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 if (!context.parse_error && !analyse_funcs(&context)) {
246 fprintf(stderr, "oceani: type error in program - not running.\n");
247 context.parse_error = 1;
255 if (doexec && !context.parse_error)
256 interp_main(&context, argc - optind, argv + optind);
259 struct section *t = s->next;
264 // FIXME parser should pop scope even on error
265 while (context.scope_depth > 0)
268 ## free context types
269 ## free context storage
270 exit(context.parse_error ? 1 : 0);
275 The four requirements of parse, analyse, print, interpret apply to
276 each language element individually so that is how most of the code
279 Three of the four are fairly self explanatory. The one that requires
280 a little explanation is the analysis step.
282 The current language design does not require the types of variables to
283 be declared, but they must still have a single type. Different
284 operations impose different requirements on the variables, for example
285 addition requires both arguments to be numeric, and assignment
286 requires the variable on the left to have the same type as the
287 expression on the right.
289 Analysis involves propagating these type requirements around and
290 consequently setting the type of each variable. If any requirements
291 are violated (e.g. a string is compared with a number) or if a
292 variable needs to have two different types, then an error is raised
293 and the program will not run.
295 If the same variable is declared in both branchs of an 'if/else', or
296 in all cases of a 'switch' then the multiple instances may be merged
297 into just one variable if the variable is referenced after the
298 conditional statement. When this happens, the types must naturally be
299 consistent across all the branches. When the variable is not used
300 outside the if, the variables in the different branches are distinct
301 and can be of different types.
303 Undeclared names may only appear in "use" statements and "case" expressions.
304 These names are given a type of "label" and a unique value.
305 This allows them to fill the role of a name in an enumerated type, which
306 is useful for testing the `switch` statement.
308 As we will see, the condition part of a `while` statement can return
309 either a Boolean or some other type. This requires that the expected
310 type that gets passed around comprises a type and a flag to indicate
311 that `Tbool` is also permitted.
313 As there are, as yet, no distinct types that are compatible, there
314 isn't much subtlety in the analysis. When we have distinct number
315 types, this will become more interesting.
319 When analysis discovers an inconsistency it needs to report an error;
320 just refusing to run the code ensures that the error doesn't cascade,
321 but by itself it isn't very useful. A clear understanding of the sort
322 of error message that are useful will help guide the process of
325 At a simplistic level, the only sort of error that type analysis can
326 report is that the type of some construct doesn't match a contextual
327 requirement. For example, in `4 + "hello"` the addition provides a
328 contextual requirement for numbers, but `"hello"` is not a number. In
329 this particular example no further information is needed as the types
330 are obvious from local information. When a variable is involved that
331 isn't the case. It may be helpful to explain why the variable has a
332 particular type, by indicating the location where the type was set,
333 whether by declaration or usage.
335 Using a recursive-descent analysis we can easily detect a problem at
336 multiple locations. In "`hello:= "there"; 4 + hello`" the addition
337 will detect that one argument is not a number and the usage of `hello`
338 will detect that a number was wanted, but not provided. In this
339 (early) version of the language, we will generate error reports at
340 multiple locations, so the use of `hello` will report an error and
341 explain were the value was set, and the addition will report an error
342 and say why numbers are needed. To be able to report locations for
343 errors, each language element will need to record a file location
344 (line and column) and each variable will need to record the language
345 element where its type was set. For now we will assume that each line
346 of an error message indicates one location in the file, and up to 2
347 types. So we provide a `printf`-like function which takes a format, a
348 location (a `struct exec` which has not yet been introduced), and 2
349 types. "`%1`" reports the first type, "`%2`" reports the second. We
350 will need a function to print the location, once we know how that is
351 stored. e As will be explained later, there are sometimes extra rules for
352 type matching and they might affect error messages, we need to pass those
355 As well as type errors, we sometimes need to report problems with
356 tokens, which might be unexpected or might name a type that has not
357 been defined. For these we have `tok_err()` which reports an error
358 with a given token. Each of the error functions sets the flag in the
359 context so indicate that parsing failed.
363 static void fput_loc(struct exec *loc, FILE *f);
364 static void type_err(struct parse_context *c,
365 char *fmt, struct exec *loc,
366 struct type *t1, int rules, struct type *t2);
368 ###### core functions
370 static void type_err(struct parse_context *c,
371 char *fmt, struct exec *loc,
372 struct type *t1, int rules, struct type *t2)
374 fprintf(stderr, "%s:", c->file_name);
375 fput_loc(loc, stderr);
376 for (; *fmt ; fmt++) {
383 case '%': fputc(*fmt, stderr); break; // NOTEST
384 default: fputc('?', stderr); break; // NOTEST
386 type_print(t1, stderr);
389 type_print(t2, stderr);
398 static void tok_err(struct parse_context *c, char *fmt, struct token *t)
400 fprintf(stderr, "%s:%d:%d: %s: %.*s\n", c->file_name, t->line, t->col, fmt,
401 t->txt.len, t->txt.txt);
405 ## Entities: declared and predeclared.
407 There are various "things" that the language and/or the interpreter
408 needs to know about to parse and execute a program. These include
409 types, variables, values, and executable code. These are all lumped
410 together under the term "entities" (calling them "objects" would be
411 confusing) and introduced here. The following section will present the
412 different specific code elements which comprise or manipulate these
417 Executables can be lots of different things. In many cases an
418 executable is just an operation combined with one or two other
419 executables. This allows for expressions and lists etc. Other times an
420 executable is something quite specific like a constant or variable name.
421 So we define a `struct exec` to be a general executable with a type, and
422 a `struct binode` which is a subclass of `exec`, forms a node in a
423 binary tree, and holds an operation. There will be other subclasses,
424 and to access these we need to be able to `cast` the `exec` into the
425 various other types. The first field in any `struct exec` is the type
426 from the `exec_types` enum.
429 #define cast(structname, pointer) ({ \
430 const typeof( ((struct structname *)0)->type) *__mptr = &(pointer)->type; \
431 if (__mptr && *__mptr != X##structname) abort(); \
432 (struct structname *)( (char *)__mptr);})
434 #define new(structname) ({ \
435 struct structname *__ptr = ((struct structname *)calloc(1,sizeof(struct structname))); \
436 __ptr->type = X##structname; \
437 __ptr->line = -1; __ptr->column = -1; \
440 #define new_pos(structname, token) ({ \
441 struct structname *__ptr = ((struct structname *)calloc(1,sizeof(struct structname))); \
442 __ptr->type = X##structname; \
443 __ptr->line = token.line; __ptr->column = token.col; \
452 enum exec_types type;
461 struct exec *left, *right;
466 static int __fput_loc(struct exec *loc, FILE *f)
470 if (loc->line >= 0) {
471 fprintf(f, "%d:%d: ", loc->line, loc->column);
474 if (loc->type == Xbinode)
475 return __fput_loc(cast(binode,loc)->left, f) ||
476 __fput_loc(cast(binode,loc)->right, f); // NOTEST
479 static void fput_loc(struct exec *loc, FILE *f)
481 if (!__fput_loc(loc, f))
482 fprintf(f, "??:??: ");
485 Each different type of `exec` node needs a number of functions defined,
486 a bit like methods. We must be able to free it, print it, analyse it
487 and execute it. Once we have specific `exec` types we will need to
488 parse them too. Let's take this a bit more slowly.
492 The parser generator requires a `free_foo` function for each struct
493 that stores attributes and they will often be `exec`s and subtypes
494 there-of. So we need `free_exec` which can handle all the subtypes,
495 and we need `free_binode`.
499 static void free_binode(struct binode *b)
508 ###### core functions
509 static void free_exec(struct exec *e)
520 static void free_exec(struct exec *e);
522 ###### free exec cases
523 case Xbinode: free_binode(cast(binode, e)); break;
527 Printing an `exec` requires that we know the current indent level for
528 printing line-oriented components. As will become clear later, we
529 also want to know what sort of bracketing to use.
533 static void do_indent(int i, char *str)
540 ###### core functions
541 static void print_binode(struct binode *b, int indent, int bracket)
545 ## print binode cases
549 static void print_exec(struct exec *e, int indent, int bracket)
555 print_binode(cast(binode, e), indent, bracket); break;
560 do_indent(indent, "/* FREE");
561 for (v = e->to_free; v; v = v->next_free) {
562 printf(" %.*s", v->name->name.len, v->name->name.txt);
563 printf("[%d,%d]", v->scope_start, v->scope_end);
564 if (v->frame_pos >= 0)
565 printf("(%d+%d)", v->frame_pos,
566 v->type ? v->type->size:0);
574 static void print_exec(struct exec *e, int indent, int bracket);
578 As discussed, analysis involves propagating type requirements around the
579 program and looking for errors.
581 So `propagate_types` is passed an expected type (being a `struct type`
582 pointer together with some `val_rules` flags) that the `exec` is
583 expected to return, and returns the type that it does return, either
584 of which can be `NULL` signifying "unknown". An `ok` flag is passed
585 by reference. It is set to `0` when an error is found, and `2` when
586 any change is made. If it remains unchanged at `1`, then no more
587 propagation is needed.
591 enum val_rules {Rnolabel = 1<<0, Rboolok = 1<<1, Rnoconstant = 1<<2};
595 if (rules & Rnolabel)
596 fputs(" (labels not permitted)", stderr);
600 static struct type *propagate_types(struct exec *prog, struct parse_context *c, int *ok,
601 struct type *type, int rules);
602 ###### core functions
604 static struct type *__propagate_types(struct exec *prog, struct parse_context *c, int *ok,
605 struct type *type, int rules)
612 switch (prog->type) {
615 struct binode *b = cast(binode, prog);
617 ## propagate binode cases
621 ## propagate exec cases
626 static struct type *propagate_types(struct exec *prog, struct parse_context *c, int *ok,
627 struct type *type, int rules)
629 struct type *ret = __propagate_types(prog, c, ok, type, rules);
638 Interpreting an `exec` doesn't require anything but the `exec`. State
639 is stored in variables and each variable will be directly linked from
640 within the `exec` tree. The exception to this is the `main` function
641 which needs to look at command line arguments. This function will be
642 interpreted separately.
644 Each `exec` can return a value combined with a type in `struct lrval`.
645 The type may be `Tnone` but must be non-NULL. Some `exec`s will return
646 the location of a value, which can be updated, in `lval`. Others will
647 set `lval` to NULL indicating that there is a value of appropriate type
651 static struct value interp_exec(struct parse_context *c, struct exec *e,
652 struct type **typeret);
653 ###### core functions
657 struct value rval, *lval;
660 /* If dest is passed, dtype must give the expected type, and
661 * result can go there, in which case type is returned as NULL.
663 static struct lrval _interp_exec(struct parse_context *c, struct exec *e,
664 struct value *dest, struct type *dtype);
666 static struct value interp_exec(struct parse_context *c, struct exec *e,
667 struct type **typeret)
669 struct lrval ret = _interp_exec(c, e, NULL, NULL);
671 if (!ret.type) abort();
675 dup_value(ret.type, ret.lval, &ret.rval);
679 static struct value *linterp_exec(struct parse_context *c, struct exec *e,
680 struct type **typeret)
682 struct lrval ret = _interp_exec(c, e, NULL, NULL);
684 if (!ret.type) abort();
688 free_value(ret.type, &ret.rval);
692 /* dinterp_exec is used when the destination type is certain and
693 * the value has a place to go.
695 static void dinterp_exec(struct parse_context *c, struct exec *e,
696 struct value *dest, struct type *dtype,
699 struct lrval ret = _interp_exec(c, e, dest, dtype);
703 free_value(dtype, dest);
705 dup_value(dtype, ret.lval, dest);
707 memcpy(dest, &ret.rval, dtype->size);
710 static struct lrval _interp_exec(struct parse_context *c, struct exec *e,
711 struct value *dest, struct type *dtype)
713 /* If the result is copied to dest, ret.type is set to NULL */
715 struct value rv = {}, *lrv = NULL;
718 rvtype = ret.type = Tnone;
728 struct binode *b = cast(binode, e);
729 struct value left, right, *lleft;
730 struct type *ltype, *rtype;
731 ltype = rtype = Tnone;
733 ## interp binode cases
735 free_value(ltype, &left);
736 free_value(rtype, &right);
746 ## interp exec cleanup
752 Values come in a wide range of types, with more likely to be added.
753 Each type needs to be able to print its own values (for convenience at
754 least) as well as to compare two values, at least for equality and
755 possibly for order. For now, values might need to be duplicated and
756 freed, though eventually such manipulations will be better integrated
759 Rather than requiring every numeric type to support all numeric
760 operations (add, multiply, etc), we allow types to be able to present
761 as one of a few standard types: integer, float, and fraction. The
762 existence of these conversion functions eventually enable types to
763 determine if they are compatible with other types, though such types
764 have not yet been implemented.
766 Named type are stored in a simple linked list. Objects of each type are
767 "values" which are often passed around by value.
769 There are both explicitly named types, and anonymous types. Anonymous
770 cannot be accessed by name, but are used internally and have a name
771 which might be reported in error messages.
778 ## value union fields
787 void (*init)(struct type *type, struct value *val);
788 void (*prepare_type)(struct parse_context *c, struct type *type, int parse_time);
789 void (*print)(struct type *type, struct value *val, FILE *f);
790 void (*print_type)(struct type *type, FILE *f);
791 int (*cmp_order)(struct type *t1, struct type *t2,
792 struct value *v1, struct value *v2);
793 int (*cmp_eq)(struct type *t1, struct type *t2,
794 struct value *v1, struct value *v2);
795 void (*dup)(struct type *type, struct value *vold, struct value *vnew);
796 void (*free)(struct type *type, struct value *val);
797 void (*free_type)(struct type *t);
798 long long (*to_int)(struct value *v);
799 double (*to_float)(struct value *v);
800 int (*to_mpq)(mpq_t *q, struct value *v);
809 struct type *typelist;
816 static struct type *find_type(struct parse_context *c, struct text s)
818 struct type *t = c->typelist;
820 while (t && (t->anon ||
821 text_cmp(t->name, s) != 0))
826 static struct type *_add_type(struct parse_context *c, struct text s,
827 struct type *proto, int anon)
831 n = calloc(1, sizeof(*n));
835 n->next = c->typelist;
840 static struct type *add_type(struct parse_context *c, struct text s,
843 return _add_type(c, s, proto, 0);
846 static struct type *add_anon_type(struct parse_context *c,
847 struct type *proto, char *name, ...)
853 vasprintf(&t.txt, name, ap);
855 t.len = strlen(name);
856 return _add_type(c, t, proto, 1);
859 static void free_type(struct type *t)
861 /* The type is always a reference to something in the
862 * context, so we don't need to free anything.
866 static void free_value(struct type *type, struct value *v)
870 memset(v, 0x5a, type->size);
874 static void type_print(struct type *type, FILE *f)
877 fputs("*unknown*type*", f); // NOTEST
878 else if (type->name.len && !type->anon)
879 fprintf(f, "%.*s", type->name.len, type->name.txt);
880 else if (type->print_type)
881 type->print_type(type, f);
883 fputs("*invalid*type*", f);
886 static void val_init(struct type *type, struct value *val)
888 if (type && type->init)
889 type->init(type, val);
892 static void dup_value(struct type *type,
893 struct value *vold, struct value *vnew)
895 if (type && type->dup)
896 type->dup(type, vold, vnew);
899 static int value_cmp(struct type *tl, struct type *tr,
900 struct value *left, struct value *right)
902 if (tl && tl->cmp_order)
903 return tl->cmp_order(tl, tr, left, right);
904 if (tl && tl->cmp_eq) // NOTEST
905 return tl->cmp_eq(tl, tr, left, right); // NOTEST
909 static void print_value(struct type *type, struct value *v, FILE *f)
911 if (type && type->print)
912 type->print(type, v, f);
914 fprintf(f, "*Unknown*"); // NOTEST
919 static void free_value(struct type *type, struct value *v);
920 static int type_compat(struct type *require, struct type *have, int rules);
921 static void type_print(struct type *type, FILE *f);
922 static void val_init(struct type *type, struct value *v);
923 static void dup_value(struct type *type,
924 struct value *vold, struct value *vnew);
925 static int value_cmp(struct type *tl, struct type *tr,
926 struct value *left, struct value *right);
927 static void print_value(struct type *type, struct value *v, FILE *f);
929 ###### free context types
931 while (context.typelist) {
932 struct type *t = context.typelist;
934 context.typelist = t->next;
942 Type can be specified for local variables, for fields in a structure,
943 for formal parameters to functions, and possibly elsewhere. Different
944 rules may apply in different contexts. As a minimum, a named type may
945 always be used. Currently the type of a formal parameter can be
946 different from types in other contexts, so we have a separate grammar
952 Type -> IDENTIFIER ${
953 $0 = find_type(c, $1.txt);
956 "error: undefined type", &$1);
963 FormalType -> Type ${ $0 = $<1; }$
964 ## formal type grammar
968 Values of the base types can be numbers, which we represent as
969 multi-precision fractions, strings, Booleans and labels. When
970 analysing the program we also need to allow for places where no value
971 is meaningful (type `Tnone`) and where we don't know what type to
972 expect yet (type is `NULL`).
974 Values are never shared, they are always copied when used, and freed
975 when no longer needed.
977 When propagating type information around the program, we need to
978 determine if two types are compatible, where type `NULL` is compatible
979 with anything. There are two special cases with type compatibility,
980 both related to the Conditional Statement which will be described
981 later. In some cases a Boolean can be accepted as well as some other
982 primary type, and in others any type is acceptable except a label (`Vlabel`).
983 A separate function encoding these cases will simplify some code later.
985 ###### type functions
987 int (*compat)(struct type *this, struct type *other);
991 static int type_compat(struct type *require, struct type *have, int rules)
993 if ((rules & Rboolok) && have == Tbool)
995 if ((rules & Rnolabel) && have == Tlabel)
997 if (!require || !have)
1000 if (require->compat)
1001 return require->compat(require, have);
1003 return require == have;
1008 #include "parse_string.h"
1009 #include "parse_number.h"
1012 myLDLIBS := libnumber.o libstring.o -lgmp
1013 LDLIBS := $(filter-out $(myLDLIBS),$(LDLIBS)) $(myLDLIBS)
1015 ###### type union fields
1016 enum vtype {Vnone, Vstr, Vnum, Vbool, Vlabel} vtype;
1018 ###### value union fields
1024 ###### ast functions
1025 static void _free_value(struct type *type, struct value *v)
1029 switch (type->vtype) {
1031 case Vstr: free(v->str.txt); break;
1032 case Vnum: mpq_clear(v->num); break;
1038 ###### value functions
1040 static void _val_init(struct type *type, struct value *val)
1042 switch(type->vtype) {
1043 case Vnone: // NOTEST
1046 mpq_init(val->num); break;
1048 val->str.txt = malloc(1);
1060 static void _dup_value(struct type *type,
1061 struct value *vold, struct value *vnew)
1063 switch (type->vtype) {
1064 case Vnone: // NOTEST
1067 vnew->label = vold->label;
1070 vnew->bool = vold->bool;
1073 mpq_init(vnew->num);
1074 mpq_set(vnew->num, vold->num);
1077 vnew->str.len = vold->str.len;
1078 vnew->str.txt = malloc(vnew->str.len);
1079 memcpy(vnew->str.txt, vold->str.txt, vnew->str.len);
1084 static int _value_cmp(struct type *tl, struct type *tr,
1085 struct value *left, struct value *right)
1089 return tl - tr; // NOTEST
1090 switch (tl->vtype) {
1091 case Vlabel: cmp = left->label == right->label ? 0 : 1; break;
1092 case Vnum: cmp = mpq_cmp(left->num, right->num); break;
1093 case Vstr: cmp = text_cmp(left->str, right->str); break;
1094 case Vbool: cmp = left->bool - right->bool; break;
1095 case Vnone: cmp = 0; // NOTEST
1100 static void _print_value(struct type *type, struct value *v, FILE *f)
1102 switch (type->vtype) {
1103 case Vnone: // NOTEST
1104 fprintf(f, "*no-value*"); break; // NOTEST
1105 case Vlabel: // NOTEST
1106 fprintf(f, "*label-%p*", v->label); break; // NOTEST
1108 fprintf(f, "%.*s", v->str.len, v->str.txt); break;
1110 fprintf(f, "%s", v->bool ? "True":"False"); break;
1115 mpf_set_q(fl, v->num);
1116 gmp_fprintf(f, "%Fg", fl);
1123 static void _free_value(struct type *type, struct value *v);
1125 static struct type base_prototype = {
1127 .print = _print_value,
1128 .cmp_order = _value_cmp,
1129 .cmp_eq = _value_cmp,
1131 .free = _free_value,
1134 static struct type *Tbool, *Tstr, *Tnum, *Tnone, *Tlabel;
1136 ###### ast functions
1137 static struct type *add_base_type(struct parse_context *c, char *n,
1138 enum vtype vt, int size)
1140 struct text txt = { n, strlen(n) };
1143 t = add_type(c, txt, &base_prototype);
1146 t->align = size > sizeof(void*) ? sizeof(void*) : size;
1147 if (t->size & (t->align - 1))
1148 t->size = (t->size | (t->align - 1)) + 1; // NOTEST
1152 ###### context initialization
1154 Tbool = add_base_type(&context, "Boolean", Vbool, sizeof(char));
1155 Tstr = add_base_type(&context, "string", Vstr, sizeof(struct text));
1156 Tnum = add_base_type(&context, "number", Vnum, sizeof(mpq_t));
1157 Tnone = add_base_type(&context, "none", Vnone, 0);
1158 Tlabel = add_base_type(&context, "label", Vlabel, sizeof(void*));
1162 We have already met values as separate objects. When manifest constants
1163 appear in the program text, that must result in an executable which has
1164 a constant value. So the `val` structure embeds a value in an
1177 ###### ast functions
1178 struct val *new_val(struct type *T, struct token tk)
1180 struct val *v = new_pos(val, tk);
1191 $0 = new_val(Tbool, $1);
1195 $0 = new_val(Tbool, $1);
1200 $0 = new_val(Tnum, $1);
1201 if (number_parse($0->val.num, tail, $1.txt) == 0)
1202 mpq_init($0->val.num); // UNTESTED
1204 tok_err(c, "error: unsupported number suffix",
1209 $0 = new_val(Tstr, $1);
1210 string_parse(&$1, '\\', &$0->val.str, tail);
1212 tok_err(c, "error: unsupported string suffix",
1217 $0 = new_val(Tstr, $1);
1218 string_parse(&$1, '\\', &$0->val.str, tail);
1220 tok_err(c, "error: unsupported string suffix",
1224 ###### print exec cases
1227 struct val *v = cast(val, e);
1228 if (v->vtype == Tstr)
1230 print_value(v->vtype, &v->val, stdout);
1231 if (v->vtype == Tstr)
1236 ###### propagate exec cases
1239 struct val *val = cast(val, prog);
1240 if (!type_compat(type, val->vtype, rules))
1241 type_err(c, "error: expected %1%r found %2",
1242 prog, type, rules, val->vtype);
1246 ###### interp exec cases
1248 rvtype = cast(val, e)->vtype;
1249 dup_value(rvtype, &cast(val, e)->val, &rv);
1252 ###### ast functions
1253 static void free_val(struct val *v)
1256 free_value(v->vtype, &v->val);
1260 ###### free exec cases
1261 case Xval: free_val(cast(val, e)); break;
1263 ###### ast functions
1264 // Move all nodes from 'b' to 'rv', reversing their order.
1265 // In 'b' 'left' is a list, and 'right' is the last node.
1266 // In 'rv', left' is the first node and 'right' is a list.
1267 static struct binode *reorder_bilist(struct binode *b)
1269 struct binode *rv = NULL;
1272 struct exec *t = b->right;
1276 b = cast(binode, b->left);
1286 Variables are scoped named values. We store the names in a linked list
1287 of "bindings" sorted in lexical order, and use sequential search and
1294 struct binding *next; // in lexical order
1298 This linked list is stored in the parse context so that "reduce"
1299 functions can find or add variables, and so the analysis phase can
1300 ensure that every variable gets a type.
1302 ###### parse context
1304 struct binding *varlist; // In lexical order
1306 ###### ast functions
1308 static struct binding *find_binding(struct parse_context *c, struct text s)
1310 struct binding **l = &c->varlist;
1315 (cmp = text_cmp((*l)->name, s)) < 0)
1319 n = calloc(1, sizeof(*n));
1326 Each name can be linked to multiple variables defined in different
1327 scopes. Each scope starts where the name is declared and continues
1328 until the end of the containing code block. Scopes of a given name
1329 cannot nest, so a declaration while a name is in-scope is an error.
1331 ###### binding fields
1332 struct variable *var;
1336 struct variable *previous;
1338 struct binding *name;
1339 struct exec *where_decl;// where name was declared
1340 struct exec *where_set; // where type was set
1344 When a scope closes, the values of the variables might need to be freed.
1345 This happens in the context of some `struct exec` and each `exec` will
1346 need to know which variables need to be freed when it completes.
1349 struct variable *to_free;
1351 ####### variable fields
1352 struct exec *cleanup_exec;
1353 struct variable *next_free;
1355 ####### interp exec cleanup
1358 for (v = e->to_free; v; v = v->next_free) {
1359 struct value *val = var_value(c, v);
1360 free_value(v->type, val);
1364 ###### ast functions
1365 static void variable_unlink_exec(struct variable *v)
1367 struct variable **vp;
1368 if (!v->cleanup_exec)
1370 for (vp = &v->cleanup_exec->to_free;
1371 *vp; vp = &(*vp)->next_free) {
1375 v->cleanup_exec = NULL;
1380 While the naming seems strange, we include local constants in the
1381 definition of variables. A name declared `var := value` can
1382 subsequently be changed, but a name declared `var ::= value` cannot -
1385 ###### variable fields
1388 Scopes in parallel branches can be partially merged. More
1389 specifically, if a given name is declared in both branches of an
1390 if/else then its scope is a candidate for merging. Similarly if
1391 every branch of an exhaustive switch (e.g. has an "else" clause)
1392 declares a given name, then the scopes from the branches are
1393 candidates for merging.
1395 Note that names declared inside a loop (which is only parallel to
1396 itself) are never visible after the loop. Similarly names defined in
1397 scopes which are not parallel, such as those started by `for` and
1398 `switch`, are never visible after the scope. Only variables defined in
1399 both `then` and `else` (including the implicit then after an `if`, and
1400 excluding `then` used with `for`) and in all `case`s and `else` of a
1401 `switch` or `while` can be visible beyond the `if`/`switch`/`while`.
1403 Labels, which are a bit like variables, follow different rules.
1404 Labels are not explicitly declared, but if an undeclared name appears
1405 in a context where a label is legal, that effectively declares the
1406 name as a label. The declaration remains in force (or in scope) at
1407 least to the end of the immediately containing block and conditionally
1408 in any larger containing block which does not declare the name in some
1409 other way. Importantly, the conditional scope extension happens even
1410 if the label is only used in one parallel branch of a conditional --
1411 when used in one branch it is treated as having been declared in all
1414 Merge candidates are tentatively visible beyond the end of the
1415 branching statement which creates them. If the name is used, the
1416 merge is affirmed and they become a single variable visible at the
1417 outer layer. If not - if it is redeclared first - the merge lapses.
1419 To track scopes we have an extra stack, implemented as a linked list,
1420 which roughly parallels the parse stack and which is used exclusively
1421 for scoping. When a new scope is opened, a new frame is pushed and
1422 the child-count of the parent frame is incremented. This child-count
1423 is used to distinguish between the first of a set of parallel scopes,
1424 in which declared variables must not be in scope, and subsequent
1425 branches, whether they may already be conditionally scoped.
1427 We need a total ordering of scopes so we can easily compare to variables
1428 to see if they are concurrently in scope. To achieve this we record a
1429 `scope_count` which is actually a count of both beginnings and endings
1430 of scopes. Then each variable has a record of the scope count where it
1431 enters scope, and where it leaves.
1433 To push a new frame *before* any code in the frame is parsed, we need a
1434 grammar reduction. This is most easily achieved with a grammar
1435 element which derives the empty string, and creates the new scope when
1436 it is recognised. This can be placed, for example, between a keyword
1437 like "if" and the code following it.
1441 struct scope *parent;
1445 ###### parse context
1448 struct scope *scope_stack;
1450 ###### variable fields
1451 int scope_start, scope_end;
1453 ###### ast functions
1454 static void scope_pop(struct parse_context *c)
1456 struct scope *s = c->scope_stack;
1458 c->scope_stack = s->parent;
1460 c->scope_depth -= 1;
1461 c->scope_count += 1;
1464 static void scope_push(struct parse_context *c)
1466 struct scope *s = calloc(1, sizeof(*s));
1468 c->scope_stack->child_count += 1;
1469 s->parent = c->scope_stack;
1471 c->scope_depth += 1;
1472 c->scope_count += 1;
1478 OpenScope -> ${ scope_push(c); }$
1480 Each variable records a scope depth and is in one of four states:
1482 - "in scope". This is the case between the declaration of the
1483 variable and the end of the containing block, and also between
1484 the usage with affirms a merge and the end of that block.
1486 The scope depth is not greater than the current parse context scope
1487 nest depth. When the block of that depth closes, the state will
1488 change. To achieve this, all "in scope" variables are linked
1489 together as a stack in nesting order.
1491 - "pending". The "in scope" block has closed, but other parallel
1492 scopes are still being processed. So far, every parallel block at
1493 the same level that has closed has declared the name.
1495 The scope depth is the depth of the last parallel block that
1496 enclosed the declaration, and that has closed.
1498 - "conditionally in scope". The "in scope" block and all parallel
1499 scopes have closed, and no further mention of the name has been seen.
1500 This state includes a secondary nest depth (`min_depth`) which records
1501 the outermost scope seen since the variable became conditionally in
1502 scope. If a use of the name is found, the variable becomes "in scope"
1503 and that secondary depth becomes the recorded scope depth. If the
1504 name is declared as a new variable, the old variable becomes "out of
1505 scope" and the recorded scope depth stays unchanged.
1507 - "out of scope". The variable is neither in scope nor conditionally
1508 in scope. It is permanently out of scope now and can be removed from
1509 the "in scope" stack. When a variable becomes out-of-scope it is
1510 moved to a separate list (`out_scope`) of variables which have fully
1511 known scope. This will be used at the end of each function to assign
1512 each variable a place in the stack frame.
1514 ###### variable fields
1515 int depth, min_depth;
1516 enum { OutScope, PendingScope, CondScope, InScope } scope;
1517 struct variable *in_scope;
1519 ###### parse context
1521 struct variable *in_scope;
1522 struct variable *out_scope;
1524 All variables with the same name are linked together using the
1525 'previous' link. Those variable that have been affirmatively merged all
1526 have a 'merged' pointer that points to one primary variable - the most
1527 recently declared instance. When merging variables, we need to also
1528 adjust the 'merged' pointer on any other variables that had previously
1529 been merged with the one that will no longer be primary.
1531 A variable that is no longer the most recent instance of a name may
1532 still have "pending" scope, if it might still be merged with most
1533 recent instance. These variables don't really belong in the
1534 "in_scope" list, but are not immediately removed when a new instance
1535 is found. Instead, they are detected and ignored when considering the
1536 list of in_scope names.
1538 The storage of the value of a variable will be described later. For now
1539 we just need to know that when a variable goes out of scope, it might
1540 need to be freed. For this we need to be able to find it, so assume that
1541 `var_value()` will provide that.
1543 ###### variable fields
1544 struct variable *merged;
1546 ###### ast functions
1548 static void variable_merge(struct variable *primary, struct variable *secondary)
1552 primary = primary->merged;
1554 for (v = primary->previous; v; v=v->previous)
1555 if (v == secondary || v == secondary->merged ||
1556 v->merged == secondary ||
1557 v->merged == secondary->merged) {
1558 v->scope = OutScope;
1559 v->merged = primary;
1560 if (v->scope_start < primary->scope_start)
1561 primary->scope_start = v->scope_start;
1562 if (v->scope_end > primary->scope_end)
1563 primary->scope_end = v->scope_end; // NOTEST
1564 variable_unlink_exec(v);
1568 ###### forward decls
1569 static struct value *var_value(struct parse_context *c, struct variable *v);
1571 ###### free global vars
1573 while (context.varlist) {
1574 struct binding *b = context.varlist;
1575 struct variable *v = b->var;
1576 context.varlist = b->next;
1579 struct variable *next = v->previous;
1582 free_value(v->type, var_value(&context, v));
1584 // This is a global constant
1585 free_exec(v->where_decl);
1592 #### Manipulating Bindings
1594 When a name is conditionally visible, a new declaration discards the old
1595 binding - the condition lapses. Similarly when we reach the end of a
1596 function (outermost non-global scope) any conditional scope must lapse.
1597 Conversely a usage of the name affirms the visibility and extends it to
1598 the end of the containing block - i.e. the block that contains both the
1599 original declaration and the latest usage. This is determined from
1600 `min_depth`. When a conditionally visible variable gets affirmed like
1601 this, it is also merged with other conditionally visible variables with
1604 When we parse a variable declaration we either report an error if the
1605 name is currently bound, or create a new variable at the current nest
1606 depth if the name is unbound or bound to a conditionally scoped or
1607 pending-scope variable. If the previous variable was conditionally
1608 scoped, it and its homonyms becomes out-of-scope.
1610 When we parse a variable reference (including non-declarative assignment
1611 "foo = bar") we report an error if the name is not bound or is bound to
1612 a pending-scope variable; update the scope if the name is bound to a
1613 conditionally scoped variable; or just proceed normally if the named
1614 variable is in scope.
1616 When we exit a scope, any variables bound at this level are either
1617 marked out of scope or pending-scoped, depending on whether the scope
1618 was sequential or parallel. Here a "parallel" scope means the "then"
1619 or "else" part of a conditional, or any "case" or "else" branch of a
1620 switch. Other scopes are "sequential".
1622 When exiting a parallel scope we check if there are any variables that
1623 were previously pending and are still visible. If there are, then
1624 they weren't redeclared in the most recent scope, so they cannot be
1625 merged and must become out-of-scope. If it is not the first of
1626 parallel scopes (based on `child_count`), we check that there was a
1627 previous binding that is still pending-scope. If there isn't, the new
1628 variable must now be out-of-scope.
1630 When exiting a sequential scope that immediately enclosed parallel
1631 scopes, we need to resolve any pending-scope variables. If there was
1632 no `else` clause, and we cannot determine that the `switch` was exhaustive,
1633 we need to mark all pending-scope variable as out-of-scope. Otherwise
1634 all pending-scope variables become conditionally scoped.
1637 enum closetype { CloseSequential, CloseFunction, CloseParallel, CloseElse };
1639 ###### ast functions
1641 static struct variable *var_decl(struct parse_context *c, struct text s)
1643 struct binding *b = find_binding(c, s);
1644 struct variable *v = b->var;
1646 switch (v ? v->scope : OutScope) {
1648 /* Caller will report the error */
1652 v && v->scope == CondScope;
1654 v->scope = OutScope;
1658 v = calloc(1, sizeof(*v));
1659 v->previous = b->var;
1663 v->min_depth = v->depth = c->scope_depth;
1665 v->in_scope = c->in_scope;
1666 v->scope_start = c->scope_count;
1672 static struct variable *var_ref(struct parse_context *c, struct text s)
1674 struct binding *b = find_binding(c, s);
1675 struct variable *v = b->var;
1676 struct variable *v2;
1678 switch (v ? v->scope : OutScope) {
1681 /* Caller will report the error */
1684 /* All CondScope variables of this name need to be merged
1685 * and become InScope
1687 v->depth = v->min_depth;
1689 for (v2 = v->previous;
1690 v2 && v2->scope == CondScope;
1692 variable_merge(v, v2);
1700 static int var_refile(struct parse_context *c, struct variable *v)
1702 /* Variable just went out of scope. Add it to the out_scope
1703 * list, sorted by ->scope_start
1705 struct variable **vp = &c->out_scope;
1706 while ((*vp) && (*vp)->scope_start < v->scope_start)
1707 vp = &(*vp)->in_scope;
1713 static void var_block_close(struct parse_context *c, enum closetype ct,
1716 /* Close off all variables that are in_scope.
1717 * Some variables in c->scope may already be not-in-scope,
1718 * such as when a PendingScope variable is hidden by a new
1719 * variable with the same name.
1720 * So we check for v->name->var != v and drop them.
1721 * If we choose to make a variable OutScope, we drop it
1724 struct variable *v, **vp, *v2;
1727 for (vp = &c->in_scope;
1728 (v = *vp) && v->min_depth > c->scope_depth;
1729 (v->scope == OutScope || v->name->var != v)
1730 ? (*vp = v->in_scope, var_refile(c, v))
1731 : ( vp = &v->in_scope, 0)) {
1732 v->min_depth = c->scope_depth;
1733 if (v->name->var != v)
1734 /* This is still in scope, but we haven't just
1738 v->min_depth = c->scope_depth;
1739 if (v->scope == InScope)
1740 v->scope_end = c->scope_count;
1741 if (v->scope == InScope && e && !v->global) {
1742 /* This variable gets cleaned up when 'e' finishes */
1743 variable_unlink_exec(v);
1744 v->cleanup_exec = e;
1745 v->next_free = e->to_free;
1750 case CloseParallel: /* handle PendingScope */
1754 if (c->scope_stack->child_count == 1)
1755 /* first among parallel branches */
1756 v->scope = PendingScope;
1757 else if (v->previous &&
1758 v->previous->scope == PendingScope)
1759 /* all previous branches used name */
1760 v->scope = PendingScope;
1761 else if (v->type == Tlabel)
1762 /* Labels remain pending even when not used */
1763 v->scope = PendingScope; // UNTESTED
1765 v->scope = OutScope;
1766 if (ct == CloseElse) {
1767 /* All Pending variables with this name
1768 * are now Conditional */
1770 v2 && v2->scope == PendingScope;
1772 v2->scope = CondScope;
1776 /* Not possible as it would require
1777 * parallel scope to be nested immediately
1778 * in a parallel scope, and that never
1782 /* Not possible as we already tested for
1789 if (v->scope == CondScope)
1790 /* Condition cannot continue past end of function */
1793 case CloseSequential:
1794 if (v->type == Tlabel)
1795 v->scope = PendingScope;
1798 v->scope = OutScope;
1801 /* There was no 'else', so we can only become
1802 * conditional if we know the cases were exhaustive,
1803 * and that doesn't mean anything yet.
1804 * So only labels become conditional..
1807 v2 && v2->scope == PendingScope;
1809 if (v2->type == Tlabel)
1810 v2->scope = CondScope;
1812 v2->scope = OutScope;
1815 case OutScope: break;
1824 The value of a variable is store separately from the variable, on an
1825 analogue of a stack frame. There are (currently) two frames that can be
1826 active. A global frame which currently only stores constants, and a
1827 stacked frame which stores local variables. Each variable knows if it
1828 is global or not, and what its index into the frame is.
1830 Values in the global frame are known immediately they are relevant, so
1831 the frame needs to be reallocated as it grows so it can store those
1832 values. The local frame doesn't get values until the interpreted phase
1833 is started, so there is no need to allocate until the size is known.
1835 We initialize the `frame_pos` to an impossible value, so that we can
1836 tell if it was set or not later.
1838 ###### variable fields
1842 ###### variable init
1845 ###### parse context
1847 short global_size, global_alloc;
1849 void *global, *local;
1851 ###### forward decls
1852 static struct value *global_alloc(struct parse_context *c, struct type *t,
1853 struct variable *v, struct value *init);
1855 ###### ast functions
1857 static struct value *var_value(struct parse_context *c, struct variable *v)
1860 if (!c->local || !v->type)
1861 return NULL; // NOTEST
1862 if (v->frame_pos + v->type->size > c->local_size) {
1863 printf("INVALID frame_pos\n"); // NOTEST
1866 return c->local + v->frame_pos;
1868 if (c->global_size > c->global_alloc) {
1869 int old = c->global_alloc;
1870 c->global_alloc = (c->global_size | 1023) + 1024;
1871 c->global = realloc(c->global, c->global_alloc);
1872 memset(c->global + old, 0, c->global_alloc - old);
1874 return c->global + v->frame_pos;
1877 static struct value *global_alloc(struct parse_context *c, struct type *t,
1878 struct variable *v, struct value *init)
1881 struct variable scratch;
1883 if (t->prepare_type)
1884 t->prepare_type(c, t, 1); // NOTEST
1886 if (c->global_size & (t->align - 1))
1887 c->global_size = (c->global_size + t->align) & ~(t->align-1);
1892 v->frame_pos = c->global_size;
1894 c->global_size += v->type->size;
1895 ret = var_value(c, v);
1897 memcpy(ret, init, t->size);
1903 As global values are found -- struct field initializers, labels etc --
1904 `global_alloc()` is called to record the value in the global frame.
1906 When the program is fully parsed, each function is analysed, we need to
1907 walk the list of variables local to that function and assign them an
1908 offset in the stack frame. For this we have `scope_finalize()`.
1910 We keep the stack from dense by re-using space for between variables
1911 that are not in scope at the same time. The `out_scope` list is sorted
1912 by `scope_start` and as we process a varible, we move it to an FIFO
1913 stack. For each variable we consider, we first discard any from the
1914 stack anything that went out of scope before the new variable came in.
1915 Then we place the new variable just after the one at the top of the
1918 ###### ast functions
1920 static void scope_finalize(struct parse_context *c, struct type *ft)
1922 int size = ft->function.local_size;
1923 struct variable *next = ft->function.scope;
1924 struct variable *done = NULL;
1927 struct variable *v = next;
1928 struct type *t = v->type;
1935 if (v->frame_pos >= 0)
1937 while (done && done->scope_end < v->scope_start)
1938 done = done->in_scope;
1940 pos = done->frame_pos + done->type->size;
1942 pos = ft->function.local_size;
1943 if (pos & (t->align - 1))
1944 pos = (pos + t->align) & ~(t->align-1);
1946 if (size < pos + v->type->size)
1947 size = pos + v->type->size;
1951 c->out_scope = NULL;
1952 ft->function.local_size = size;
1955 ###### free context storage
1956 free(context.global);
1958 #### Variables as executables
1960 Just as we used a `val` to wrap a value into an `exec`, we similarly
1961 need a `var` to wrap a `variable` into an exec. While each `val`
1962 contained a copy of the value, each `var` holds a link to the variable
1963 because it really is the same variable no matter where it appears.
1964 When a variable is used, we need to remember to follow the `->merged`
1965 link to find the primary instance.
1967 When a variable is declared, it may or may not be given an explicit
1968 type. We need to record which so that we can report the parsed code
1977 struct variable *var;
1980 ###### variable fields
1988 VariableDecl -> IDENTIFIER : ${ {
1989 struct variable *v = var_decl(c, $1.txt);
1990 $0 = new_pos(var, $1);
1995 v = var_ref(c, $1.txt);
1997 type_err(c, "error: variable '%v' redeclared",
1999 type_err(c, "info: this is where '%v' was first declared",
2000 v->where_decl, NULL, 0, NULL);
2003 | IDENTIFIER :: ${ {
2004 struct variable *v = var_decl(c, $1.txt);
2005 $0 = new_pos(var, $1);
2011 v = var_ref(c, $1.txt);
2013 type_err(c, "error: variable '%v' redeclared",
2015 type_err(c, "info: this is where '%v' was first declared",
2016 v->where_decl, NULL, 0, NULL);
2019 | IDENTIFIER : Type ${ {
2020 struct variable *v = var_decl(c, $1.txt);
2021 $0 = new_pos(var, $1);
2027 v->explicit_type = 1;
2029 v = var_ref(c, $1.txt);
2031 type_err(c, "error: variable '%v' redeclared",
2033 type_err(c, "info: this is where '%v' was first declared",
2034 v->where_decl, NULL, 0, NULL);
2037 | IDENTIFIER :: Type ${ {
2038 struct variable *v = var_decl(c, $1.txt);
2039 $0 = new_pos(var, $1);
2046 v->explicit_type = 1;
2048 v = var_ref(c, $1.txt);
2050 type_err(c, "error: variable '%v' redeclared",
2052 type_err(c, "info: this is where '%v' was first declared",
2053 v->where_decl, NULL, 0, NULL);
2058 Variable -> IDENTIFIER ${ {
2059 struct variable *v = var_ref(c, $1.txt);
2060 $0 = new_pos(var, $1);
2062 /* This might be a label - allocate a var just in case */
2063 v = var_decl(c, $1.txt);
2070 cast(var, $0)->var = v;
2073 ###### print exec cases
2076 struct var *v = cast(var, e);
2078 struct binding *b = v->var->name;
2079 printf("%.*s", b->name.len, b->name.txt);
2086 if (loc && loc->type == Xvar) {
2087 struct var *v = cast(var, loc);
2089 struct binding *b = v->var->name;
2090 fprintf(stderr, "%.*s", b->name.len, b->name.txt);
2092 fputs("???", stderr); // NOTEST
2094 fputs("NOTVAR", stderr);
2097 ###### propagate exec cases
2101 struct var *var = cast(var, prog);
2102 struct variable *v = var->var;
2104 type_err(c, "%d:BUG: no variable!!", prog, NULL, 0, NULL); // NOTEST
2105 return Tnone; // NOTEST
2108 if (v->constant && (rules & Rnoconstant)) {
2109 type_err(c, "error: Cannot assign to a constant: %v",
2110 prog, NULL, 0, NULL);
2111 type_err(c, "info: name was defined as a constant here",
2112 v->where_decl, NULL, 0, NULL);
2115 if (v->type == Tnone && v->where_decl == prog)
2116 type_err(c, "error: variable used but not declared: %v",
2117 prog, NULL, 0, NULL);
2118 if (v->type == NULL) {
2119 if (type && *ok != 0) {
2121 v->where_set = prog;
2126 if (!type_compat(type, v->type, rules)) {
2127 type_err(c, "error: expected %1%r but variable '%v' is %2", prog,
2128 type, rules, v->type);
2129 type_err(c, "info: this is where '%v' was set to %1", v->where_set,
2130 v->type, rules, NULL);
2137 ###### interp exec cases
2140 struct var *var = cast(var, e);
2141 struct variable *v = var->var;
2144 lrv = var_value(c, v);
2149 ###### ast functions
2151 static void free_var(struct var *v)
2156 ###### free exec cases
2157 case Xvar: free_var(cast(var, e)); break;
2162 Now that we have the shape of the interpreter in place we can add some
2163 complex types and connected them in to the data structures and the
2164 different phases of parse, analyse, print, interpret.
2166 Being "complex" the language will naturally have syntax to access
2167 specifics of objects of these types. These will fit into the grammar as
2168 "Terms" which are the things that are combined with various operators to
2169 form "Expression". Where a Term is formed by some operation on another
2170 Term, the subordinate Term will always come first, so for example a
2171 member of an array will be expressed as the Term for the array followed
2172 by an index in square brackets. The strict rule of using postfix
2173 operations makes precedence irrelevant within terms. To provide a place
2174 to put the grammar for each terms of each type, we will start out by
2175 introducing the "Term" grammar production, with contains at least a
2176 simple "Value" (to be explained later).
2180 Term -> Value ${ $0 = $<1; }$
2181 | Variable ${ $0 = $<1; }$
2184 Thus far the complex types we have are arrays and structs.
2188 Arrays can be declared by giving a size and a type, as `[size]type' so
2189 `freq:[26]number` declares `freq` to be an array of 26 numbers. The
2190 size can be either a literal number, or a named constant. Some day an
2191 arbitrary expression will be supported.
2193 As a formal parameter to a function, the array can be declared with a
2194 new variable as the size: `name:[size::number]string`. The `size`
2195 variable is set to the size of the array and must be a constant. As
2196 `number` is the only supported type, it can be left out:
2197 `name:[size::]string`.
2199 Arrays cannot be assigned. When pointers are introduced we will also
2200 introduce array slices which can refer to part or all of an array -
2201 the assignment syntax will create a slice. For now, an array can only
2202 ever be referenced by the name it is declared with. It is likely that
2203 a "`copy`" primitive will eventually be define which can be used to
2204 make a copy of an array with controllable recursive depth.
2206 For now we have two sorts of array, those with fixed size either because
2207 it is given as a literal number or because it is a struct member (which
2208 cannot have a runtime-changing size), and those with a size that is
2209 determined at runtime - local variables with a const size. The former
2210 have their size calculated at parse time, the latter at run time.
2212 For the latter type, the `size` field of the type is the size of a
2213 pointer, and the array is reallocated every time it comes into scope.
2215 We differentiate struct fields with a const size from local variables
2216 with a const size by whether they are prepared at parse time or not.
2218 ###### type union fields
2221 int unspec; // size is unspecified - vsize must be set.
2224 struct variable *vsize;
2225 struct type *member;
2228 ###### value union fields
2229 void *array; // used if not static_size
2231 ###### value functions
2233 static void array_prepare_type(struct parse_context *c, struct type *type,
2236 struct value *vsize;
2238 if (type->array.static_size)
2240 if (type->array.unspec && parse_time)
2243 if (type->array.vsize) {
2244 vsize = var_value(c, type->array.vsize);
2248 mpz_tdiv_q(q, mpq_numref(vsize->num), mpq_denref(vsize->num));
2249 type->array.size = mpz_get_si(q);
2253 if (parse_time && type->array.member->size) {
2254 type->array.static_size = 1;
2255 type->size = type->array.size * type->array.member->size;
2256 type->align = type->array.member->align;
2260 static void array_init(struct type *type, struct value *val)
2263 void *ptr = val->ptr;
2267 if (!type->array.static_size) {
2268 val->array = calloc(type->array.size,
2269 type->array.member->size);
2272 for (i = 0; i < type->array.size; i++) {
2274 v = (void*)ptr + i * type->array.member->size;
2275 val_init(type->array.member, v);
2279 static void array_free(struct type *type, struct value *val)
2282 void *ptr = val->ptr;
2284 if (!type->array.static_size)
2286 for (i = 0; i < type->array.size; i++) {
2288 v = (void*)ptr + i * type->array.member->size;
2289 free_value(type->array.member, v);
2291 if (!type->array.static_size)
2295 static int array_compat(struct type *require, struct type *have)
2297 if (have->compat != require->compat)
2299 /* Both are arrays, so we can look at details */
2300 if (!type_compat(require->array.member, have->array.member, 0))
2302 if (have->array.unspec && require->array.unspec) {
2303 if (have->array.vsize && require->array.vsize &&
2304 have->array.vsize != require->array.vsize) // UNTESTED
2305 /* sizes might not be the same */
2306 return 0; // UNTESTED
2309 if (have->array.unspec || require->array.unspec)
2310 return 1; // UNTESTED
2311 if (require->array.vsize == NULL && have->array.vsize == NULL)
2312 return require->array.size == have->array.size;
2314 return require->array.vsize == have->array.vsize; // UNTESTED
2317 static void array_print_type(struct type *type, FILE *f)
2320 if (type->array.vsize) {
2321 struct binding *b = type->array.vsize->name;
2322 fprintf(f, "%.*s%s]", b->name.len, b->name.txt,
2323 type->array.unspec ? "::" : "");
2324 } else if (type->array.size)
2325 fprintf(f, "%d]", type->array.size);
2328 type_print(type->array.member, f);
2331 static struct type array_prototype = {
2333 .prepare_type = array_prepare_type,
2334 .print_type = array_print_type,
2335 .compat = array_compat,
2337 .size = sizeof(void*),
2338 .align = sizeof(void*),
2341 ###### declare terminals
2346 | [ NUMBER ] Type ${ {
2352 if (number_parse(num, tail, $2.txt) == 0)
2353 tok_err(c, "error: unrecognised number", &$2);
2355 tok_err(c, "error: unsupported number suffix", &$2);
2358 elements = mpz_get_ui(mpq_numref(num));
2359 if (mpz_cmp_ui(mpq_denref(num), 1) != 0) {
2360 tok_err(c, "error: array size must be an integer",
2362 } else if (mpz_cmp_ui(mpq_numref(num), 1UL << 30) >= 0)
2363 tok_err(c, "error: array size is too large",
2368 $0 = t = add_anon_type(c, &array_prototype, "array[%d]", elements );
2369 t->array.size = elements;
2370 t->array.member = $<4;
2371 t->array.vsize = NULL;
2374 | [ IDENTIFIER ] Type ${ {
2375 struct variable *v = var_ref(c, $2.txt);
2378 tok_err(c, "error: name undeclared", &$2);
2379 else if (!v->constant)
2380 tok_err(c, "error: array size must be a constant", &$2);
2382 $0 = add_anon_type(c, &array_prototype, "array[%.*s]", $2.txt.len, $2.txt.txt);
2383 $0->array.member = $<4;
2385 $0->array.vsize = v;
2390 OptType -> Type ${ $0 = $<1; }$
2393 ###### formal type grammar
2395 | [ IDENTIFIER :: OptType ] Type ${ {
2396 struct variable *v = var_decl(c, $ID.txt);
2402 $0 = add_anon_type(c, &array_prototype, "array[var]");
2403 $0->array.member = $<6;
2405 $0->array.unspec = 1;
2406 $0->array.vsize = v;
2414 | Term [ Expression ] ${ {
2415 struct binode *b = new(binode);
2422 ###### print binode cases
2424 print_exec(b->left, -1, bracket);
2426 print_exec(b->right, -1, bracket);
2430 ###### propagate binode cases
2432 /* left must be an array, right must be a number,
2433 * result is the member type of the array
2435 propagate_types(b->right, c, ok, Tnum, 0);
2436 t = propagate_types(b->left, c, ok, NULL, rules & Rnoconstant);
2437 if (!t || t->compat != array_compat) {
2438 type_err(c, "error: %1 cannot be indexed", prog, t, 0, NULL);
2441 if (!type_compat(type, t->array.member, rules)) {
2442 type_err(c, "error: have %1 but need %2", prog,
2443 t->array.member, rules, type);
2445 return t->array.member;
2449 ###### interp binode cases
2455 lleft = linterp_exec(c, b->left, <ype);
2456 right = interp_exec(c, b->right, &rtype);
2458 mpz_tdiv_q(q, mpq_numref(right.num), mpq_denref(right.num));
2462 if (ltype->array.static_size)
2465 ptr = *(void**)lleft;
2466 rvtype = ltype->array.member;
2467 if (i >= 0 && i < ltype->array.size)
2468 lrv = ptr + i * rvtype->size;
2470 val_init(ltype->array.member, &rv); // UNSAFE
2477 A `struct` is a data-type that contains one or more other data-types.
2478 It differs from an array in that each member can be of a different
2479 type, and they are accessed by name rather than by number. Thus you
2480 cannot choose an element by calculation, you need to know what you
2483 The language makes no promises about how a given structure will be
2484 stored in memory - it is free to rearrange fields to suit whatever
2485 criteria seems important.
2487 Structs are declared separately from program code - they cannot be
2488 declared in-line in a variable declaration like arrays can. A struct
2489 is given a name and this name is used to identify the type - the name
2490 is not prefixed by the word `struct` as it would be in C.
2492 Structs are only treated as the same if they have the same name.
2493 Simply having the same fields in the same order is not enough. This
2494 might change once we can create structure initializers from a list of
2497 Each component datum is identified much like a variable is declared,
2498 with a name, one or two colons, and a type. The type cannot be omitted
2499 as there is no opportunity to deduce the type from usage. An initial
2500 value can be given following an equals sign, so
2502 ##### Example: a struct type
2508 would declare a type called "complex" which has two number fields,
2509 each initialised to zero.
2511 Struct will need to be declared separately from the code that uses
2512 them, so we will need to be able to print out the declaration of a
2513 struct when reprinting the whole program. So a `print_type_decl` type
2514 function will be needed.
2516 ###### type union fields
2525 } *fields; // This is created when field_list is analysed.
2527 struct fieldlist *prev;
2530 } *field_list; // This is created during parsing
2533 ###### type functions
2534 void (*print_type_decl)(struct type *type, FILE *f);
2536 ###### value functions
2538 static void structure_init(struct type *type, struct value *val)
2542 for (i = 0; i < type->structure.nfields; i++) {
2544 v = (void*) val->ptr + type->structure.fields[i].offset;
2545 if (type->structure.fields[i].init)
2546 dup_value(type->structure.fields[i].type,
2547 type->structure.fields[i].init,
2550 val_init(type->structure.fields[i].type, v);
2554 static void structure_free(struct type *type, struct value *val)
2558 for (i = 0; i < type->structure.nfields; i++) {
2560 v = (void*)val->ptr + type->structure.fields[i].offset;
2561 free_value(type->structure.fields[i].type, v);
2565 static void free_fieldlist(struct fieldlist *f)
2569 free_fieldlist(f->prev);
2574 static void structure_free_type(struct type *t)
2577 for (i = 0; i < t->structure.nfields; i++)
2578 if (t->structure.fields[i].init) {
2579 free_value(t->structure.fields[i].type,
2580 t->structure.fields[i].init);
2582 free(t->structure.fields);
2583 free_fieldlist(t->structure.field_list);
2586 static void structure_prepare_type(struct parse_context *c,
2587 struct type *t, int parse_time)
2590 struct fieldlist *f;
2592 if (!parse_time || t->structure.fields)
2595 for (f = t->structure.field_list; f; f=f->prev) {
2599 if (f->f.type->prepare_type)
2600 f->f.type->prepare_type(c, f->f.type, 1);
2601 if (f->init == NULL)
2605 propagate_types(f->init, c, &ok, f->f.type, 0);
2608 c->parse_error = 1; // NOTEST
2611 t->structure.nfields = cnt;
2612 t->structure.fields = calloc(cnt, sizeof(struct field));
2613 f = t->structure.field_list;
2615 int a = f->f.type->align;
2617 t->structure.fields[cnt] = f->f;
2618 if (t->size & (a-1))
2619 t->size = (t->size | (a-1)) + 1;
2620 t->structure.fields[cnt].offset = t->size;
2621 t->size += ((f->f.type->size - 1) | (a-1)) + 1;
2625 if (f->init && !c->parse_error) {
2626 struct value vl = interp_exec(c, f->init, NULL);
2627 t->structure.fields[cnt].init =
2628 global_alloc(c, f->f.type, NULL, &vl);
2635 static struct type structure_prototype = {
2636 .init = structure_init,
2637 .free = structure_free,
2638 .free_type = structure_free_type,
2639 .print_type_decl = structure_print_type,
2640 .prepare_type = structure_prepare_type,
2654 ###### free exec cases
2656 free_exec(cast(fieldref, e)->left);
2660 ###### declare terminals
2665 | Term . IDENTIFIER ${ {
2666 struct fieldref *fr = new_pos(fieldref, $2);
2673 ###### print exec cases
2677 struct fieldref *f = cast(fieldref, e);
2678 print_exec(f->left, -1, bracket);
2679 printf(".%.*s", f->name.len, f->name.txt);
2683 ###### ast functions
2684 static int find_struct_index(struct type *type, struct text field)
2687 for (i = 0; i < type->structure.nfields; i++)
2688 if (text_cmp(type->structure.fields[i].name, field) == 0)
2693 ###### propagate exec cases
2697 struct fieldref *f = cast(fieldref, prog);
2698 struct type *st = propagate_types(f->left, c, ok, NULL, 0);
2701 type_err(c, "error: unknown type for field access", f->left, // UNTESTED
2703 else if (st->init != structure_init)
2704 type_err(c, "error: field reference attempted on %1, not a struct",
2705 f->left, st, 0, NULL);
2706 else if (f->index == -2) {
2707 f->index = find_struct_index(st, f->name);
2709 type_err(c, "error: cannot find requested field in %1",
2710 f->left, st, 0, NULL);
2712 if (f->index >= 0) {
2713 struct type *ft = st->structure.fields[f->index].type;
2714 if (!type_compat(type, ft, rules))
2715 type_err(c, "error: have %1 but need %2", prog,
2722 ###### interp exec cases
2725 struct fieldref *f = cast(fieldref, e);
2727 struct value *lleft = linterp_exec(c, f->left, <ype);
2728 lrv = (void*)lleft->ptr + ltype->structure.fields[f->index].offset;
2729 rvtype = ltype->structure.fields[f->index].type;
2733 ###### top level grammar
2734 DeclareStruct -> struct IDENTIFIER FieldBlock Newlines ${ {
2736 add_type(c, $2.txt, &structure_prototype);
2737 t->structure.field_list = $<FB;
2738 if (t->prepare_type)
2739 t->prepare_type(c, t, 1);
2744 FieldBlock -> { IN OptNL FieldLines OUT OptNL } ${ $0 = $<FL; }$
2745 | { SimpleFieldList } ${ $0 = $<SFL; }$
2746 | IN OptNL FieldLines OUT ${ $0 = $<FL; }$
2747 | SimpleFieldList EOL ${ $0 = $<SFL; }$
2749 FieldLines -> SimpleFieldList Newlines ${ $0 = $<SFL; }$
2750 | FieldLines SimpleFieldList Newlines ${
2755 SimpleFieldList -> Field ${ $0 = $<F; }$
2756 | SimpleFieldList ; Field ${
2760 | SimpleFieldList ; ${
2763 | ERROR ${ tok_err(c, "Syntax error in struct field", &$1); }$
2765 Field -> IDENTIFIER : Type = Expression ${ {
2766 $0 = calloc(1, sizeof(struct fieldlist));
2767 $0->f.name = $ID.txt;
2768 $0->f.type = $<Type;
2772 | IDENTIFIER : Type ${
2773 $0 = calloc(1, sizeof(struct fieldlist));
2774 $0->f.name = $ID.txt;
2775 $0->f.type = $<Type;
2778 ###### forward decls
2779 static void structure_print_type(struct type *t, FILE *f);
2781 ###### value functions
2782 static void structure_print_type(struct type *t, FILE *f)
2786 fprintf(f, "struct %.*s\n", t->name.len, t->name.txt);
2788 for (i = 0; i < t->structure.nfields; i++) {
2789 struct field *fl = t->structure.fields + i;
2790 fprintf(f, " %.*s : ", fl->name.len, fl->name.txt);
2791 type_print(fl->type, f);
2792 if (fl->type->print && fl->init) {
2794 if (fl->type == Tstr)
2795 fprintf(f, "\""); // UNTESTED
2796 print_value(fl->type, fl->init, f);
2797 if (fl->type == Tstr)
2798 fprintf(f, "\""); // UNTESTED
2804 ###### print type decls
2809 while (target != 0) {
2811 for (t = context.typelist; t ; t=t->next)
2812 if (!t->anon && t->print_type_decl &&
2822 t->print_type_decl(t, stdout);
2830 A function is a chunk of code which can be passed parameters and can
2831 return results. Each function has a type which includes the set of
2832 parameters and the return value. As yet these types cannot be declared
2833 separately from the function itself.
2835 The parameters can be specified either in parentheses as a ';' separated
2838 ##### Example: function 1
2840 func main(av:[ac::number]string; env:[envc::number]string)
2843 or as an indented list of one parameter per line (though each line can
2844 be a ';' separated list)
2846 ##### Example: function 2
2849 argv:[argc::number]string
2850 env:[envc::number]string
2854 In the first case a return type can follow the parentheses after a colon,
2855 in the second it is given on a line starting with the word `return`.
2857 ##### Example: functions that return
2859 func add(a:number; b:number): number
2869 Rather than returning a type, the function can specify a set of local
2870 variables to return as a struct. The values of these variables when the
2871 function exits will be provided to the caller. For this the return type
2872 is replaced with a block of result declarations, either in parentheses
2873 or bracketed by `return` and `do`.
2875 ##### Example: functions returning multiple variables
2877 func to_cartesian(rho:number; theta:number):(x:number; y:number)
2890 For constructing the lists we use a `List` binode, which will be
2891 further detailed when Expression Lists are introduced.
2893 ###### type union fields
2896 struct binode *params;
2897 struct type *return_type;
2898 struct variable *scope;
2899 int inline_result; // return value is at start of 'local'
2903 ###### value union fields
2904 struct exec *function;
2906 ###### type functions
2907 void (*check_args)(struct parse_context *c, int *ok,
2908 struct type *require, struct exec *args);
2910 ###### value functions
2912 static void function_free(struct type *type, struct value *val)
2914 free_exec(val->function);
2915 val->function = NULL;
2918 static int function_compat(struct type *require, struct type *have)
2920 // FIXME can I do anything here yet?
2924 static void function_check_args(struct parse_context *c, int *ok,
2925 struct type *require, struct exec *args)
2927 /* This should be 'compat', but we don't have a 'tuple' type to
2928 * hold the type of 'args'
2930 struct binode *arg = cast(binode, args);
2931 struct binode *param = require->function.params;
2934 struct var *pv = cast(var, param->left);
2936 type_err(c, "error: insufficient arguments to function.",
2937 args, NULL, 0, NULL);
2941 propagate_types(arg->left, c, ok, pv->var->type, 0);
2942 param = cast(binode, param->right);
2943 arg = cast(binode, arg->right);
2946 type_err(c, "error: too many arguments to function.",
2947 args, NULL, 0, NULL);
2950 static void function_print(struct type *type, struct value *val, FILE *f)
2952 print_exec(val->function, 1, 0);
2955 static void function_print_type_decl(struct type *type, FILE *f)
2959 for (b = type->function.params; b; b = cast(binode, b->right)) {
2960 struct variable *v = cast(var, b->left)->var;
2961 fprintf(f, "%.*s%s", v->name->name.len, v->name->name.txt,
2962 v->constant ? "::" : ":");
2963 type_print(v->type, f);
2968 if (type->function.return_type != Tnone) {
2970 if (type->function.inline_result) {
2972 struct type *t = type->function.return_type;
2974 for (i = 0; i < t->structure.nfields; i++) {
2975 struct field *fl = t->structure.fields + i;
2978 fprintf(f, "%.*s:", fl->name.len, fl->name.txt);
2979 type_print(fl->type, f);
2983 type_print(type->function.return_type, f);
2988 static void function_free_type(struct type *t)
2990 free_exec(t->function.params);
2993 static struct type function_prototype = {
2994 .size = sizeof(void*),
2995 .align = sizeof(void*),
2996 .free = function_free,
2997 .compat = function_compat,
2998 .check_args = function_check_args,
2999 .print = function_print,
3000 .print_type_decl = function_print_type_decl,
3001 .free_type = function_free_type,
3004 ###### declare terminals
3014 FuncName -> IDENTIFIER ${ {
3015 struct variable *v = var_decl(c, $1.txt);
3016 struct var *e = new_pos(var, $1);
3022 v = var_ref(c, $1.txt);
3024 type_err(c, "error: function '%v' redeclared",
3026 type_err(c, "info: this is where '%v' was first declared",
3027 v->where_decl, NULL, 0, NULL);
3033 Args -> ArgsLine NEWLINE ${ $0 = $<AL; }$
3034 | Args ArgsLine NEWLINE ${ {
3035 struct binode *b = $<AL;
3036 struct binode **bp = &b;
3038 bp = (struct binode **)&(*bp)->left;
3043 ArgsLine -> ${ $0 = NULL; }$
3044 | Varlist ${ $0 = $<1; }$
3045 | Varlist ; ${ $0 = $<1; }$
3047 Varlist -> Varlist ; ArgDecl ${
3061 ArgDecl -> IDENTIFIER : FormalType ${ {
3062 struct variable *v = var_decl(c, $1.txt);
3068 ##### Function calls
3070 A function call can appear either as an expression or as a statement.
3071 We use a new 'Funcall' binode type to link the function with a list of
3072 arguments, form with the 'List' nodes.
3074 We have already seen the "Term" which is how a function call can appear
3075 in an expression. To parse a function call into a statement we include
3076 it in the "SimpleStatement Grammar" which will be described later.
3082 | Term ( ExpressionList ) ${ {
3083 struct binode *b = new(binode);
3086 b->right = reorder_bilist($<EL);
3090 struct binode *b = new(binode);
3097 ###### SimpleStatement Grammar
3099 | Term ( ExpressionList ) ${ {
3100 struct binode *b = new(binode);
3103 b->right = reorder_bilist($<EL);
3107 ###### print binode cases
3110 do_indent(indent, "");
3111 print_exec(b->left, -1, bracket);
3113 for (b = cast(binode, b->right); b; b = cast(binode, b->right)) {
3116 print_exec(b->left, -1, bracket);
3126 ###### propagate binode cases
3129 /* Every arg must match formal parameter, and result
3130 * is return type of function
3132 struct binode *args = cast(binode, b->right);
3133 struct var *v = cast(var, b->left);
3135 if (!v->var->type || v->var->type->check_args == NULL) {
3136 type_err(c, "error: attempt to call a non-function.",
3137 prog, NULL, 0, NULL);
3140 v->var->type->check_args(c, ok, v->var->type, args);
3141 return v->var->type->function.return_type;
3144 ###### interp binode cases
3147 struct var *v = cast(var, b->left);
3148 struct type *t = v->var->type;
3149 void *oldlocal = c->local;
3150 int old_size = c->local_size;
3151 void *local = calloc(1, t->function.local_size);
3152 struct value *fbody = var_value(c, v->var);
3153 struct binode *arg = cast(binode, b->right);
3154 struct binode *param = t->function.params;
3157 struct var *pv = cast(var, param->left);
3158 struct type *vtype = NULL;
3159 struct value val = interp_exec(c, arg->left, &vtype);
3161 c->local = local; c->local_size = t->function.local_size;
3162 lval = var_value(c, pv->var);
3163 c->local = oldlocal; c->local_size = old_size;
3164 memcpy(lval, &val, vtype->size);
3165 param = cast(binode, param->right);
3166 arg = cast(binode, arg->right);
3168 c->local = local; c->local_size = t->function.local_size;
3169 if (t->function.inline_result && dtype) {
3170 _interp_exec(c, fbody->function, NULL, NULL);
3171 memcpy(dest, local, dtype->size);
3172 rvtype = ret.type = NULL;
3174 rv = interp_exec(c, fbody->function, &rvtype);
3175 c->local = oldlocal; c->local_size = old_size;
3180 ## Complex executables: statements and expressions
3182 Now that we have types and values and variables and most of the basic
3183 Terms which provide access to these, we can explore the more complex
3184 code that combine all of these to get useful work done. Specifically
3185 statements and expressions.
3187 Expressions are various combinations of Terms. We will use operator
3188 precedence to ensure correct parsing. The simplest Expression is just a
3189 Term - others will follow.
3194 Expression -> Term ${ $0 = $<Term; }$
3195 ## expression grammar
3197 ### Expressions: Conditional
3199 Our first user of the `binode` will be conditional expressions, which
3200 is a bit odd as they actually have three components. That will be
3201 handled by having 2 binodes for each expression. The conditional
3202 expression is the lowest precedence operator which is why we define it
3203 first - to start the precedence list.
3205 Conditional expressions are of the form "value `if` condition `else`
3206 other_value". They associate to the right, so everything to the right
3207 of `else` is part of an else value, while only a higher-precedence to
3208 the left of `if` is the if values. Between `if` and `else` there is no
3209 room for ambiguity, so a full conditional expression is allowed in
3215 ###### declare terminals
3219 ###### expression grammar
3221 | Expression if Expression else Expression $$ifelse ${ {
3222 struct binode *b1 = new(binode);
3223 struct binode *b2 = new(binode);
3233 ###### print binode cases
3236 b2 = cast(binode, b->right);
3237 if (bracket) printf("(");
3238 print_exec(b2->left, -1, bracket);
3240 print_exec(b->left, -1, bracket);
3242 print_exec(b2->right, -1, bracket);
3243 if (bracket) printf(")");
3246 ###### propagate binode cases
3249 /* cond must be Tbool, others must match */
3250 struct binode *b2 = cast(binode, b->right);
3253 propagate_types(b->left, c, ok, Tbool, 0);
3254 t = propagate_types(b2->left, c, ok, type, Rnolabel);
3255 t2 = propagate_types(b2->right, c, ok, type ?: t, Rnolabel);
3259 ###### interp binode cases
3262 struct binode *b2 = cast(binode, b->right);
3263 left = interp_exec(c, b->left, <ype);
3265 rv = interp_exec(c, b2->left, &rvtype); // UNTESTED
3267 rv = interp_exec(c, b2->right, &rvtype);
3273 We take a brief detour, now that we have expressions, to describe lists
3274 of expressions. These will be needed for function parameters and
3275 possibly other situations. They seem generic enough to introduce here
3276 to be used elsewhere.
3278 And ExpressionList will use the `List` type of `binode`, building up at
3279 the end. And place where they are used will probably call
3280 `reorder_bilist()` to get a more normal first/next arrangement.
3282 ###### declare terminals
3285 `List` execs have no implicit semantics, so they are never propagated or
3286 interpreted. The can be printed as a comma separate list, which is how
3287 they are parsed. Note they are also used for function formal parameter
3288 lists. In that case a separate function is used to print them.
3290 ###### print binode cases
3294 print_exec(b->left, -1, bracket);
3297 b = cast(binode, b->right);
3301 ###### propagate binode cases
3302 case List: abort(); // NOTEST
3303 ###### interp binode cases
3304 case List: abort(); // NOTEST
3309 ExpressionList -> ExpressionList , Expression ${
3322 ### Expressions: Boolean
3324 The next class of expressions to use the `binode` will be Boolean
3325 expressions. "`and then`" and "`or else`" are similar to `and` and `or`
3326 have same corresponding precendence. The difference is that they don't
3327 evaluate the second expression if not necessary.
3336 ###### declare terminals
3341 ###### expression grammar
3342 | Expression or Expression ${ {
3343 struct binode *b = new(binode);
3349 | Expression or else Expression ${ {
3350 struct binode *b = new(binode);
3357 | Expression and Expression ${ {
3358 struct binode *b = new(binode);
3364 | Expression and then Expression ${ {
3365 struct binode *b = new(binode);
3372 | not Expression ${ {
3373 struct binode *b = new(binode);
3379 ###### print binode cases
3381 if (bracket) printf("(");
3382 print_exec(b->left, -1, bracket);
3384 print_exec(b->right, -1, bracket);
3385 if (bracket) printf(")");
3388 if (bracket) printf("(");
3389 print_exec(b->left, -1, bracket);
3390 printf(" and then ");
3391 print_exec(b->right, -1, bracket);
3392 if (bracket) printf(")");
3395 if (bracket) printf("(");
3396 print_exec(b->left, -1, bracket);
3398 print_exec(b->right, -1, bracket);
3399 if (bracket) printf(")");
3402 if (bracket) printf("(");
3403 print_exec(b->left, -1, bracket);
3404 printf(" or else ");
3405 print_exec(b->right, -1, bracket);
3406 if (bracket) printf(")");
3409 if (bracket) printf("(");
3411 print_exec(b->right, -1, bracket);
3412 if (bracket) printf(")");
3415 ###### propagate binode cases
3421 /* both must be Tbool, result is Tbool */
3422 propagate_types(b->left, c, ok, Tbool, 0);
3423 propagate_types(b->right, c, ok, Tbool, 0);
3424 if (type && type != Tbool)
3425 type_err(c, "error: %1 operation found where %2 expected", prog,
3429 ###### interp binode cases
3431 rv = interp_exec(c, b->left, &rvtype);
3432 right = interp_exec(c, b->right, &rtype);
3433 rv.bool = rv.bool && right.bool;
3436 rv = interp_exec(c, b->left, &rvtype);
3438 rv = interp_exec(c, b->right, NULL);
3441 rv = interp_exec(c, b->left, &rvtype);
3442 right = interp_exec(c, b->right, &rtype);
3443 rv.bool = rv.bool || right.bool;
3446 rv = interp_exec(c, b->left, &rvtype);
3448 rv = interp_exec(c, b->right, NULL);
3451 rv = interp_exec(c, b->right, &rvtype);
3455 ### Expressions: Comparison
3457 Of slightly higher precedence that Boolean expressions are Comparisons.
3458 A comparison takes arguments of any comparable type, but the two types
3461 To simplify the parsing we introduce an `eop` which can record an
3462 expression operator, and the `CMPop` non-terminal will match one of them.
3469 ###### ast functions
3470 static void free_eop(struct eop *e)
3484 ###### declare terminals
3485 $LEFT < > <= >= == != CMPop
3487 ###### expression grammar
3488 | Expression CMPop Expression ${ {
3489 struct binode *b = new(binode);
3499 CMPop -> < ${ $0.op = Less; }$
3500 | > ${ $0.op = Gtr; }$
3501 | <= ${ $0.op = LessEq; }$
3502 | >= ${ $0.op = GtrEq; }$
3503 | == ${ $0.op = Eql; }$
3504 | != ${ $0.op = NEql; }$
3506 ###### print binode cases
3514 if (bracket) printf("(");
3515 print_exec(b->left, -1, bracket);
3517 case Less: printf(" < "); break;
3518 case LessEq: printf(" <= "); break;
3519 case Gtr: printf(" > "); break;
3520 case GtrEq: printf(" >= "); break;
3521 case Eql: printf(" == "); break;
3522 case NEql: printf(" != "); break;
3523 default: abort(); // NOTEST
3525 print_exec(b->right, -1, bracket);
3526 if (bracket) printf(")");
3529 ###### propagate binode cases
3536 /* Both must match but not be labels, result is Tbool */
3537 t = propagate_types(b->left, c, ok, NULL, Rnolabel);
3539 propagate_types(b->right, c, ok, t, 0);
3541 t = propagate_types(b->right, c, ok, NULL, Rnolabel); // UNTESTED
3543 t = propagate_types(b->left, c, ok, t, 0); // UNTESTED
3545 if (!type_compat(type, Tbool, 0))
3546 type_err(c, "error: Comparison returns %1 but %2 expected", prog,
3547 Tbool, rules, type);
3550 ###### interp binode cases
3559 left = interp_exec(c, b->left, <ype);
3560 right = interp_exec(c, b->right, &rtype);
3561 cmp = value_cmp(ltype, rtype, &left, &right);
3564 case Less: rv.bool = cmp < 0; break;
3565 case LessEq: rv.bool = cmp <= 0; break;
3566 case Gtr: rv.bool = cmp > 0; break;
3567 case GtrEq: rv.bool = cmp >= 0; break;
3568 case Eql: rv.bool = cmp == 0; break;
3569 case NEql: rv.bool = cmp != 0; break;
3570 default: rv.bool = 0; break; // NOTEST
3575 ### Expressions: Arithmetic etc.
3577 The remaining expressions with the highest precedence are arithmetic,
3578 string concatenation, and string conversion. String concatenation
3579 (`++`) has the same precedence as multiplication and division, but lower
3582 String conversion is a temporary feature until I get a better type
3583 system. `$` is a prefix operator which expects a string and returns
3586 `+` and `-` are both infix and prefix operations (where they are
3587 absolute value and negation). These have different operator names.
3589 We also have a 'Bracket' operator which records where parentheses were
3590 found. This makes it easy to reproduce these when printing. Possibly I
3591 should only insert brackets were needed for precedence. Putting
3592 parentheses around an expression converts it into a Term,
3602 ###### declare terminals
3608 ###### expression grammar
3609 | Expression Eop Expression ${ {
3610 struct binode *b = new(binode);
3617 | Expression Top Expression ${ {
3618 struct binode *b = new(binode);
3625 | Uop Expression ${ {
3626 struct binode *b = new(binode);
3634 | ( Expression ) ${ {
3635 struct binode *b = new_pos(binode, $1);
3644 Eop -> + ${ $0.op = Plus; }$
3645 | - ${ $0.op = Minus; }$
3647 Uop -> + ${ $0.op = Absolute; }$
3648 | - ${ $0.op = Negate; }$
3649 | $ ${ $0.op = StringConv; }$
3651 Top -> * ${ $0.op = Times; }$
3652 | / ${ $0.op = Divide; }$
3653 | % ${ $0.op = Rem; }$
3654 | ++ ${ $0.op = Concat; }$
3656 ###### print binode cases
3663 if (bracket) printf("(");
3664 print_exec(b->left, indent, bracket);
3666 case Plus: fputs(" + ", stdout); break;
3667 case Minus: fputs(" - ", stdout); break;
3668 case Times: fputs(" * ", stdout); break;
3669 case Divide: fputs(" / ", stdout); break;
3670 case Rem: fputs(" % ", stdout); break;
3671 case Concat: fputs(" ++ ", stdout); break;
3672 default: abort(); // NOTEST
3674 print_exec(b->right, indent, bracket);
3675 if (bracket) printf(")");
3680 if (bracket) printf("(");
3682 case Absolute: fputs("+", stdout); break;
3683 case Negate: fputs("-", stdout); break;
3684 case StringConv: fputs("$", stdout); break;
3685 default: abort(); // NOTEST
3687 print_exec(b->right, indent, bracket);
3688 if (bracket) printf(")");
3692 print_exec(b->right, indent, bracket);
3696 ###### propagate binode cases
3702 /* both must be numbers, result is Tnum */
3705 /* as propagate_types ignores a NULL,
3706 * unary ops fit here too */
3707 propagate_types(b->left, c, ok, Tnum, 0);
3708 propagate_types(b->right, c, ok, Tnum, 0);
3709 if (!type_compat(type, Tnum, 0))
3710 type_err(c, "error: Arithmetic returns %1 but %2 expected", prog,
3715 /* both must be Tstr, result is Tstr */
3716 propagate_types(b->left, c, ok, Tstr, 0);
3717 propagate_types(b->right, c, ok, Tstr, 0);
3718 if (!type_compat(type, Tstr, 0))
3719 type_err(c, "error: Concat returns %1 but %2 expected", prog,
3724 /* op must be string, result is number */
3725 propagate_types(b->left, c, ok, Tstr, 0);
3726 if (!type_compat(type, Tnum, 0))
3727 type_err(c, // UNTESTED
3728 "error: Can only convert string to number, not %1",
3729 prog, type, 0, NULL);
3733 return propagate_types(b->right, c, ok, type, 0);
3735 ###### interp binode cases
3738 rv = interp_exec(c, b->left, &rvtype);
3739 right = interp_exec(c, b->right, &rtype);
3740 mpq_add(rv.num, rv.num, right.num);
3743 rv = interp_exec(c, b->left, &rvtype);
3744 right = interp_exec(c, b->right, &rtype);
3745 mpq_sub(rv.num, rv.num, right.num);
3748 rv = interp_exec(c, b->left, &rvtype);
3749 right = interp_exec(c, b->right, &rtype);
3750 mpq_mul(rv.num, rv.num, right.num);
3753 rv = interp_exec(c, b->left, &rvtype);
3754 right = interp_exec(c, b->right, &rtype);
3755 mpq_div(rv.num, rv.num, right.num);
3760 left = interp_exec(c, b->left, <ype);
3761 right = interp_exec(c, b->right, &rtype);
3762 mpz_init(l); mpz_init(r); mpz_init(rem);
3763 mpz_tdiv_q(l, mpq_numref(left.num), mpq_denref(left.num));
3764 mpz_tdiv_q(r, mpq_numref(right.num), mpq_denref(right.num));
3765 mpz_tdiv_r(rem, l, r);
3766 val_init(Tnum, &rv);
3767 mpq_set_z(rv.num, rem);
3768 mpz_clear(r); mpz_clear(l); mpz_clear(rem);
3773 rv = interp_exec(c, b->right, &rvtype);
3774 mpq_neg(rv.num, rv.num);
3777 rv = interp_exec(c, b->right, &rvtype);
3778 mpq_abs(rv.num, rv.num);
3781 rv = interp_exec(c, b->right, &rvtype);
3784 left = interp_exec(c, b->left, <ype);
3785 right = interp_exec(c, b->right, &rtype);
3787 rv.str = text_join(left.str, right.str);
3790 right = interp_exec(c, b->right, &rvtype);
3794 struct text tx = right.str;
3797 if (tx.txt[0] == '-') {
3798 neg = 1; // UNTESTED
3799 tx.txt++; // UNTESTED
3800 tx.len--; // UNTESTED
3802 if (number_parse(rv.num, tail, tx) == 0)
3803 mpq_init(rv.num); // UNTESTED
3805 mpq_neg(rv.num, rv.num); // UNTESTED
3807 printf("Unsupported suffix: %.*s\n", tx.len, tx.txt); // UNTESTED
3811 ###### value functions
3813 static struct text text_join(struct text a, struct text b)
3816 rv.len = a.len + b.len;
3817 rv.txt = malloc(rv.len);
3818 memcpy(rv.txt, a.txt, a.len);
3819 memcpy(rv.txt+a.len, b.txt, b.len);
3823 ### Blocks, Statements, and Statement lists.
3825 Now that we have expressions out of the way we need to turn to
3826 statements. There are simple statements and more complex statements.
3827 Simple statements do not contain (syntactic) newlines, complex statements do.
3829 Statements often come in sequences and we have corresponding simple
3830 statement lists and complex statement lists.
3831 The former comprise only simple statements separated by semicolons.
3832 The later comprise complex statements and simple statement lists. They are
3833 separated by newlines. Thus the semicolon is only used to separate
3834 simple statements on the one line. This may be overly restrictive,
3835 but I'm not sure I ever want a complex statement to share a line with
3838 Note that a simple statement list can still use multiple lines if
3839 subsequent lines are indented, so
3841 ###### Example: wrapped simple statement list
3846 is a single simple statement list. This might allow room for
3847 confusion, so I'm not set on it yet.
3849 A simple statement list needs no extra syntax. A complex statement
3850 list has two syntactic forms. It can be enclosed in braces (much like
3851 C blocks), or it can be introduced by an indent and continue until an
3852 unindented newline (much like Python blocks). With this extra syntax
3853 it is referred to as a block.
3855 Note that a block does not have to include any newlines if it only
3856 contains simple statements. So both of:
3858 if condition: a=b; d=f
3860 if condition { a=b; print f }
3864 In either case the list is constructed from a `binode` list with
3865 `Block` as the operator. When parsing the list it is most convenient
3866 to append to the end, so a list is a list and a statement. When using
3867 the list it is more convenient to consider a list to be a statement
3868 and a list. So we need a function to re-order a list.
3869 `reorder_bilist` serves this purpose.
3871 The only stand-alone statement we introduce at this stage is `pass`
3872 which does nothing and is represented as a `NULL` pointer in a `Block`
3873 list. Other stand-alone statements will follow once the infrastructure
3876 As many statements will use binodes, we declare a binode pointer 'b' in
3877 the common header for all reductions to use.
3879 ###### Parser: reduce
3890 Block -> { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3891 | { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3892 | SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
3893 | SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
3894 | IN OptNL Statementlist OUT ${ $0 = $<Sl; }$
3896 OpenBlock -> OpenScope { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3897 | OpenScope { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3898 | OpenScope SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
3899 | OpenScope SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
3900 | IN OpenScope OptNL Statementlist OUT ${ $0 = $<Sl; }$
3902 UseBlock -> { OpenScope IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3903 | { OpenScope SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3904 | IN OpenScope OptNL Statementlist OUT ${ $0 = $<Sl; }$
3906 ColonBlock -> { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3907 | { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3908 | : SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
3909 | : SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
3910 | : IN OptNL Statementlist OUT ${ $0 = $<Sl; }$
3912 Statementlist -> ComplexStatements ${ $0 = reorder_bilist($<CS); }$
3914 ComplexStatements -> ComplexStatements ComplexStatement ${
3924 | ComplexStatement ${
3936 ComplexStatement -> SimpleStatements Newlines ${
3937 $0 = reorder_bilist($<SS);
3939 | SimpleStatements ; Newlines ${
3940 $0 = reorder_bilist($<SS);
3942 ## ComplexStatement Grammar
3945 SimpleStatements -> SimpleStatements ; SimpleStatement ${
3951 | SimpleStatement ${
3960 SimpleStatement -> pass ${ $0 = NULL; }$
3961 | ERROR ${ tok_err(c, "Syntax error in statement", &$1); }$
3962 ## SimpleStatement Grammar
3964 ###### print binode cases
3968 if (b->left == NULL) // UNTESTED
3969 printf("pass"); // UNTESTED
3971 print_exec(b->left, indent, bracket); // UNTESTED
3972 if (b->right) { // UNTESTED
3973 printf("; "); // UNTESTED
3974 print_exec(b->right, indent, bracket); // UNTESTED
3977 // block, one per line
3978 if (b->left == NULL)
3979 do_indent(indent, "pass\n");
3981 print_exec(b->left, indent, bracket);
3983 print_exec(b->right, indent, bracket);
3987 ###### propagate binode cases
3990 /* If any statement returns something other than Tnone
3991 * or Tbool then all such must return same type.
3992 * As each statement may be Tnone or something else,
3993 * we must always pass NULL (unknown) down, otherwise an incorrect
3994 * error might occur. We never return Tnone unless it is
3999 for (e = b; e; e = cast(binode, e->right)) {
4000 t = propagate_types(e->left, c, ok, NULL, rules);
4001 if ((rules & Rboolok) && (t == Tbool || t == Tnone))
4003 if (t == Tnone && e->right)
4004 /* Only the final statement *must* return a value
4012 type_err(c, "error: expected %1%r, found %2",
4013 e->left, type, rules, t);
4019 ###### interp binode cases
4021 while (rvtype == Tnone &&
4024 rv = interp_exec(c, b->left, &rvtype);
4025 b = cast(binode, b->right);
4029 ### The Print statement
4031 `print` is a simple statement that takes a comma-separated list of
4032 expressions and prints the values separated by spaces and terminated
4033 by a newline. No control of formatting is possible.
4035 `print` uses `ExpressionList` to collect the expressions and stores them
4036 on the left side of a `Print` binode unlessthere is a trailing comma
4037 when the list is stored on the `right` side and no trailing newline is
4043 ##### declare terminals
4046 ###### SimpleStatement Grammar
4048 | print ExpressionList ${
4049 $0 = b = new(binode);
4052 b->left = reorder_bilist($<EL);
4054 | print ExpressionList , ${ {
4055 $0 = b = new(binode);
4057 b->right = reorder_bilist($<EL);
4061 $0 = b = new(binode);
4067 ###### print binode cases
4070 do_indent(indent, "print");
4072 print_exec(b->right, -1, bracket);
4075 print_exec(b->left, -1, bracket);
4080 ###### propagate binode cases
4083 /* don't care but all must be consistent */
4085 b = cast(binode, b->left);
4087 b = cast(binode, b->right);
4089 propagate_types(b->left, c, ok, NULL, Rnolabel);
4090 b = cast(binode, b->right);
4094 ###### interp binode cases
4098 struct binode *b2 = cast(binode, b->left);
4100 b2 = cast(binode, b->right);
4101 for (; b2; b2 = cast(binode, b2->right)) {
4102 left = interp_exec(c, b2->left, <ype);
4103 print_value(ltype, &left, stdout);
4104 free_value(ltype, &left);
4108 if (b->right == NULL)
4114 ###### Assignment statement
4116 An assignment will assign a value to a variable, providing it hasn't
4117 been declared as a constant. The analysis phase ensures that the type
4118 will be correct so the interpreter just needs to perform the
4119 calculation. There is a form of assignment which declares a new
4120 variable as well as assigning a value. If a name is assigned before
4121 it is declared, and error will be raised as the name is created as
4122 `Tlabel` and it is illegal to assign to such names.
4128 ###### declare terminals
4131 ###### SimpleStatement Grammar
4132 | Term = Expression ${
4133 $0 = b= new(binode);
4138 | VariableDecl = Expression ${
4139 $0 = b= new(binode);
4146 if ($1->var->where_set == NULL) {
4148 "Variable declared with no type or value: %v",
4152 $0 = b = new(binode);
4159 ###### print binode cases
4162 do_indent(indent, "");
4163 print_exec(b->left, indent, bracket);
4165 print_exec(b->right, indent, bracket);
4172 struct variable *v = cast(var, b->left)->var;
4173 do_indent(indent, "");
4174 print_exec(b->left, indent, bracket);
4175 if (cast(var, b->left)->var->constant) {
4177 if (v->explicit_type) {
4178 type_print(v->type, stdout);
4183 if (v->explicit_type) {
4184 type_print(v->type, stdout);
4190 print_exec(b->right, indent, bracket);
4197 ###### propagate binode cases
4201 /* Both must match and not be labels,
4202 * Type must support 'dup',
4203 * For Assign, left must not be constant.
4206 t = propagate_types(b->left, c, ok, NULL,
4207 Rnolabel | (b->op == Assign ? Rnoconstant : 0));
4212 if (propagate_types(b->right, c, ok, t, 0) != t)
4213 if (b->left->type == Xvar)
4214 type_err(c, "info: variable '%v' was set as %1 here.",
4215 cast(var, b->left)->var->where_set, t, rules, NULL);
4217 t = propagate_types(b->right, c, ok, NULL, Rnolabel);
4219 propagate_types(b->left, c, ok, t,
4220 (b->op == Assign ? Rnoconstant : 0));
4222 if (t && t->dup == NULL && t->name.txt[0] != ' ') // HACK
4223 type_err(c, "error: cannot assign value of type %1", b, t, 0, NULL);
4228 ###### interp binode cases
4231 lleft = linterp_exec(c, b->left, <ype);
4233 dinterp_exec(c, b->right, lleft, ltype, 1);
4239 struct variable *v = cast(var, b->left)->var;
4242 val = var_value(c, v);
4243 if (v->type->prepare_type)
4244 v->type->prepare_type(c, v->type, 0);
4246 dinterp_exec(c, b->right, val, v->type, 0);
4248 val_init(v->type, val);
4252 ### The `use` statement
4254 The `use` statement is the last "simple" statement. It is needed when a
4255 statement block can return a value. This includes the body of a
4256 function which has a return type, and the "condition" code blocks in
4257 `if`, `while`, and `switch` statements.
4262 ###### declare terminals
4265 ###### SimpleStatement Grammar
4267 $0 = b = new_pos(binode, $1);
4270 if (b->right->type == Xvar) {
4271 struct var *v = cast(var, b->right);
4272 if (v->var->type == Tnone) {
4273 /* Convert this to a label */
4276 v->var->type = Tlabel;
4277 val = global_alloc(c, Tlabel, v->var, NULL);
4283 ###### print binode cases
4286 do_indent(indent, "use ");
4287 print_exec(b->right, -1, bracket);
4292 ###### propagate binode cases
4295 /* result matches value */
4296 return propagate_types(b->right, c, ok, type, 0);
4298 ###### interp binode cases
4301 rv = interp_exec(c, b->right, &rvtype);
4304 ### The Conditional Statement
4306 This is the biggy and currently the only complex statement. This
4307 subsumes `if`, `while`, `do/while`, `switch`, and some parts of `for`.
4308 It is comprised of a number of parts, all of which are optional though
4309 set combinations apply. Each part is (usually) a key word (`then` is
4310 sometimes optional) followed by either an expression or a code block,
4311 except the `casepart` which is a "key word and an expression" followed
4312 by a code block. The code-block option is valid for all parts and,
4313 where an expression is also allowed, the code block can use the `use`
4314 statement to report a value. If the code block does not report a value
4315 the effect is similar to reporting `True`.
4317 The `else` and `case` parts, as well as `then` when combined with
4318 `if`, can contain a `use` statement which will apply to some
4319 containing conditional statement. `for` parts, `do` parts and `then`
4320 parts used with `for` can never contain a `use`, except in some
4321 subordinate conditional statement.
4323 If there is a `forpart`, it is executed first, only once.
4324 If there is a `dopart`, then it is executed repeatedly providing
4325 always that the `condpart` or `cond`, if present, does not return a non-True
4326 value. `condpart` can fail to return any value if it simply executes
4327 to completion. This is treated the same as returning `True`.
4329 If there is a `thenpart` it will be executed whenever the `condpart`
4330 or `cond` returns True (or does not return any value), but this will happen
4331 *after* `dopart` (when present).
4333 If `elsepart` is present it will be executed at most once when the
4334 condition returns `False` or some value that isn't `True` and isn't
4335 matched by any `casepart`. If there are any `casepart`s, they will be
4336 executed when the condition returns a matching value.
4338 The particular sorts of values allowed in case parts has not yet been
4339 determined in the language design, so nothing is prohibited.
4341 The various blocks in this complex statement potentially provide scope
4342 for variables as described earlier. Each such block must include the
4343 "OpenScope" nonterminal before parsing the block, and must call
4344 `var_block_close()` when closing the block.
4346 The code following "`if`", "`switch`" and "`for`" does not get its own
4347 scope, but is in a scope covering the whole statement, so names
4348 declared there cannot be redeclared elsewhere. Similarly the
4349 condition following "`while`" is in a scope the covers the body
4350 ("`do`" part) of the loop, and which does not allow conditional scope
4351 extension. Code following "`then`" (both looping and non-looping),
4352 "`else`" and "`case`" each get their own local scope.
4354 The type requirements on the code block in a `whilepart` are quite
4355 unusal. It is allowed to return a value of some identifiable type, in
4356 which case the loop aborts and an appropriate `casepart` is run, or it
4357 can return a Boolean, in which case the loop either continues to the
4358 `dopart` (on `True`) or aborts and runs the `elsepart` (on `False`).
4359 This is different both from the `ifpart` code block which is expected to
4360 return a Boolean, or the `switchpart` code block which is expected to
4361 return the same type as the casepart values. The correct analysis of
4362 the type of the `whilepart` code block is the reason for the
4363 `Rboolok` flag which is passed to `propagate_types()`.
4365 The `cond_statement` cannot fit into a `binode` so a new `exec` is
4366 defined. As there are two scopes which cover multiple parts - one for
4367 the whole statement and one for "while" and "do" - and as we will use
4368 the 'struct exec' to track scopes, we actually need two new types of
4369 exec. One is a `binode` for the looping part, the rest is the
4370 `cond_statement`. The `cond_statement` will use an auxilliary `struct
4371 casepart` to track a list of case parts.
4382 struct exec *action;
4383 struct casepart *next;
4385 struct cond_statement {
4387 struct exec *forpart, *condpart, *thenpart, *elsepart;
4388 struct binode *looppart;
4389 struct casepart *casepart;
4392 ###### ast functions
4394 static void free_casepart(struct casepart *cp)
4398 free_exec(cp->value);
4399 free_exec(cp->action);
4406 static void free_cond_statement(struct cond_statement *s)
4410 free_exec(s->forpart);
4411 free_exec(s->condpart);
4412 free_exec(s->looppart);
4413 free_exec(s->thenpart);
4414 free_exec(s->elsepart);
4415 free_casepart(s->casepart);
4419 ###### free exec cases
4420 case Xcond_statement: free_cond_statement(cast(cond_statement, e)); break;
4422 ###### ComplexStatement Grammar
4423 | CondStatement ${ $0 = $<1; }$
4425 ###### declare terminals
4426 $TERM for then while do
4433 // A CondStatement must end with EOL, as does CondSuffix and
4435 // ForPart, ThenPart, SwitchPart, CasePart are non-empty and
4436 // may or may not end with EOL
4437 // WhilePart and IfPart include an appropriate Suffix
4439 // ForPart, SwitchPart, and IfPart open scopes, o we have to close
4440 // them. WhilePart opens and closes its own scope.
4441 CondStatement -> ForPart OptNL ThenPart OptNL WhilePart CondSuffix ${
4444 $0->thenpart = $<TP;
4445 $0->looppart = $<WP;
4446 var_block_close(c, CloseSequential, $0);
4448 | ForPart OptNL WhilePart CondSuffix ${
4451 $0->looppart = $<WP;
4452 var_block_close(c, CloseSequential, $0);
4454 | WhilePart CondSuffix ${
4456 $0->looppart = $<WP;
4458 | SwitchPart OptNL CasePart CondSuffix ${
4460 $0->condpart = $<SP;
4461 $CP->next = $0->casepart;
4462 $0->casepart = $<CP;
4463 var_block_close(c, CloseSequential, $0);
4465 | SwitchPart : IN OptNL CasePart CondSuffix OUT Newlines ${
4467 $0->condpart = $<SP;
4468 $CP->next = $0->casepart;
4469 $0->casepart = $<CP;
4470 var_block_close(c, CloseSequential, $0);
4472 | IfPart IfSuffix ${
4474 $0->condpart = $IP.condpart; $IP.condpart = NULL;
4475 $0->thenpart = $IP.thenpart; $IP.thenpart = NULL;
4476 // This is where we close an "if" statement
4477 var_block_close(c, CloseSequential, $0);
4480 CondSuffix -> IfSuffix ${
4483 | Newlines CasePart CondSuffix ${
4485 $CP->next = $0->casepart;
4486 $0->casepart = $<CP;
4488 | CasePart CondSuffix ${
4490 $CP->next = $0->casepart;
4491 $0->casepart = $<CP;
4494 IfSuffix -> Newlines ${ $0 = new(cond_statement); }$
4495 | Newlines ElsePart ${ $0 = $<EP; }$
4496 | ElsePart ${$0 = $<EP; }$
4498 ElsePart -> else OpenBlock Newlines ${
4499 $0 = new(cond_statement);
4500 $0->elsepart = $<OB;
4501 var_block_close(c, CloseElse, $0->elsepart);
4503 | else OpenScope CondStatement ${
4504 $0 = new(cond_statement);
4505 $0->elsepart = $<CS;
4506 var_block_close(c, CloseElse, $0->elsepart);
4510 CasePart -> case Expression OpenScope ColonBlock ${
4511 $0 = calloc(1,sizeof(struct casepart));
4514 var_block_close(c, CloseParallel, $0->action);
4518 // These scopes are closed in CondStatement
4519 ForPart -> for OpenBlock ${
4523 ThenPart -> then OpenBlock ${
4525 var_block_close(c, CloseSequential, $0);
4529 // This scope is closed in CondStatement
4530 WhilePart -> while UseBlock OptNL do OpenBlock ${
4535 var_block_close(c, CloseSequential, $0->right);
4536 var_block_close(c, CloseSequential, $0);
4538 | while OpenScope Expression OpenScope ColonBlock ${
4543 var_block_close(c, CloseSequential, $0->right);
4544 var_block_close(c, CloseSequential, $0);
4548 IfPart -> if UseBlock OptNL then OpenBlock ${
4551 var_block_close(c, CloseParallel, $0.thenpart);
4553 | if OpenScope Expression OpenScope ColonBlock ${
4556 var_block_close(c, CloseParallel, $0.thenpart);
4558 | if OpenScope Expression OpenScope OptNL then Block ${
4561 var_block_close(c, CloseParallel, $0.thenpart);
4565 // This scope is closed in CondStatement
4566 SwitchPart -> switch OpenScope Expression ${
4569 | switch UseBlock ${
4573 ###### print binode cases
4575 if (b->left && b->left->type == Xbinode &&
4576 cast(binode, b->left)->op == Block) {
4578 do_indent(indent, "while {\n");
4580 do_indent(indent, "while\n");
4581 print_exec(b->left, indent+1, bracket);
4583 do_indent(indent, "} do {\n");
4585 do_indent(indent, "do\n");
4586 print_exec(b->right, indent+1, bracket);
4588 do_indent(indent, "}\n");
4590 do_indent(indent, "while ");
4591 print_exec(b->left, 0, bracket);
4596 print_exec(b->right, indent+1, bracket);
4598 do_indent(indent, "}\n");
4602 ###### print exec cases
4604 case Xcond_statement:
4606 struct cond_statement *cs = cast(cond_statement, e);
4607 struct casepart *cp;
4609 do_indent(indent, "for");
4610 if (bracket) printf(" {\n"); else printf("\n");
4611 print_exec(cs->forpart, indent+1, bracket);
4614 do_indent(indent, "} then {\n");
4616 do_indent(indent, "then\n");
4617 print_exec(cs->thenpart, indent+1, bracket);
4619 if (bracket) do_indent(indent, "}\n");
4622 print_exec(cs->looppart, indent, bracket);
4626 do_indent(indent, "switch");
4628 do_indent(indent, "if");
4629 if (cs->condpart && cs->condpart->type == Xbinode &&
4630 cast(binode, cs->condpart)->op == Block) {
4635 print_exec(cs->condpart, indent+1, bracket);
4637 do_indent(indent, "}\n");
4639 do_indent(indent, "then\n");
4640 print_exec(cs->thenpart, indent+1, bracket);
4644 print_exec(cs->condpart, 0, bracket);
4650 print_exec(cs->thenpart, indent+1, bracket);
4652 do_indent(indent, "}\n");
4657 for (cp = cs->casepart; cp; cp = cp->next) {
4658 do_indent(indent, "case ");
4659 print_exec(cp->value, -1, 0);
4664 print_exec(cp->action, indent+1, bracket);
4666 do_indent(indent, "}\n");
4669 do_indent(indent, "else");
4674 print_exec(cs->elsepart, indent+1, bracket);
4676 do_indent(indent, "}\n");
4681 ###### propagate binode cases
4683 t = propagate_types(b->right, c, ok, Tnone, 0);
4684 if (!type_compat(Tnone, t, 0))
4685 *ok = 0; // UNTESTED
4686 return propagate_types(b->left, c, ok, type, rules);
4688 ###### propagate exec cases
4689 case Xcond_statement:
4691 // forpart and looppart->right must return Tnone
4692 // thenpart must return Tnone if there is a loopart,
4693 // otherwise it is like elsepart.
4695 // be bool if there is no casepart
4696 // match casepart->values if there is a switchpart
4697 // either be bool or match casepart->value if there
4699 // elsepart and casepart->action must match the return type
4700 // expected of this statement.
4701 struct cond_statement *cs = cast(cond_statement, prog);
4702 struct casepart *cp;
4704 t = propagate_types(cs->forpart, c, ok, Tnone, 0);
4705 if (!type_compat(Tnone, t, 0))
4706 *ok = 0; // UNTESTED
4709 t = propagate_types(cs->thenpart, c, ok, Tnone, 0);
4710 if (!type_compat(Tnone, t, 0))
4711 *ok = 0; // UNTESTED
4713 if (cs->casepart == NULL) {
4714 propagate_types(cs->condpart, c, ok, Tbool, 0);
4715 propagate_types(cs->looppart, c, ok, Tbool, 0);
4717 /* Condpart must match case values, with bool permitted */
4719 for (cp = cs->casepart;
4720 cp && !t; cp = cp->next)
4721 t = propagate_types(cp->value, c, ok, NULL, 0);
4722 if (!t && cs->condpart)
4723 t = propagate_types(cs->condpart, c, ok, NULL, Rboolok); // UNTESTED
4724 if (!t && cs->looppart)
4725 t = propagate_types(cs->looppart, c, ok, NULL, Rboolok); // UNTESTED
4726 // Now we have a type (I hope) push it down
4728 for (cp = cs->casepart; cp; cp = cp->next)
4729 propagate_types(cp->value, c, ok, t, 0);
4730 propagate_types(cs->condpart, c, ok, t, Rboolok);
4731 propagate_types(cs->looppart, c, ok, t, Rboolok);
4734 // (if)then, else, and case parts must return expected type.
4735 if (!cs->looppart && !type)
4736 type = propagate_types(cs->thenpart, c, ok, NULL, rules);
4738 type = propagate_types(cs->elsepart, c, ok, NULL, rules);
4739 for (cp = cs->casepart;
4741 cp = cp->next) // UNTESTED
4742 type = propagate_types(cp->action, c, ok, NULL, rules); // UNTESTED
4745 propagate_types(cs->thenpart, c, ok, type, rules);
4746 propagate_types(cs->elsepart, c, ok, type, rules);
4747 for (cp = cs->casepart; cp ; cp = cp->next)
4748 propagate_types(cp->action, c, ok, type, rules);
4754 ###### interp binode cases
4756 // This just performs one iterration of the loop
4757 rv = interp_exec(c, b->left, &rvtype);
4758 if (rvtype == Tnone ||
4759 (rvtype == Tbool && rv.bool != 0))
4760 // rvtype is Tnone or Tbool, doesn't need to be freed
4761 interp_exec(c, b->right, NULL);
4764 ###### interp exec cases
4765 case Xcond_statement:
4767 struct value v, cnd;
4768 struct type *vtype, *cndtype;
4769 struct casepart *cp;
4770 struct cond_statement *cs = cast(cond_statement, e);
4773 interp_exec(c, cs->forpart, NULL);
4775 while ((cnd = interp_exec(c, cs->looppart, &cndtype)),
4776 cndtype == Tnone || (cndtype == Tbool && cnd.bool != 0))
4777 interp_exec(c, cs->thenpart, NULL);
4779 cnd = interp_exec(c, cs->condpart, &cndtype);
4780 if ((cndtype == Tnone ||
4781 (cndtype == Tbool && cnd.bool != 0))) {
4782 // cnd is Tnone or Tbool, doesn't need to be freed
4783 rv = interp_exec(c, cs->thenpart, &rvtype);
4784 // skip else (and cases)
4788 for (cp = cs->casepart; cp; cp = cp->next) {
4789 v = interp_exec(c, cp->value, &vtype);
4790 if (value_cmp(cndtype, vtype, &v, &cnd) == 0) {
4791 free_value(vtype, &v);
4792 free_value(cndtype, &cnd);
4793 rv = interp_exec(c, cp->action, &rvtype);
4796 free_value(vtype, &v);
4798 free_value(cndtype, &cnd);
4800 rv = interp_exec(c, cs->elsepart, &rvtype);
4807 ### Top level structure
4809 All the language elements so far can be used in various places. Now
4810 it is time to clarify what those places are.
4812 At the top level of a file there will be a number of declarations.
4813 Many of the things that can be declared haven't been described yet,
4814 such as functions, procedures, imports, and probably more.
4815 For now there are two sorts of things that can appear at the top
4816 level. They are predefined constants, `struct` types, and the `main`
4817 function. While the syntax will allow the `main` function to appear
4818 multiple times, that will trigger an error if it is actually attempted.
4820 The various declarations do not return anything. They store the
4821 various declarations in the parse context.
4823 ###### Parser: grammar
4826 Ocean -> OptNL DeclarationList
4828 ## declare terminals
4836 DeclarationList -> Declaration
4837 | DeclarationList Declaration
4839 Declaration -> ERROR Newlines ${
4840 tok_err(c, // UNTESTED
4841 "error: unhandled parse error", &$1);
4847 ## top level grammar
4851 ### The `const` section
4853 As well as being defined in with the code that uses them, constants
4854 can be declared at the top level. These have full-file scope, so they
4855 are always `InScope`. The value of a top level constant can be given
4856 as an expression, and this is evaluated immediately rather than in the
4857 later interpretation stage. Once we add functions to the language, we
4858 will need rules concern which, if any, can be used to define a top
4861 Constants are defined in a section that starts with the reserved word
4862 `const` and then has a block with a list of assignment statements.
4863 For syntactic consistency, these must use the double-colon syntax to
4864 make it clear that they are constants. Type can also be given: if
4865 not, the type will be determined during analysis, as with other
4868 As the types constants are inserted at the head of a list, printing
4869 them in the same order that they were read is not straight forward.
4870 We take a quadratic approach here and count the number of constants
4871 (variables of depth 0), then count down from there, each time
4872 searching through for the Nth constant for decreasing N.
4874 ###### top level grammar
4878 DeclareConstant -> const { IN OptNL ConstList OUT OptNL } Newlines
4879 | const { SimpleConstList } Newlines
4880 | const IN OptNL ConstList OUT Newlines
4881 | const SimpleConstList Newlines
4883 ConstList -> ConstList SimpleConstLine
4886 SimpleConstList -> SimpleConstList ; Const
4890 SimpleConstLine -> SimpleConstList Newlines
4891 | ERROR Newlines ${ tok_err(c, "Syntax error in constant", &$1); }$
4894 CType -> Type ${ $0 = $<1; }$
4898 Const -> IDENTIFIER :: CType = Expression ${ {
4902 v = var_decl(c, $1.txt);
4904 struct var *var = new_pos(var, $1);
4905 v->where_decl = var;
4911 struct variable *vorig = var_ref(c, $1.txt);
4912 tok_err(c, "error: name already declared", &$1);
4913 type_err(c, "info: this is where '%v' was first declared",
4914 vorig->where_decl, NULL, 0, NULL);
4918 propagate_types($5, c, &ok, $3, 0);
4923 struct value res = interp_exec(c, $5, &v->type);
4924 global_alloc(c, v->type, v, &res);
4928 ###### print const decls
4933 while (target != 0) {
4935 for (v = context.in_scope; v; v=v->in_scope)
4936 if (v->depth == 0 && v->constant) {
4947 struct value *val = var_value(&context, v);
4948 printf(" %.*s :: ", v->name->name.len, v->name->name.txt);
4949 type_print(v->type, stdout);
4951 if (v->type == Tstr)
4953 print_value(v->type, val, stdout);
4954 if (v->type == Tstr)
4962 ### Function declarations
4964 The code in an Ocean program is all stored in function declarations.
4965 One of the functions must be named `main` and it must accept an array of
4966 strings as a parameter - the command line arguments.
4968 As this is the top level, several things are handled a bit differently.
4969 The function is not interpreted by `interp_exec` as that isn't passed
4970 the argument list which the program requires. Similarly type analysis
4971 is a bit more interesting at this level.
4973 ###### ast functions
4975 static struct type *handle_results(struct parse_context *c,
4976 struct binode *results)
4978 /* Create a 'struct' type from the results list, which
4979 * is a list for 'struct var'
4981 struct type *t = add_anon_type(c, &structure_prototype,
4982 " function result");
4986 for (b = results; b; b = cast(binode, b->right))
4988 t->structure.nfields = cnt;
4989 t->structure.fields = calloc(cnt, sizeof(struct field));
4991 for (b = results; b; b = cast(binode, b->right)) {
4992 struct var *v = cast(var, b->left);
4993 struct field *f = &t->structure.fields[cnt++];
4994 int a = v->var->type->align;
4995 f->name = v->var->name->name;
4996 f->type = v->var->type;
4998 f->offset = t->size;
4999 v->var->frame_pos = f->offset;
5000 t->size += ((f->type->size - 1) | (a-1)) + 1;
5003 variable_unlink_exec(v->var);
5005 free_binode(results);
5009 static struct variable *declare_function(struct parse_context *c,
5010 struct variable *name,
5011 struct binode *args,
5013 struct binode *results,
5017 struct value fn = {.function = code};
5019 var_block_close(c, CloseFunction, code);
5020 t = add_anon_type(c, &function_prototype,
5021 "func %.*s", name->name->name.len,
5022 name->name->name.txt);
5024 t->function.params = reorder_bilist(args);
5026 ret = handle_results(c, reorder_bilist(results));
5027 t->function.inline_result = 1;
5028 t->function.local_size = ret->size;
5030 t->function.return_type = ret;
5031 global_alloc(c, t, name, &fn);
5032 name->type->function.scope = c->out_scope;
5037 var_block_close(c, CloseFunction, NULL);
5039 c->out_scope = NULL;
5043 ###### declare terminals
5046 ###### top level grammar
5049 DeclareFunction -> func FuncName ( OpenScope ArgsLine ) Block Newlines ${
5050 $0 = declare_function(c, $<FN, $<Ar, Tnone, NULL, $<Bl);
5052 | func FuncName IN OpenScope Args OUT OptNL do Block Newlines ${
5053 $0 = declare_function(c, $<FN, $<Ar, Tnone, NULL, $<Bl);
5055 | func FuncName NEWLINE OpenScope OptNL do Block Newlines ${
5056 $0 = declare_function(c, $<FN, NULL, Tnone, NULL, $<Bl);
5058 | func FuncName ( OpenScope ArgsLine ) : Type Block Newlines ${
5059 $0 = declare_function(c, $<FN, $<Ar, $<Ty, NULL, $<Bl);
5061 | func FuncName ( OpenScope ArgsLine ) : ( ArgsLine ) Block Newlines ${
5062 $0 = declare_function(c, $<FN, $<AL, NULL, $<AL2, $<Bl);
5064 | func FuncName IN OpenScope Args OUT OptNL return Type Newlines do Block Newlines ${
5065 $0 = declare_function(c, $<FN, $<Ar, $<Ty, NULL, $<Bl);
5067 | func FuncName NEWLINE OpenScope return Type Newlines do Block Newlines ${
5068 $0 = declare_function(c, $<FN, NULL, $<Ty, NULL, $<Bl);
5070 | func FuncName IN OpenScope Args OUT OptNL return IN Args OUT OptNL do Block Newlines ${
5071 $0 = declare_function(c, $<FN, $<Ar, NULL, $<Ar2, $<Bl);
5073 | func FuncName NEWLINE OpenScope return IN Args OUT OptNL do Block Newlines ${
5074 $0 = declare_function(c, $<FN, NULL, NULL, $<Ar, $<Bl);
5077 ###### print func decls
5082 while (target != 0) {
5084 for (v = context.in_scope; v; v=v->in_scope)
5085 if (v->depth == 0 && v->type && v->type->check_args) {
5094 struct value *val = var_value(&context, v);
5095 printf("func %.*s", v->name->name.len, v->name->name.txt);
5096 v->type->print_type_decl(v->type, stdout);
5098 print_exec(val->function, 0, brackets);
5100 print_value(v->type, val, stdout);
5101 printf("/* frame size %d */\n", v->type->function.local_size);
5107 ###### core functions
5109 static int analyse_funcs(struct parse_context *c)
5113 for (v = c->in_scope; v; v = v->in_scope) {
5117 if (v->depth != 0 || !v->type || !v->type->check_args)
5119 ret = v->type->function.inline_result ?
5120 Tnone : v->type->function.return_type;
5121 val = var_value(c, v);
5124 propagate_types(val->function, c, &ok, ret, 0);
5127 /* Make sure everything is still consistent */
5128 propagate_types(val->function, c, &ok, ret, 0);
5131 if (!v->type->function.inline_result &&
5132 !v->type->function.return_type->dup) {
5133 type_err(c, "error: function cannot return value of type %1",
5134 v->where_decl, v->type->function.return_type, 0, NULL);
5137 scope_finalize(c, v->type);
5142 static int analyse_main(struct type *type, struct parse_context *c)
5144 struct binode *bp = type->function.params;
5148 struct type *argv_type;
5150 argv_type = add_anon_type(c, &array_prototype, "argv");
5151 argv_type->array.member = Tstr;
5152 argv_type->array.unspec = 1;
5154 for (b = bp; b; b = cast(binode, b->right)) {
5158 propagate_types(b->left, c, &ok, argv_type, 0);
5160 default: /* invalid */ // NOTEST
5161 propagate_types(b->left, c, &ok, Tnone, 0); // NOTEST
5167 return !c->parse_error;
5170 static void interp_main(struct parse_context *c, int argc, char **argv)
5172 struct value *progp = NULL;
5173 struct text main_name = { "main", 4 };
5174 struct variable *mainv;
5180 mainv = var_ref(c, main_name);
5182 progp = var_value(c, mainv);
5183 if (!progp || !progp->function) {
5184 fprintf(stderr, "oceani: no main function found.\n");
5188 if (!analyse_main(mainv->type, c)) {
5189 fprintf(stderr, "oceani: main has wrong type.\n");
5193 al = mainv->type->function.params;
5195 c->local_size = mainv->type->function.local_size;
5196 c->local = calloc(1, c->local_size);
5198 struct var *v = cast(var, al->left);
5199 struct value *vl = var_value(c, v->var);
5209 mpq_set_ui(argcq, argc, 1);
5210 memcpy(var_value(c, t->array.vsize), &argcq, sizeof(argcq));
5211 t->prepare_type(c, t, 0);
5212 array_init(v->var->type, vl);
5213 for (i = 0; i < argc; i++) {
5214 struct value *vl2 = vl->array + i * v->var->type->array.member->size;
5216 arg.str.txt = argv[i];
5217 arg.str.len = strlen(argv[i]);
5218 free_value(Tstr, vl2);
5219 dup_value(Tstr, &arg, vl2);
5223 al = cast(binode, al->right);
5225 v = interp_exec(c, progp->function, &vtype);
5226 free_value(vtype, &v);
5231 ###### ast functions
5232 void free_variable(struct variable *v)
5236 ## And now to test it out.
5238 Having a language requires having a "hello world" program. I'll
5239 provide a little more than that: a program that prints "Hello world"
5240 finds the GCD of two numbers, prints the first few elements of
5241 Fibonacci, performs a binary search for a number, and a few other
5242 things which will likely grow as the languages grows.
5244 ###### File: oceani.mk
5247 @echo "===== DEMO ====="
5248 ./oceani --section "demo: hello" oceani.mdc 55 33
5254 four ::= 2 + 2 ; five ::= 10/2
5255 const pie ::= "I like Pie";
5256 cake ::= "The cake is"
5264 func main(argv:[argc::]string)
5265 print "Hello World, what lovely oceans you have!"
5266 print "Are there", five, "?"
5267 print pi, pie, "but", cake
5269 A := $argv[1]; B := $argv[2]
5271 /* When a variable is defined in both branches of an 'if',
5272 * and used afterwards, the variables are merged.
5278 print "Is", A, "bigger than", B,"? ", bigger
5279 /* If a variable is not used after the 'if', no
5280 * merge happens, so types can be different
5283 double:string = "yes"
5284 print A, "is more than twice", B, "?", double
5287 print "double", B, "is", double
5292 if a > 0 and then b > 0:
5298 print "GCD of", A, "and", B,"is", a
5300 print a, "is not positive, cannot calculate GCD"
5302 print b, "is not positive, cannot calculate GCD"
5307 print "Fibonacci:", f1,f2,
5308 then togo = togo - 1
5316 /* Binary search... */
5321 mid := (lo + hi) / 2
5334 print "Yay, I found", target
5336 print "Closest I found was", lo
5341 // "middle square" PRNG. Not particularly good, but one my
5342 // Dad taught me - the first one I ever heard of.
5343 for i:=1; then i = i + 1; while i < size:
5344 n := list[i-1] * list[i-1]
5345 list[i] = (n / 100) % 10 000
5347 print "Before sort:",
5348 for i:=0; then i = i + 1; while i < size:
5352 for i := 1; then i=i+1; while i < size:
5353 for j:=i-1; then j=j-1; while j >= 0:
5354 if list[j] > list[j+1]:
5358 print " After sort:",
5359 for i:=0; then i = i + 1; while i < size:
5363 if 1 == 2 then print "yes"; else print "no"
5367 bob.alive = (bob.name == "Hello")
5368 print "bob", "is" if bob.alive else "isn't", "alive"