222 argv_type->array.unspec = 1;
5224 for (b = bp; b; b = cast(binode, b->right)) {
5228 propagate_types(b->left, c, &perr, argv_type, 0);
5230 default: /* invalid */ // NOTEST
5231 propagate_types(b->left, c, &perr, Tnone, 0); // NOTEST
5234 c->parse_error += 1;
5237 return !c->parse_error;
5240 static void interp_main(struct parse_context *c, int argc, char **argv)
5242 struct value *progp = NULL;
5243 struct text main_name = { "main", 4 };
5244 struct variable *mainv;
5250 mainv = var_ref(c, main_name);
5252 progp = var_value(c, mainv);
5253 if (!progp || !progp->function) {
5254 fprintf(stderr, "oceani: no main function found.\n");
5255 c->parse_error += 1;
5258 if (!analyse_main(mainv->type, c)) {
5259 fprintf(stderr, "oceani: main has wrong type.\n");
5260 c->parse_error += 1;
5263 al = mainv->type->function.params;
5265 c->local_size = mainv->type->function.local_size;
5266 c->local = calloc(1, c->local_size);
5268 struct var *v = cast(var, al->left);
5269 struct value *vl = var_value(c, v->var);
5279 mpq_set_ui(argcq, argc, 1);
5280 memcpy(var_value(c, t->array.vsize), &argcq, sizeof(argcq));
5281 t->prepare_type(c, t, 0);
5282 array_init(v->var->type, vl);
5283 for (i = 0; i < argc; i++) {
5284 struct value *vl2 = vl->array + i * v->var->type->array.member->size;
5286 arg.str.txt = argv[i];
5287 arg.str.len = strlen(argv[i]);
5288 free_value(Tstr, vl2);
5289 dup_value(Tstr, &arg, vl2);
5293 al = cast(binode, al->right);
5295 v = interp_exec(c, progp->function, &vtype);
5296 free_value(vtype, &v);
5301 ###### ast functions
5302 void free_variable(struct variable *v)
5306 ## And now to test it out.
5308 Having a language requires having a "hello world" program. I'll
5309 provide a little more than that: a program that prints "Hello world"
5310 finds the GCD of two numbers, prints the first few elements of
5311 Fibonacci, performs a binary search for a number, and a few other
5312 things which will likely grow as the languages grows.
5314 ###### File: oceani.mk
5317 @echo "===== DEMO ====="
5318 ./oceani --section "demo: hello" oceani.mdc 55 33
5324 four ::= 2 + 2 ; five ::= 10/2
5325 const pie ::= "I like Pie";
5326 cake ::= "The cake is"
5334 func main(argv:[argc::]string)
5335 print "Hello World, what lovely oceans you have!"
5336 print "Are there", five, "?"
5337 print pi, pie, "but", cake
5339 A := $argv[1]; B := $argv[2]
5341 /* When a variable is defined in both branches of an 'if',
5342 * and used afterwards, the variables are merged.
5348 print "Is", A, "bigger than", B,"? ", bigger
5349 /* If a variable is not used after the 'if', no
5350 * merge happens, so types can be different
5353 double:string = "yes"
5354 print A, "is more than twice", B, "?", double
5357 print "double", B, "is", double
5362 if a > 0 and then b > 0:
5368 print "GCD of", A, "and", B,"is", a
5370 print a, "is not positive, cannot calculate GCD"
5372 print b, "is not positive, cannot calculate GCD"
5377 print "Fibonacci:", f1,f2,
5378 then togo = togo - 1
5386 /* Binary search... */
5391 mid := (lo + hi) / 2
5404 print "Yay, I found", target
5406 print "Closest I found was", lo
5411 // "middle square" PRNG. Not particularly good, but one my
5412 // Dad taught me - the first one I ever heard of.
5413 for i:=1; then i = i + 1; while i < size:
5414 n := list[i-1] * list[i-1]
5415 list[i] = (n / 100) % 10 000
5417 print "Before sort:",
5418 for i:=0; then i = i + 1; while i < size:
5422 for i := 1; then i=i+1; while i < size:
5423 for j:=i-1; then j=j-1; while j >= 0:
5424 if list[j] > list[j+1]:
5428 print " After sort:",
5429 for i:=0; then i = i + 1; while i < size:
5433 if 1 == 2 then print "yes"; else print "no"
5437 bob.alive = (bob.name == "Hello")
5438 print "bob", "is" if bob.alive else "isn't", "alive"