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 prepare_types(&context);
246 if (!context.parse_error && !analyse_funcs(&context)) {
247 fprintf(stderr, "oceani: type error in program - not running.\n");
248 context.parse_error = 1;
256 if (doexec && !context.parse_error)
257 interp_main(&context, argc - optind, argv + optind);
260 struct section *t = s->next;
265 // FIXME parser should pop scope even on error
266 while (context.scope_depth > 0)
269 ## free context types
270 ## free context storage
271 exit(context.parse_error ? 1 : 0);
276 The four requirements of parse, analyse, print, interpret apply to
277 each language element individually so that is how most of the code
280 Three of the four are fairly self explanatory. The one that requires
281 a little explanation is the analysis step.
283 The current language design does not require the types of variables to
284 be declared, but they must still have a single type. Different
285 operations impose different requirements on the variables, for example
286 addition requires both arguments to be numeric, and assignment
287 requires the variable on the left to have the same type as the
288 expression on the right.
290 Analysis involves propagating these type requirements around and
291 consequently setting the type of each variable. If any requirements
292 are violated (e.g. a string is compared with a number) or if a
293 variable needs to have two different types, then an error is raised
294 and the program will not run.
296 If the same variable is declared in both branchs of an 'if/else', or
297 in all cases of a 'switch' then the multiple instances may be merged
298 into just one variable if the variable is referenced after the
299 conditional statement. When this happens, the types must naturally be
300 consistent across all the branches. When the variable is not used
301 outside the if, the variables in the different branches are distinct
302 and can be of different types.
304 Undeclared names may only appear in "use" statements and "case" expressions.
305 These names are given a type of "label" and a unique value.
306 This allows them to fill the role of a name in an enumerated type, which
307 is useful for testing the `switch` statement.
309 As we will see, the condition part of a `while` statement can return
310 either a Boolean or some other type. This requires that the expected
311 type that gets passed around comprises a type and a flag to indicate
312 that `Tbool` is also permitted.
314 As there are, as yet, no distinct types that are compatible, there
315 isn't much subtlety in the analysis. When we have distinct number
316 types, this will become more interesting.
320 When analysis discovers an inconsistency it needs to report an error;
321 just refusing to run the code ensures that the error doesn't cascade,
322 but by itself it isn't very useful. A clear understanding of the sort
323 of error message that are useful will help guide the process of
326 At a simplistic level, the only sort of error that type analysis can
327 report is that the type of some construct doesn't match a contextual
328 requirement. For example, in `4 + "hello"` the addition provides a
329 contextual requirement for numbers, but `"hello"` is not a number. In
330 this particular example no further information is needed as the types
331 are obvious from local information. When a variable is involved that
332 isn't the case. It may be helpful to explain why the variable has a
333 particular type, by indicating the location where the type was set,
334 whether by declaration or usage.
336 Using a recursive-descent analysis we can easily detect a problem at
337 multiple locations. In "`hello:= "there"; 4 + hello`" the addition
338 will detect that one argument is not a number and the usage of `hello`
339 will detect that a number was wanted, but not provided. In this
340 (early) version of the language, we will generate error reports at
341 multiple locations, so the use of `hello` will report an error and
342 explain were the value was set, and the addition will report an error
343 and say why numbers are needed. To be able to report locations for
344 errors, each language element will need to record a file location
345 (line and column) and each variable will need to record the language
346 element where its type was set. For now we will assume that each line
347 of an error message indicates one location in the file, and up to 2
348 types. So we provide a `printf`-like function which takes a format, a
349 location (a `struct exec` which has not yet been introduced), and 2
350 types. "`%1`" reports the first type, "`%2`" reports the second. We
351 will need a function to print the location, once we know how that is
352 stored. e As will be explained later, there are sometimes extra rules for
353 type matching and they might affect error messages, we need to pass those
356 As well as type errors, we sometimes need to report problems with
357 tokens, which might be unexpected or might name a type that has not
358 been defined. For these we have `tok_err()` which reports an error
359 with a given token. Each of the error functions sets the flag in the
360 context so indicate that parsing failed.
364 static void fput_loc(struct exec *loc, FILE *f);
365 static void type_err(struct parse_context *c,
366 char *fmt, struct exec *loc,
367 struct type *t1, int rules, struct type *t2);
369 ###### core functions
371 static void type_err(struct parse_context *c,
372 char *fmt, struct exec *loc,
373 struct type *t1, int rules, struct type *t2)
375 fprintf(stderr, "%s:", c->file_name);
376 fput_loc(loc, stderr);
377 for (; *fmt ; fmt++) {
384 case '%': fputc(*fmt, stderr); break; // NOTEST
385 default: fputc('?', stderr); break; // NOTEST
387 type_print(t1, stderr);
390 type_print(t2, stderr);
399 static void tok_err(struct parse_context *c, char *fmt, struct token *t)
401 fprintf(stderr, "%s:%d:%d: %s: %.*s\n", c->file_name, t->line, t->col, fmt,
402 t->txt.len, t->txt.txt);
406 ## Entities: declared and predeclared.
408 There are various "things" that the language and/or the interpreter
409 needs to know about to parse and execute a program. These include
410 types, variables, values, and executable code. These are all lumped
411 together under the term "entities" (calling them "objects" would be
412 confusing) and introduced here. The following section will present the
413 different specific code elements which comprise or manipulate these
418 Executables can be lots of different things. In many cases an
419 executable is just an operation combined with one or two other
420 executables. This allows for expressions and lists etc. Other times an
421 executable is something quite specific like a constant or variable name.
422 So we define a `struct exec` to be a general executable with a type, and
423 a `struct binode` which is a subclass of `exec`, forms a node in a
424 binary tree, and holds an operation. There will be other subclasses,
425 and to access these we need to be able to `cast` the `exec` into the
426 various other types. The first field in any `struct exec` is the type
427 from the `exec_types` enum.
430 #define cast(structname, pointer) ({ \
431 const typeof( ((struct structname *)0)->type) *__mptr = &(pointer)->type; \
432 if (__mptr && *__mptr != X##structname) abort(); \
433 (struct structname *)( (char *)__mptr);})
435 #define new(structname) ({ \
436 struct structname *__ptr = ((struct structname *)calloc(1,sizeof(struct structname))); \
437 __ptr->type = X##structname; \
438 __ptr->line = -1; __ptr->column = -1; \
441 #define new_pos(structname, token) ({ \
442 struct structname *__ptr = ((struct structname *)calloc(1,sizeof(struct structname))); \
443 __ptr->type = X##structname; \
444 __ptr->line = token.line; __ptr->column = token.col; \
453 enum exec_types type;
462 struct exec *left, *right;
467 static int __fput_loc(struct exec *loc, FILE *f)
471 if (loc->line >= 0) {
472 fprintf(f, "%d:%d: ", loc->line, loc->column);
475 if (loc->type == Xbinode)
476 return __fput_loc(cast(binode,loc)->left, f) ||
477 __fput_loc(cast(binode,loc)->right, f); // NOTEST
480 static void fput_loc(struct exec *loc, FILE *f)
482 if (!__fput_loc(loc, f))
483 fprintf(f, "??:??: ");
486 Each different type of `exec` node needs a number of functions defined,
487 a bit like methods. We must be able to free it, print it, analyse it
488 and execute it. Once we have specific `exec` types we will need to
489 parse them too. Let's take this a bit more slowly.
493 The parser generator requires a `free_foo` function for each struct
494 that stores attributes and they will often be `exec`s and subtypes
495 there-of. So we need `free_exec` which can handle all the subtypes,
496 and we need `free_binode`.
500 static void free_binode(struct binode *b)
509 ###### core functions
510 static void free_exec(struct exec *e)
521 static void free_exec(struct exec *e);
523 ###### free exec cases
524 case Xbinode: free_binode(cast(binode, e)); break;
528 Printing an `exec` requires that we know the current indent level for
529 printing line-oriented components. As will become clear later, we
530 also want to know what sort of bracketing to use.
534 static void do_indent(int i, char *str)
541 ###### core functions
542 static void print_binode(struct binode *b, int indent, int bracket)
546 ## print binode cases
550 static void print_exec(struct exec *e, int indent, int bracket)
556 print_binode(cast(binode, e), indent, bracket); break;
561 do_indent(indent, "/* FREE");
562 for (v = e->to_free; v; v = v->next_free) {
563 printf(" %.*s", v->name->name.len, v->name->name.txt);
564 printf("[%d,%d]", v->scope_start, v->scope_end);
565 if (v->frame_pos >= 0)
566 printf("(%d+%d)", v->frame_pos,
567 v->type ? v->type->size:0);
575 static void print_exec(struct exec *e, int indent, int bracket);
579 As discussed, analysis involves propagating type requirements around the
580 program and looking for errors.
582 So `propagate_types` is passed an expected type (being a `struct type`
583 pointer together with some `val_rules` flags) that the `exec` is
584 expected to return, and returns the type that it does return, either
585 of which can be `NULL` signifying "unknown". An `ok` flag is passed
586 by reference. It is set to `0` when an error is found, and `2` when
587 any change is made. If it remains unchanged at `1`, then no more
588 propagation is needed.
592 enum val_rules {Rnolabel = 1<<0, Rboolok = 1<<1, Rnoconstant = 1<<2};
596 if (rules & Rnolabel)
597 fputs(" (labels not permitted)", stderr);
601 static struct type *propagate_types(struct exec *prog, struct parse_context *c, int *ok,
602 struct type *type, int rules);
603 ###### core functions
605 static struct type *__propagate_types(struct exec *prog, struct parse_context *c, int *ok,
606 struct type *type, int rules)
613 switch (prog->type) {
616 struct binode *b = cast(binode, prog);
618 ## propagate binode cases
622 ## propagate exec cases
627 static struct type *propagate_types(struct exec *prog, struct parse_context *c, int *ok,
628 struct type *type, int rules)
630 struct type *ret = __propagate_types(prog, c, ok, type, rules);
639 Interpreting an `exec` doesn't require anything but the `exec`. State
640 is stored in variables and each variable will be directly linked from
641 within the `exec` tree. The exception to this is the `main` function
642 which needs to look at command line arguments. This function will be
643 interpreted separately.
645 Each `exec` can return a value combined with a type in `struct lrval`.
646 The type may be `Tnone` but must be non-NULL. Some `exec`s will return
647 the location of a value, which can be updated, in `lval`. Others will
648 set `lval` to NULL indicating that there is a value of appropriate type
652 static struct value interp_exec(struct parse_context *c, struct exec *e,
653 struct type **typeret);
654 ###### core functions
658 struct value rval, *lval;
661 /* If dest is passed, dtype must give the expected type, and
662 * result can go there, in which case type is returned as NULL.
664 static struct lrval _interp_exec(struct parse_context *c, struct exec *e,
665 struct value *dest, struct type *dtype);
667 static struct value interp_exec(struct parse_context *c, struct exec *e,
668 struct type **typeret)
670 struct lrval ret = _interp_exec(c, e, NULL, NULL);
672 if (!ret.type) abort();
676 dup_value(ret.type, ret.lval, &ret.rval);
680 static struct value *linterp_exec(struct parse_context *c, struct exec *e,
681 struct type **typeret)
683 struct lrval ret = _interp_exec(c, e, NULL, NULL);
685 if (!ret.type) abort();
689 free_value(ret.type, &ret.rval);
693 /* dinterp_exec is used when the destination type is certain and
694 * the value has a place to go.
696 static void dinterp_exec(struct parse_context *c, struct exec *e,
697 struct value *dest, struct type *dtype,
700 struct lrval ret = _interp_exec(c, e, dest, dtype);
704 free_value(dtype, dest);
706 dup_value(dtype, ret.lval, dest);
708 memcpy(dest, &ret.rval, dtype->size);
711 static struct lrval _interp_exec(struct parse_context *c, struct exec *e,
712 struct value *dest, struct type *dtype)
714 /* If the result is copied to dest, ret.type is set to NULL */
716 struct value rv = {}, *lrv = NULL;
719 rvtype = ret.type = Tnone;
729 struct binode *b = cast(binode, e);
730 struct value left, right, *lleft;
731 struct type *ltype, *rtype;
732 ltype = rtype = Tnone;
734 ## interp binode cases
736 free_value(ltype, &left);
737 free_value(rtype, &right);
747 ## interp exec cleanup
753 Values come in a wide range of types, with more likely to be added.
754 Each type needs to be able to print its own values (for convenience at
755 least) as well as to compare two values, at least for equality and
756 possibly for order. For now, values might need to be duplicated and
757 freed, though eventually such manipulations will be better integrated
760 Rather than requiring every numeric type to support all numeric
761 operations (add, multiply, etc), we allow types to be able to present
762 as one of a few standard types: integer, float, and fraction. The
763 existence of these conversion functions eventually enable types to
764 determine if they are compatible with other types, though such types
765 have not yet been implemented.
767 Named type are stored in a simple linked list. Objects of each type are
768 "values" which are often passed around by value.
770 There are both explicitly named types, and anonymous types. Anonymous
771 cannot be accessed by name, but are used internally and have a name
772 which might be reported in error messages.
779 ## value union fields
788 void (*init)(struct type *type, struct value *val);
789 void (*prepare_type)(struct parse_context *c, struct type *type, int parse_time);
790 void (*print)(struct type *type, struct value *val, FILE *f);
791 void (*print_type)(struct type *type, FILE *f);
792 int (*cmp_order)(struct type *t1, struct type *t2,
793 struct value *v1, struct value *v2);
794 int (*cmp_eq)(struct type *t1, struct type *t2,
795 struct value *v1, struct value *v2);
796 void (*dup)(struct type *type, struct value *vold, struct value *vnew);
797 void (*free)(struct type *type, struct value *val);
798 void (*free_type)(struct type *t);
799 long long (*to_int)(struct value *v);
800 double (*to_float)(struct value *v);
801 int (*to_mpq)(mpq_t *q, struct value *v);
810 struct type *typelist;
817 static struct type *find_type(struct parse_context *c, struct text s)
819 struct type *t = c->typelist;
821 while (t && (t->anon ||
822 text_cmp(t->name, s) != 0))
827 static struct type *_add_type(struct parse_context *c, struct text s,
828 struct type *proto, int anon)
832 n = calloc(1, sizeof(*n));
836 n->next = c->typelist;
841 static struct type *add_type(struct parse_context *c, struct text s,
844 return _add_type(c, s, proto, 0);
847 static struct type *add_anon_type(struct parse_context *c,
848 struct type *proto, char *name, ...)
854 vasprintf(&t.txt, name, ap);
856 t.len = strlen(name);
857 return _add_type(c, t, proto, 1);
860 static void free_type(struct type *t)
862 /* The type is always a reference to something in the
863 * context, so we don't need to free anything.
867 static void free_value(struct type *type, struct value *v)
871 memset(v, 0x5a, type->size);
875 static void type_print(struct type *type, FILE *f)
878 fputs("*unknown*type*", f); // NOTEST
879 else if (type->name.len && !type->anon)
880 fprintf(f, "%.*s", type->name.len, type->name.txt);
881 else if (type->print_type)
882 type->print_type(type, f);
884 fputs("*invalid*type*", f);
887 static void val_init(struct type *type, struct value *val)
889 if (type && type->init)
890 type->init(type, val);
893 static void dup_value(struct type *type,
894 struct value *vold, struct value *vnew)
896 if (type && type->dup)
897 type->dup(type, vold, vnew);
900 static int value_cmp(struct type *tl, struct type *tr,
901 struct value *left, struct value *right)
903 if (tl && tl->cmp_order)
904 return tl->cmp_order(tl, tr, left, right);
905 if (tl && tl->cmp_eq) // NOTEST
906 return tl->cmp_eq(tl, tr, left, right); // NOTEST
910 static void print_value(struct type *type, struct value *v, FILE *f)
912 if (type && type->print)
913 type->print(type, v, f);
915 fprintf(f, "*Unknown*"); // NOTEST
918 static void prepare_types(struct parse_context *c)
922 for (t = c->typelist; t; t = t->next)
924 t->prepare_type(c, t, 1);
929 static void free_value(struct type *type, struct value *v);
930 static int type_compat(struct type *require, struct type *have, int rules);
931 static void type_print(struct type *type, FILE *f);
932 static void val_init(struct type *type, struct value *v);
933 static void dup_value(struct type *type,
934 struct value *vold, struct value *vnew);
935 static int value_cmp(struct type *tl, struct type *tr,
936 struct value *left, struct value *right);
937 static void print_value(struct type *type, struct value *v, FILE *f);
939 ###### free context types
941 while (context.typelist) {
942 struct type *t = context.typelist;
944 context.typelist = t->next;
952 Type can be specified for local variables, for fields in a structure,
953 for formal parameters to functions, and possibly elsewhere. Different
954 rules may apply in different contexts. As a minimum, a named type may
955 always be used. Currently the type of a formal parameter can be
956 different from types in other contexts, so we have a separate grammar
962 Type -> IDENTIFIER ${
963 $0 = find_type(c, $1.txt);
966 "error: undefined type", &$1);
973 FormalType -> Type ${ $0 = $<1; }$
974 ## formal type grammar
978 Values of the base types can be numbers, which we represent as
979 multi-precision fractions, strings, Booleans and labels. When
980 analysing the program we also need to allow for places where no value
981 is meaningful (type `Tnone`) and where we don't know what type to
982 expect yet (type is `NULL`).
984 Values are never shared, they are always copied when used, and freed
985 when no longer needed.
987 When propagating type information around the program, we need to
988 determine if two types are compatible, where type `NULL` is compatible
989 with anything. There are two special cases with type compatibility,
990 both related to the Conditional Statement which will be described
991 later. In some cases a Boolean can be accepted as well as some other
992 primary type, and in others any type is acceptable except a label (`Vlabel`).
993 A separate function encoding these cases will simplify some code later.
995 ###### type functions
997 int (*compat)(struct type *this, struct type *other);
1001 static int type_compat(struct type *require, struct type *have, int rules)
1003 if ((rules & Rboolok) && have == Tbool)
1005 if ((rules & Rnolabel) && have == Tlabel)
1007 if (!require || !have)
1010 if (require->compat)
1011 return require->compat(require, have);
1013 return require == have;
1018 #include "parse_string.h"
1019 #include "parse_number.h"
1022 myLDLIBS := libnumber.o libstring.o -lgmp
1023 LDLIBS := $(filter-out $(myLDLIBS),$(LDLIBS)) $(myLDLIBS)
1025 ###### type union fields
1026 enum vtype {Vnone, Vstr, Vnum, Vbool, Vlabel} vtype;
1028 ###### value union fields
1034 ###### ast functions
1035 static void _free_value(struct type *type, struct value *v)
1039 switch (type->vtype) {
1041 case Vstr: free(v->str.txt); break;
1042 case Vnum: mpq_clear(v->num); break;
1048 ###### value functions
1050 static void _val_init(struct type *type, struct value *val)
1052 switch(type->vtype) {
1053 case Vnone: // NOTEST
1056 mpq_init(val->num); break;
1058 val->str.txt = malloc(1);
1070 static void _dup_value(struct type *type,
1071 struct value *vold, struct value *vnew)
1073 switch (type->vtype) {
1074 case Vnone: // NOTEST
1077 vnew->label = vold->label;
1080 vnew->bool = vold->bool;
1083 mpq_init(vnew->num);
1084 mpq_set(vnew->num, vold->num);
1087 vnew->str.len = vold->str.len;
1088 vnew->str.txt = malloc(vnew->str.len);
1089 memcpy(vnew->str.txt, vold->str.txt, vnew->str.len);
1094 static int _value_cmp(struct type *tl, struct type *tr,
1095 struct value *left, struct value *right)
1099 return tl - tr; // NOTEST
1100 switch (tl->vtype) {
1101 case Vlabel: cmp = left->label == right->label ? 0 : 1; break;
1102 case Vnum: cmp = mpq_cmp(left->num, right->num); break;
1103 case Vstr: cmp = text_cmp(left->str, right->str); break;
1104 case Vbool: cmp = left->bool - right->bool; break;
1105 case Vnone: cmp = 0; // NOTEST
1110 static void _print_value(struct type *type, struct value *v, FILE *f)
1112 switch (type->vtype) {
1113 case Vnone: // NOTEST
1114 fprintf(f, "*no-value*"); break; // NOTEST
1115 case Vlabel: // NOTEST
1116 fprintf(f, "*label-%p*", v->label); break; // NOTEST
1118 fprintf(f, "%.*s", v->str.len, v->str.txt); break;
1120 fprintf(f, "%s", v->bool ? "True":"False"); break;
1125 mpf_set_q(fl, v->num);
1126 gmp_fprintf(f, "%Fg", fl);
1133 static void _free_value(struct type *type, struct value *v);
1135 static struct type base_prototype = {
1137 .print = _print_value,
1138 .cmp_order = _value_cmp,
1139 .cmp_eq = _value_cmp,
1141 .free = _free_value,
1144 static struct type *Tbool, *Tstr, *Tnum, *Tnone, *Tlabel;
1146 ###### ast functions
1147 static struct type *add_base_type(struct parse_context *c, char *n,
1148 enum vtype vt, int size)
1150 struct text txt = { n, strlen(n) };
1153 t = add_type(c, txt, &base_prototype);
1156 t->align = size > sizeof(void*) ? sizeof(void*) : size;
1157 if (t->size & (t->align - 1))
1158 t->size = (t->size | (t->align - 1)) + 1; // NOTEST
1162 ###### context initialization
1164 Tbool = add_base_type(&context, "Boolean", Vbool, sizeof(char));
1165 Tstr = add_base_type(&context, "string", Vstr, sizeof(struct text));
1166 Tnum = add_base_type(&context, "number", Vnum, sizeof(mpq_t));
1167 Tnone = add_base_type(&context, "none", Vnone, 0);
1168 Tlabel = add_base_type(&context, "label", Vlabel, sizeof(void*));
1172 We have already met values as separate objects. When manifest constants
1173 appear in the program text, that must result in an executable which has
1174 a constant value. So the `val` structure embeds a value in an
1187 ###### ast functions
1188 struct val *new_val(struct type *T, struct token tk)
1190 struct val *v = new_pos(val, tk);
1201 $0 = new_val(Tbool, $1);
1205 $0 = new_val(Tbool, $1);
1210 $0 = new_val(Tnum, $1);
1211 if (number_parse($0->val.num, tail, $1.txt) == 0)
1212 mpq_init($0->val.num); // UNTESTED
1214 tok_err(c, "error: unsupported number suffix",
1219 $0 = new_val(Tstr, $1);
1220 string_parse(&$1, '\\', &$0->val.str, tail);
1222 tok_err(c, "error: unsupported string suffix",
1227 $0 = new_val(Tstr, $1);
1228 string_parse(&$1, '\\', &$0->val.str, tail);
1230 tok_err(c, "error: unsupported string suffix",
1234 ###### print exec cases
1237 struct val *v = cast(val, e);
1238 if (v->vtype == Tstr)
1240 print_value(v->vtype, &v->val, stdout);
1241 if (v->vtype == Tstr)
1246 ###### propagate exec cases
1249 struct val *val = cast(val, prog);
1250 if (!type_compat(type, val->vtype, rules))
1251 type_err(c, "error: expected %1%r found %2",
1252 prog, type, rules, val->vtype);
1256 ###### interp exec cases
1258 rvtype = cast(val, e)->vtype;
1259 dup_value(rvtype, &cast(val, e)->val, &rv);
1262 ###### ast functions
1263 static void free_val(struct val *v)
1266 free_value(v->vtype, &v->val);
1270 ###### free exec cases
1271 case Xval: free_val(cast(val, e)); break;
1273 ###### ast functions
1274 // Move all nodes from 'b' to 'rv', reversing their order.
1275 // In 'b' 'left' is a list, and 'right' is the last node.
1276 // In 'rv', left' is the first node and 'right' is a list.
1277 static struct binode *reorder_bilist(struct binode *b)
1279 struct binode *rv = NULL;
1282 struct exec *t = b->right;
1286 b = cast(binode, b->left);
1296 Variables are scoped named values. We store the names in a linked list
1297 of "bindings" sorted in lexical order, and use sequential search and
1304 struct binding *next; // in lexical order
1308 This linked list is stored in the parse context so that "reduce"
1309 functions can find or add variables, and so the analysis phase can
1310 ensure that every variable gets a type.
1312 ###### parse context
1314 struct binding *varlist; // In lexical order
1316 ###### ast functions
1318 static struct binding *find_binding(struct parse_context *c, struct text s)
1320 struct binding **l = &c->varlist;
1325 (cmp = text_cmp((*l)->name, s)) < 0)
1329 n = calloc(1, sizeof(*n));
1336 Each name can be linked to multiple variables defined in different
1337 scopes. Each scope starts where the name is declared and continues
1338 until the end of the containing code block. Scopes of a given name
1339 cannot nest, so a declaration while a name is in-scope is an error.
1341 ###### binding fields
1342 struct variable *var;
1346 struct variable *previous;
1348 struct binding *name;
1349 struct exec *where_decl;// where name was declared
1350 struct exec *where_set; // where type was set
1354 When a scope closes, the values of the variables might need to be freed.
1355 This happens in the context of some `struct exec` and each `exec` will
1356 need to know which variables need to be freed when it completes.
1359 struct variable *to_free;
1361 ####### variable fields
1362 struct exec *cleanup_exec;
1363 struct variable *next_free;
1365 ####### interp exec cleanup
1368 for (v = e->to_free; v; v = v->next_free) {
1369 struct value *val = var_value(c, v);
1370 free_value(v->type, val);
1374 ###### ast functions
1375 static void variable_unlink_exec(struct variable *v)
1377 struct variable **vp;
1378 if (!v->cleanup_exec)
1380 for (vp = &v->cleanup_exec->to_free;
1381 *vp; vp = &(*vp)->next_free) {
1385 v->cleanup_exec = NULL;
1390 While the naming seems strange, we include local constants in the
1391 definition of variables. A name declared `var := value` can
1392 subsequently be changed, but a name declared `var ::= value` cannot -
1395 ###### variable fields
1398 Scopes in parallel branches can be partially merged. More
1399 specifically, if a given name is declared in both branches of an
1400 if/else then its scope is a candidate for merging. Similarly if
1401 every branch of an exhaustive switch (e.g. has an "else" clause)
1402 declares a given name, then the scopes from the branches are
1403 candidates for merging.
1405 Note that names declared inside a loop (which is only parallel to
1406 itself) are never visible after the loop. Similarly names defined in
1407 scopes which are not parallel, such as those started by `for` and
1408 `switch`, are never visible after the scope. Only variables defined in
1409 both `then` and `else` (including the implicit then after an `if`, and
1410 excluding `then` used with `for`) and in all `case`s and `else` of a
1411 `switch` or `while` can be visible beyond the `if`/`switch`/`while`.
1413 Labels, which are a bit like variables, follow different rules.
1414 Labels are not explicitly declared, but if an undeclared name appears
1415 in a context where a label is legal, that effectively declares the
1416 name as a label. The declaration remains in force (or in scope) at
1417 least to the end of the immediately containing block and conditionally
1418 in any larger containing block which does not declare the name in some
1419 other way. Importantly, the conditional scope extension happens even
1420 if the label is only used in one parallel branch of a conditional --
1421 when used in one branch it is treated as having been declared in all
1424 Merge candidates are tentatively visible beyond the end of the
1425 branching statement which creates them. If the name is used, the
1426 merge is affirmed and they become a single variable visible at the
1427 outer layer. If not - if it is redeclared first - the merge lapses.
1429 To track scopes we have an extra stack, implemented as a linked list,
1430 which roughly parallels the parse stack and which is used exclusively
1431 for scoping. When a new scope is opened, a new frame is pushed and
1432 the child-count of the parent frame is incremented. This child-count
1433 is used to distinguish between the first of a set of parallel scopes,
1434 in which declared variables must not be in scope, and subsequent
1435 branches, whether they may already be conditionally scoped.
1437 We need a total ordering of scopes so we can easily compare to variables
1438 to see if they are concurrently in scope. To achieve this we record a
1439 `scope_count` which is actually a count of both beginnings and endings
1440 of scopes. Then each variable has a record of the scope count where it
1441 enters scope, and where it leaves.
1443 To push a new frame *before* any code in the frame is parsed, we need a
1444 grammar reduction. This is most easily achieved with a grammar
1445 element which derives the empty string, and creates the new scope when
1446 it is recognised. This can be placed, for example, between a keyword
1447 like "if" and the code following it.
1451 struct scope *parent;
1455 ###### parse context
1458 struct scope *scope_stack;
1460 ###### variable fields
1461 int scope_start, scope_end;
1463 ###### ast functions
1464 static void scope_pop(struct parse_context *c)
1466 struct scope *s = c->scope_stack;
1468 c->scope_stack = s->parent;
1470 c->scope_depth -= 1;
1471 c->scope_count += 1;
1474 static void scope_push(struct parse_context *c)
1476 struct scope *s = calloc(1, sizeof(*s));
1478 c->scope_stack->child_count += 1;
1479 s->parent = c->scope_stack;
1481 c->scope_depth += 1;
1482 c->scope_count += 1;
1488 OpenScope -> ${ scope_push(c); }$
1490 Each variable records a scope depth and is in one of four states:
1492 - "in scope". This is the case between the declaration of the
1493 variable and the end of the containing block, and also between
1494 the usage with affirms a merge and the end of that block.
1496 The scope depth is not greater than the current parse context scope
1497 nest depth. When the block of that depth closes, the state will
1498 change. To achieve this, all "in scope" variables are linked
1499 together as a stack in nesting order.
1501 - "pending". The "in scope" block has closed, but other parallel
1502 scopes are still being processed. So far, every parallel block at
1503 the same level that has closed has declared the name.
1505 The scope depth is the depth of the last parallel block that
1506 enclosed the declaration, and that has closed.
1508 - "conditionally in scope". The "in scope" block and all parallel
1509 scopes have closed, and no further mention of the name has been seen.
1510 This state includes a secondary nest depth (`min_depth`) which records
1511 the outermost scope seen since the variable became conditionally in
1512 scope. If a use of the name is found, the variable becomes "in scope"
1513 and that secondary depth becomes the recorded scope depth. If the
1514 name is declared as a new variable, the old variable becomes "out of
1515 scope" and the recorded scope depth stays unchanged.
1517 - "out of scope". The variable is neither in scope nor conditionally
1518 in scope. It is permanently out of scope now and can be removed from
1519 the "in scope" stack. When a variable becomes out-of-scope it is
1520 moved to a separate list (`out_scope`) of variables which have fully
1521 known scope. This will be used at the end of each function to assign
1522 each variable a place in the stack frame.
1524 ###### variable fields
1525 int depth, min_depth;
1526 enum { OutScope, PendingScope, CondScope, InScope } scope;
1527 struct variable *in_scope;
1529 ###### parse context
1531 struct variable *in_scope;
1532 struct variable *out_scope;
1534 All variables with the same name are linked together using the
1535 'previous' link. Those variable that have been affirmatively merged all
1536 have a 'merged' pointer that points to one primary variable - the most
1537 recently declared instance. When merging variables, we need to also
1538 adjust the 'merged' pointer on any other variables that had previously
1539 been merged with the one that will no longer be primary.
1541 A variable that is no longer the most recent instance of a name may
1542 still have "pending" scope, if it might still be merged with most
1543 recent instance. These variables don't really belong in the
1544 "in_scope" list, but are not immediately removed when a new instance
1545 is found. Instead, they are detected and ignored when considering the
1546 list of in_scope names.
1548 The storage of the value of a variable will be described later. For now
1549 we just need to know that when a variable goes out of scope, it might
1550 need to be freed. For this we need to be able to find it, so assume that
1551 `var_value()` will provide that.
1553 ###### variable fields
1554 struct variable *merged;
1556 ###### ast functions
1558 static void variable_merge(struct variable *primary, struct variable *secondary)
1562 primary = primary->merged;
1564 for (v = primary->previous; v; v=v->previous)
1565 if (v == secondary || v == secondary->merged ||
1566 v->merged == secondary ||
1567 v->merged == secondary->merged) {
1568 v->scope = OutScope;
1569 v->merged = primary;
1570 if (v->scope_start < primary->scope_start)
1571 primary->scope_start = v->scope_start;
1572 if (v->scope_end > primary->scope_end)
1573 primary->scope_end = v->scope_end; // NOTEST
1574 variable_unlink_exec(v);
1578 ###### forward decls
1579 static struct value *var_value(struct parse_context *c, struct variable *v);
1581 ###### free global vars
1583 while (context.varlist) {
1584 struct binding *b = context.varlist;
1585 struct variable *v = b->var;
1586 context.varlist = b->next;
1589 struct variable *next = v->previous;
1592 free_value(v->type, var_value(&context, v));
1594 // This is a global constant
1595 free_exec(v->where_decl);
1602 #### Manipulating Bindings
1604 When a name is conditionally visible, a new declaration discards the old
1605 binding - the condition lapses. Similarly when we reach the end of a
1606 function (outermost non-global scope) any conditional scope must lapse.
1607 Conversely a usage of the name affirms the visibility and extends it to
1608 the end of the containing block - i.e. the block that contains both the
1609 original declaration and the latest usage. This is determined from
1610 `min_depth`. When a conditionally visible variable gets affirmed like
1611 this, it is also merged with other conditionally visible variables with
1614 When we parse a variable declaration we either report an error if the
1615 name is currently bound, or create a new variable at the current nest
1616 depth if the name is unbound or bound to a conditionally scoped or
1617 pending-scope variable. If the previous variable was conditionally
1618 scoped, it and its homonyms becomes out-of-scope.
1620 When we parse a variable reference (including non-declarative assignment
1621 "foo = bar") we report an error if the name is not bound or is bound to
1622 a pending-scope variable; update the scope if the name is bound to a
1623 conditionally scoped variable; or just proceed normally if the named
1624 variable is in scope.
1626 When we exit a scope, any variables bound at this level are either
1627 marked out of scope or pending-scoped, depending on whether the scope
1628 was sequential or parallel. Here a "parallel" scope means the "then"
1629 or "else" part of a conditional, or any "case" or "else" branch of a
1630 switch. Other scopes are "sequential".
1632 When exiting a parallel scope we check if there are any variables that
1633 were previously pending and are still visible. If there are, then
1634 they weren't redeclared in the most recent scope, so they cannot be
1635 merged and must become out-of-scope. If it is not the first of
1636 parallel scopes (based on `child_count`), we check that there was a
1637 previous binding that is still pending-scope. If there isn't, the new
1638 variable must now be out-of-scope.
1640 When exiting a sequential scope that immediately enclosed parallel
1641 scopes, we need to resolve any pending-scope variables. If there was
1642 no `else` clause, and we cannot determine that the `switch` was exhaustive,
1643 we need to mark all pending-scope variable as out-of-scope. Otherwise
1644 all pending-scope variables become conditionally scoped.
1647 enum closetype { CloseSequential, CloseFunction, CloseParallel, CloseElse };
1649 ###### ast functions
1651 static struct variable *var_decl(struct parse_context *c, struct text s)
1653 struct binding *b = find_binding(c, s);
1654 struct variable *v = b->var;
1656 switch (v ? v->scope : OutScope) {
1658 /* Caller will report the error */
1662 v && v->scope == CondScope;
1664 v->scope = OutScope;
1668 v = calloc(1, sizeof(*v));
1669 v->previous = b->var;
1673 v->min_depth = v->depth = c->scope_depth;
1675 v->in_scope = c->in_scope;
1676 v->scope_start = c->scope_count;
1682 static struct variable *var_ref(struct parse_context *c, struct text s)
1684 struct binding *b = find_binding(c, s);
1685 struct variable *v = b->var;
1686 struct variable *v2;
1688 switch (v ? v->scope : OutScope) {
1691 /* Caller will report the error */
1694 /* All CondScope variables of this name need to be merged
1695 * and become InScope
1697 v->depth = v->min_depth;
1699 for (v2 = v->previous;
1700 v2 && v2->scope == CondScope;
1702 variable_merge(v, v2);
1710 static int var_refile(struct parse_context *c, struct variable *v)
1712 /* Variable just went out of scope. Add it to the out_scope
1713 * list, sorted by ->scope_start
1715 struct variable **vp = &c->out_scope;
1716 while ((*vp) && (*vp)->scope_start < v->scope_start)
1717 vp = &(*vp)->in_scope;
1723 static void var_block_close(struct parse_context *c, enum closetype ct,
1726 /* Close off all variables that are in_scope.
1727 * Some variables in c->scope may already be not-in-scope,
1728 * such as when a PendingScope variable is hidden by a new
1729 * variable with the same name.
1730 * So we check for v->name->var != v and drop them.
1731 * If we choose to make a variable OutScope, we drop it
1734 struct variable *v, **vp, *v2;
1737 for (vp = &c->in_scope;
1738 (v = *vp) && v->min_depth > c->scope_depth;
1739 (v->scope == OutScope || v->name->var != v)
1740 ? (*vp = v->in_scope, var_refile(c, v))
1741 : ( vp = &v->in_scope, 0)) {
1742 v->min_depth = c->scope_depth;
1743 if (v->name->var != v)
1744 /* This is still in scope, but we haven't just
1748 v->min_depth = c->scope_depth;
1749 if (v->scope == InScope)
1750 v->scope_end = c->scope_count;
1751 if (v->scope == InScope && e && !v->global) {
1752 /* This variable gets cleaned up when 'e' finishes */
1753 variable_unlink_exec(v);
1754 v->cleanup_exec = e;
1755 v->next_free = e->to_free;
1760 case CloseParallel: /* handle PendingScope */
1764 if (c->scope_stack->child_count == 1)
1765 /* first among parallel branches */
1766 v->scope = PendingScope;
1767 else if (v->previous &&
1768 v->previous->scope == PendingScope)
1769 /* all previous branches used name */
1770 v->scope = PendingScope;
1771 else if (v->type == Tlabel)
1772 /* Labels remain pending even when not used */
1773 v->scope = PendingScope; // UNTESTED
1775 v->scope = OutScope;
1776 if (ct == CloseElse) {
1777 /* All Pending variables with this name
1778 * are now Conditional */
1780 v2 && v2->scope == PendingScope;
1782 v2->scope = CondScope;
1786 /* Not possible as it would require
1787 * parallel scope to be nested immediately
1788 * in a parallel scope, and that never
1792 /* Not possible as we already tested for
1799 if (v->scope == CondScope)
1800 /* Condition cannot continue past end of function */
1803 case CloseSequential:
1804 if (v->type == Tlabel)
1805 v->scope = PendingScope;
1808 v->scope = OutScope;
1811 /* There was no 'else', so we can only become
1812 * conditional if we know the cases were exhaustive,
1813 * and that doesn't mean anything yet.
1814 * So only labels become conditional..
1817 v2 && v2->scope == PendingScope;
1819 if (v2->type == Tlabel)
1820 v2->scope = CondScope;
1822 v2->scope = OutScope;
1825 case OutScope: break;
1834 The value of a variable is store separately from the variable, on an
1835 analogue of a stack frame. There are (currently) two frames that can be
1836 active. A global frame which currently only stores constants, and a
1837 stacked frame which stores local variables. Each variable knows if it
1838 is global or not, and what its index into the frame is.
1840 Values in the global frame are known immediately they are relevant, so
1841 the frame needs to be reallocated as it grows so it can store those
1842 values. The local frame doesn't get values until the interpreted phase
1843 is started, so there is no need to allocate until the size is known.
1845 We initialize the `frame_pos` to an impossible value, so that we can
1846 tell if it was set or not later.
1848 ###### variable fields
1852 ###### variable init
1855 ###### parse context
1857 short global_size, global_alloc;
1859 void *global, *local;
1861 ###### forward decls
1862 static struct value *global_alloc(struct parse_context *c, struct type *t,
1863 struct variable *v, struct value *init);
1865 ###### ast functions
1867 static struct value *var_value(struct parse_context *c, struct variable *v)
1870 if (!c->local || !v->type)
1872 if (v->frame_pos + v->type->size > c->local_size) {
1873 printf("INVALID frame_pos\n"); // NOTEST
1876 return c->local + v->frame_pos;
1878 if (c->global_size > c->global_alloc) {
1879 int old = c->global_alloc;
1880 c->global_alloc = (c->global_size | 1023) + 1024;
1881 c->global = realloc(c->global, c->global_alloc);
1882 memset(c->global + old, 0, c->global_alloc - old);
1884 return c->global + v->frame_pos;
1887 static struct value *global_alloc(struct parse_context *c, struct type *t,
1888 struct variable *v, struct value *init)
1891 struct variable scratch;
1893 if (t->prepare_type)
1894 t->prepare_type(c, t, 1); // NOTEST
1896 if (c->global_size & (t->align - 1))
1897 c->global_size = (c->global_size + t->align) & ~(t->align-1); // NOTEST
1902 v->frame_pos = c->global_size;
1904 c->global_size += v->type->size;
1905 ret = var_value(c, v);
1907 memcpy(ret, init, t->size);
1913 As global values are found -- struct field initializers, labels etc --
1914 `global_alloc()` is called to record the value in the global frame.
1916 When the program is fully parsed, each function is analysed, we need to
1917 walk the list of variables local to that function and assign them an
1918 offset in the stack frame. For this we have `scope_finalize()`.
1920 We keep the stack from dense by re-using space for between variables
1921 that are not in scope at the same time. The `out_scope` list is sorted
1922 by `scope_start` and as we process a varible, we move it to an FIFO
1923 stack. For each variable we consider, we first discard any from the
1924 stack anything that went out of scope before the new variable came in.
1925 Then we place the new variable just after the one at the top of the
1928 ###### ast functions
1930 static void scope_finalize(struct parse_context *c, struct type *ft)
1932 int size = ft->function.local_size;
1933 struct variable *next = ft->function.scope;
1934 struct variable *done = NULL;
1937 struct variable *v = next;
1938 struct type *t = v->type;
1945 if (v->frame_pos >= 0)
1947 while (done && done->scope_end < v->scope_start)
1948 done = done->in_scope;
1950 pos = done->frame_pos + done->type->size;
1952 pos = ft->function.local_size;
1953 if (pos & (t->align - 1))
1954 pos = (pos + t->align) & ~(t->align-1);
1956 if (size < pos + v->type->size)
1957 size = pos + v->type->size;
1961 c->out_scope = NULL;
1962 ft->function.local_size = size;
1965 ###### free context storage
1966 free(context.global);
1968 #### Variables as executables
1970 Just as we used a `val` to wrap a value into an `exec`, we similarly
1971 need a `var` to wrap a `variable` into an exec. While each `val`
1972 contained a copy of the value, each `var` holds a link to the variable
1973 because it really is the same variable no matter where it appears.
1974 When a variable is used, we need to remember to follow the `->merged`
1975 link to find the primary instance.
1977 When a variable is declared, it may or may not be given an explicit
1978 type. We need to record which so that we can report the parsed code
1987 struct variable *var;
1990 ###### variable fields
1998 VariableDecl -> IDENTIFIER : ${ {
1999 struct variable *v = var_decl(c, $1.txt);
2000 $0 = new_pos(var, $1);
2005 v = var_ref(c, $1.txt);
2007 type_err(c, "error: variable '%v' redeclared",
2009 type_err(c, "info: this is where '%v' was first declared",
2010 v->where_decl, NULL, 0, NULL);
2013 | IDENTIFIER :: ${ {
2014 struct variable *v = var_decl(c, $1.txt);
2015 $0 = new_pos(var, $1);
2021 v = var_ref(c, $1.txt);
2023 type_err(c, "error: variable '%v' redeclared",
2025 type_err(c, "info: this is where '%v' was first declared",
2026 v->where_decl, NULL, 0, NULL);
2029 | IDENTIFIER : Type ${ {
2030 struct variable *v = var_decl(c, $1.txt);
2031 $0 = new_pos(var, $1);
2037 v->explicit_type = 1;
2039 v = var_ref(c, $1.txt);
2041 type_err(c, "error: variable '%v' redeclared",
2043 type_err(c, "info: this is where '%v' was first declared",
2044 v->where_decl, NULL, 0, NULL);
2047 | IDENTIFIER :: Type ${ {
2048 struct variable *v = var_decl(c, $1.txt);
2049 $0 = new_pos(var, $1);
2056 v->explicit_type = 1;
2058 v = var_ref(c, $1.txt);
2060 type_err(c, "error: variable '%v' redeclared",
2062 type_err(c, "info: this is where '%v' was first declared",
2063 v->where_decl, NULL, 0, NULL);
2068 Variable -> IDENTIFIER ${ {
2069 struct variable *v = var_ref(c, $1.txt);
2070 $0 = new_pos(var, $1);
2072 /* This might be a label - allocate a var just in case */
2073 v = var_decl(c, $1.txt);
2080 cast(var, $0)->var = v;
2083 ###### print exec cases
2086 struct var *v = cast(var, e);
2088 struct binding *b = v->var->name;
2089 printf("%.*s", b->name.len, b->name.txt);
2096 if (loc && loc->type == Xvar) {
2097 struct var *v = cast(var, loc);
2099 struct binding *b = v->var->name;
2100 fprintf(stderr, "%.*s", b->name.len, b->name.txt);
2102 fputs("???", stderr); // NOTEST
2104 fputs("NOTVAR", stderr);
2107 ###### propagate exec cases
2111 struct var *var = cast(var, prog);
2112 struct variable *v = var->var;
2114 type_err(c, "%d:BUG: no variable!!", prog, NULL, 0, NULL); // NOTEST
2115 return Tnone; // NOTEST
2118 if (v->constant && (rules & Rnoconstant)) {
2119 type_err(c, "error: Cannot assign to a constant: %v",
2120 prog, NULL, 0, NULL);
2121 type_err(c, "info: name was defined as a constant here",
2122 v->where_decl, NULL, 0, NULL);
2125 if (v->type == Tnone && v->where_decl == prog)
2126 type_err(c, "error: variable used but not declared: %v",
2127 prog, NULL, 0, NULL);
2128 if (v->type == NULL) {
2129 if (type && *ok != 0) {
2131 v->where_set = prog;
2136 if (!type_compat(type, v->type, rules)) {
2137 type_err(c, "error: expected %1%r but variable '%v' is %2", prog,
2138 type, rules, v->type);
2139 type_err(c, "info: this is where '%v' was set to %1", v->where_set,
2140 v->type, rules, NULL);
2147 ###### interp exec cases
2150 struct var *var = cast(var, e);
2151 struct variable *v = var->var;
2154 lrv = var_value(c, v);
2159 ###### ast functions
2161 static void free_var(struct var *v)
2166 ###### free exec cases
2167 case Xvar: free_var(cast(var, e)); break;
2172 Now that we have the shape of the interpreter in place we can add some
2173 complex types and connected them in to the data structures and the
2174 different phases of parse, analyse, print, interpret.
2176 Being "complex" the language will naturally have syntax to access
2177 specifics of objects of these types. These will fit into the grammar as
2178 "Terms" which are the things that are combined with various operators to
2179 form "Expression". Where a Term is formed by some operation on another
2180 Term, the subordinate Term will always come first, so for example a
2181 member of an array will be expressed as the Term for the array followed
2182 by an index in square brackets. The strict rule of using postfix
2183 operations makes precedence irrelevant within terms. To provide a place
2184 to put the grammar for each terms of each type, we will start out by
2185 introducing the "Term" grammar production, with contains at least a
2186 simple "Value" (to be explained later).
2190 Term -> Value ${ $0 = $<1; }$
2191 | Variable ${ $0 = $<1; }$
2194 Thus far the complex types we have are arrays and structs.
2198 Arrays can be declared by giving a size and a type, as `[size]type' so
2199 `freq:[26]number` declares `freq` to be an array of 26 numbers. The
2200 size can be either a literal number, or a named constant. Some day an
2201 arbitrary expression will be supported.
2203 As a formal parameter to a function, the array can be declared with a
2204 new variable as the size: `name:[size::number]string`. The `size`
2205 variable is set to the size of the array and must be a constant. As
2206 `number` is the only supported type, it can be left out:
2207 `name:[size::]string`.
2209 Arrays cannot be assigned. When pointers are introduced we will also
2210 introduce array slices which can refer to part or all of an array -
2211 the assignment syntax will create a slice. For now, an array can only
2212 ever be referenced by the name it is declared with. It is likely that
2213 a "`copy`" primitive will eventually be define which can be used to
2214 make a copy of an array with controllable recursive depth.
2216 For now we have two sorts of array, those with fixed size either because
2217 it is given as a literal number or because it is a struct member (which
2218 cannot have a runtime-changing size), and those with a size that is
2219 determined at runtime - local variables with a const size. The former
2220 have their size calculated at parse time, the latter at run time.
2222 For the latter type, the `size` field of the type is the size of a
2223 pointer, and the array is reallocated every time it comes into scope.
2225 We differentiate struct fields with a const size from local variables
2226 with a const size by whether they are prepared at parse time or not.
2228 ###### type union fields
2231 int unspec; // size is unspecified - vsize must be set.
2234 struct variable *vsize;
2235 struct type *member;
2238 ###### value union fields
2239 void *array; // used if not static_size
2241 ###### value functions
2243 static void array_prepare_type(struct parse_context *c, struct type *type,
2246 struct value *vsize;
2248 if (type->array.static_size)
2250 if (type->array.unspec && parse_time)
2253 if (type->array.vsize) {
2254 vsize = var_value(c, type->array.vsize);
2258 mpz_tdiv_q(q, mpq_numref(vsize->num), mpq_denref(vsize->num));
2259 type->array.size = mpz_get_si(q);
2263 if (parse_time && type->array.member->size) {
2264 type->array.static_size = 1;
2265 type->size = type->array.size * type->array.member->size;
2266 type->align = type->array.member->align;
2270 static void array_init(struct type *type, struct value *val)
2273 void *ptr = val->ptr;
2277 if (!type->array.static_size) {
2278 val->array = calloc(type->array.size,
2279 type->array.member->size);
2282 for (i = 0; i < type->array.size; i++) {
2284 v = (void*)ptr + i * type->array.member->size;
2285 val_init(type->array.member, v);
2289 static void array_free(struct type *type, struct value *val)
2292 void *ptr = val->ptr;
2294 if (!type->array.static_size)
2296 for (i = 0; i < type->array.size; i++) {
2298 v = (void*)ptr + i * type->array.member->size;
2299 free_value(type->array.member, v);
2301 if (!type->array.static_size)
2305 static int array_compat(struct type *require, struct type *have)
2307 if (have->compat != require->compat)
2309 /* Both are arrays, so we can look at details */
2310 if (!type_compat(require->array.member, have->array.member, 0))
2312 if (have->array.unspec && require->array.unspec) {
2313 if (have->array.vsize && require->array.vsize &&
2314 have->array.vsize != require->array.vsize) // UNTESTED
2315 /* sizes might not be the same */
2316 return 0; // UNTESTED
2319 if (have->array.unspec || require->array.unspec)
2320 return 1; // UNTESTED
2321 if (require->array.vsize == NULL && have->array.vsize == NULL)
2322 return require->array.size == have->array.size;
2324 return require->array.vsize == have->array.vsize; // UNTESTED
2327 static void array_print_type(struct type *type, FILE *f)
2330 if (type->array.vsize) {
2331 struct binding *b = type->array.vsize->name;
2332 fprintf(f, "%.*s%s]", b->name.len, b->name.txt,
2333 type->array.unspec ? "::" : "");
2334 } else if (type->array.size)
2335 fprintf(f, "%d]", type->array.size);
2338 type_print(type->array.member, f);
2341 static struct type array_prototype = {
2343 .prepare_type = array_prepare_type,
2344 .print_type = array_print_type,
2345 .compat = array_compat,
2347 .size = sizeof(void*),
2348 .align = sizeof(void*),
2351 ###### declare terminals
2356 | [ NUMBER ] Type ${ {
2362 if (number_parse(num, tail, $2.txt) == 0)
2363 tok_err(c, "error: unrecognised number", &$2);
2365 tok_err(c, "error: unsupported number suffix", &$2);
2368 elements = mpz_get_ui(mpq_numref(num));
2369 if (mpz_cmp_ui(mpq_denref(num), 1) != 0) {
2370 tok_err(c, "error: array size must be an integer",
2372 } else if (mpz_cmp_ui(mpq_numref(num), 1UL << 30) >= 0)
2373 tok_err(c, "error: array size is too large",
2378 $0 = t = add_anon_type(c, &array_prototype, "array[%d]", elements );
2379 t->array.size = elements;
2380 t->array.member = $<4;
2381 t->array.vsize = NULL;
2384 | [ IDENTIFIER ] Type ${ {
2385 struct variable *v = var_ref(c, $2.txt);
2388 tok_err(c, "error: name undeclared", &$2);
2389 else if (!v->constant)
2390 tok_err(c, "error: array size must be a constant", &$2);
2392 $0 = add_anon_type(c, &array_prototype, "array[%.*s]", $2.txt.len, $2.txt.txt);
2393 $0->array.member = $<4;
2395 $0->array.vsize = v;
2400 OptType -> Type ${ $0 = $<1; }$
2403 ###### formal type grammar
2405 | [ IDENTIFIER :: OptType ] Type ${ {
2406 struct variable *v = var_decl(c, $ID.txt);
2412 $0 = add_anon_type(c, &array_prototype, "array[var]");
2413 $0->array.member = $<6;
2415 $0->array.unspec = 1;
2416 $0->array.vsize = v;
2424 | Term [ Expression ] ${ {
2425 struct binode *b = new(binode);
2432 ###### print binode cases
2434 print_exec(b->left, -1, bracket);
2436 print_exec(b->right, -1, bracket);
2440 ###### propagate binode cases
2442 /* left must be an array, right must be a number,
2443 * result is the member type of the array
2445 propagate_types(b->right, c, ok, Tnum, 0);
2446 t = propagate_types(b->left, c, ok, NULL, rules & Rnoconstant);
2447 if (!t || t->compat != array_compat) {
2448 type_err(c, "error: %1 cannot be indexed", prog, t, 0, NULL);
2451 if (!type_compat(type, t->array.member, rules)) {
2452 type_err(c, "error: have %1 but need %2", prog,
2453 t->array.member, rules, type);
2455 return t->array.member;
2459 ###### interp binode cases
2465 lleft = linterp_exec(c, b->left, <ype);
2466 right = interp_exec(c, b->right, &rtype);
2468 mpz_tdiv_q(q, mpq_numref(right.num), mpq_denref(right.num));
2472 if (ltype->array.static_size)
2475 ptr = *(void**)lleft;
2476 rvtype = ltype->array.member;
2477 if (i >= 0 && i < ltype->array.size)
2478 lrv = ptr + i * rvtype->size;
2480 val_init(ltype->array.member, &rv); // UNSAFE
2487 A `struct` is a data-type that contains one or more other data-types.
2488 It differs from an array in that each member can be of a different
2489 type, and they are accessed by name rather than by number. Thus you
2490 cannot choose an element by calculation, you need to know what you
2493 The language makes no promises about how a given structure will be
2494 stored in memory - it is free to rearrange fields to suit whatever
2495 criteria seems important.
2497 Structs are declared separately from program code - they cannot be
2498 declared in-line in a variable declaration like arrays can. A struct
2499 is given a name and this name is used to identify the type - the name
2500 is not prefixed by the word `struct` as it would be in C.
2502 Structs are only treated as the same if they have the same name.
2503 Simply having the same fields in the same order is not enough. This
2504 might change once we can create structure initializers from a list of
2507 Each component datum is identified much like a variable is declared,
2508 with a name, one or two colons, and a type. The type cannot be omitted
2509 as there is no opportunity to deduce the type from usage. An initial
2510 value can be given following an equals sign, so
2512 ##### Example: a struct type
2518 would declare a type called "complex" which has two number fields,
2519 each initialised to zero.
2521 Struct will need to be declared separately from the code that uses
2522 them, so we will need to be able to print out the declaration of a
2523 struct when reprinting the whole program. So a `print_type_decl` type
2524 function will be needed.
2526 ###### type union fields
2535 } *fields; // This is created when field_list is analysed.
2537 struct fieldlist *prev;
2540 } *field_list; // This is created during parsing
2543 ###### type functions
2544 void (*print_type_decl)(struct type *type, FILE *f);
2546 ###### value functions
2548 static void structure_init(struct type *type, struct value *val)
2552 for (i = 0; i < type->structure.nfields; i++) {
2554 v = (void*) val->ptr + type->structure.fields[i].offset;
2555 if (type->structure.fields[i].init)
2556 dup_value(type->structure.fields[i].type,
2557 type->structure.fields[i].init,
2560 val_init(type->structure.fields[i].type, v);
2564 static void structure_free(struct type *type, struct value *val)
2568 for (i = 0; i < type->structure.nfields; i++) {
2570 v = (void*)val->ptr + type->structure.fields[i].offset;
2571 free_value(type->structure.fields[i].type, v);
2575 static void free_fieldlist(struct fieldlist *f)
2579 free_fieldlist(f->prev);
2584 static void structure_free_type(struct type *t)
2587 for (i = 0; i < t->structure.nfields; i++)
2588 if (t->structure.fields[i].init) {
2589 free_value(t->structure.fields[i].type,
2590 t->structure.fields[i].init);
2592 free(t->structure.fields);
2593 free_fieldlist(t->structure.field_list);
2596 static void structure_prepare_type(struct parse_context *c,
2597 struct type *t, int parse_time)
2600 struct fieldlist *f;
2602 if (!parse_time || t->structure.fields)
2605 for (f = t->structure.field_list; f; f=f->prev) {
2609 if (f->f.type->prepare_type)
2610 f->f.type->prepare_type(c, f->f.type, 1);
2611 if (f->init == NULL)
2615 propagate_types(f->init, c, &ok, f->f.type, 0);
2618 c->parse_error = 1; // NOTEST
2621 t->structure.nfields = cnt;
2622 t->structure.fields = calloc(cnt, sizeof(struct field));
2623 f = t->structure.field_list;
2625 int a = f->f.type->align;
2627 t->structure.fields[cnt] = f->f;
2628 if (t->size & (a-1))
2629 t->size = (t->size | (a-1)) + 1;
2630 t->structure.fields[cnt].offset = t->size;
2631 t->size += ((f->f.type->size - 1) | (a-1)) + 1;
2635 if (f->init && !c->parse_error) {
2636 struct value vl = interp_exec(c, f->init, NULL);
2637 t->structure.fields[cnt].init =
2638 global_alloc(c, f->f.type, NULL, &vl);
2645 static struct type structure_prototype = {
2646 .init = structure_init,
2647 .free = structure_free,
2648 .free_type = structure_free_type,
2649 .print_type_decl = structure_print_type,
2650 .prepare_type = structure_prepare_type,
2664 ###### free exec cases
2666 free_exec(cast(fieldref, e)->left);
2670 ###### declare terminals
2675 | Term . IDENTIFIER ${ {
2676 struct fieldref *fr = new_pos(fieldref, $2);
2683 ###### print exec cases
2687 struct fieldref *f = cast(fieldref, e);
2688 print_exec(f->left, -1, bracket);
2689 printf(".%.*s", f->name.len, f->name.txt);
2693 ###### ast functions
2694 static int find_struct_index(struct type *type, struct text field)
2697 for (i = 0; i < type->structure.nfields; i++)
2698 if (text_cmp(type->structure.fields[i].name, field) == 0)
2703 ###### propagate exec cases
2707 struct fieldref *f = cast(fieldref, prog);
2708 struct type *st = propagate_types(f->left, c, ok, NULL, 0);
2711 type_err(c, "error: unknown type for field access", f->left, // UNTESTED
2713 else if (st->init != structure_init)
2714 type_err(c, "error: field reference attempted on %1, not a struct",
2715 f->left, st, 0, NULL);
2716 else if (f->index == -2) {
2717 f->index = find_struct_index(st, f->name);
2719 type_err(c, "error: cannot find requested field in %1",
2720 f->left, st, 0, NULL);
2722 if (f->index >= 0) {
2723 struct type *ft = st->structure.fields[f->index].type;
2724 if (!type_compat(type, ft, rules))
2725 type_err(c, "error: have %1 but need %2", prog,
2732 ###### interp exec cases
2735 struct fieldref *f = cast(fieldref, e);
2737 struct value *lleft = linterp_exec(c, f->left, <ype);
2738 lrv = (void*)lleft->ptr + ltype->structure.fields[f->index].offset;
2739 rvtype = ltype->structure.fields[f->index].type;
2743 ###### top level grammar
2744 DeclareStruct -> struct IDENTIFIER FieldBlock Newlines ${ {
2746 add_type(c, $2.txt, &structure_prototype);
2747 t->structure.field_list = $<FB;
2751 FieldBlock -> { IN OptNL FieldLines OUT OptNL } ${ $0 = $<FL; }$
2752 | { SimpleFieldList } ${ $0 = $<SFL; }$
2753 | IN OptNL FieldLines OUT ${ $0 = $<FL; }$
2754 | SimpleFieldList EOL ${ $0 = $<SFL; }$
2756 FieldLines -> SimpleFieldList Newlines ${ $0 = $<SFL; }$
2757 | FieldLines SimpleFieldList Newlines ${
2762 SimpleFieldList -> Field ${ $0 = $<F; }$
2763 | SimpleFieldList ; Field ${
2767 | SimpleFieldList ; ${
2770 | ERROR ${ tok_err(c, "Syntax error in struct field", &$1); }$
2772 Field -> IDENTIFIER : Type = Expression ${ {
2773 $0 = calloc(1, sizeof(struct fieldlist));
2774 $0->f.name = $ID.txt;
2775 $0->f.type = $<Type;
2779 | IDENTIFIER : Type ${
2780 $0 = calloc(1, sizeof(struct fieldlist));
2781 $0->f.name = $ID.txt;
2782 $0->f.type = $<Type;
2785 ###### forward decls
2786 static void structure_print_type(struct type *t, FILE *f);
2788 ###### value functions
2789 static void structure_print_type(struct type *t, FILE *f)
2793 fprintf(f, "struct %.*s\n", t->name.len, t->name.txt);
2795 for (i = 0; i < t->structure.nfields; i++) {
2796 struct field *fl = t->structure.fields + i;
2797 fprintf(f, " %.*s : ", fl->name.len, fl->name.txt);
2798 type_print(fl->type, f);
2799 if (fl->type->print && fl->init) {
2801 if (fl->type == Tstr)
2802 fprintf(f, "\""); // UNTESTED
2803 print_value(fl->type, fl->init, f);
2804 if (fl->type == Tstr)
2805 fprintf(f, "\""); // UNTESTED
2811 ###### print type decls
2816 while (target != 0) {
2818 for (t = context.typelist; t ; t=t->next)
2819 if (!t->anon && t->print_type_decl &&
2829 t->print_type_decl(t, stdout);
2837 A function is a chunk of code which can be passed parameters and can
2838 return results. Each function has a type which includes the set of
2839 parameters and the return value. As yet these types cannot be declared
2840 separately from the function itself.
2842 The parameters can be specified either in parentheses as a ';' separated
2845 ##### Example: function 1
2847 func main(av:[ac::number]string; env:[envc::number]string)
2850 or as an indented list of one parameter per line (though each line can
2851 be a ';' separated list)
2853 ##### Example: function 2
2856 argv:[argc::number]string
2857 env:[envc::number]string
2861 In the first case a return type can follow the parentheses after a colon,
2862 in the second it is given on a line starting with the word `return`.
2864 ##### Example: functions that return
2866 func add(a:number; b:number): number
2876 Rather than returning a type, the function can specify a set of local
2877 variables to return as a struct. The values of these variables when the
2878 function exits will be provided to the caller. For this the return type
2879 is replaced with a block of result declarations, either in parentheses
2880 or bracketed by `return` and `do`.
2882 ##### Example: functions returning multiple variables
2884 func to_cartesian(rho:number; theta:number):(x:number; y:number)
2897 For constructing the lists we use a `List` binode, which will be
2898 further detailed when Expression Lists are introduced.
2900 ###### type union fields
2903 struct binode *params;
2904 struct type *return_type;
2905 struct variable *scope;
2906 int inline_result; // return value is at start of 'local'
2910 ###### value union fields
2911 struct exec *function;
2913 ###### type functions
2914 void (*check_args)(struct parse_context *c, int *ok,
2915 struct type *require, struct exec *args);
2917 ###### value functions
2919 static void function_free(struct type *type, struct value *val)
2921 free_exec(val->function);
2922 val->function = NULL;
2925 static int function_compat(struct type *require, struct type *have)
2927 // FIXME can I do anything here yet?
2931 static void function_check_args(struct parse_context *c, int *ok,
2932 struct type *require, struct exec *args)
2934 /* This should be 'compat', but we don't have a 'tuple' type to
2935 * hold the type of 'args'
2937 struct binode *arg = cast(binode, args);
2938 struct binode *param = require->function.params;
2941 struct var *pv = cast(var, param->left);
2943 type_err(c, "error: insufficient arguments to function.",
2944 args, NULL, 0, NULL);
2948 propagate_types(arg->left, c, ok, pv->var->type, 0);
2949 param = cast(binode, param->right);
2950 arg = cast(binode, arg->right);
2953 type_err(c, "error: too many arguments to function.",
2954 args, NULL, 0, NULL);
2957 static void function_print(struct type *type, struct value *val, FILE *f)
2959 print_exec(val->function, 1, 0);
2962 static void function_print_type_decl(struct type *type, FILE *f)
2966 for (b = type->function.params; b; b = cast(binode, b->right)) {
2967 struct variable *v = cast(var, b->left)->var;
2968 fprintf(f, "%.*s%s", v->name->name.len, v->name->name.txt,
2969 v->constant ? "::" : ":");
2970 type_print(v->type, f);
2975 if (type->function.return_type != Tnone) {
2977 if (type->function.inline_result) {
2979 struct type *t = type->function.return_type;
2981 for (i = 0; i < t->structure.nfields; i++) {
2982 struct field *fl = t->structure.fields + i;
2985 fprintf(f, "%.*s:", fl->name.len, fl->name.txt);
2986 type_print(fl->type, f);
2990 type_print(type->function.return_type, f);
2995 static void function_free_type(struct type *t)
2997 free_exec(t->function.params);
3000 static struct type function_prototype = {
3001 .size = sizeof(void*),
3002 .align = sizeof(void*),
3003 .free = function_free,
3004 .compat = function_compat,
3005 .check_args = function_check_args,
3006 .print = function_print,
3007 .print_type_decl = function_print_type_decl,
3008 .free_type = function_free_type,
3011 ###### declare terminals
3021 FuncName -> IDENTIFIER ${ {
3022 struct variable *v = var_decl(c, $1.txt);
3023 struct var *e = new_pos(var, $1);
3029 v = var_ref(c, $1.txt);
3031 type_err(c, "error: function '%v' redeclared",
3033 type_err(c, "info: this is where '%v' was first declared",
3034 v->where_decl, NULL, 0, NULL);
3040 Args -> ArgsLine NEWLINE ${ $0 = $<AL; }$
3041 | Args ArgsLine NEWLINE ${ {
3042 struct binode *b = $<AL;
3043 struct binode **bp = &b;
3045 bp = (struct binode **)&(*bp)->left;
3050 ArgsLine -> ${ $0 = NULL; }$
3051 | Varlist ${ $0 = $<1; }$
3052 | Varlist ; ${ $0 = $<1; }$
3054 Varlist -> Varlist ; ArgDecl ${
3068 ArgDecl -> IDENTIFIER : FormalType ${ {
3069 struct variable *v = var_decl(c, $1.txt);
3075 ##### Function calls
3077 A function call can appear either as an expression or as a statement.
3078 We use a new 'Funcall' binode type to link the function with a list of
3079 arguments, form with the 'List' nodes.
3081 We have already seen the "Term" which is how a function call can appear
3082 in an expression. To parse a function call into a statement we include
3083 it in the "SimpleStatement Grammar" which will be described later.
3089 | Term ( ExpressionList ) ${ {
3090 struct binode *b = new(binode);
3093 b->right = reorder_bilist($<EL);
3097 struct binode *b = new(binode);
3104 ###### SimpleStatement Grammar
3106 | Term ( ExpressionList ) ${ {
3107 struct binode *b = new(binode);
3110 b->right = reorder_bilist($<EL);
3114 ###### print binode cases
3117 do_indent(indent, "");
3118 print_exec(b->left, -1, bracket);
3120 for (b = cast(binode, b->right); b; b = cast(binode, b->right)) {
3123 print_exec(b->left, -1, bracket);
3133 ###### propagate binode cases
3136 /* Every arg must match formal parameter, and result
3137 * is return type of function
3139 struct binode *args = cast(binode, b->right);
3140 struct var *v = cast(var, b->left);
3142 if (!v->var->type || v->var->type->check_args == NULL) {
3143 type_err(c, "error: attempt to call a non-function.",
3144 prog, NULL, 0, NULL);
3147 v->var->type->check_args(c, ok, v->var->type, args);
3148 return v->var->type->function.return_type;
3151 ###### interp binode cases
3154 struct var *v = cast(var, b->left);
3155 struct type *t = v->var->type;
3156 void *oldlocal = c->local;
3157 int old_size = c->local_size;
3158 void *local = calloc(1, t->function.local_size);
3159 struct value *fbody = var_value(c, v->var);
3160 struct binode *arg = cast(binode, b->right);
3161 struct binode *param = t->function.params;
3164 struct var *pv = cast(var, param->left);
3165 struct type *vtype = NULL;
3166 struct value val = interp_exec(c, arg->left, &vtype);
3168 c->local = local; c->local_size = t->function.local_size;
3169 lval = var_value(c, pv->var);
3170 c->local = oldlocal; c->local_size = old_size;
3171 memcpy(lval, &val, vtype->size);
3172 param = cast(binode, param->right);
3173 arg = cast(binode, arg->right);
3175 c->local = local; c->local_size = t->function.local_size;
3176 if (t->function.inline_result && dtype) {
3177 _interp_exec(c, fbody->function, NULL, NULL);
3178 memcpy(dest, local, dtype->size);
3179 rvtype = ret.type = NULL;
3181 rv = interp_exec(c, fbody->function, &rvtype);
3182 c->local = oldlocal; c->local_size = old_size;
3187 ## Complex executables: statements and expressions
3189 Now that we have types and values and variables and most of the basic
3190 Terms which provide access to these, we can explore the more complex
3191 code that combine all of these to get useful work done. Specifically
3192 statements and expressions.
3194 Expressions are various combinations of Terms. We will use operator
3195 precedence to ensure correct parsing. The simplest Expression is just a
3196 Term - others will follow.
3201 Expression -> Term ${ $0 = $<Term; }$
3202 ## expression grammar
3204 ### Expressions: Conditional
3206 Our first user of the `binode` will be conditional expressions, which
3207 is a bit odd as they actually have three components. That will be
3208 handled by having 2 binodes for each expression. The conditional
3209 expression is the lowest precedence operator which is why we define it
3210 first - to start the precedence list.
3212 Conditional expressions are of the form "value `if` condition `else`
3213 other_value". They associate to the right, so everything to the right
3214 of `else` is part of an else value, while only a higher-precedence to
3215 the left of `if` is the if values. Between `if` and `else` there is no
3216 room for ambiguity, so a full conditional expression is allowed in
3222 ###### declare terminals
3226 ###### expression grammar
3228 | Expression if Expression else Expression $$ifelse ${ {
3229 struct binode *b1 = new(binode);
3230 struct binode *b2 = new(binode);
3240 ###### print binode cases
3243 b2 = cast(binode, b->right);
3244 if (bracket) printf("(");
3245 print_exec(b2->left, -1, bracket);
3247 print_exec(b->left, -1, bracket);
3249 print_exec(b2->right, -1, bracket);
3250 if (bracket) printf(")");
3253 ###### propagate binode cases
3256 /* cond must be Tbool, others must match */
3257 struct binode *b2 = cast(binode, b->right);
3260 propagate_types(b->left, c, ok, Tbool, 0);
3261 t = propagate_types(b2->left, c, ok, type, Rnolabel);
3262 t2 = propagate_types(b2->right, c, ok, type ?: t, Rnolabel);
3266 ###### interp binode cases
3269 struct binode *b2 = cast(binode, b->right);
3270 left = interp_exec(c, b->left, <ype);
3272 rv = interp_exec(c, b2->left, &rvtype); // UNTESTED
3274 rv = interp_exec(c, b2->right, &rvtype);
3280 We take a brief detour, now that we have expressions, to describe lists
3281 of expressions. These will be needed for function parameters and
3282 possibly other situations. They seem generic enough to introduce here
3283 to be used elsewhere.
3285 And ExpressionList will use the `List` type of `binode`, building up at
3286 the end. And place where they are used will probably call
3287 `reorder_bilist()` to get a more normal first/next arrangement.
3289 ###### declare terminals
3292 `List` execs have no implicit semantics, so they are never propagated or
3293 interpreted. The can be printed as a comma separate list, which is how
3294 they are parsed. Note they are also used for function formal parameter
3295 lists. In that case a separate function is used to print them.
3297 ###### print binode cases
3301 print_exec(b->left, -1, bracket);
3304 b = cast(binode, b->right);
3308 ###### propagate binode cases
3309 case List: abort(); // NOTEST
3310 ###### interp binode cases
3311 case List: abort(); // NOTEST
3316 ExpressionList -> ExpressionList , Expression ${
3329 ### Expressions: Boolean
3331 The next class of expressions to use the `binode` will be Boolean
3332 expressions. "`and then`" and "`or else`" are similar to `and` and `or`
3333 have same corresponding precendence. The difference is that they don't
3334 evaluate the second expression if not necessary.
3343 ###### declare terminals
3348 ###### expression grammar
3349 | Expression or Expression ${ {
3350 struct binode *b = new(binode);
3356 | Expression or else Expression ${ {
3357 struct binode *b = new(binode);
3364 | Expression and Expression ${ {
3365 struct binode *b = new(binode);
3371 | Expression and then Expression ${ {
3372 struct binode *b = new(binode);
3379 | not Expression ${ {
3380 struct binode *b = new(binode);
3386 ###### print binode cases
3388 if (bracket) printf("(");
3389 print_exec(b->left, -1, bracket);
3391 print_exec(b->right, -1, bracket);
3392 if (bracket) printf(")");
3395 if (bracket) printf("(");
3396 print_exec(b->left, -1, bracket);
3397 printf(" and then ");
3398 print_exec(b->right, -1, bracket);
3399 if (bracket) printf(")");
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(" or else ");
3412 print_exec(b->right, -1, bracket);
3413 if (bracket) printf(")");
3416 if (bracket) printf("(");
3418 print_exec(b->right, -1, bracket);
3419 if (bracket) printf(")");
3422 ###### propagate binode cases
3428 /* both must be Tbool, result is Tbool */
3429 propagate_types(b->left, c, ok, Tbool, 0);
3430 propagate_types(b->right, c, ok, Tbool, 0);
3431 if (type && type != Tbool)
3432 type_err(c, "error: %1 operation found where %2 expected", prog,
3436 ###### interp binode cases
3438 rv = interp_exec(c, b->left, &rvtype);
3439 right = interp_exec(c, b->right, &rtype);
3440 rv.bool = rv.bool && right.bool;
3443 rv = interp_exec(c, b->left, &rvtype);
3445 rv = interp_exec(c, b->right, NULL);
3448 rv = interp_exec(c, b->left, &rvtype);
3449 right = interp_exec(c, b->right, &rtype);
3450 rv.bool = rv.bool || right.bool;
3453 rv = interp_exec(c, b->left, &rvtype);
3455 rv = interp_exec(c, b->right, NULL);
3458 rv = interp_exec(c, b->right, &rvtype);
3462 ### Expressions: Comparison
3464 Of slightly higher precedence that Boolean expressions are Comparisons.
3465 A comparison takes arguments of any comparable type, but the two types
3468 To simplify the parsing we introduce an `eop` which can record an
3469 expression operator, and the `CMPop` non-terminal will match one of them.
3476 ###### ast functions
3477 static void free_eop(struct eop *e)
3491 ###### declare terminals
3492 $LEFT < > <= >= == != CMPop
3494 ###### expression grammar
3495 | Expression CMPop Expression ${ {
3496 struct binode *b = new(binode);
3506 CMPop -> < ${ $0.op = Less; }$
3507 | > ${ $0.op = Gtr; }$
3508 | <= ${ $0.op = LessEq; }$
3509 | >= ${ $0.op = GtrEq; }$
3510 | == ${ $0.op = Eql; }$
3511 | != ${ $0.op = NEql; }$
3513 ###### print binode cases
3521 if (bracket) printf("(");
3522 print_exec(b->left, -1, bracket);
3524 case Less: printf(" < "); break;
3525 case LessEq: printf(" <= "); break;
3526 case Gtr: printf(" > "); break;
3527 case GtrEq: printf(" >= "); break;
3528 case Eql: printf(" == "); break;
3529 case NEql: printf(" != "); break;
3530 default: abort(); // NOTEST
3532 print_exec(b->right, -1, bracket);
3533 if (bracket) printf(")");
3536 ###### propagate binode cases
3543 /* Both must match but not be labels, result is Tbool */
3544 t = propagate_types(b->left, c, ok, NULL, Rnolabel);
3546 propagate_types(b->right, c, ok, t, 0);
3548 t = propagate_types(b->right, c, ok, NULL, Rnolabel); // UNTESTED
3550 t = propagate_types(b->left, c, ok, t, 0); // UNTESTED
3552 if (!type_compat(type, Tbool, 0))
3553 type_err(c, "error: Comparison returns %1 but %2 expected", prog,
3554 Tbool, rules, type);
3557 ###### interp binode cases
3566 left = interp_exec(c, b->left, <ype);
3567 right = interp_exec(c, b->right, &rtype);
3568 cmp = value_cmp(ltype, rtype, &left, &right);
3571 case Less: rv.bool = cmp < 0; break;
3572 case LessEq: rv.bool = cmp <= 0; break;
3573 case Gtr: rv.bool = cmp > 0; break;
3574 case GtrEq: rv.bool = cmp >= 0; break;
3575 case Eql: rv.bool = cmp == 0; break;
3576 case NEql: rv.bool = cmp != 0; break;
3577 default: rv.bool = 0; break; // NOTEST
3582 ### Expressions: Arithmetic etc.
3584 The remaining expressions with the highest precedence are arithmetic,
3585 string concatenation, and string conversion. String concatenation
3586 (`++`) has the same precedence as multiplication and division, but lower
3589 String conversion is a temporary feature until I get a better type
3590 system. `$` is a prefix operator which expects a string and returns
3593 `+` and `-` are both infix and prefix operations (where they are
3594 absolute value and negation). These have different operator names.
3596 We also have a 'Bracket' operator which records where parentheses were
3597 found. This makes it easy to reproduce these when printing. Possibly I
3598 should only insert brackets were needed for precedence. Putting
3599 parentheses around an expression converts it into a Term,
3609 ###### declare terminals
3615 ###### expression grammar
3616 | Expression Eop Expression ${ {
3617 struct binode *b = new(binode);
3624 | Expression Top Expression ${ {
3625 struct binode *b = new(binode);
3632 | Uop Expression ${ {
3633 struct binode *b = new(binode);
3641 | ( Expression ) ${ {
3642 struct binode *b = new_pos(binode, $1);
3651 Eop -> + ${ $0.op = Plus; }$
3652 | - ${ $0.op = Minus; }$
3654 Uop -> + ${ $0.op = Absolute; }$
3655 | - ${ $0.op = Negate; }$
3656 | $ ${ $0.op = StringConv; }$
3658 Top -> * ${ $0.op = Times; }$
3659 | / ${ $0.op = Divide; }$
3660 | % ${ $0.op = Rem; }$
3661 | ++ ${ $0.op = Concat; }$
3663 ###### print binode cases
3670 if (bracket) printf("(");
3671 print_exec(b->left, indent, bracket);
3673 case Plus: fputs(" + ", stdout); break;
3674 case Minus: fputs(" - ", stdout); break;
3675 case Times: fputs(" * ", stdout); break;
3676 case Divide: fputs(" / ", stdout); break;
3677 case Rem: fputs(" % ", stdout); break;
3678 case Concat: fputs(" ++ ", stdout); break;
3679 default: abort(); // NOTEST
3681 print_exec(b->right, indent, bracket);
3682 if (bracket) printf(")");
3687 if (bracket) printf("(");
3689 case Absolute: fputs("+", stdout); break;
3690 case Negate: fputs("-", stdout); break;
3691 case StringConv: fputs("$", stdout); break;
3692 default: abort(); // NOTEST
3694 print_exec(b->right, indent, bracket);
3695 if (bracket) printf(")");
3699 print_exec(b->right, indent, bracket);
3703 ###### propagate binode cases
3709 /* both must be numbers, result is Tnum */
3712 /* as propagate_types ignores a NULL,
3713 * unary ops fit here too */
3714 propagate_types(b->left, c, ok, Tnum, 0);
3715 propagate_types(b->right, c, ok, Tnum, 0);
3716 if (!type_compat(type, Tnum, 0))
3717 type_err(c, "error: Arithmetic returns %1 but %2 expected", prog,
3722 /* both must be Tstr, result is Tstr */
3723 propagate_types(b->left, c, ok, Tstr, 0);
3724 propagate_types(b->right, c, ok, Tstr, 0);
3725 if (!type_compat(type, Tstr, 0))
3726 type_err(c, "error: Concat returns %1 but %2 expected", prog,
3731 /* op must be string, result is number */
3732 propagate_types(b->left, c, ok, Tstr, 0);
3733 if (!type_compat(type, Tnum, 0))
3734 type_err(c, // UNTESTED
3735 "error: Can only convert string to number, not %1",
3736 prog, type, 0, NULL);
3740 return propagate_types(b->right, c, ok, type, 0);
3742 ###### interp binode cases
3745 rv = interp_exec(c, b->left, &rvtype);
3746 right = interp_exec(c, b->right, &rtype);
3747 mpq_add(rv.num, rv.num, right.num);
3750 rv = interp_exec(c, b->left, &rvtype);
3751 right = interp_exec(c, b->right, &rtype);
3752 mpq_sub(rv.num, rv.num, right.num);
3755 rv = interp_exec(c, b->left, &rvtype);
3756 right = interp_exec(c, b->right, &rtype);
3757 mpq_mul(rv.num, rv.num, right.num);
3760 rv = interp_exec(c, b->left, &rvtype);
3761 right = interp_exec(c, b->right, &rtype);
3762 mpq_div(rv.num, rv.num, right.num);
3767 left = interp_exec(c, b->left, <ype);
3768 right = interp_exec(c, b->right, &rtype);
3769 mpz_init(l); mpz_init(r); mpz_init(rem);
3770 mpz_tdiv_q(l, mpq_numref(left.num), mpq_denref(left.num));
3771 mpz_tdiv_q(r, mpq_numref(right.num), mpq_denref(right.num));
3772 mpz_tdiv_r(rem, l, r);
3773 val_init(Tnum, &rv);
3774 mpq_set_z(rv.num, rem);
3775 mpz_clear(r); mpz_clear(l); mpz_clear(rem);
3780 rv = interp_exec(c, b->right, &rvtype);
3781 mpq_neg(rv.num, rv.num);
3784 rv = interp_exec(c, b->right, &rvtype);
3785 mpq_abs(rv.num, rv.num);
3788 rv = interp_exec(c, b->right, &rvtype);
3791 left = interp_exec(c, b->left, <ype);
3792 right = interp_exec(c, b->right, &rtype);
3794 rv.str = text_join(left.str, right.str);
3797 right = interp_exec(c, b->right, &rvtype);
3801 struct text tx = right.str;
3804 if (tx.txt[0] == '-') {
3805 neg = 1; // UNTESTED
3806 tx.txt++; // UNTESTED
3807 tx.len--; // UNTESTED
3809 if (number_parse(rv.num, tail, tx) == 0)
3810 mpq_init(rv.num); // UNTESTED
3812 mpq_neg(rv.num, rv.num); // UNTESTED
3814 printf("Unsupported suffix: %.*s\n", tx.len, tx.txt); // UNTESTED
3818 ###### value functions
3820 static struct text text_join(struct text a, struct text b)
3823 rv.len = a.len + b.len;
3824 rv.txt = malloc(rv.len);
3825 memcpy(rv.txt, a.txt, a.len);
3826 memcpy(rv.txt+a.len, b.txt, b.len);
3830 ### Blocks, Statements, and Statement lists.
3832 Now that we have expressions out of the way we need to turn to
3833 statements. There are simple statements and more complex statements.
3834 Simple statements do not contain (syntactic) newlines, complex statements do.
3836 Statements often come in sequences and we have corresponding simple
3837 statement lists and complex statement lists.
3838 The former comprise only simple statements separated by semicolons.
3839 The later comprise complex statements and simple statement lists. They are
3840 separated by newlines. Thus the semicolon is only used to separate
3841 simple statements on the one line. This may be overly restrictive,
3842 but I'm not sure I ever want a complex statement to share a line with
3845 Note that a simple statement list can still use multiple lines if
3846 subsequent lines are indented, so
3848 ###### Example: wrapped simple statement list
3853 is a single simple statement list. This might allow room for
3854 confusion, so I'm not set on it yet.
3856 A simple statement list needs no extra syntax. A complex statement
3857 list has two syntactic forms. It can be enclosed in braces (much like
3858 C blocks), or it can be introduced by an indent and continue until an
3859 unindented newline (much like Python blocks). With this extra syntax
3860 it is referred to as a block.
3862 Note that a block does not have to include any newlines if it only
3863 contains simple statements. So both of:
3865 if condition: a=b; d=f
3867 if condition { a=b; print f }
3871 In either case the list is constructed from a `binode` list with
3872 `Block` as the operator. When parsing the list it is most convenient
3873 to append to the end, so a list is a list and a statement. When using
3874 the list it is more convenient to consider a list to be a statement
3875 and a list. So we need a function to re-order a list.
3876 `reorder_bilist` serves this purpose.
3878 The only stand-alone statement we introduce at this stage is `pass`
3879 which does nothing and is represented as a `NULL` pointer in a `Block`
3880 list. Other stand-alone statements will follow once the infrastructure
3883 As many statements will use binodes, we declare a binode pointer 'b' in
3884 the common header for all reductions to use.
3886 ###### Parser: reduce
3897 Block -> { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3898 | { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3899 | SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
3900 | SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
3901 | IN OptNL Statementlist OUT ${ $0 = $<Sl; }$
3903 OpenBlock -> OpenScope { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3904 | OpenScope { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3905 | OpenScope SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
3906 | OpenScope SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
3907 | IN OpenScope OptNL Statementlist OUT ${ $0 = $<Sl; }$
3909 UseBlock -> { OpenScope IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3910 | { OpenScope SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3911 | IN OpenScope OptNL Statementlist OUT ${ $0 = $<Sl; }$
3913 ColonBlock -> { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3914 | { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3915 | : SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
3916 | : SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
3917 | : IN OptNL Statementlist OUT ${ $0 = $<Sl; }$
3919 Statementlist -> ComplexStatements ${ $0 = reorder_bilist($<CS); }$
3921 ComplexStatements -> ComplexStatements ComplexStatement ${
3931 | ComplexStatement ${
3943 ComplexStatement -> SimpleStatements Newlines ${
3944 $0 = reorder_bilist($<SS);
3946 | SimpleStatements ; Newlines ${
3947 $0 = reorder_bilist($<SS);
3949 ## ComplexStatement Grammar
3952 SimpleStatements -> SimpleStatements ; SimpleStatement ${
3958 | SimpleStatement ${
3967 SimpleStatement -> pass ${ $0 = NULL; }$
3968 | ERROR ${ tok_err(c, "Syntax error in statement", &$1); }$
3969 ## SimpleStatement Grammar
3971 ###### print binode cases
3975 if (b->left == NULL) // UNTESTED
3976 printf("pass"); // UNTESTED
3978 print_exec(b->left, indent, bracket); // UNTESTED
3979 if (b->right) { // UNTESTED
3980 printf("; "); // UNTESTED
3981 print_exec(b->right, indent, bracket); // UNTESTED
3984 // block, one per line
3985 if (b->left == NULL)
3986 do_indent(indent, "pass\n");
3988 print_exec(b->left, indent, bracket);
3990 print_exec(b->right, indent, bracket);
3994 ###### propagate binode cases
3997 /* If any statement returns something other than Tnone
3998 * or Tbool then all such must return same type.
3999 * As each statement may be Tnone or something else,
4000 * we must always pass NULL (unknown) down, otherwise an incorrect
4001 * error might occur. We never return Tnone unless it is
4006 for (e = b; e; e = cast(binode, e->right)) {
4007 t = propagate_types(e->left, c, ok, NULL, rules);
4008 if ((rules & Rboolok) && (t == Tbool || t == Tnone))
4010 if (t == Tnone && e->right)
4011 /* Only the final statement *must* return a value
4019 type_err(c, "error: expected %1%r, found %2",
4020 e->left, type, rules, t);
4026 ###### interp binode cases
4028 while (rvtype == Tnone &&
4031 rv = interp_exec(c, b->left, &rvtype);
4032 b = cast(binode, b->right);
4036 ### The Print statement
4038 `print` is a simple statement that takes a comma-separated list of
4039 expressions and prints the values separated by spaces and terminated
4040 by a newline. No control of formatting is possible.
4042 `print` uses `ExpressionList` to collect the expressions and stores them
4043 on the left side of a `Print` binode unlessthere is a trailing comma
4044 when the list is stored on the `right` side and no trailing newline is
4050 ##### declare terminals
4053 ###### SimpleStatement Grammar
4055 | print ExpressionList ${
4056 $0 = b = new(binode);
4059 b->left = reorder_bilist($<EL);
4061 | print ExpressionList , ${ {
4062 $0 = b = new(binode);
4064 b->right = reorder_bilist($<EL);
4068 $0 = b = new(binode);
4074 ###### print binode cases
4077 do_indent(indent, "print");
4079 print_exec(b->right, -1, bracket);
4082 print_exec(b->left, -1, bracket);
4087 ###### propagate binode cases
4090 /* don't care but all must be consistent */
4092 b = cast(binode, b->left);
4094 b = cast(binode, b->right);
4096 propagate_types(b->left, c, ok, NULL, Rnolabel);
4097 b = cast(binode, b->right);
4101 ###### interp binode cases
4105 struct binode *b2 = cast(binode, b->left);
4107 b2 = cast(binode, b->right);
4108 for (; b2; b2 = cast(binode, b2->right)) {
4109 left = interp_exec(c, b2->left, <ype);
4110 print_value(ltype, &left, stdout);
4111 free_value(ltype, &left);
4115 if (b->right == NULL)
4121 ###### Assignment statement
4123 An assignment will assign a value to a variable, providing it hasn't
4124 been declared as a constant. The analysis phase ensures that the type
4125 will be correct so the interpreter just needs to perform the
4126 calculation. There is a form of assignment which declares a new
4127 variable as well as assigning a value. If a name is assigned before
4128 it is declared, and error will be raised as the name is created as
4129 `Tlabel` and it is illegal to assign to such names.
4135 ###### declare terminals
4138 ###### SimpleStatement Grammar
4139 | Term = Expression ${
4140 $0 = b= new(binode);
4145 | VariableDecl = Expression ${
4146 $0 = b= new(binode);
4153 if ($1->var->where_set == NULL) {
4155 "Variable declared with no type or value: %v",
4159 $0 = b = new(binode);
4166 ###### print binode cases
4169 do_indent(indent, "");
4170 print_exec(b->left, indent, bracket);
4172 print_exec(b->right, indent, bracket);
4179 struct variable *v = cast(var, b->left)->var;
4180 do_indent(indent, "");
4181 print_exec(b->left, indent, bracket);
4182 if (cast(var, b->left)->var->constant) {
4184 if (v->explicit_type) {
4185 type_print(v->type, stdout);
4190 if (v->explicit_type) {
4191 type_print(v->type, stdout);
4197 print_exec(b->right, indent, bracket);
4204 ###### propagate binode cases
4208 /* Both must match and not be labels,
4209 * Type must support 'dup',
4210 * For Assign, left must not be constant.
4213 t = propagate_types(b->left, c, ok, NULL,
4214 Rnolabel | (b->op == Assign ? Rnoconstant : 0));
4219 if (propagate_types(b->right, c, ok, t, 0) != t)
4220 if (b->left->type == Xvar)
4221 type_err(c, "info: variable '%v' was set as %1 here.",
4222 cast(var, b->left)->var->where_set, t, rules, NULL);
4224 t = propagate_types(b->right, c, ok, NULL, Rnolabel);
4226 propagate_types(b->left, c, ok, t,
4227 (b->op == Assign ? Rnoconstant : 0));
4229 if (t && t->dup == NULL && t->name.txt[0] != ' ') // HACK
4230 type_err(c, "error: cannot assign value of type %1", b, t, 0, NULL);
4235 ###### interp binode cases
4238 lleft = linterp_exec(c, b->left, <ype);
4240 dinterp_exec(c, b->right, lleft, ltype, 1);
4246 struct variable *v = cast(var, b->left)->var;
4249 val = var_value(c, v);
4250 if (v->type->prepare_type)
4251 v->type->prepare_type(c, v->type, 0);
4253 dinterp_exec(c, b->right, val, v->type, 0);
4255 val_init(v->type, val);
4259 ### The `use` statement
4261 The `use` statement is the last "simple" statement. It is needed when a
4262 statement block can return a value. This includes the body of a
4263 function which has a return type, and the "condition" code blocks in
4264 `if`, `while`, and `switch` statements.
4269 ###### declare terminals
4272 ###### SimpleStatement Grammar
4274 $0 = b = new_pos(binode, $1);
4277 if (b->right->type == Xvar) {
4278 struct var *v = cast(var, b->right);
4279 if (v->var->type == Tnone) {
4280 /* Convert this to a label */
4283 v->var->type = Tlabel;
4284 val = global_alloc(c, Tlabel, v->var, NULL);
4290 ###### print binode cases
4293 do_indent(indent, "use ");
4294 print_exec(b->right, -1, bracket);
4299 ###### propagate binode cases
4302 /* result matches value */
4303 return propagate_types(b->right, c, ok, type, 0);
4305 ###### interp binode cases
4308 rv = interp_exec(c, b->right, &rvtype);
4311 ### The Conditional Statement
4313 This is the biggy and currently the only complex statement. This
4314 subsumes `if`, `while`, `do/while`, `switch`, and some parts of `for`.
4315 It is comprised of a number of parts, all of which are optional though
4316 set combinations apply. Each part is (usually) a key word (`then` is
4317 sometimes optional) followed by either an expression or a code block,
4318 except the `casepart` which is a "key word and an expression" followed
4319 by a code block. The code-block option is valid for all parts and,
4320 where an expression is also allowed, the code block can use the `use`
4321 statement to report a value. If the code block does not report a value
4322 the effect is similar to reporting `True`.
4324 The `else` and `case` parts, as well as `then` when combined with
4325 `if`, can contain a `use` statement which will apply to some
4326 containing conditional statement. `for` parts, `do` parts and `then`
4327 parts used with `for` can never contain a `use`, except in some
4328 subordinate conditional statement.
4330 If there is a `forpart`, it is executed first, only once.
4331 If there is a `dopart`, then it is executed repeatedly providing
4332 always that the `condpart` or `cond`, if present, does not return a non-True
4333 value. `condpart` can fail to return any value if it simply executes
4334 to completion. This is treated the same as returning `True`.
4336 If there is a `thenpart` it will be executed whenever the `condpart`
4337 or `cond` returns True (or does not return any value), but this will happen
4338 *after* `dopart` (when present).
4340 If `elsepart` is present it will be executed at most once when the
4341 condition returns `False` or some value that isn't `True` and isn't
4342 matched by any `casepart`. If there are any `casepart`s, they will be
4343 executed when the condition returns a matching value.
4345 The particular sorts of values allowed in case parts has not yet been
4346 determined in the language design, so nothing is prohibited.
4348 The various blocks in this complex statement potentially provide scope
4349 for variables as described earlier. Each such block must include the
4350 "OpenScope" nonterminal before parsing the block, and must call
4351 `var_block_close()` when closing the block.
4353 The code following "`if`", "`switch`" and "`for`" does not get its own
4354 scope, but is in a scope covering the whole statement, so names
4355 declared there cannot be redeclared elsewhere. Similarly the
4356 condition following "`while`" is in a scope the covers the body
4357 ("`do`" part) of the loop, and which does not allow conditional scope
4358 extension. Code following "`then`" (both looping and non-looping),
4359 "`else`" and "`case`" each get their own local scope.
4361 The type requirements on the code block in a `whilepart` are quite
4362 unusal. It is allowed to return a value of some identifiable type, in
4363 which case the loop aborts and an appropriate `casepart` is run, or it
4364 can return a Boolean, in which case the loop either continues to the
4365 `dopart` (on `True`) or aborts and runs the `elsepart` (on `False`).
4366 This is different both from the `ifpart` code block which is expected to
4367 return a Boolean, or the `switchpart` code block which is expected to
4368 return the same type as the casepart values. The correct analysis of
4369 the type of the `whilepart` code block is the reason for the
4370 `Rboolok` flag which is passed to `propagate_types()`.
4372 The `cond_statement` cannot fit into a `binode` so a new `exec` is
4373 defined. As there are two scopes which cover multiple parts - one for
4374 the whole statement and one for "while" and "do" - and as we will use
4375 the 'struct exec' to track scopes, we actually need two new types of
4376 exec. One is a `binode` for the looping part, the rest is the
4377 `cond_statement`. The `cond_statement` will use an auxilliary `struct
4378 casepart` to track a list of case parts.
4389 struct exec *action;
4390 struct casepart *next;
4392 struct cond_statement {
4394 struct exec *forpart, *condpart, *thenpart, *elsepart;
4395 struct binode *looppart;
4396 struct casepart *casepart;
4399 ###### ast functions
4401 static void free_casepart(struct casepart *cp)
4405 free_exec(cp->value);
4406 free_exec(cp->action);
4413 static void free_cond_statement(struct cond_statement *s)
4417 free_exec(s->forpart);
4418 free_exec(s->condpart);
4419 free_exec(s->looppart);
4420 free_exec(s->thenpart);
4421 free_exec(s->elsepart);
4422 free_casepart(s->casepart);
4426 ###### free exec cases
4427 case Xcond_statement: free_cond_statement(cast(cond_statement, e)); break;
4429 ###### ComplexStatement Grammar
4430 | CondStatement ${ $0 = $<1; }$
4432 ###### declare terminals
4433 $TERM for then while do
4440 // A CondStatement must end with EOL, as does CondSuffix and
4442 // ForPart, ThenPart, SwitchPart, CasePart are non-empty and
4443 // may or may not end with EOL
4444 // WhilePart and IfPart include an appropriate Suffix
4446 // ForPart, SwitchPart, and IfPart open scopes, o we have to close
4447 // them. WhilePart opens and closes its own scope.
4448 CondStatement -> ForPart OptNL ThenPart OptNL WhilePart CondSuffix ${
4451 $0->thenpart = $<TP;
4452 $0->looppart = $<WP;
4453 var_block_close(c, CloseSequential, $0);
4455 | ForPart OptNL WhilePart CondSuffix ${
4458 $0->looppart = $<WP;
4459 var_block_close(c, CloseSequential, $0);
4461 | WhilePart CondSuffix ${
4463 $0->looppart = $<WP;
4465 | SwitchPart OptNL CasePart CondSuffix ${
4467 $0->condpart = $<SP;
4468 $CP->next = $0->casepart;
4469 $0->casepart = $<CP;
4470 var_block_close(c, CloseSequential, $0);
4472 | SwitchPart : IN OptNL CasePart CondSuffix OUT Newlines ${
4474 $0->condpart = $<SP;
4475 $CP->next = $0->casepart;
4476 $0->casepart = $<CP;
4477 var_block_close(c, CloseSequential, $0);
4479 | IfPart IfSuffix ${
4481 $0->condpart = $IP.condpart; $IP.condpart = NULL;
4482 $0->thenpart = $IP.thenpart; $IP.thenpart = NULL;
4483 // This is where we close an "if" statement
4484 var_block_close(c, CloseSequential, $0);
4487 CondSuffix -> IfSuffix ${
4490 | Newlines CasePart CondSuffix ${
4492 $CP->next = $0->casepart;
4493 $0->casepart = $<CP;
4495 | CasePart CondSuffix ${
4497 $CP->next = $0->casepart;
4498 $0->casepart = $<CP;
4501 IfSuffix -> Newlines ${ $0 = new(cond_statement); }$
4502 | Newlines ElsePart ${ $0 = $<EP; }$
4503 | ElsePart ${$0 = $<EP; }$
4505 ElsePart -> else OpenBlock Newlines ${
4506 $0 = new(cond_statement);
4507 $0->elsepart = $<OB;
4508 var_block_close(c, CloseElse, $0->elsepart);
4510 | else OpenScope CondStatement ${
4511 $0 = new(cond_statement);
4512 $0->elsepart = $<CS;
4513 var_block_close(c, CloseElse, $0->elsepart);
4517 CasePart -> case Expression OpenScope ColonBlock ${
4518 $0 = calloc(1,sizeof(struct casepart));
4521 var_block_close(c, CloseParallel, $0->action);
4525 // These scopes are closed in CondStatement
4526 ForPart -> for OpenBlock ${
4530 ThenPart -> then OpenBlock ${
4532 var_block_close(c, CloseSequential, $0);
4536 // This scope is closed in CondStatement
4537 WhilePart -> while UseBlock OptNL do OpenBlock ${
4542 var_block_close(c, CloseSequential, $0->right);
4543 var_block_close(c, CloseSequential, $0);
4545 | while OpenScope Expression OpenScope ColonBlock ${
4550 var_block_close(c, CloseSequential, $0->right);
4551 var_block_close(c, CloseSequential, $0);
4555 IfPart -> if UseBlock OptNL then OpenBlock ${
4558 var_block_close(c, CloseParallel, $0.thenpart);
4560 | if OpenScope Expression OpenScope ColonBlock ${
4563 var_block_close(c, CloseParallel, $0.thenpart);
4565 | if OpenScope Expression OpenScope OptNL then Block ${
4568 var_block_close(c, CloseParallel, $0.thenpart);
4572 // This scope is closed in CondStatement
4573 SwitchPart -> switch OpenScope Expression ${
4576 | switch UseBlock ${
4580 ###### print binode cases
4582 if (b->left && b->left->type == Xbinode &&
4583 cast(binode, b->left)->op == Block) {
4585 do_indent(indent, "while {\n");
4587 do_indent(indent, "while\n");
4588 print_exec(b->left, indent+1, bracket);
4590 do_indent(indent, "} do {\n");
4592 do_indent(indent, "do\n");
4593 print_exec(b->right, indent+1, bracket);
4595 do_indent(indent, "}\n");
4597 do_indent(indent, "while ");
4598 print_exec(b->left, 0, bracket);
4603 print_exec(b->right, indent+1, bracket);
4605 do_indent(indent, "}\n");
4609 ###### print exec cases
4611 case Xcond_statement:
4613 struct cond_statement *cs = cast(cond_statement, e);
4614 struct casepart *cp;
4616 do_indent(indent, "for");
4617 if (bracket) printf(" {\n"); else printf("\n");
4618 print_exec(cs->forpart, indent+1, bracket);
4621 do_indent(indent, "} then {\n");
4623 do_indent(indent, "then\n");
4624 print_exec(cs->thenpart, indent+1, bracket);
4626 if (bracket) do_indent(indent, "}\n");
4629 print_exec(cs->looppart, indent, bracket);
4633 do_indent(indent, "switch");
4635 do_indent(indent, "if");
4636 if (cs->condpart && cs->condpart->type == Xbinode &&
4637 cast(binode, cs->condpart)->op == Block) {
4642 print_exec(cs->condpart, indent+1, bracket);
4644 do_indent(indent, "}\n");
4646 do_indent(indent, "then\n");
4647 print_exec(cs->thenpart, indent+1, bracket);
4651 print_exec(cs->condpart, 0, bracket);
4657 print_exec(cs->thenpart, indent+1, bracket);
4659 do_indent(indent, "}\n");
4664 for (cp = cs->casepart; cp; cp = cp->next) {
4665 do_indent(indent, "case ");
4666 print_exec(cp->value, -1, 0);
4671 print_exec(cp->action, indent+1, bracket);
4673 do_indent(indent, "}\n");
4676 do_indent(indent, "else");
4681 print_exec(cs->elsepart, indent+1, bracket);
4683 do_indent(indent, "}\n");
4688 ###### propagate binode cases
4690 t = propagate_types(b->right, c, ok, Tnone, 0);
4691 if (!type_compat(Tnone, t, 0))
4692 *ok = 0; // UNTESTED
4693 return propagate_types(b->left, c, ok, type, rules);
4695 ###### propagate exec cases
4696 case Xcond_statement:
4698 // forpart and looppart->right must return Tnone
4699 // thenpart must return Tnone if there is a loopart,
4700 // otherwise it is like elsepart.
4702 // be bool if there is no casepart
4703 // match casepart->values if there is a switchpart
4704 // either be bool or match casepart->value if there
4706 // elsepart and casepart->action must match the return type
4707 // expected of this statement.
4708 struct cond_statement *cs = cast(cond_statement, prog);
4709 struct casepart *cp;
4711 t = propagate_types(cs->forpart, c, ok, Tnone, 0);
4712 if (!type_compat(Tnone, t, 0))
4713 *ok = 0; // UNTESTED
4716 t = propagate_types(cs->thenpart, c, ok, Tnone, 0);
4717 if (!type_compat(Tnone, t, 0))
4718 *ok = 0; // UNTESTED
4720 if (cs->casepart == NULL) {
4721 propagate_types(cs->condpart, c, ok, Tbool, 0);
4722 propagate_types(cs->looppart, c, ok, Tbool, 0);
4724 /* Condpart must match case values, with bool permitted */
4726 for (cp = cs->casepart;
4727 cp && !t; cp = cp->next)
4728 t = propagate_types(cp->value, c, ok, NULL, 0);
4729 if (!t && cs->condpart)
4730 t = propagate_types(cs->condpart, c, ok, NULL, Rboolok); // UNTESTED
4731 if (!t && cs->looppart)
4732 t = propagate_types(cs->looppart, c, ok, NULL, Rboolok); // UNTESTED
4733 // Now we have a type (I hope) push it down
4735 for (cp = cs->casepart; cp; cp = cp->next)
4736 propagate_types(cp->value, c, ok, t, 0);
4737 propagate_types(cs->condpart, c, ok, t, Rboolok);
4738 propagate_types(cs->looppart, c, ok, t, Rboolok);
4741 // (if)then, else, and case parts must return expected type.
4742 if (!cs->looppart && !type)
4743 type = propagate_types(cs->thenpart, c, ok, NULL, rules);
4745 type = propagate_types(cs->elsepart, c, ok, NULL, rules);
4746 for (cp = cs->casepart;
4748 cp = cp->next) // UNTESTED
4749 type = propagate_types(cp->action, c, ok, NULL, rules); // UNTESTED
4752 propagate_types(cs->thenpart, c, ok, type, rules);
4753 propagate_types(cs->elsepart, c, ok, type, rules);
4754 for (cp = cs->casepart; cp ; cp = cp->next)
4755 propagate_types(cp->action, c, ok, type, rules);
4761 ###### interp binode cases
4763 // This just performs one iterration of the loop
4764 rv = interp_exec(c, b->left, &rvtype);
4765 if (rvtype == Tnone ||
4766 (rvtype == Tbool && rv.bool != 0))
4767 // rvtype is Tnone or Tbool, doesn't need to be freed
4768 interp_exec(c, b->right, NULL);
4771 ###### interp exec cases
4772 case Xcond_statement:
4774 struct value v, cnd;
4775 struct type *vtype, *cndtype;
4776 struct casepart *cp;
4777 struct cond_statement *cs = cast(cond_statement, e);
4780 interp_exec(c, cs->forpart, NULL);
4782 while ((cnd = interp_exec(c, cs->looppart, &cndtype)),
4783 cndtype == Tnone || (cndtype == Tbool && cnd.bool != 0))
4784 interp_exec(c, cs->thenpart, NULL);
4786 cnd = interp_exec(c, cs->condpart, &cndtype);
4787 if ((cndtype == Tnone ||
4788 (cndtype == Tbool && cnd.bool != 0))) {
4789 // cnd is Tnone or Tbool, doesn't need to be freed
4790 rv = interp_exec(c, cs->thenpart, &rvtype);
4791 // skip else (and cases)
4795 for (cp = cs->casepart; cp; cp = cp->next) {
4796 v = interp_exec(c, cp->value, &vtype);
4797 if (value_cmp(cndtype, vtype, &v, &cnd) == 0) {
4798 free_value(vtype, &v);
4799 free_value(cndtype, &cnd);
4800 rv = interp_exec(c, cp->action, &rvtype);
4803 free_value(vtype, &v);
4805 free_value(cndtype, &cnd);
4807 rv = interp_exec(c, cs->elsepart, &rvtype);
4814 ### Top level structure
4816 All the language elements so far can be used in various places. Now
4817 it is time to clarify what those places are.
4819 At the top level of a file there will be a number of declarations.
4820 Many of the things that can be declared haven't been described yet,
4821 such as functions, procedures, imports, and probably more.
4822 For now there are two sorts of things that can appear at the top
4823 level. They are predefined constants, `struct` types, and the `main`
4824 function. While the syntax will allow the `main` function to appear
4825 multiple times, that will trigger an error if it is actually attempted.
4827 The various declarations do not return anything. They store the
4828 various declarations in the parse context.
4830 ###### Parser: grammar
4833 Ocean -> OptNL DeclarationList
4835 ## declare terminals
4843 DeclarationList -> Declaration
4844 | DeclarationList Declaration
4846 Declaration -> ERROR Newlines ${
4847 tok_err(c, // UNTESTED
4848 "error: unhandled parse error", &$1);
4854 ## top level grammar
4858 ### The `const` section
4860 As well as being defined in with the code that uses them, constants
4861 can be declared at the top level. These have full-file scope, so they
4862 are always `InScope`. The value of a top level constant can be given
4863 as an expression, and this is evaluated immediately rather than in the
4864 later interpretation stage. Once we add functions to the language, we
4865 will need rules concern which, if any, can be used to define a top
4868 Constants are defined in a section that starts with the reserved word
4869 `const` and then has a block with a list of assignment statements.
4870 For syntactic consistency, these must use the double-colon syntax to
4871 make it clear that they are constants. Type can also be given: if
4872 not, the type will be determined during analysis, as with other
4875 As the types constants are inserted at the head of a list, printing
4876 them in the same order that they were read is not straight forward.
4877 We take a quadratic approach here and count the number of constants
4878 (variables of depth 0), then count down from there, each time
4879 searching through for the Nth constant for decreasing N.
4881 ###### top level grammar
4885 DeclareConstant -> const { IN OptNL ConstList OUT OptNL } Newlines
4886 | const { SimpleConstList } Newlines
4887 | const IN OptNL ConstList OUT Newlines
4888 | const SimpleConstList Newlines
4890 ConstList -> ConstList SimpleConstLine
4893 SimpleConstList -> SimpleConstList ; Const
4897 SimpleConstLine -> SimpleConstList Newlines
4898 | ERROR Newlines ${ tok_err(c, "Syntax error in constant", &$1); }$
4901 CType -> Type ${ $0 = $<1; }$
4905 Const -> IDENTIFIER :: CType = Expression ${ {
4909 v = var_decl(c, $1.txt);
4911 struct var *var = new_pos(var, $1);
4912 v->where_decl = var;
4918 struct variable *vorig = var_ref(c, $1.txt);
4919 tok_err(c, "error: name already declared", &$1);
4920 type_err(c, "info: this is where '%v' was first declared",
4921 vorig->where_decl, NULL, 0, NULL);
4925 propagate_types($5, c, &ok, $3, 0);
4930 struct value res = interp_exec(c, $5, &v->type);
4931 global_alloc(c, v->type, v, &res);
4935 ###### print const decls
4940 while (target != 0) {
4942 for (v = context.in_scope; v; v=v->in_scope)
4943 if (v->depth == 0 && v->constant) {
4954 struct value *val = var_value(&context, v);
4955 printf(" %.*s :: ", v->name->name.len, v->name->name.txt);
4956 type_print(v->type, stdout);
4958 if (v->type == Tstr)
4960 print_value(v->type, val, stdout);
4961 if (v->type == Tstr)
4969 ### Function declarations
4971 The code in an Ocean program is all stored in function declarations.
4972 One of the functions must be named `main` and it must accept an array of
4973 strings as a parameter - the command line arguments.
4975 As this is the top level, several things are handled a bit differently.
4976 The function is not interpreted by `interp_exec` as that isn't passed
4977 the argument list which the program requires. Similarly type analysis
4978 is a bit more interesting at this level.
4980 ###### ast functions
4982 static struct type *handle_results(struct parse_context *c,
4983 struct binode *results)
4985 /* Create a 'struct' type from the results list, which
4986 * is a list for 'struct var'
4988 struct type *t = add_anon_type(c, &structure_prototype,
4989 " function result");
4993 for (b = results; b; b = cast(binode, b->right))
4995 t->structure.nfields = cnt;
4996 t->structure.fields = calloc(cnt, sizeof(struct field));
4998 for (b = results; b; b = cast(binode, b->right)) {
4999 struct var *v = cast(var, b->left);
5000 struct field *f = &t->structure.fields[cnt++];
5001 int a = v->var->type->align;
5002 f->name = v->var->name->name;
5003 f->type = v->var->type;
5005 f->offset = t->size;
5006 v->var->frame_pos = f->offset;
5007 t->size += ((f->type->size - 1) | (a-1)) + 1;
5010 variable_unlink_exec(v->var);
5012 free_binode(results);
5016 static struct variable *declare_function(struct parse_context *c,
5017 struct variable *name,
5018 struct binode *args,
5020 struct binode *results,
5024 struct value fn = {.function = code};
5026 var_block_close(c, CloseFunction, code);
5027 t = add_anon_type(c, &function_prototype,
5028 "func %.*s", name->name->name.len,
5029 name->name->name.txt);
5031 t->function.params = reorder_bilist(args);
5033 ret = handle_results(c, reorder_bilist(results));
5034 t->function.inline_result = 1;
5035 t->function.local_size = ret->size;
5037 t->function.return_type = ret;
5038 global_alloc(c, t, name, &fn);
5039 name->type->function.scope = c->out_scope;
5044 var_block_close(c, CloseFunction, NULL);
5046 c->out_scope = NULL;
5050 ###### declare terminals
5053 ###### top level grammar
5056 DeclareFunction -> func FuncName ( OpenScope ArgsLine ) Block Newlines ${
5057 $0 = declare_function(c, $<FN, $<Ar, Tnone, NULL, $<Bl);
5059 | func FuncName IN OpenScope Args OUT OptNL do Block Newlines ${
5060 $0 = declare_function(c, $<FN, $<Ar, Tnone, NULL, $<Bl);
5062 | func FuncName NEWLINE OpenScope OptNL do Block Newlines ${
5063 $0 = declare_function(c, $<FN, NULL, Tnone, NULL, $<Bl);
5065 | func FuncName ( OpenScope ArgsLine ) : Type Block Newlines ${
5066 $0 = declare_function(c, $<FN, $<Ar, $<Ty, NULL, $<Bl);
5068 | func FuncName ( OpenScope ArgsLine ) : ( ArgsLine ) Block Newlines ${
5069 $0 = declare_function(c, $<FN, $<AL, NULL, $<AL2, $<Bl);
5071 | func FuncName IN OpenScope Args OUT OptNL return Type Newlines do Block Newlines ${
5072 $0 = declare_function(c, $<FN, $<Ar, $<Ty, NULL, $<Bl);
5074 | func FuncName NEWLINE OpenScope return Type Newlines do Block Newlines ${
5075 $0 = declare_function(c, $<FN, NULL, $<Ty, NULL, $<Bl);
5077 | func FuncName IN OpenScope Args OUT OptNL return IN Args OUT OptNL do Block Newlines ${
5078 $0 = declare_function(c, $<FN, $<Ar, NULL, $<Ar2, $<Bl);
5080 | func FuncName NEWLINE OpenScope return IN Args OUT OptNL do Block Newlines ${
5081 $0 = declare_function(c, $<FN, NULL, NULL, $<Ar, $<Bl);
5084 ###### print func decls
5089 while (target != 0) {
5091 for (v = context.in_scope; v; v=v->in_scope)
5092 if (v->depth == 0 && v->type && v->type->check_args) {
5101 struct value *val = var_value(&context, v);
5102 printf("func %.*s", v->name->name.len, v->name->name.txt);
5103 v->type->print_type_decl(v->type, stdout);
5105 print_exec(val->function, 0, brackets);
5107 print_value(v->type, val, stdout);
5108 printf("/* frame size %d */\n", v->type->function.local_size);
5114 ###### core functions
5116 static int analyse_funcs(struct parse_context *c)
5120 for (v = c->in_scope; v; v = v->in_scope) {
5124 if (v->depth != 0 || !v->type || !v->type->check_args)
5126 ret = v->type->function.inline_result ?
5127 Tnone : v->type->function.return_type;
5128 val = var_value(c, v);
5131 propagate_types(val->function, c, &ok, ret, 0);
5134 /* Make sure everything is still consistent */
5135 propagate_types(val->function, c, &ok, ret, 0);
5138 if (!v->type->function.inline_result &&
5139 !v->type->function.return_type->dup) {
5140 type_err(c, "error: function cannot return value of type %1",
5141 v->where_decl, v->type->function.return_type, 0, NULL);
5144 scope_finalize(c, v->type);
5149 static int analyse_main(struct type *type, struct parse_context *c)
5151 struct binode *bp = type->function.params;
5155 struct type *argv_type;
5157 argv_type = add_anon_type(c, &array_prototype, "argv");
5158 argv_type->array.member = Tstr;
5159 argv_type->array.unspec = 1;
5161 for (b = bp; b; b = cast(binode, b->right)) {
5165 propagate_types(b->left, c, &ok, argv_type, 0);
5167 default: /* invalid */ // NOTEST
5168 propagate_types(b->left, c, &ok, Tnone, 0); // NOTEST
5174 return !c->parse_error;
5177 static void interp_main(struct parse_context *c, int argc, char **argv)
5179 struct value *progp = NULL;
5180 struct text main_name = { "main", 4 };
5181 struct variable *mainv;
5187 mainv = var_ref(c, main_name);
5189 progp = var_value(c, mainv);
5190 if (!progp || !progp->function) {
5191 fprintf(stderr, "oceani: no main function found.\n");
5195 if (!analyse_main(mainv->type, c)) {
5196 fprintf(stderr, "oceani: main has wrong type.\n");
5200 al = mainv->type->function.params;
5202 c->local_size = mainv->type->function.local_size;
5203 c->local = calloc(1, c->local_size);
5205 struct var *v = cast(var, al->left);
5206 struct value *vl = var_value(c, v->var);
5216 mpq_set_ui(argcq, argc, 1);
5217 memcpy(var_value(c, t->array.vsize), &argcq, sizeof(argcq));
5218 t->prepare_type(c, t, 0);
5219 array_init(v->var->type, vl);
5220 for (i = 0; i < argc; i++) {
5221 struct value *vl2 = vl->array + i * v->var->type->array.member->size;
5223 arg.str.txt = argv[i];
5224 arg.str.len = strlen(argv[i]);
5225 free_value(Tstr, vl2);
5226 dup_value(Tstr, &arg, vl2);
5230 al = cast(binode, al->right);
5232 v = interp_exec(c, progp->function, &vtype);
5233 free_value(vtype, &v);
5238 ###### ast functions
5239 void free_variable(struct variable *v)
5243 ## And now to test it out.
5245 Having a language requires having a "hello world" program. I'll
5246 provide a little more than that: a program that prints "Hello world"
5247 finds the GCD of two numbers, prints the first few elements of
5248 Fibonacci, performs a binary search for a number, and a few other
5249 things which will likely grow as the languages grows.
5251 ###### File: oceani.mk
5254 @echo "===== DEMO ====="
5255 ./oceani --section "demo: hello" oceani.mdc 55 33
5261 four ::= 2 + 2 ; five ::= 10/2
5262 const pie ::= "I like Pie";
5263 cake ::= "The cake is"
5271 func main(argv:[argc::]string)
5272 print "Hello World, what lovely oceans you have!"
5273 print "Are there", five, "?"
5274 print pi, pie, "but", cake
5276 A := $argv[1]; B := $argv[2]
5278 /* When a variable is defined in both branches of an 'if',
5279 * and used afterwards, the variables are merged.
5285 print "Is", A, "bigger than", B,"? ", bigger
5286 /* If a variable is not used after the 'if', no
5287 * merge happens, so types can be different
5290 double:string = "yes"
5291 print A, "is more than twice", B, "?", double
5294 print "double", B, "is", double
5299 if a > 0 and then b > 0:
5305 print "GCD of", A, "and", B,"is", a
5307 print a, "is not positive, cannot calculate GCD"
5309 print b, "is not positive, cannot calculate GCD"
5314 print "Fibonacci:", f1,f2,
5315 then togo = togo - 1
5323 /* Binary search... */
5328 mid := (lo + hi) / 2
5341 print "Yay, I found", target
5343 print "Closest I found was", lo
5348 // "middle square" PRNG. Not particularly good, but one my
5349 // Dad taught me - the first one I ever heard of.
5350 for i:=1; then i = i + 1; while i < size:
5351 n := list[i-1] * list[i-1]
5352 list[i] = (n / 100) % 10 000
5354 print "Before sort:",
5355 for i:=0; then i = i + 1; while i < size:
5359 for i := 1; then i=i+1; while i < size:
5360 for j:=i-1; then j=j-1; while j >= 0:
5361 if list[j] > list[j+1]:
5365 print " After sort:",
5366 for i:=0; then i = i + 1; while i < size:
5370 if 1 == 2 then print "yes"; else print "no"
5374 bob.alive = (bob.name == "Hello")
5375 print "bob", "is" if bob.alive else "isn't", "alive"