1 # Ocean Interpreter - Jamison Creek version
3 Ocean is intended to be a compiled language, so this interpreter is
4 not targeted at being the final product. It is, rather, an intermediate
5 stage and fills that role in two distinct ways.
7 Firstly, it exists as a platform to experiment with the early language
8 design. An interpreter is easy to write and easy to get working, so
9 the barrier for entry is lower if I aim to start with an interpreter.
11 Secondly, the plan for the Ocean compiler is to write it in the
12 [Ocean language](http://ocean-lang.org). To achieve this we naturally
13 need some sort of boot-strap process and this interpreter - written in
14 portable C - will fill that role. It will be used to bootstrap the
17 Two features that are not needed to fill either of these roles are
18 performance and completeness. The interpreter only needs to be fast
19 enough to run small test programs and occasionally to run the compiler
20 on itself. It only needs to be complete enough to test aspects of the
21 design which are developed before the compiler is working, and to run
22 the compiler on itself. Any features not used by the compiler when
23 compiling itself are superfluous. They may be included anyway, but
26 Nonetheless, the interpreter should end up being reasonably complete,
27 and any performance bottlenecks which appear and are easily fixed, will
32 This third version of the interpreter exists to test out some initial
33 ideas relating to types. Particularly it adds arrays (indexed from
34 zero) and simple structures. Basic control flow and variable scoping
35 are already fairly well established, as are basic numerical and
38 Some operators that have only recently been added, and so have not
39 generated all that much experience yet are "and then" and "or else" as
40 short-circuit Boolean operators, and the "if ... else" trinary
41 operator which can select between two expressions based on a third
42 (which appears syntactically in the middle).
44 The "func" clause currently only allows a "main" function to be
45 declared. That will be extended when proper function support is added.
47 An element that is present purely to make a usable language, and
48 without any expectation that they will remain, is the "print" statement
49 which performs simple output.
51 The current scalar types are "number", "Boolean", and "string".
52 Boolean will likely stay in its current form, the other two might, but
53 could just as easily be changed.
57 Versions of the interpreter which obviously do not support a complete
58 language will be named after creeks and streams. This one is Jamison
61 Once we have something reasonably resembling a complete language, the
62 names of rivers will be used.
63 Early versions of the compiler will be named after seas. Major
64 releases of the compiler will be named after oceans. Hopefully I will
65 be finished once I get to the Pacific Ocean release.
69 As well as parsing and executing a program, the interpreter can print
70 out the program from the parsed internal structure. This is useful
71 for validating the parsing.
72 So the main requirements of the interpreter are:
74 - Parse the program, possibly with tracing,
75 - Analyse the parsed program to ensure consistency,
77 - Execute the "main" function in the program, if no parsing or
78 consistency errors were found.
80 This is all performed by a single C program extracted with
83 There will be two formats for printing the program: a default and one
84 that uses bracketing. So a `--bracket` command line option is needed
85 for that. Normally the first code section found is used, however an
86 alternate section can be requested so that a file (such as this one)
87 can contain multiple programs. This is effected with the `--section`
90 This code must be compiled with `-fplan9-extensions` so that anonymous
91 structures can be used.
93 ###### File: oceani.mk
95 myCFLAGS := -Wall -g -fplan9-extensions
96 CFLAGS := $(filter-out $(myCFLAGS),$(CFLAGS)) $(myCFLAGS)
97 myLDLIBS:= libparser.o libscanner.o libmdcode.o -licuuc
98 LDLIBS := $(filter-out $(myLDLIBS),$(LDLIBS)) $(myLDLIBS)
100 all :: $(LDLIBS) oceani
101 oceani.c oceani.h : oceani.mdc parsergen
102 ./parsergen -o oceani --LALR --tag Parser oceani.mdc
103 oceani.mk: oceani.mdc md2c
106 oceani: oceani.o $(LDLIBS)
107 $(CC) $(CFLAGS) -o oceani oceani.o $(LDLIBS)
109 ###### Parser: header
111 struct parse_context;
113 struct parse_context {
114 struct token_config config;
122 #define container_of(ptr, type, member) ({ \
123 const typeof( ((type *)0)->member ) *__mptr = (ptr); \
124 (type *)( (char *)__mptr - offsetof(type,member) );})
126 #define config2context(_conf) container_of(_conf, struct parse_context, \
129 ###### Parser: reduce
130 struct parse_context *c = config2context(config);
138 #include <sys/mman.h>
157 static char Usage[] =
158 "Usage: oceani --trace --print --noexec --brackets --section=SectionName prog.ocn\n";
159 static const struct option long_options[] = {
160 {"trace", 0, NULL, 't'},
161 {"print", 0, NULL, 'p'},
162 {"noexec", 0, NULL, 'n'},
163 {"brackets", 0, NULL, 'b'},
164 {"section", 1, NULL, 's'},
167 const char *options = "tpnbs";
169 static void pr_err(char *msg) // NOTEST
171 fprintf(stderr, "%s\n", msg); // NOTEST
174 int main(int argc, char *argv[])
179 struct section *s = NULL, *ss;
180 char *section = NULL;
181 struct parse_context context = {
183 .ignored = (1 << TK_mark),
184 .number_chars = ".,_+- ",
189 int doprint=0, dotrace=0, doexec=1, brackets=0;
191 while ((opt = getopt_long(argc, argv, options, long_options, NULL))
194 case 't': dotrace=1; break;
195 case 'p': doprint=1; break;
196 case 'n': doexec=0; break;
197 case 'b': brackets=1; break;
198 case 's': section = optarg; break;
199 default: fprintf(stderr, Usage);
203 if (optind >= argc) {
204 fprintf(stderr, "oceani: no input file given\n");
207 fd = open(argv[optind], O_RDONLY);
209 fprintf(stderr, "oceani: cannot open %s\n", argv[optind]);
212 context.file_name = argv[optind];
213 len = lseek(fd, 0, 2);
214 file = mmap(NULL, len, PROT_READ, MAP_SHARED, fd, 0);
215 s = code_extract(file, file+len, pr_err);
217 fprintf(stderr, "oceani: could not find any code in %s\n",
222 ## context initialization
225 for (ss = s; ss; ss = ss->next) {
226 struct text sec = ss->section;
227 if (sec.len == strlen(section) &&
228 strncmp(sec.txt, section, sec.len) == 0)
232 fprintf(stderr, "oceani: cannot find section %s\n",
239 fprintf(stderr, "oceani: no code found in requested section\n"); // NOTEST
240 goto cleanup; // NOTEST
243 parse_oceani(ss->code, &context.config, dotrace ? stderr : NULL);
245 resolve_consts(&context);
246 prepare_types(&context);
247 if (!context.parse_error && !analyse_funcs(&context)) {
248 fprintf(stderr, "oceani: type error in program - not running.\n");
249 context.parse_error = 1;
257 if (doexec && !context.parse_error)
258 interp_main(&context, argc - optind, argv + optind);
261 struct section *t = s->next;
266 // FIXME parser should pop scope even on error
267 while (context.scope_depth > 0)
271 ## free context types
272 ## free context storage
273 exit(context.parse_error ? 1 : 0);
278 The four requirements of parse, analyse, print, interpret apply to
279 each language element individually so that is how most of the code
282 Three of the four are fairly self explanatory. The one that requires
283 a little explanation is the analysis step.
285 The current language design does not require the types of variables to
286 be declared, but they must still have a single type. Different
287 operations impose different requirements on the variables, for example
288 addition requires both arguments to be numeric, and assignment
289 requires the variable on the left to have the same type as the
290 expression on the right.
292 Analysis involves propagating these type requirements around and
293 consequently setting the type of each variable. If any requirements
294 are violated (e.g. a string is compared with a number) or if a
295 variable needs to have two different types, then an error is raised
296 and the program will not run.
298 If the same variable is declared in both branchs of an 'if/else', or
299 in all cases of a 'switch' then the multiple instances may be merged
300 into just one variable if the variable is referenced after the
301 conditional statement. When this happens, the types must naturally be
302 consistent across all the branches. When the variable is not used
303 outside the if, the variables in the different branches are distinct
304 and can be of different types.
306 Undeclared names may only appear in "use" statements and "case" expressions.
307 These names are given a type of "label" and a unique value.
308 This allows them to fill the role of a name in an enumerated type, which
309 is useful for testing the `switch` statement.
311 As we will see, the condition part of a `while` statement can return
312 either a Boolean or some other type. This requires that the expected
313 type that gets passed around comprises a type and a flag to indicate
314 that `Tbool` is also permitted.
316 As there are, as yet, no distinct types that are compatible, there
317 isn't much subtlety in the analysis. When we have distinct number
318 types, this will become more interesting.
322 When analysis discovers an inconsistency it needs to report an error;
323 just refusing to run the code ensures that the error doesn't cascade,
324 but by itself it isn't very useful. A clear understanding of the sort
325 of error message that are useful will help guide the process of
328 At a simplistic level, the only sort of error that type analysis can
329 report is that the type of some construct doesn't match a contextual
330 requirement. For example, in `4 + "hello"` the addition provides a
331 contextual requirement for numbers, but `"hello"` is not a number. In
332 this particular example no further information is needed as the types
333 are obvious from local information. When a variable is involved that
334 isn't the case. It may be helpful to explain why the variable has a
335 particular type, by indicating the location where the type was set,
336 whether by declaration or usage.
338 Using a recursive-descent analysis we can easily detect a problem at
339 multiple locations. In "`hello:= "there"; 4 + hello`" the addition
340 will detect that one argument is not a number and the usage of `hello`
341 will detect that a number was wanted, but not provided. In this
342 (early) version of the language, we will generate error reports at
343 multiple locations, so the use of `hello` will report an error and
344 explain were the value was set, and the addition will report an error
345 and say why numbers are needed. To be able to report locations for
346 errors, each language element will need to record a file location
347 (line and column) and each variable will need to record the language
348 element where its type was set. For now we will assume that each line
349 of an error message indicates one location in the file, and up to 2
350 types. So we provide a `printf`-like function which takes a format, a
351 location (a `struct exec` which has not yet been introduced), and 2
352 types. "`%1`" reports the first type, "`%2`" reports the second. We
353 will need a function to print the location, once we know how that is
354 stored. e As will be explained later, there are sometimes extra rules for
355 type matching and they might affect error messages, we need to pass those
358 As well as type errors, we sometimes need to report problems with
359 tokens, which might be unexpected or might name a type that has not
360 been defined. For these we have `tok_err()` which reports an error
361 with a given token. Each of the error functions sets the flag in the
362 context so indicate that parsing failed.
366 static void fput_loc(struct exec *loc, FILE *f);
367 static void type_err(struct parse_context *c,
368 char *fmt, struct exec *loc,
369 struct type *t1, int rules, struct type *t2);
371 ###### core functions
373 static void type_err(struct parse_context *c,
374 char *fmt, struct exec *loc,
375 struct type *t1, int rules, struct type *t2)
377 fprintf(stderr, "%s:", c->file_name);
378 fput_loc(loc, stderr);
379 for (; *fmt ; fmt++) {
386 case '%': fputc(*fmt, stderr); break; // NOTEST
387 default: fputc('?', stderr); break; // NOTEST
389 type_print(t1, stderr);
392 type_print(t2, stderr);
401 static void tok_err(struct parse_context *c, char *fmt, struct token *t)
403 fprintf(stderr, "%s:%d:%d: %s: %.*s\n", c->file_name, t->line, t->col, fmt,
404 t->txt.len, t->txt.txt);
408 ## Entities: declared and predeclared.
410 There are various "things" that the language and/or the interpreter
411 needs to know about to parse and execute a program. These include
412 types, variables, values, and executable code. These are all lumped
413 together under the term "entities" (calling them "objects" would be
414 confusing) and introduced here. The following section will present the
415 different specific code elements which comprise or manipulate these
420 Executables can be lots of different things. In many cases an
421 executable is just an operation combined with one or two other
422 executables. This allows for expressions and lists etc. Other times an
423 executable is something quite specific like a constant or variable name.
424 So we define a `struct exec` to be a general executable with a type, and
425 a `struct binode` which is a subclass of `exec`, forms a node in a
426 binary tree, and holds an operation. There will be other subclasses,
427 and to access these we need to be able to `cast` the `exec` into the
428 various other types. The first field in any `struct exec` is the type
429 from the `exec_types` enum.
432 #define cast(structname, pointer) ({ \
433 const typeof( ((struct structname *)0)->type) *__mptr = &(pointer)->type; \
434 if (__mptr && *__mptr != X##structname) abort(); \
435 (struct structname *)( (char *)__mptr);})
437 #define new(structname) ({ \
438 struct structname *__ptr = ((struct structname *)calloc(1,sizeof(struct structname))); \
439 __ptr->type = X##structname; \
440 __ptr->line = -1; __ptr->column = -1; \
443 #define new_pos(structname, token) ({ \
444 struct structname *__ptr = ((struct structname *)calloc(1,sizeof(struct structname))); \
445 __ptr->type = X##structname; \
446 __ptr->line = token.line; __ptr->column = token.col; \
455 enum exec_types type;
464 struct exec *left, *right;
469 static int __fput_loc(struct exec *loc, FILE *f)
473 if (loc->line >= 0) {
474 fprintf(f, "%d:%d: ", loc->line, loc->column);
477 if (loc->type == Xbinode)
478 return __fput_loc(cast(binode,loc)->left, f) ||
479 __fput_loc(cast(binode,loc)->right, f); // NOTEST
482 static void fput_loc(struct exec *loc, FILE *f)
484 if (!__fput_loc(loc, f))
485 fprintf(f, "??:??: ");
488 Each different type of `exec` node needs a number of functions defined,
489 a bit like methods. We must be able to free it, print it, analyse it
490 and execute it. Once we have specific `exec` types we will need to
491 parse them too. Let's take this a bit more slowly.
495 The parser generator requires a `free_foo` function for each struct
496 that stores attributes and they will often be `exec`s and subtypes
497 there-of. So we need `free_exec` which can handle all the subtypes,
498 and we need `free_binode`.
502 static void free_binode(struct binode *b)
511 ###### core functions
512 static void free_exec(struct exec *e)
523 static void free_exec(struct exec *e);
525 ###### free exec cases
526 case Xbinode: free_binode(cast(binode, e)); break;
530 Printing an `exec` requires that we know the current indent level for
531 printing line-oriented components. As will become clear later, we
532 also want to know what sort of bracketing to use.
536 static void do_indent(int i, char *str)
543 ###### core functions
544 static void print_binode(struct binode *b, int indent, int bracket)
548 ## print binode cases
552 static void print_exec(struct exec *e, int indent, int bracket)
558 print_binode(cast(binode, e), indent, bracket); break;
563 do_indent(indent, "/* FREE");
564 for (v = e->to_free; v; v = v->next_free) {
565 printf(" %.*s", v->name->name.len, v->name->name.txt);
566 printf("[%d,%d]", v->scope_start, v->scope_end);
567 if (v->frame_pos >= 0)
568 printf("(%d+%d)", v->frame_pos,
569 v->type ? v->type->size:0);
577 static void print_exec(struct exec *e, int indent, int bracket);
581 As discussed, analysis involves propagating type requirements around the
582 program and looking for errors.
584 So `propagate_types` is passed an expected type (being a `struct type`
585 pointer together with some `val_rules` flags) that the `exec` is
586 expected to return, and returns the type that it does return, either
587 of which can be `NULL` signifying "unknown". An `ok` flag is passed
588 by reference. It is set to `0` when an error is found, and `2` when
589 any change is made. If it remains unchanged at `1`, then no more
590 propagation is needed.
594 enum val_rules {Rnolabel = 1<<0, Rboolok = 1<<1, Rnoconstant = 1<<2};
598 if (rules & Rnolabel)
599 fputs(" (labels not permitted)", stderr);
603 static struct type *propagate_types(struct exec *prog, struct parse_context *c, int *ok,
604 struct type *type, int rules);
605 ###### core functions
607 static struct type *__propagate_types(struct exec *prog, struct parse_context *c, int *ok,
608 struct type *type, int rules)
615 switch (prog->type) {
618 struct binode *b = cast(binode, prog);
620 ## propagate binode cases
624 ## propagate exec cases
629 static struct type *propagate_types(struct exec *prog, struct parse_context *c, int *ok,
630 struct type *type, int rules)
632 struct type *ret = __propagate_types(prog, c, ok, type, rules);
641 Interpreting an `exec` doesn't require anything but the `exec`. State
642 is stored in variables and each variable will be directly linked from
643 within the `exec` tree. The exception to this is the `main` function
644 which needs to look at command line arguments. This function will be
645 interpreted separately.
647 Each `exec` can return a value combined with a type in `struct lrval`.
648 The type may be `Tnone` but must be non-NULL. Some `exec`s will return
649 the location of a value, which can be updated, in `lval`. Others will
650 set `lval` to NULL indicating that there is a value of appropriate type
654 static struct value interp_exec(struct parse_context *c, struct exec *e,
655 struct type **typeret);
656 ###### core functions
660 struct value rval, *lval;
663 /* If dest is passed, dtype must give the expected type, and
664 * result can go there, in which case type is returned as NULL.
666 static struct lrval _interp_exec(struct parse_context *c, struct exec *e,
667 struct value *dest, struct type *dtype);
669 static struct value interp_exec(struct parse_context *c, struct exec *e,
670 struct type **typeret)
672 struct lrval ret = _interp_exec(c, e, NULL, NULL);
674 if (!ret.type) abort();
678 dup_value(ret.type, ret.lval, &ret.rval);
682 static struct value *linterp_exec(struct parse_context *c, struct exec *e,
683 struct type **typeret)
685 struct lrval ret = _interp_exec(c, e, NULL, NULL);
687 if (!ret.type) abort();
691 free_value(ret.type, &ret.rval);
695 /* dinterp_exec is used when the destination type is certain and
696 * the value has a place to go.
698 static void dinterp_exec(struct parse_context *c, struct exec *e,
699 struct value *dest, struct type *dtype,
702 struct lrval ret = _interp_exec(c, e, dest, dtype);
706 free_value(dtype, dest);
708 dup_value(dtype, ret.lval, dest);
710 memcpy(dest, &ret.rval, dtype->size);
713 static struct lrval _interp_exec(struct parse_context *c, struct exec *e,
714 struct value *dest, struct type *dtype)
716 /* If the result is copied to dest, ret.type is set to NULL */
718 struct value rv = {}, *lrv = NULL;
721 rvtype = ret.type = Tnone;
731 struct binode *b = cast(binode, e);
732 struct value left, right, *lleft;
733 struct type *ltype, *rtype;
734 ltype = rtype = Tnone;
736 ## interp binode cases
738 free_value(ltype, &left);
739 free_value(rtype, &right);
749 ## interp exec cleanup
755 Values come in a wide range of types, with more likely to be added.
756 Each type needs to be able to print its own values (for convenience at
757 least) as well as to compare two values, at least for equality and
758 possibly for order. For now, values might need to be duplicated and
759 freed, though eventually such manipulations will be better integrated
762 Rather than requiring every numeric type to support all numeric
763 operations (add, multiply, etc), we allow types to be able to present
764 as one of a few standard types: integer, float, and fraction. The
765 existence of these conversion functions eventually enable types to
766 determine if they are compatible with other types, though such types
767 have not yet been implemented.
769 Named type are stored in a simple linked list. Objects of each type are
770 "values" which are often passed around by value.
772 There are both explicitly named types, and anonymous types. Anonymous
773 cannot be accessed by name, but are used internally and have a name
774 which might be reported in error messages.
781 ## value union fields
790 void (*init)(struct type *type, struct value *val);
791 void (*prepare_type)(struct parse_context *c, struct type *type, int parse_time);
792 void (*print)(struct type *type, struct value *val, FILE *f);
793 void (*print_type)(struct type *type, FILE *f);
794 int (*cmp_order)(struct type *t1, struct type *t2,
795 struct value *v1, struct value *v2);
796 int (*cmp_eq)(struct type *t1, struct type *t2,
797 struct value *v1, struct value *v2);
798 void (*dup)(struct type *type, struct value *vold, struct value *vnew);
799 void (*free)(struct type *type, struct value *val);
800 void (*free_type)(struct type *t);
801 long long (*to_int)(struct value *v);
802 double (*to_float)(struct value *v);
803 int (*to_mpq)(mpq_t *q, struct value *v);
812 struct type *typelist;
819 static struct type *find_type(struct parse_context *c, struct text s)
821 struct type *t = c->typelist;
823 while (t && (t->anon ||
824 text_cmp(t->name, s) != 0))
829 static struct type *_add_type(struct parse_context *c, struct text s,
830 struct type *proto, int anon)
834 n = calloc(1, sizeof(*n));
838 n->next = c->typelist;
843 static struct type *add_type(struct parse_context *c, struct text s,
846 return _add_type(c, s, proto, 0);
849 static struct type *add_anon_type(struct parse_context *c,
850 struct type *proto, char *name, ...)
856 vasprintf(&t.txt, name, ap);
858 t.len = strlen(name);
859 return _add_type(c, t, proto, 1);
862 static void free_type(struct type *t)
864 /* The type is always a reference to something in the
865 * context, so we don't need to free anything.
869 static void free_value(struct type *type, struct value *v)
873 memset(v, 0x5a, type->size);
877 static void type_print(struct type *type, FILE *f)
880 fputs("*unknown*type*", f); // NOTEST
881 else if (type->name.len && !type->anon)
882 fprintf(f, "%.*s", type->name.len, type->name.txt);
883 else if (type->print_type)
884 type->print_type(type, f);
886 fputs("*invalid*type*", f);
889 static void val_init(struct type *type, struct value *val)
891 if (type && type->init)
892 type->init(type, val);
895 static void dup_value(struct type *type,
896 struct value *vold, struct value *vnew)
898 if (type && type->dup)
899 type->dup(type, vold, vnew);
902 static int value_cmp(struct type *tl, struct type *tr,
903 struct value *left, struct value *right)
905 if (tl && tl->cmp_order)
906 return tl->cmp_order(tl, tr, left, right);
907 if (tl && tl->cmp_eq) // NOTEST
908 return tl->cmp_eq(tl, tr, left, right); // NOTEST
912 static void print_value(struct type *type, struct value *v, FILE *f)
914 if (type && type->print)
915 type->print(type, v, f);
917 fprintf(f, "*Unknown*"); // NOTEST
920 static void prepare_types(struct parse_context *c)
924 for (t = c->typelist; t; t = t->next)
926 t->prepare_type(c, t, 1);
931 static void free_value(struct type *type, struct value *v);
932 static int type_compat(struct type *require, struct type *have, int rules);
933 static void type_print(struct type *type, FILE *f);
934 static void val_init(struct type *type, struct value *v);
935 static void dup_value(struct type *type,
936 struct value *vold, struct value *vnew);
937 static int value_cmp(struct type *tl, struct type *tr,
938 struct value *left, struct value *right);
939 static void print_value(struct type *type, struct value *v, FILE *f);
941 ###### free context types
943 while (context.typelist) {
944 struct type *t = context.typelist;
946 context.typelist = t->next;
954 Type can be specified for local variables, for fields in a structure,
955 for formal parameters to functions, and possibly elsewhere. Different
956 rules may apply in different contexts. As a minimum, a named type may
957 always be used. Currently the type of a formal parameter can be
958 different from types in other contexts, so we have a separate grammar
964 Type -> IDENTIFIER ${
965 $0 = find_type(c, $1.txt);
968 "error: undefined type", &$1);
975 FormalType -> Type ${ $0 = $<1; }$
976 ## formal type grammar
980 Values of the base types can be numbers, which we represent as
981 multi-precision fractions, strings, Booleans and labels. When
982 analysing the program we also need to allow for places where no value
983 is meaningful (type `Tnone`) and where we don't know what type to
984 expect yet (type is `NULL`).
986 Values are never shared, they are always copied when used, and freed
987 when no longer needed.
989 When propagating type information around the program, we need to
990 determine if two types are compatible, where type `NULL` is compatible
991 with anything. There are two special cases with type compatibility,
992 both related to the Conditional Statement which will be described
993 later. In some cases a Boolean can be accepted as well as some other
994 primary type, and in others any type is acceptable except a label (`Vlabel`).
995 A separate function encoding these cases will simplify some code later.
997 ###### type functions
999 int (*compat)(struct type *this, struct type *other);
1001 ###### ast functions
1003 static int type_compat(struct type *require, struct type *have, int rules)
1005 if ((rules & Rboolok) && have == Tbool)
1007 if ((rules & Rnolabel) && have == Tlabel)
1009 if (!require || !have)
1012 if (require->compat)
1013 return require->compat(require, have);
1015 return require == have;
1020 #include "parse_string.h"
1021 #include "parse_number.h"
1024 myLDLIBS := libnumber.o libstring.o -lgmp
1025 LDLIBS := $(filter-out $(myLDLIBS),$(LDLIBS)) $(myLDLIBS)
1027 ###### type union fields
1028 enum vtype {Vnone, Vstr, Vnum, Vbool, Vlabel} vtype;
1030 ###### value union fields
1036 ###### ast functions
1037 static void _free_value(struct type *type, struct value *v)
1041 switch (type->vtype) {
1043 case Vstr: free(v->str.txt); break;
1044 case Vnum: mpq_clear(v->num); break;
1050 ###### value functions
1052 static void _val_init(struct type *type, struct value *val)
1054 switch(type->vtype) {
1055 case Vnone: // NOTEST
1058 mpq_init(val->num); break;
1060 val->str.txt = malloc(1);
1072 static void _dup_value(struct type *type,
1073 struct value *vold, struct value *vnew)
1075 switch (type->vtype) {
1076 case Vnone: // NOTEST
1079 vnew->label = vold->label;
1082 vnew->bool = vold->bool;
1085 mpq_init(vnew->num);
1086 mpq_set(vnew->num, vold->num);
1089 vnew->str.len = vold->str.len;
1090 vnew->str.txt = malloc(vnew->str.len);
1091 memcpy(vnew->str.txt, vold->str.txt, vnew->str.len);
1096 static int _value_cmp(struct type *tl, struct type *tr,
1097 struct value *left, struct value *right)
1101 return tl - tr; // NOTEST
1102 switch (tl->vtype) {
1103 case Vlabel: cmp = left->label == right->label ? 0 : 1; break;
1104 case Vnum: cmp = mpq_cmp(left->num, right->num); break;
1105 case Vstr: cmp = text_cmp(left->str, right->str); break;
1106 case Vbool: cmp = left->bool - right->bool; break;
1107 case Vnone: cmp = 0; // NOTEST
1112 static void _print_value(struct type *type, struct value *v, FILE *f)
1114 switch (type->vtype) {
1115 case Vnone: // NOTEST
1116 fprintf(f, "*no-value*"); break; // NOTEST
1117 case Vlabel: // NOTEST
1118 fprintf(f, "*label-%p*", v->label); break; // NOTEST
1120 fprintf(f, "%.*s", v->str.len, v->str.txt); break;
1122 fprintf(f, "%s", v->bool ? "True":"False"); break;
1127 mpf_set_q(fl, v->num);
1128 gmp_fprintf(f, "%Fg", fl);
1135 static void _free_value(struct type *type, struct value *v);
1137 static struct type base_prototype = {
1139 .print = _print_value,
1140 .cmp_order = _value_cmp,
1141 .cmp_eq = _value_cmp,
1143 .free = _free_value,
1146 static struct type *Tbool, *Tstr, *Tnum, *Tnone, *Tlabel;
1148 ###### ast functions
1149 static struct type *add_base_type(struct parse_context *c, char *n,
1150 enum vtype vt, int size)
1152 struct text txt = { n, strlen(n) };
1155 t = add_type(c, txt, &base_prototype);
1158 t->align = size > sizeof(void*) ? sizeof(void*) : size;
1159 if (t->size & (t->align - 1))
1160 t->size = (t->size | (t->align - 1)) + 1; // NOTEST
1164 ###### context initialization
1166 Tbool = add_base_type(&context, "Boolean", Vbool, sizeof(char));
1167 Tstr = add_base_type(&context, "string", Vstr, sizeof(struct text));
1168 Tnum = add_base_type(&context, "number", Vnum, sizeof(mpq_t));
1169 Tnone = add_base_type(&context, "none", Vnone, 0);
1170 Tlabel = add_base_type(&context, "label", Vlabel, sizeof(void*));
1174 We have already met values as separate objects. When manifest constants
1175 appear in the program text, that must result in an executable which has
1176 a constant value. So the `val` structure embeds a value in an
1189 ###### ast functions
1190 struct val *new_val(struct type *T, struct token tk)
1192 struct val *v = new_pos(val, tk);
1203 $0 = new_val(Tbool, $1);
1207 $0 = new_val(Tbool, $1);
1212 $0 = new_val(Tnum, $1);
1213 if (number_parse($0->val.num, tail, $1.txt) == 0)
1214 mpq_init($0->val.num); // UNTESTED
1216 tok_err(c, "error: unsupported number suffix",
1221 $0 = new_val(Tstr, $1);
1222 string_parse(&$1, '\\', &$0->val.str, tail);
1224 tok_err(c, "error: unsupported string suffix",
1229 $0 = new_val(Tstr, $1);
1230 string_parse(&$1, '\\', &$0->val.str, tail);
1232 tok_err(c, "error: unsupported string suffix",
1236 ###### print exec cases
1239 struct val *v = cast(val, e);
1240 if (v->vtype == Tstr)
1242 print_value(v->vtype, &v->val, stdout);
1243 if (v->vtype == Tstr)
1248 ###### propagate exec cases
1251 struct val *val = cast(val, prog);
1252 if (!type_compat(type, val->vtype, rules))
1253 type_err(c, "error: expected %1%r found %2",
1254 prog, type, rules, val->vtype);
1258 ###### interp exec cases
1260 rvtype = cast(val, e)->vtype;
1261 dup_value(rvtype, &cast(val, e)->val, &rv);
1264 ###### ast functions
1265 static void free_val(struct val *v)
1268 free_value(v->vtype, &v->val);
1272 ###### free exec cases
1273 case Xval: free_val(cast(val, e)); break;
1275 ###### ast functions
1276 // Move all nodes from 'b' to 'rv', reversing their order.
1277 // In 'b' 'left' is a list, and 'right' is the last node.
1278 // In 'rv', left' is the first node and 'right' is a list.
1279 static struct binode *reorder_bilist(struct binode *b)
1281 struct binode *rv = NULL;
1284 struct exec *t = b->right;
1288 b = cast(binode, b->left);
1298 Variables are scoped named values. We store the names in a linked list
1299 of "bindings" sorted in lexical order, and use sequential search and
1306 struct binding *next; // in lexical order
1310 This linked list is stored in the parse context so that "reduce"
1311 functions can find or add variables, and so the analysis phase can
1312 ensure that every variable gets a type.
1314 ###### parse context
1316 struct binding *varlist; // In lexical order
1318 ###### ast functions
1320 static struct binding *find_binding(struct parse_context *c, struct text s)
1322 struct binding **l = &c->varlist;
1327 (cmp = text_cmp((*l)->name, s)) < 0)
1331 n = calloc(1, sizeof(*n));
1338 Each name can be linked to multiple variables defined in different
1339 scopes. Each scope starts where the name is declared and continues
1340 until the end of the containing code block. Scopes of a given name
1341 cannot nest, so a declaration while a name is in-scope is an error.
1343 ###### binding fields
1344 struct variable *var;
1348 struct variable *previous;
1350 struct binding *name;
1351 struct exec *where_decl;// where name was declared
1352 struct exec *where_set; // where type was set
1356 When a scope closes, the values of the variables might need to be freed.
1357 This happens in the context of some `struct exec` and each `exec` will
1358 need to know which variables need to be freed when it completes.
1361 struct variable *to_free;
1363 ####### variable fields
1364 struct exec *cleanup_exec;
1365 struct variable *next_free;
1367 ####### interp exec cleanup
1370 for (v = e->to_free; v; v = v->next_free) {
1371 struct value *val = var_value(c, v);
1372 free_value(v->type, val);
1376 ###### ast functions
1377 static void variable_unlink_exec(struct variable *v)
1379 struct variable **vp;
1380 if (!v->cleanup_exec)
1382 for (vp = &v->cleanup_exec->to_free;
1383 *vp; vp = &(*vp)->next_free) {
1387 v->cleanup_exec = NULL;
1392 While the naming seems strange, we include local constants in the
1393 definition of variables. A name declared `var := value` can
1394 subsequently be changed, but a name declared `var ::= value` cannot -
1397 ###### variable fields
1400 Scopes in parallel branches can be partially merged. More
1401 specifically, if a given name is declared in both branches of an
1402 if/else then its scope is a candidate for merging. Similarly if
1403 every branch of an exhaustive switch (e.g. has an "else" clause)
1404 declares a given name, then the scopes from the branches are
1405 candidates for merging.
1407 Note that names declared inside a loop (which is only parallel to
1408 itself) are never visible after the loop. Similarly names defined in
1409 scopes which are not parallel, such as those started by `for` and
1410 `switch`, are never visible after the scope. Only variables defined in
1411 both `then` and `else` (including the implicit then after an `if`, and
1412 excluding `then` used with `for`) and in all `case`s and `else` of a
1413 `switch` or `while` can be visible beyond the `if`/`switch`/`while`.
1415 Labels, which are a bit like variables, follow different rules.
1416 Labels are not explicitly declared, but if an undeclared name appears
1417 in a context where a label is legal, that effectively declares the
1418 name as a label. The declaration remains in force (or in scope) at
1419 least to the end of the immediately containing block and conditionally
1420 in any larger containing block which does not declare the name in some
1421 other way. Importantly, the conditional scope extension happens even
1422 if the label is only used in one parallel branch of a conditional --
1423 when used in one branch it is treated as having been declared in all
1426 Merge candidates are tentatively visible beyond the end of the
1427 branching statement which creates them. If the name is used, the
1428 merge is affirmed and they become a single variable visible at the
1429 outer layer. If not - if it is redeclared first - the merge lapses.
1431 To track scopes we have an extra stack, implemented as a linked list,
1432 which roughly parallels the parse stack and which is used exclusively
1433 for scoping. When a new scope is opened, a new frame is pushed and
1434 the child-count of the parent frame is incremented. This child-count
1435 is used to distinguish between the first of a set of parallel scopes,
1436 in which declared variables must not be in scope, and subsequent
1437 branches, whether they may already be conditionally scoped.
1439 We need a total ordering of scopes so we can easily compare to variables
1440 to see if they are concurrently in scope. To achieve this we record a
1441 `scope_count` which is actually a count of both beginnings and endings
1442 of scopes. Then each variable has a record of the scope count where it
1443 enters scope, and where it leaves.
1445 To push a new frame *before* any code in the frame is parsed, we need a
1446 grammar reduction. This is most easily achieved with a grammar
1447 element which derives the empty string, and creates the new scope when
1448 it is recognised. This can be placed, for example, between a keyword
1449 like "if" and the code following it.
1453 struct scope *parent;
1457 ###### parse context
1460 struct scope *scope_stack;
1462 ###### variable fields
1463 int scope_start, scope_end;
1465 ###### ast functions
1466 static void scope_pop(struct parse_context *c)
1468 struct scope *s = c->scope_stack;
1470 c->scope_stack = s->parent;
1472 c->scope_depth -= 1;
1473 c->scope_count += 1;
1476 static void scope_push(struct parse_context *c)
1478 struct scope *s = calloc(1, sizeof(*s));
1480 c->scope_stack->child_count += 1;
1481 s->parent = c->scope_stack;
1483 c->scope_depth += 1;
1484 c->scope_count += 1;
1490 OpenScope -> ${ scope_push(c); }$
1492 Each variable records a scope depth and is in one of four states:
1494 - "in scope". This is the case between the declaration of the
1495 variable and the end of the containing block, and also between
1496 the usage with affirms a merge and the end of that block.
1498 The scope depth is not greater than the current parse context scope
1499 nest depth. When the block of that depth closes, the state will
1500 change. To achieve this, all "in scope" variables are linked
1501 together as a stack in nesting order.
1503 - "pending". The "in scope" block has closed, but other parallel
1504 scopes are still being processed. So far, every parallel block at
1505 the same level that has closed has declared the name.
1507 The scope depth is the depth of the last parallel block that
1508 enclosed the declaration, and that has closed.
1510 - "conditionally in scope". The "in scope" block and all parallel
1511 scopes have closed, and no further mention of the name has been seen.
1512 This state includes a secondary nest depth (`min_depth`) which records
1513 the outermost scope seen since the variable became conditionally in
1514 scope. If a use of the name is found, the variable becomes "in scope"
1515 and that secondary depth becomes the recorded scope depth. If the
1516 name is declared as a new variable, the old variable becomes "out of
1517 scope" and the recorded scope depth stays unchanged.
1519 - "out of scope". The variable is neither in scope nor conditionally
1520 in scope. It is permanently out of scope now and can be removed from
1521 the "in scope" stack. When a variable becomes out-of-scope it is
1522 moved to a separate list (`out_scope`) of variables which have fully
1523 known scope. This will be used at the end of each function to assign
1524 each variable a place in the stack frame.
1526 ###### variable fields
1527 int depth, min_depth;
1528 enum { OutScope, PendingScope, CondScope, InScope } scope;
1529 struct variable *in_scope;
1531 ###### parse context
1533 struct variable *in_scope;
1534 struct variable *out_scope;
1536 All variables with the same name are linked together using the
1537 'previous' link. Those variable that have been affirmatively merged all
1538 have a 'merged' pointer that points to one primary variable - the most
1539 recently declared instance. When merging variables, we need to also
1540 adjust the 'merged' pointer on any other variables that had previously
1541 been merged with the one that will no longer be primary.
1543 A variable that is no longer the most recent instance of a name may
1544 still have "pending" scope, if it might still be merged with most
1545 recent instance. These variables don't really belong in the
1546 "in_scope" list, but are not immediately removed when a new instance
1547 is found. Instead, they are detected and ignored when considering the
1548 list of in_scope names.
1550 The storage of the value of a variable will be described later. For now
1551 we just need to know that when a variable goes out of scope, it might
1552 need to be freed. For this we need to be able to find it, so assume that
1553 `var_value()` will provide that.
1555 ###### variable fields
1556 struct variable *merged;
1558 ###### ast functions
1560 static void variable_merge(struct variable *primary, struct variable *secondary)
1564 primary = primary->merged;
1566 for (v = primary->previous; v; v=v->previous)
1567 if (v == secondary || v == secondary->merged ||
1568 v->merged == secondary ||
1569 v->merged == secondary->merged) {
1570 v->scope = OutScope;
1571 v->merged = primary;
1572 if (v->scope_start < primary->scope_start)
1573 primary->scope_start = v->scope_start;
1574 if (v->scope_end > primary->scope_end)
1575 primary->scope_end = v->scope_end; // NOTEST
1576 variable_unlink_exec(v);
1580 ###### forward decls
1581 static struct value *var_value(struct parse_context *c, struct variable *v);
1583 ###### free global vars
1585 while (context.varlist) {
1586 struct binding *b = context.varlist;
1587 struct variable *v = b->var;
1588 context.varlist = b->next;
1591 struct variable *next = v->previous;
1593 if (v->global && v->frame_pos >= 0) {
1594 free_value(v->type, var_value(&context, v));
1595 if (v->depth == 0 && v->type->free == function_free)
1596 // This is a function constant
1597 free_exec(v->where_decl);
1604 #### Manipulating Bindings
1606 When a name is conditionally visible, a new declaration discards the old
1607 binding - the condition lapses. Similarly when we reach the end of a
1608 function (outermost non-global scope) any conditional scope must lapse.
1609 Conversely a usage of the name affirms the visibility and extends it to
1610 the end of the containing block - i.e. the block that contains both the
1611 original declaration and the latest usage. This is determined from
1612 `min_depth`. When a conditionally visible variable gets affirmed like
1613 this, it is also merged with other conditionally visible variables with
1616 When we parse a variable declaration we either report an error if the
1617 name is currently bound, or create a new variable at the current nest
1618 depth if the name is unbound or bound to a conditionally scoped or
1619 pending-scope variable. If the previous variable was conditionally
1620 scoped, it and its homonyms becomes out-of-scope.
1622 When we parse a variable reference (including non-declarative assignment
1623 "foo = bar") we report an error if the name is not bound or is bound to
1624 a pending-scope variable; update the scope if the name is bound to a
1625 conditionally scoped variable; or just proceed normally if the named
1626 variable is in scope.
1628 When we exit a scope, any variables bound at this level are either
1629 marked out of scope or pending-scoped, depending on whether the scope
1630 was sequential or parallel. Here a "parallel" scope means the "then"
1631 or "else" part of a conditional, or any "case" or "else" branch of a
1632 switch. Other scopes are "sequential".
1634 When exiting a parallel scope we check if there are any variables that
1635 were previously pending and are still visible. If there are, then
1636 they weren't redeclared in the most recent scope, so they cannot be
1637 merged and must become out-of-scope. If it is not the first of
1638 parallel scopes (based on `child_count`), we check that there was a
1639 previous binding that is still pending-scope. If there isn't, the new
1640 variable must now be out-of-scope.
1642 When exiting a sequential scope that immediately enclosed parallel
1643 scopes, we need to resolve any pending-scope variables. If there was
1644 no `else` clause, and we cannot determine that the `switch` was exhaustive,
1645 we need to mark all pending-scope variable as out-of-scope. Otherwise
1646 all pending-scope variables become conditionally scoped.
1649 enum closetype { CloseSequential, CloseFunction, CloseParallel, CloseElse };
1651 ###### ast functions
1653 static struct variable *var_decl(struct parse_context *c, struct text s)
1655 struct binding *b = find_binding(c, s);
1656 struct variable *v = b->var;
1658 switch (v ? v->scope : OutScope) {
1660 /* Caller will report the error */
1664 v && v->scope == CondScope;
1666 v->scope = OutScope;
1670 v = calloc(1, sizeof(*v));
1671 v->previous = b->var;
1675 v->min_depth = v->depth = c->scope_depth;
1677 v->in_scope = c->in_scope;
1678 v->scope_start = c->scope_count;
1684 static struct variable *var_ref(struct parse_context *c, struct text s)
1686 struct binding *b = find_binding(c, s);
1687 struct variable *v = b->var;
1688 struct variable *v2;
1690 switch (v ? v->scope : OutScope) {
1693 /* Caller will report the error */
1696 /* All CondScope variables of this name need to be merged
1697 * and become InScope
1699 v->depth = v->min_depth;
1701 for (v2 = v->previous;
1702 v2 && v2->scope == CondScope;
1704 variable_merge(v, v2);
1712 static int var_refile(struct parse_context *c, struct variable *v)
1714 /* Variable just went out of scope. Add it to the out_scope
1715 * list, sorted by ->scope_start
1717 struct variable **vp = &c->out_scope;
1718 while ((*vp) && (*vp)->scope_start < v->scope_start)
1719 vp = &(*vp)->in_scope;
1725 static void var_block_close(struct parse_context *c, enum closetype ct,
1728 /* Close off all variables that are in_scope.
1729 * Some variables in c->scope may already be not-in-scope,
1730 * such as when a PendingScope variable is hidden by a new
1731 * variable with the same name.
1732 * So we check for v->name->var != v and drop them.
1733 * If we choose to make a variable OutScope, we drop it
1736 struct variable *v, **vp, *v2;
1739 for (vp = &c->in_scope;
1740 (v = *vp) && v->min_depth > c->scope_depth;
1741 (v->scope == OutScope || v->name->var != v)
1742 ? (*vp = v->in_scope, var_refile(c, v))
1743 : ( vp = &v->in_scope, 0)) {
1744 v->min_depth = c->scope_depth;
1745 if (v->name->var != v)
1746 /* This is still in scope, but we haven't just
1750 v->min_depth = c->scope_depth;
1751 if (v->scope == InScope)
1752 v->scope_end = c->scope_count;
1753 if (v->scope == InScope && e && !v->global) {
1754 /* This variable gets cleaned up when 'e' finishes */
1755 variable_unlink_exec(v);
1756 v->cleanup_exec = e;
1757 v->next_free = e->to_free;
1762 case CloseParallel: /* handle PendingScope */
1766 if (c->scope_stack->child_count == 1)
1767 /* first among parallel branches */
1768 v->scope = PendingScope;
1769 else if (v->previous &&
1770 v->previous->scope == PendingScope)
1771 /* all previous branches used name */
1772 v->scope = PendingScope;
1773 else if (v->type == Tlabel)
1774 /* Labels remain pending even when not used */
1775 v->scope = PendingScope; // UNTESTED
1777 v->scope = OutScope;
1778 if (ct == CloseElse) {
1779 /* All Pending variables with this name
1780 * are now Conditional */
1782 v2 && v2->scope == PendingScope;
1784 v2->scope = CondScope;
1788 /* Not possible as it would require
1789 * parallel scope to be nested immediately
1790 * in a parallel scope, and that never
1794 /* Not possible as we already tested for
1801 if (v->scope == CondScope)
1802 /* Condition cannot continue past end of function */
1805 case CloseSequential:
1806 if (v->type == Tlabel)
1807 v->scope = PendingScope;
1810 v->scope = OutScope;
1813 /* There was no 'else', so we can only become
1814 * conditional if we know the cases were exhaustive,
1815 * and that doesn't mean anything yet.
1816 * So only labels become conditional..
1819 v2 && v2->scope == PendingScope;
1821 if (v2->type == Tlabel)
1822 v2->scope = CondScope;
1824 v2->scope = OutScope;
1827 case OutScope: break;
1836 The value of a variable is store separately from the variable, on an
1837 analogue of a stack frame. There are (currently) two frames that can be
1838 active. A global frame which currently only stores constants, and a
1839 stacked frame which stores local variables. Each variable knows if it
1840 is global or not, and what its index into the frame is.
1842 Values in the global frame are known immediately they are relevant, so
1843 the frame needs to be reallocated as it grows so it can store those
1844 values. The local frame doesn't get values until the interpreted phase
1845 is started, so there is no need to allocate until the size is known.
1847 We initialize the `frame_pos` to an impossible value, so that we can
1848 tell if it was set or not later.
1850 ###### variable fields
1854 ###### variable init
1857 ###### parse context
1859 short global_size, global_alloc;
1861 void *global, *local;
1863 ###### forward decls
1864 static struct value *global_alloc(struct parse_context *c, struct type *t,
1865 struct variable *v, struct value *init);
1867 ###### ast functions
1869 static struct value *var_value(struct parse_context *c, struct variable *v)
1872 if (!c->local || !v->type)
1874 if (v->frame_pos + v->type->size > c->local_size) {
1875 printf("INVALID frame_pos\n"); // NOTEST
1878 return c->local + v->frame_pos;
1880 if (c->global_size > c->global_alloc) {
1881 int old = c->global_alloc;
1882 c->global_alloc = (c->global_size | 1023) + 1024;
1883 c->global = realloc(c->global, c->global_alloc);
1884 memset(c->global + old, 0, c->global_alloc - old);
1886 return c->global + v->frame_pos;
1889 static struct value *global_alloc(struct parse_context *c, struct type *t,
1890 struct variable *v, struct value *init)
1893 struct variable scratch;
1895 if (t->prepare_type)
1896 t->prepare_type(c, t, 1); // NOTEST
1898 if (c->global_size & (t->align - 1))
1899 c->global_size = (c->global_size + t->align) & ~(t->align-1); // NOTEST
1904 v->frame_pos = c->global_size;
1906 c->global_size += v->type->size;
1907 ret = var_value(c, v);
1909 memcpy(ret, init, t->size);
1915 As global values are found -- struct field initializers, labels etc --
1916 `global_alloc()` is called to record the value in the global frame.
1918 When the program is fully parsed, each function is analysed, we need to
1919 walk the list of variables local to that function and assign them an
1920 offset in the stack frame. For this we have `scope_finalize()`.
1922 We keep the stack from dense by re-using space for between variables
1923 that are not in scope at the same time. The `out_scope` list is sorted
1924 by `scope_start` and as we process a varible, we move it to an FIFO
1925 stack. For each variable we consider, we first discard any from the
1926 stack anything that went out of scope before the new variable came in.
1927 Then we place the new variable just after the one at the top of the
1930 ###### ast functions
1932 static void scope_finalize(struct parse_context *c, struct type *ft)
1934 int size = ft->function.local_size;
1935 struct variable *next = ft->function.scope;
1936 struct variable *done = NULL;
1939 struct variable *v = next;
1940 struct type *t = v->type;
1947 if (v->frame_pos >= 0)
1949 while (done && done->scope_end < v->scope_start)
1950 done = done->in_scope;
1952 pos = done->frame_pos + done->type->size;
1954 pos = ft->function.local_size;
1955 if (pos & (t->align - 1))
1956 pos = (pos + t->align) & ~(t->align-1);
1958 if (size < pos + v->type->size)
1959 size = pos + v->type->size;
1963 c->out_scope = NULL;
1964 ft->function.local_size = size;
1967 ###### free context storage
1968 free(context.global);
1970 #### Variables as executables
1972 Just as we used a `val` to wrap a value into an `exec`, we similarly
1973 need a `var` to wrap a `variable` into an exec. While each `val`
1974 contained a copy of the value, each `var` holds a link to the variable
1975 because it really is the same variable no matter where it appears.
1976 When a variable is used, we need to remember to follow the `->merged`
1977 link to find the primary instance.
1979 When a variable is declared, it may or may not be given an explicit
1980 type. We need to record which so that we can report the parsed code
1989 struct variable *var;
1992 ###### variable fields
2000 VariableDecl -> IDENTIFIER : ${ {
2001 struct variable *v = var_decl(c, $1.txt);
2002 $0 = new_pos(var, $1);
2007 v = var_ref(c, $1.txt);
2009 type_err(c, "error: variable '%v' redeclared",
2011 type_err(c, "info: this is where '%v' was first declared",
2012 v->where_decl, NULL, 0, NULL);
2015 | IDENTIFIER :: ${ {
2016 struct variable *v = var_decl(c, $1.txt);
2017 $0 = new_pos(var, $1);
2023 v = var_ref(c, $1.txt);
2025 type_err(c, "error: variable '%v' redeclared",
2027 type_err(c, "info: this is where '%v' was first declared",
2028 v->where_decl, NULL, 0, NULL);
2031 | IDENTIFIER : Type ${ {
2032 struct variable *v = var_decl(c, $1.txt);
2033 $0 = new_pos(var, $1);
2039 v->explicit_type = 1;
2041 v = var_ref(c, $1.txt);
2043 type_err(c, "error: variable '%v' redeclared",
2045 type_err(c, "info: this is where '%v' was first declared",
2046 v->where_decl, NULL, 0, NULL);
2049 | IDENTIFIER :: Type ${ {
2050 struct variable *v = var_decl(c, $1.txt);
2051 $0 = new_pos(var, $1);
2058 v->explicit_type = 1;
2060 v = var_ref(c, $1.txt);
2062 type_err(c, "error: variable '%v' redeclared",
2064 type_err(c, "info: this is where '%v' was first declared",
2065 v->where_decl, NULL, 0, NULL);
2070 Variable -> IDENTIFIER ${ {
2071 struct variable *v = var_ref(c, $1.txt);
2072 $0 = new_pos(var, $1);
2074 /* This might be a label - allocate a var just in case */
2075 v = var_decl(c, $1.txt);
2082 cast(var, $0)->var = v;
2085 ###### print exec cases
2088 struct var *v = cast(var, e);
2090 struct binding *b = v->var->name;
2091 printf("%.*s", b->name.len, b->name.txt);
2098 if (loc && loc->type == Xvar) {
2099 struct var *v = cast(var, loc);
2101 struct binding *b = v->var->name;
2102 fprintf(stderr, "%.*s", b->name.len, b->name.txt);
2104 fputs("???", stderr); // NOTEST
2106 fputs("NOTVAR", stderr);
2109 ###### propagate exec cases
2113 struct var *var = cast(var, prog);
2114 struct variable *v = var->var;
2116 type_err(c, "%d:BUG: no variable!!", prog, NULL, 0, NULL); // NOTEST
2117 return Tnone; // NOTEST
2120 if (v->constant && (rules & Rnoconstant)) {
2121 type_err(c, "error: Cannot assign to a constant: %v",
2122 prog, NULL, 0, NULL);
2123 type_err(c, "info: name was defined as a constant here",
2124 v->where_decl, NULL, 0, NULL);
2127 if (v->type == Tnone && v->where_decl == prog)
2128 type_err(c, "error: variable used but not declared: %v",
2129 prog, NULL, 0, NULL);
2130 if (v->type == NULL) {
2131 if (type && *ok != 0) {
2133 v->where_set = prog;
2138 if (!type_compat(type, v->type, rules)) {
2139 type_err(c, "error: expected %1%r but variable '%v' is %2", prog,
2140 type, rules, v->type);
2141 type_err(c, "info: this is where '%v' was set to %1", v->where_set,
2142 v->type, rules, NULL);
2149 ###### interp exec cases
2152 struct var *var = cast(var, e);
2153 struct variable *v = var->var;
2156 lrv = var_value(c, v);
2161 ###### ast functions
2163 static void free_var(struct var *v)
2168 ###### free exec cases
2169 case Xvar: free_var(cast(var, e)); break;
2174 Now that we have the shape of the interpreter in place we can add some
2175 complex types and connected them in to the data structures and the
2176 different phases of parse, analyse, print, interpret.
2178 Being "complex" the language will naturally have syntax to access
2179 specifics of objects of these types. These will fit into the grammar as
2180 "Terms" which are the things that are combined with various operators to
2181 form "Expression". Where a Term is formed by some operation on another
2182 Term, the subordinate Term will always come first, so for example a
2183 member of an array will be expressed as the Term for the array followed
2184 by an index in square brackets. The strict rule of using postfix
2185 operations makes precedence irrelevant within terms. To provide a place
2186 to put the grammar for each terms of each type, we will start out by
2187 introducing the "Term" grammar production, with contains at least a
2188 simple "Value" (to be explained later).
2192 Term -> Value ${ $0 = $<1; }$
2193 | Variable ${ $0 = $<1; }$
2196 Thus far the complex types we have are arrays and structs.
2200 Arrays can be declared by giving a size and a type, as `[size]type' so
2201 `freq:[26]number` declares `freq` to be an array of 26 numbers. The
2202 size can be either a literal number, or a named constant. Some day an
2203 arbitrary expression will be supported.
2205 As a formal parameter to a function, the array can be declared with a
2206 new variable as the size: `name:[size::number]string`. The `size`
2207 variable is set to the size of the array and must be a constant. As
2208 `number` is the only supported type, it can be left out:
2209 `name:[size::]string`.
2211 Arrays cannot be assigned. When pointers are introduced we will also
2212 introduce array slices which can refer to part or all of an array -
2213 the assignment syntax will create a slice. For now, an array can only
2214 ever be referenced by the name it is declared with. It is likely that
2215 a "`copy`" primitive will eventually be define which can be used to
2216 make a copy of an array with controllable recursive depth.
2218 For now we have two sorts of array, those with fixed size either because
2219 it is given as a literal number or because it is a struct member (which
2220 cannot have a runtime-changing size), and those with a size that is
2221 determined at runtime - local variables with a const size. The former
2222 have their size calculated at parse time, the latter at run time.
2224 For the latter type, the `size` field of the type is the size of a
2225 pointer, and the array is reallocated every time it comes into scope.
2227 We differentiate struct fields with a const size from local variables
2228 with a const size by whether they are prepared at parse time or not.
2230 ###### type union fields
2233 int unspec; // size is unspecified - vsize must be set.
2236 struct variable *vsize;
2237 struct type *member;
2240 ###### value union fields
2241 void *array; // used if not static_size
2243 ###### value functions
2245 static void array_prepare_type(struct parse_context *c, struct type *type,
2248 struct value *vsize;
2250 if (type->array.static_size)
2252 if (type->array.unspec && parse_time)
2255 if (type->array.vsize) {
2256 vsize = var_value(c, type->array.vsize);
2260 mpz_tdiv_q(q, mpq_numref(vsize->num), mpq_denref(vsize->num));
2261 type->array.size = mpz_get_si(q);
2265 if (parse_time && type->array.member->size) {
2266 type->array.static_size = 1;
2267 type->size = type->array.size * type->array.member->size;
2268 type->align = type->array.member->align;
2272 static void array_init(struct type *type, struct value *val)
2275 void *ptr = val->ptr;
2279 if (!type->array.static_size) {
2280 val->array = calloc(type->array.size,
2281 type->array.member->size);
2284 for (i = 0; i < type->array.size; i++) {
2286 v = (void*)ptr + i * type->array.member->size;
2287 val_init(type->array.member, v);
2291 static void array_free(struct type *type, struct value *val)
2294 void *ptr = val->ptr;
2296 if (!type->array.static_size)
2298 for (i = 0; i < type->array.size; i++) {
2300 v = (void*)ptr + i * type->array.member->size;
2301 free_value(type->array.member, v);
2303 if (!type->array.static_size)
2307 static int array_compat(struct type *require, struct type *have)
2309 if (have->compat != require->compat)
2311 /* Both are arrays, so we can look at details */
2312 if (!type_compat(require->array.member, have->array.member, 0))
2314 if (have->array.unspec && require->array.unspec) {
2315 if (have->array.vsize && require->array.vsize &&
2316 have->array.vsize != require->array.vsize) // UNTESTED
2317 /* sizes might not be the same */
2318 return 0; // UNTESTED
2321 if (have->array.unspec || require->array.unspec)
2322 return 1; // UNTESTED
2323 if (require->array.vsize == NULL && have->array.vsize == NULL)
2324 return require->array.size == have->array.size;
2326 return require->array.vsize == have->array.vsize; // UNTESTED
2329 static void array_print_type(struct type *type, FILE *f)
2332 if (type->array.vsize) {
2333 struct binding *b = type->array.vsize->name;
2334 fprintf(f, "%.*s%s]", b->name.len, b->name.txt,
2335 type->array.unspec ? "::" : "");
2336 } else if (type->array.size)
2337 fprintf(f, "%d]", type->array.size);
2340 type_print(type->array.member, f);
2343 static struct type array_prototype = {
2345 .prepare_type = array_prepare_type,
2346 .print_type = array_print_type,
2347 .compat = array_compat,
2349 .size = sizeof(void*),
2350 .align = sizeof(void*),
2353 ###### declare terminals
2358 | [ NUMBER ] Type ${ {
2364 if (number_parse(num, tail, $2.txt) == 0)
2365 tok_err(c, "error: unrecognised number", &$2);
2367 tok_err(c, "error: unsupported number suffix", &$2);
2370 elements = mpz_get_ui(mpq_numref(num));
2371 if (mpz_cmp_ui(mpq_denref(num), 1) != 0) {
2372 tok_err(c, "error: array size must be an integer",
2374 } else if (mpz_cmp_ui(mpq_numref(num), 1UL << 30) >= 0)
2375 tok_err(c, "error: array size is too large",
2380 $0 = t = add_anon_type(c, &array_prototype, "array[%d]", elements );
2381 t->array.size = elements;
2382 t->array.member = $<4;
2383 t->array.vsize = NULL;
2386 | [ IDENTIFIER ] Type ${ {
2387 struct variable *v = var_ref(c, $2.txt);
2390 tok_err(c, "error: name undeclared", &$2);
2391 else if (!v->constant)
2392 tok_err(c, "error: array size must be a constant", &$2);
2394 $0 = add_anon_type(c, &array_prototype, "array[%.*s]", $2.txt.len, $2.txt.txt);
2395 $0->array.member = $<4;
2397 $0->array.vsize = v;
2402 OptType -> Type ${ $0 = $<1; }$
2405 ###### formal type grammar
2407 | [ IDENTIFIER :: OptType ] Type ${ {
2408 struct variable *v = var_decl(c, $ID.txt);
2414 $0 = add_anon_type(c, &array_prototype, "array[var]");
2415 $0->array.member = $<6;
2417 $0->array.unspec = 1;
2418 $0->array.vsize = v;
2426 | Term [ Expression ] ${ {
2427 struct binode *b = new(binode);
2434 ###### print binode cases
2436 print_exec(b->left, -1, bracket);
2438 print_exec(b->right, -1, bracket);
2442 ###### propagate binode cases
2444 /* left must be an array, right must be a number,
2445 * result is the member type of the array
2447 propagate_types(b->right, c, ok, Tnum, 0);
2448 t = propagate_types(b->left, c, ok, NULL, rules & Rnoconstant);
2449 if (!t || t->compat != array_compat) {
2450 type_err(c, "error: %1 cannot be indexed", prog, t, 0, NULL);
2453 if (!type_compat(type, t->array.member, rules)) {
2454 type_err(c, "error: have %1 but need %2", prog,
2455 t->array.member, rules, type);
2457 return t->array.member;
2461 ###### interp binode cases
2467 lleft = linterp_exec(c, b->left, <ype);
2468 right = interp_exec(c, b->right, &rtype);
2470 mpz_tdiv_q(q, mpq_numref(right.num), mpq_denref(right.num));
2474 if (ltype->array.static_size)
2477 ptr = *(void**)lleft;
2478 rvtype = ltype->array.member;
2479 if (i >= 0 && i < ltype->array.size)
2480 lrv = ptr + i * rvtype->size;
2482 val_init(ltype->array.member, &rv); // UNSAFE
2489 A `struct` is a data-type that contains one or more other data-types.
2490 It differs from an array in that each member can be of a different
2491 type, and they are accessed by name rather than by number. Thus you
2492 cannot choose an element by calculation, you need to know what you
2495 The language makes no promises about how a given structure will be
2496 stored in memory - it is free to rearrange fields to suit whatever
2497 criteria seems important.
2499 Structs are declared separately from program code - they cannot be
2500 declared in-line in a variable declaration like arrays can. A struct
2501 is given a name and this name is used to identify the type - the name
2502 is not prefixed by the word `struct` as it would be in C.
2504 Structs are only treated as the same if they have the same name.
2505 Simply having the same fields in the same order is not enough. This
2506 might change once we can create structure initializers from a list of
2509 Each component datum is identified much like a variable is declared,
2510 with a name, one or two colons, and a type. The type cannot be omitted
2511 as there is no opportunity to deduce the type from usage. An initial
2512 value can be given following an equals sign, so
2514 ##### Example: a struct type
2520 would declare a type called "complex" which has two number fields,
2521 each initialised to zero.
2523 Struct will need to be declared separately from the code that uses
2524 them, so we will need to be able to print out the declaration of a
2525 struct when reprinting the whole program. So a `print_type_decl` type
2526 function will be needed.
2528 ###### type union fields
2537 } *fields; // This is created when field_list is analysed.
2539 struct fieldlist *prev;
2542 } *field_list; // This is created during parsing
2545 ###### type functions
2546 void (*print_type_decl)(struct type *type, FILE *f);
2548 ###### value functions
2550 static void structure_init(struct type *type, struct value *val)
2554 for (i = 0; i < type->structure.nfields; i++) {
2556 v = (void*) val->ptr + type->structure.fields[i].offset;
2557 if (type->structure.fields[i].init)
2558 dup_value(type->structure.fields[i].type,
2559 type->structure.fields[i].init,
2562 val_init(type->structure.fields[i].type, v);
2566 static void structure_free(struct type *type, struct value *val)
2570 for (i = 0; i < type->structure.nfields; i++) {
2572 v = (void*)val->ptr + type->structure.fields[i].offset;
2573 free_value(type->structure.fields[i].type, v);
2577 static void free_fieldlist(struct fieldlist *f)
2581 free_fieldlist(f->prev);
2586 static void structure_free_type(struct type *t)
2589 for (i = 0; i < t->structure.nfields; i++)
2590 if (t->structure.fields[i].init) {
2591 free_value(t->structure.fields[i].type,
2592 t->structure.fields[i].init);
2594 free(t->structure.fields);
2595 free_fieldlist(t->structure.field_list);
2598 static void structure_prepare_type(struct parse_context *c,
2599 struct type *t, int parse_time)
2602 struct fieldlist *f;
2604 if (!parse_time || t->structure.fields)
2607 for (f = t->structure.field_list; f; f=f->prev) {
2611 if (f->f.type->prepare_type)
2612 f->f.type->prepare_type(c, f->f.type, 1);
2613 if (f->init == NULL)
2617 propagate_types(f->init, c, &ok, f->f.type, 0);
2620 c->parse_error = 1; // NOTEST
2623 t->structure.nfields = cnt;
2624 t->structure.fields = calloc(cnt, sizeof(struct field));
2625 f = t->structure.field_list;
2627 int a = f->f.type->align;
2629 t->structure.fields[cnt] = f->f;
2630 if (t->size & (a-1))
2631 t->size = (t->size | (a-1)) + 1;
2632 t->structure.fields[cnt].offset = t->size;
2633 t->size += ((f->f.type->size - 1) | (a-1)) + 1;
2637 if (f->init && !c->parse_error) {
2638 struct value vl = interp_exec(c, f->init, NULL);
2639 t->structure.fields[cnt].init =
2640 global_alloc(c, f->f.type, NULL, &vl);
2647 static struct type structure_prototype = {
2648 .init = structure_init,
2649 .free = structure_free,
2650 .free_type = structure_free_type,
2651 .print_type_decl = structure_print_type,
2652 .prepare_type = structure_prepare_type,
2666 ###### free exec cases
2668 free_exec(cast(fieldref, e)->left);
2672 ###### declare terminals
2677 | Term . IDENTIFIER ${ {
2678 struct fieldref *fr = new_pos(fieldref, $2);
2685 ###### print exec cases
2689 struct fieldref *f = cast(fieldref, e);
2690 print_exec(f->left, -1, bracket);
2691 printf(".%.*s", f->name.len, f->name.txt);
2695 ###### ast functions
2696 static int find_struct_index(struct type *type, struct text field)
2699 for (i = 0; i < type->structure.nfields; i++)
2700 if (text_cmp(type->structure.fields[i].name, field) == 0)
2705 ###### propagate exec cases
2709 struct fieldref *f = cast(fieldref, prog);
2710 struct type *st = propagate_types(f->left, c, ok, NULL, 0);
2713 type_err(c, "error: unknown type for field access", f->left, // UNTESTED
2715 else if (st->init != structure_init)
2716 type_err(c, "error: field reference attempted on %1, not a struct",
2717 f->left, st, 0, NULL);
2718 else if (f->index == -2) {
2719 f->index = find_struct_index(st, f->name);
2721 type_err(c, "error: cannot find requested field in %1",
2722 f->left, st, 0, NULL);
2724 if (f->index >= 0) {
2725 struct type *ft = st->structure.fields[f->index].type;
2726 if (!type_compat(type, ft, rules))
2727 type_err(c, "error: have %1 but need %2", prog,
2734 ###### interp exec cases
2737 struct fieldref *f = cast(fieldref, e);
2739 struct value *lleft = linterp_exec(c, f->left, <ype);
2740 lrv = (void*)lleft->ptr + ltype->structure.fields[f->index].offset;
2741 rvtype = ltype->structure.fields[f->index].type;
2745 ###### top level grammar
2746 DeclareStruct -> struct IDENTIFIER FieldBlock Newlines ${ {
2748 add_type(c, $2.txt, &structure_prototype);
2749 t->structure.field_list = $<FB;
2753 FieldBlock -> { IN OptNL FieldLines OUT OptNL } ${ $0 = $<FL; }$
2754 | { SimpleFieldList } ${ $0 = $<SFL; }$
2755 | IN OptNL FieldLines OUT ${ $0 = $<FL; }$
2756 | SimpleFieldList EOL ${ $0 = $<SFL; }$
2758 FieldLines -> SimpleFieldList Newlines ${ $0 = $<SFL; }$
2759 | FieldLines SimpleFieldList Newlines ${
2764 SimpleFieldList -> Field ${ $0 = $<F; }$
2765 | SimpleFieldList ; Field ${
2769 | SimpleFieldList ; ${
2772 | ERROR ${ tok_err(c, "Syntax error in struct field", &$1); }$
2774 Field -> IDENTIFIER : Type = Expression ${ {
2775 $0 = calloc(1, sizeof(struct fieldlist));
2776 $0->f.name = $ID.txt;
2777 $0->f.type = $<Type;
2781 | IDENTIFIER : Type ${
2782 $0 = calloc(1, sizeof(struct fieldlist));
2783 $0->f.name = $ID.txt;
2784 $0->f.type = $<Type;
2787 ###### forward decls
2788 static void structure_print_type(struct type *t, FILE *f);
2790 ###### value functions
2791 static void structure_print_type(struct type *t, FILE *f)
2795 fprintf(f, "struct %.*s\n", t->name.len, t->name.txt);
2797 for (i = 0; i < t->structure.nfields; i++) {
2798 struct field *fl = t->structure.fields + i;
2799 fprintf(f, " %.*s : ", fl->name.len, fl->name.txt);
2800 type_print(fl->type, f);
2801 if (fl->type->print && fl->init) {
2803 if (fl->type == Tstr)
2804 fprintf(f, "\""); // UNTESTED
2805 print_value(fl->type, fl->init, f);
2806 if (fl->type == Tstr)
2807 fprintf(f, "\""); // UNTESTED
2813 ###### print type decls
2818 while (target != 0) {
2820 for (t = context.typelist; t ; t=t->next)
2821 if (!t->anon && t->print_type_decl &&
2831 t->print_type_decl(t, stdout);
2839 A function is a chunk of code which can be passed parameters and can
2840 return results. Each function has a type which includes the set of
2841 parameters and the return value. As yet these types cannot be declared
2842 separately from the function itself.
2844 The parameters can be specified either in parentheses as a ';' separated
2847 ##### Example: function 1
2849 func main(av:[ac::number]string; env:[envc::number]string)
2852 or as an indented list of one parameter per line (though each line can
2853 be a ';' separated list)
2855 ##### Example: function 2
2858 argv:[argc::number]string
2859 env:[envc::number]string
2863 In the first case a return type can follow the parentheses after a colon,
2864 in the second it is given on a line starting with the word `return`.
2866 ##### Example: functions that return
2868 func add(a:number; b:number): number
2878 Rather than returning a type, the function can specify a set of local
2879 variables to return as a struct. The values of these variables when the
2880 function exits will be provided to the caller. For this the return type
2881 is replaced with a block of result declarations, either in parentheses
2882 or bracketed by `return` and `do`.
2884 ##### Example: functions returning multiple variables
2886 func to_cartesian(rho:number; theta:number):(x:number; y:number)
2899 For constructing the lists we use a `List` binode, which will be
2900 further detailed when Expression Lists are introduced.
2902 ###### type union fields
2905 struct binode *params;
2906 struct type *return_type;
2907 struct variable *scope;
2908 int inline_result; // return value is at start of 'local'
2912 ###### value union fields
2913 struct exec *function;
2915 ###### type functions
2916 void (*check_args)(struct parse_context *c, int *ok,
2917 struct type *require, struct exec *args);
2919 ###### value functions
2921 static void function_free(struct type *type, struct value *val)
2923 free_exec(val->function);
2924 val->function = NULL;
2927 static int function_compat(struct type *require, struct type *have)
2929 // FIXME can I do anything here yet?
2933 static void function_check_args(struct parse_context *c, int *ok,
2934 struct type *require, struct exec *args)
2936 /* This should be 'compat', but we don't have a 'tuple' type to
2937 * hold the type of 'args'
2939 struct binode *arg = cast(binode, args);
2940 struct binode *param = require->function.params;
2943 struct var *pv = cast(var, param->left);
2945 type_err(c, "error: insufficient arguments to function.",
2946 args, NULL, 0, NULL);
2950 propagate_types(arg->left, c, ok, pv->var->type, 0);
2951 param = cast(binode, param->right);
2952 arg = cast(binode, arg->right);
2955 type_err(c, "error: too many arguments to function.",
2956 args, NULL, 0, NULL);
2959 static void function_print(struct type *type, struct value *val, FILE *f)
2961 print_exec(val->function, 1, 0);
2964 static void function_print_type_decl(struct type *type, FILE *f)
2968 for (b = type->function.params; b; b = cast(binode, b->right)) {
2969 struct variable *v = cast(var, b->left)->var;
2970 fprintf(f, "%.*s%s", v->name->name.len, v->name->name.txt,
2971 v->constant ? "::" : ":");
2972 type_print(v->type, f);
2977 if (type->function.return_type != Tnone) {
2979 if (type->function.inline_result) {
2981 struct type *t = type->function.return_type;
2983 for (i = 0; i < t->structure.nfields; i++) {
2984 struct field *fl = t->structure.fields + i;
2987 fprintf(f, "%.*s:", fl->name.len, fl->name.txt);
2988 type_print(fl->type, f);
2992 type_print(type->function.return_type, f);
2997 static void function_free_type(struct type *t)
2999 free_exec(t->function.params);
3002 static struct type function_prototype = {
3003 .size = sizeof(void*),
3004 .align = sizeof(void*),
3005 .free = function_free,
3006 .compat = function_compat,
3007 .check_args = function_check_args,
3008 .print = function_print,
3009 .print_type_decl = function_print_type_decl,
3010 .free_type = function_free_type,
3013 ###### declare terminals
3023 FuncName -> IDENTIFIER ${ {
3024 struct variable *v = var_decl(c, $1.txt);
3025 struct var *e = new_pos(var, $1);
3031 v = var_ref(c, $1.txt);
3033 type_err(c, "error: function '%v' redeclared",
3035 type_err(c, "info: this is where '%v' was first declared",
3036 v->where_decl, NULL, 0, NULL);
3042 Args -> ArgsLine NEWLINE ${ $0 = $<AL; }$
3043 | Args ArgsLine NEWLINE ${ {
3044 struct binode *b = $<AL;
3045 struct binode **bp = &b;
3047 bp = (struct binode **)&(*bp)->left;
3052 ArgsLine -> ${ $0 = NULL; }$
3053 | Varlist ${ $0 = $<1; }$
3054 | Varlist ; ${ $0 = $<1; }$
3056 Varlist -> Varlist ; ArgDecl ${
3070 ArgDecl -> IDENTIFIER : FormalType ${ {
3071 struct variable *v = var_decl(c, $1.txt);
3077 ##### Function calls
3079 A function call can appear either as an expression or as a statement.
3080 We use a new 'Funcall' binode type to link the function with a list of
3081 arguments, form with the 'List' nodes.
3083 We have already seen the "Term" which is how a function call can appear
3084 in an expression. To parse a function call into a statement we include
3085 it in the "SimpleStatement Grammar" which will be described later.
3091 | Term ( ExpressionList ) ${ {
3092 struct binode *b = new(binode);
3095 b->right = reorder_bilist($<EL);
3099 struct binode *b = new(binode);
3106 ###### SimpleStatement Grammar
3108 | Term ( ExpressionList ) ${ {
3109 struct binode *b = new(binode);
3112 b->right = reorder_bilist($<EL);
3116 ###### print binode cases
3119 do_indent(indent, "");
3120 print_exec(b->left, -1, bracket);
3122 for (b = cast(binode, b->right); b; b = cast(binode, b->right)) {
3125 print_exec(b->left, -1, bracket);
3135 ###### propagate binode cases
3138 /* Every arg must match formal parameter, and result
3139 * is return type of function
3141 struct binode *args = cast(binode, b->right);
3142 struct var *v = cast(var, b->left);
3144 if (!v->var->type || v->var->type->check_args == NULL) {
3145 type_err(c, "error: attempt to call a non-function.",
3146 prog, NULL, 0, NULL);
3149 v->var->type->check_args(c, ok, v->var->type, args);
3150 return v->var->type->function.return_type;
3153 ###### interp binode cases
3156 struct var *v = cast(var, b->left);
3157 struct type *t = v->var->type;
3158 void *oldlocal = c->local;
3159 int old_size = c->local_size;
3160 void *local = calloc(1, t->function.local_size);
3161 struct value *fbody = var_value(c, v->var);
3162 struct binode *arg = cast(binode, b->right);
3163 struct binode *param = t->function.params;
3166 struct var *pv = cast(var, param->left);
3167 struct type *vtype = NULL;
3168 struct value val = interp_exec(c, arg->left, &vtype);
3170 c->local = local; c->local_size = t->function.local_size;
3171 lval = var_value(c, pv->var);
3172 c->local = oldlocal; c->local_size = old_size;
3173 memcpy(lval, &val, vtype->size);
3174 param = cast(binode, param->right);
3175 arg = cast(binode, arg->right);
3177 c->local = local; c->local_size = t->function.local_size;
3178 if (t->function.inline_result && dtype) {
3179 _interp_exec(c, fbody->function, NULL, NULL);
3180 memcpy(dest, local, dtype->size);
3181 rvtype = ret.type = NULL;
3183 rv = interp_exec(c, fbody->function, &rvtype);
3184 c->local = oldlocal; c->local_size = old_size;
3189 ## Complex executables: statements and expressions
3191 Now that we have types and values and variables and most of the basic
3192 Terms which provide access to these, we can explore the more complex
3193 code that combine all of these to get useful work done. Specifically
3194 statements and expressions.
3196 Expressions are various combinations of Terms. We will use operator
3197 precedence to ensure correct parsing. The simplest Expression is just a
3198 Term - others will follow.
3203 Expression -> Term ${ $0 = $<Term; }$
3204 ## expression grammar
3206 ### Expressions: Conditional
3208 Our first user of the `binode` will be conditional expressions, which
3209 is a bit odd as they actually have three components. That will be
3210 handled by having 2 binodes for each expression. The conditional
3211 expression is the lowest precedence operator which is why we define it
3212 first - to start the precedence list.
3214 Conditional expressions are of the form "value `if` condition `else`
3215 other_value". They associate to the right, so everything to the right
3216 of `else` is part of an else value, while only a higher-precedence to
3217 the left of `if` is the if values. Between `if` and `else` there is no
3218 room for ambiguity, so a full conditional expression is allowed in
3224 ###### declare terminals
3228 ###### expression grammar
3230 | Expression if Expression else Expression $$ifelse ${ {
3231 struct binode *b1 = new(binode);
3232 struct binode *b2 = new(binode);
3242 ###### print binode cases
3245 b2 = cast(binode, b->right);
3246 if (bracket) printf("(");
3247 print_exec(b2->left, -1, bracket);
3249 print_exec(b->left, -1, bracket);
3251 print_exec(b2->right, -1, bracket);
3252 if (bracket) printf(")");
3255 ###### propagate binode cases
3258 /* cond must be Tbool, others must match */
3259 struct binode *b2 = cast(binode, b->right);
3262 propagate_types(b->left, c, ok, Tbool, 0);
3263 t = propagate_types(b2->left, c, ok, type, Rnolabel);
3264 t2 = propagate_types(b2->right, c, ok, type ?: t, Rnolabel);
3268 ###### interp binode cases
3271 struct binode *b2 = cast(binode, b->right);
3272 left = interp_exec(c, b->left, <ype);
3274 rv = interp_exec(c, b2->left, &rvtype); // UNTESTED
3276 rv = interp_exec(c, b2->right, &rvtype);
3282 We take a brief detour, now that we have expressions, to describe lists
3283 of expressions. These will be needed for function parameters and
3284 possibly other situations. They seem generic enough to introduce here
3285 to be used elsewhere.
3287 And ExpressionList will use the `List` type of `binode`, building up at
3288 the end. And place where they are used will probably call
3289 `reorder_bilist()` to get a more normal first/next arrangement.
3291 ###### declare terminals
3294 `List` execs have no implicit semantics, so they are never propagated or
3295 interpreted. The can be printed as a comma separate list, which is how
3296 they are parsed. Note they are also used for function formal parameter
3297 lists. In that case a separate function is used to print them.
3299 ###### print binode cases
3303 print_exec(b->left, -1, bracket);
3306 b = cast(binode, b->right);
3310 ###### propagate binode cases
3311 case List: abort(); // NOTEST
3312 ###### interp binode cases
3313 case List: abort(); // NOTEST
3318 ExpressionList -> ExpressionList , Expression ${
3331 ### Expressions: Boolean
3333 The next class of expressions to use the `binode` will be Boolean
3334 expressions. "`and then`" and "`or else`" are similar to `and` and `or`
3335 have same corresponding precendence. The difference is that they don't
3336 evaluate the second expression if not necessary.
3345 ###### declare terminals
3350 ###### expression grammar
3351 | Expression or Expression ${ {
3352 struct binode *b = new(binode);
3358 | Expression or else Expression ${ {
3359 struct binode *b = new(binode);
3366 | Expression and Expression ${ {
3367 struct binode *b = new(binode);
3373 | Expression and then Expression ${ {
3374 struct binode *b = new(binode);
3381 | not Expression ${ {
3382 struct binode *b = new(binode);
3388 ###### print binode cases
3390 if (bracket) printf("(");
3391 print_exec(b->left, -1, bracket);
3393 print_exec(b->right, -1, bracket);
3394 if (bracket) printf(")");
3397 if (bracket) printf("(");
3398 print_exec(b->left, -1, bracket);
3399 printf(" and then ");
3400 print_exec(b->right, -1, bracket);
3401 if (bracket) printf(")");
3404 if (bracket) printf("(");
3405 print_exec(b->left, -1, bracket);
3407 print_exec(b->right, -1, bracket);
3408 if (bracket) printf(")");
3411 if (bracket) printf("(");
3412 print_exec(b->left, -1, bracket);
3413 printf(" or else ");
3414 print_exec(b->right, -1, bracket);
3415 if (bracket) printf(")");
3418 if (bracket) printf("(");
3420 print_exec(b->right, -1, bracket);
3421 if (bracket) printf(")");
3424 ###### propagate binode cases
3430 /* both must be Tbool, result is Tbool */
3431 propagate_types(b->left, c, ok, Tbool, 0);
3432 propagate_types(b->right, c, ok, Tbool, 0);
3433 if (type && type != Tbool)
3434 type_err(c, "error: %1 operation found where %2 expected", prog,
3438 ###### interp binode cases
3440 rv = interp_exec(c, b->left, &rvtype);
3441 right = interp_exec(c, b->right, &rtype);
3442 rv.bool = rv.bool && right.bool;
3445 rv = interp_exec(c, b->left, &rvtype);
3447 rv = interp_exec(c, b->right, NULL);
3450 rv = interp_exec(c, b->left, &rvtype);
3451 right = interp_exec(c, b->right, &rtype);
3452 rv.bool = rv.bool || right.bool;
3455 rv = interp_exec(c, b->left, &rvtype);
3457 rv = interp_exec(c, b->right, NULL);
3460 rv = interp_exec(c, b->right, &rvtype);
3464 ### Expressions: Comparison
3466 Of slightly higher precedence that Boolean expressions are Comparisons.
3467 A comparison takes arguments of any comparable type, but the two types
3470 To simplify the parsing we introduce an `eop` which can record an
3471 expression operator, and the `CMPop` non-terminal will match one of them.
3478 ###### ast functions
3479 static void free_eop(struct eop *e)
3493 ###### declare terminals
3494 $LEFT < > <= >= == != CMPop
3496 ###### expression grammar
3497 | Expression CMPop Expression ${ {
3498 struct binode *b = new(binode);
3508 CMPop -> < ${ $0.op = Less; }$
3509 | > ${ $0.op = Gtr; }$
3510 | <= ${ $0.op = LessEq; }$
3511 | >= ${ $0.op = GtrEq; }$
3512 | == ${ $0.op = Eql; }$
3513 | != ${ $0.op = NEql; }$
3515 ###### print binode cases
3523 if (bracket) printf("(");
3524 print_exec(b->left, -1, bracket);
3526 case Less: printf(" < "); break;
3527 case LessEq: printf(" <= "); break;
3528 case Gtr: printf(" > "); break;
3529 case GtrEq: printf(" >= "); break;
3530 case Eql: printf(" == "); break;
3531 case NEql: printf(" != "); break;
3532 default: abort(); // NOTEST
3534 print_exec(b->right, -1, bracket);
3535 if (bracket) printf(")");
3538 ###### propagate binode cases
3545 /* Both must match but not be labels, result is Tbool */
3546 t = propagate_types(b->left, c, ok, NULL, Rnolabel);
3548 propagate_types(b->right, c, ok, t, 0);
3550 t = propagate_types(b->right, c, ok, NULL, Rnolabel); // UNTESTED
3552 t = propagate_types(b->left, c, ok, t, 0); // UNTESTED
3554 if (!type_compat(type, Tbool, 0))
3555 type_err(c, "error: Comparison returns %1 but %2 expected", prog,
3556 Tbool, rules, type);
3559 ###### interp binode cases
3568 left = interp_exec(c, b->left, <ype);
3569 right = interp_exec(c, b->right, &rtype);
3570 cmp = value_cmp(ltype, rtype, &left, &right);
3573 case Less: rv.bool = cmp < 0; break;
3574 case LessEq: rv.bool = cmp <= 0; break;
3575 case Gtr: rv.bool = cmp > 0; break;
3576 case GtrEq: rv.bool = cmp >= 0; break;
3577 case Eql: rv.bool = cmp == 0; break;
3578 case NEql: rv.bool = cmp != 0; break;
3579 default: rv.bool = 0; break; // NOTEST
3584 ### Expressions: Arithmetic etc.
3586 The remaining expressions with the highest precedence are arithmetic,
3587 string concatenation, and string conversion. String concatenation
3588 (`++`) has the same precedence as multiplication and division, but lower
3591 String conversion is a temporary feature until I get a better type
3592 system. `$` is a prefix operator which expects a string and returns
3595 `+` and `-` are both infix and prefix operations (where they are
3596 absolute value and negation). These have different operator names.
3598 We also have a 'Bracket' operator which records where parentheses were
3599 found. This makes it easy to reproduce these when printing. Possibly I
3600 should only insert brackets were needed for precedence. Putting
3601 parentheses around an expression converts it into a Term,
3611 ###### declare terminals
3617 ###### expression grammar
3618 | Expression Eop Expression ${ {
3619 struct binode *b = new(binode);
3626 | Expression Top Expression ${ {
3627 struct binode *b = new(binode);
3634 | Uop Expression ${ {
3635 struct binode *b = new(binode);
3643 | ( Expression ) ${ {
3644 struct binode *b = new_pos(binode, $1);
3653 Eop -> + ${ $0.op = Plus; }$
3654 | - ${ $0.op = Minus; }$
3656 Uop -> + ${ $0.op = Absolute; }$
3657 | - ${ $0.op = Negate; }$
3658 | $ ${ $0.op = StringConv; }$
3660 Top -> * ${ $0.op = Times; }$
3661 | / ${ $0.op = Divide; }$
3662 | % ${ $0.op = Rem; }$
3663 | ++ ${ $0.op = Concat; }$
3665 ###### print binode cases
3672 if (bracket) printf("(");
3673 print_exec(b->left, indent, bracket);
3675 case Plus: fputs(" + ", stdout); break;
3676 case Minus: fputs(" - ", stdout); break;
3677 case Times: fputs(" * ", stdout); break;
3678 case Divide: fputs(" / ", stdout); break;
3679 case Rem: fputs(" % ", stdout); break;
3680 case Concat: fputs(" ++ ", stdout); break;
3681 default: abort(); // NOTEST
3683 print_exec(b->right, indent, bracket);
3684 if (bracket) printf(")");
3689 if (bracket) printf("(");
3691 case Absolute: fputs("+", stdout); break;
3692 case Negate: fputs("-", stdout); break;
3693 case StringConv: fputs("$", stdout); break;
3694 default: abort(); // NOTEST
3696 print_exec(b->right, indent, bracket);
3697 if (bracket) printf(")");
3701 print_exec(b->right, indent, bracket);
3705 ###### propagate binode cases
3711 /* both must be numbers, result is Tnum */
3714 /* as propagate_types ignores a NULL,
3715 * unary ops fit here too */
3716 propagate_types(b->left, c, ok, Tnum, 0);
3717 propagate_types(b->right, c, ok, Tnum, 0);
3718 if (!type_compat(type, Tnum, 0))
3719 type_err(c, "error: Arithmetic returns %1 but %2 expected", prog,
3724 /* both must be Tstr, result is Tstr */
3725 propagate_types(b->left, c, ok, Tstr, 0);
3726 propagate_types(b->right, c, ok, Tstr, 0);
3727 if (!type_compat(type, Tstr, 0))
3728 type_err(c, "error: Concat returns %1 but %2 expected", prog,
3733 /* op must be string, result is number */
3734 propagate_types(b->left, c, ok, Tstr, 0);
3735 if (!type_compat(type, Tnum, 0))
3736 type_err(c, // UNTESTED
3737 "error: Can only convert string to number, not %1",
3738 prog, type, 0, NULL);
3742 return propagate_types(b->right, c, ok, type, 0);
3744 ###### interp binode cases
3747 rv = interp_exec(c, b->left, &rvtype);
3748 right = interp_exec(c, b->right, &rtype);
3749 mpq_add(rv.num, rv.num, right.num);
3752 rv = interp_exec(c, b->left, &rvtype);
3753 right = interp_exec(c, b->right, &rtype);
3754 mpq_sub(rv.num, rv.num, right.num);
3757 rv = interp_exec(c, b->left, &rvtype);
3758 right = interp_exec(c, b->right, &rtype);
3759 mpq_mul(rv.num, rv.num, right.num);
3762 rv = interp_exec(c, b->left, &rvtype);
3763 right = interp_exec(c, b->right, &rtype);
3764 mpq_div(rv.num, rv.num, right.num);
3769 left = interp_exec(c, b->left, <ype);
3770 right = interp_exec(c, b->right, &rtype);
3771 mpz_init(l); mpz_init(r); mpz_init(rem);
3772 mpz_tdiv_q(l, mpq_numref(left.num), mpq_denref(left.num));
3773 mpz_tdiv_q(r, mpq_numref(right.num), mpq_denref(right.num));
3774 mpz_tdiv_r(rem, l, r);
3775 val_init(Tnum, &rv);
3776 mpq_set_z(rv.num, rem);
3777 mpz_clear(r); mpz_clear(l); mpz_clear(rem);
3782 rv = interp_exec(c, b->right, &rvtype);
3783 mpq_neg(rv.num, rv.num);
3786 rv = interp_exec(c, b->right, &rvtype);
3787 mpq_abs(rv.num, rv.num);
3790 rv = interp_exec(c, b->right, &rvtype);
3793 left = interp_exec(c, b->left, <ype);
3794 right = interp_exec(c, b->right, &rtype);
3796 rv.str = text_join(left.str, right.str);
3799 right = interp_exec(c, b->right, &rvtype);
3803 struct text tx = right.str;
3806 if (tx.txt[0] == '-') {
3807 neg = 1; // UNTESTED
3808 tx.txt++; // UNTESTED
3809 tx.len--; // UNTESTED
3811 if (number_parse(rv.num, tail, tx) == 0)
3812 mpq_init(rv.num); // UNTESTED
3814 mpq_neg(rv.num, rv.num); // UNTESTED
3816 printf("Unsupported suffix: %.*s\n", tx.len, tx.txt); // UNTESTED
3820 ###### value functions
3822 static struct text text_join(struct text a, struct text b)
3825 rv.len = a.len + b.len;
3826 rv.txt = malloc(rv.len);
3827 memcpy(rv.txt, a.txt, a.len);
3828 memcpy(rv.txt+a.len, b.txt, b.len);
3832 ### Blocks, Statements, and Statement lists.
3834 Now that we have expressions out of the way we need to turn to
3835 statements. There are simple statements and more complex statements.
3836 Simple statements do not contain (syntactic) newlines, complex statements do.
3838 Statements often come in sequences and we have corresponding simple
3839 statement lists and complex statement lists.
3840 The former comprise only simple statements separated by semicolons.
3841 The later comprise complex statements and simple statement lists. They are
3842 separated by newlines. Thus the semicolon is only used to separate
3843 simple statements on the one line. This may be overly restrictive,
3844 but I'm not sure I ever want a complex statement to share a line with
3847 Note that a simple statement list can still use multiple lines if
3848 subsequent lines are indented, so
3850 ###### Example: wrapped simple statement list
3855 is a single simple statement list. This might allow room for
3856 confusion, so I'm not set on it yet.
3858 A simple statement list needs no extra syntax. A complex statement
3859 list has two syntactic forms. It can be enclosed in braces (much like
3860 C blocks), or it can be introduced by an indent and continue until an
3861 unindented newline (much like Python blocks). With this extra syntax
3862 it is referred to as a block.
3864 Note that a block does not have to include any newlines if it only
3865 contains simple statements. So both of:
3867 if condition: a=b; d=f
3869 if condition { a=b; print f }
3873 In either case the list is constructed from a `binode` list with
3874 `Block` as the operator. When parsing the list it is most convenient
3875 to append to the end, so a list is a list and a statement. When using
3876 the list it is more convenient to consider a list to be a statement
3877 and a list. So we need a function to re-order a list.
3878 `reorder_bilist` serves this purpose.
3880 The only stand-alone statement we introduce at this stage is `pass`
3881 which does nothing and is represented as a `NULL` pointer in a `Block`
3882 list. Other stand-alone statements will follow once the infrastructure
3885 As many statements will use binodes, we declare a binode pointer 'b' in
3886 the common header for all reductions to use.
3888 ###### Parser: reduce
3899 Block -> { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3900 | { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3901 | SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
3902 | SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
3903 | IN OptNL Statementlist OUT ${ $0 = $<Sl; }$
3905 OpenBlock -> OpenScope { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3906 | OpenScope { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3907 | OpenScope SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
3908 | OpenScope SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
3909 | IN OpenScope OptNL Statementlist OUT ${ $0 = $<Sl; }$
3911 UseBlock -> { OpenScope IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3912 | { OpenScope SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3913 | IN OpenScope OptNL Statementlist OUT ${ $0 = $<Sl; }$
3915 ColonBlock -> { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3916 | { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3917 | : SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
3918 | : SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
3919 | : IN OptNL Statementlist OUT ${ $0 = $<Sl; }$
3921 Statementlist -> ComplexStatements ${ $0 = reorder_bilist($<CS); }$
3923 ComplexStatements -> ComplexStatements ComplexStatement ${
3933 | ComplexStatement ${
3945 ComplexStatement -> SimpleStatements Newlines ${
3946 $0 = reorder_bilist($<SS);
3948 | SimpleStatements ; Newlines ${
3949 $0 = reorder_bilist($<SS);
3951 ## ComplexStatement Grammar
3954 SimpleStatements -> SimpleStatements ; SimpleStatement ${
3960 | SimpleStatement ${
3969 SimpleStatement -> pass ${ $0 = NULL; }$
3970 | ERROR ${ tok_err(c, "Syntax error in statement", &$1); }$
3971 ## SimpleStatement Grammar
3973 ###### print binode cases
3977 if (b->left == NULL) // UNTESTED
3978 printf("pass"); // UNTESTED
3980 print_exec(b->left, indent, bracket); // UNTESTED
3981 if (b->right) { // UNTESTED
3982 printf("; "); // UNTESTED
3983 print_exec(b->right, indent, bracket); // UNTESTED
3986 // block, one per line
3987 if (b->left == NULL)
3988 do_indent(indent, "pass\n");
3990 print_exec(b->left, indent, bracket);
3992 print_exec(b->right, indent, bracket);
3996 ###### propagate binode cases
3999 /* If any statement returns something other than Tnone
4000 * or Tbool then all such must return same type.
4001 * As each statement may be Tnone or something else,
4002 * we must always pass NULL (unknown) down, otherwise an incorrect
4003 * error might occur. We never return Tnone unless it is
4008 for (e = b; e; e = cast(binode, e->right)) {
4009 t = propagate_types(e->left, c, ok, NULL, rules);
4010 if ((rules & Rboolok) && (t == Tbool || t == Tnone))
4012 if (t == Tnone && e->right)
4013 /* Only the final statement *must* return a value
4021 type_err(c, "error: expected %1%r, found %2",
4022 e->left, type, rules, t);
4028 ###### interp binode cases
4030 while (rvtype == Tnone &&
4033 rv = interp_exec(c, b->left, &rvtype);
4034 b = cast(binode, b->right);
4038 ### The Print statement
4040 `print` is a simple statement that takes a comma-separated list of
4041 expressions and prints the values separated by spaces and terminated
4042 by a newline. No control of formatting is possible.
4044 `print` uses `ExpressionList` to collect the expressions and stores them
4045 on the left side of a `Print` binode unlessthere is a trailing comma
4046 when the list is stored on the `right` side and no trailing newline is
4052 ##### declare terminals
4055 ###### SimpleStatement Grammar
4057 | print ExpressionList ${
4058 $0 = b = new(binode);
4061 b->left = reorder_bilist($<EL);
4063 | print ExpressionList , ${ {
4064 $0 = b = new(binode);
4066 b->right = reorder_bilist($<EL);
4070 $0 = b = new(binode);
4076 ###### print binode cases
4079 do_indent(indent, "print");
4081 print_exec(b->right, -1, bracket);
4084 print_exec(b->left, -1, bracket);
4089 ###### propagate binode cases
4092 /* don't care but all must be consistent */
4094 b = cast(binode, b->left);
4096 b = cast(binode, b->right);
4098 propagate_types(b->left, c, ok, NULL, Rnolabel);
4099 b = cast(binode, b->right);
4103 ###### interp binode cases
4107 struct binode *b2 = cast(binode, b->left);
4109 b2 = cast(binode, b->right);
4110 for (; b2; b2 = cast(binode, b2->right)) {
4111 left = interp_exec(c, b2->left, <ype);
4112 print_value(ltype, &left, stdout);
4113 free_value(ltype, &left);
4117 if (b->right == NULL)
4123 ###### Assignment statement
4125 An assignment will assign a value to a variable, providing it hasn't
4126 been declared as a constant. The analysis phase ensures that the type
4127 will be correct so the interpreter just needs to perform the
4128 calculation. There is a form of assignment which declares a new
4129 variable as well as assigning a value. If a name is assigned before
4130 it is declared, and error will be raised as the name is created as
4131 `Tlabel` and it is illegal to assign to such names.
4137 ###### declare terminals
4140 ###### SimpleStatement Grammar
4141 | Term = Expression ${
4142 $0 = b= new(binode);
4147 | VariableDecl = Expression ${
4148 $0 = b= new(binode);
4155 if ($1->var->where_set == NULL) {
4157 "Variable declared with no type or value: %v",
4161 $0 = b = new(binode);
4168 ###### print binode cases
4171 do_indent(indent, "");
4172 print_exec(b->left, indent, bracket);
4174 print_exec(b->right, indent, bracket);
4181 struct variable *v = cast(var, b->left)->var;
4182 do_indent(indent, "");
4183 print_exec(b->left, indent, bracket);
4184 if (cast(var, b->left)->var->constant) {
4186 if (v->explicit_type) {
4187 type_print(v->type, stdout);
4192 if (v->explicit_type) {
4193 type_print(v->type, stdout);
4199 print_exec(b->right, indent, bracket);
4206 ###### propagate binode cases
4210 /* Both must match and not be labels,
4211 * Type must support 'dup',
4212 * For Assign, left must not be constant.
4215 t = propagate_types(b->left, c, ok, NULL,
4216 Rnolabel | (b->op == Assign ? Rnoconstant : 0));
4221 if (propagate_types(b->right, c, ok, t, 0) != t)
4222 if (b->left->type == Xvar)
4223 type_err(c, "info: variable '%v' was set as %1 here.",
4224 cast(var, b->left)->var->where_set, t, rules, NULL);
4226 t = propagate_types(b->right, c, ok, NULL, Rnolabel);
4228 propagate_types(b->left, c, ok, t,
4229 (b->op == Assign ? Rnoconstant : 0));
4231 if (t && t->dup == NULL && t->name.txt[0] != ' ') // HACK
4232 type_err(c, "error: cannot assign value of type %1", b, t, 0, NULL);
4237 ###### interp binode cases
4240 lleft = linterp_exec(c, b->left, <ype);
4242 dinterp_exec(c, b->right, lleft, ltype, 1);
4248 struct variable *v = cast(var, b->left)->var;
4251 val = var_value(c, v);
4252 if (v->type->prepare_type)
4253 v->type->prepare_type(c, v->type, 0);
4255 dinterp_exec(c, b->right, val, v->type, 0);
4257 val_init(v->type, val);
4261 ### The `use` statement
4263 The `use` statement is the last "simple" statement. It is needed when a
4264 statement block can return a value. This includes the body of a
4265 function which has a return type, and the "condition" code blocks in
4266 `if`, `while`, and `switch` statements.
4271 ###### declare terminals
4274 ###### SimpleStatement Grammar
4276 $0 = b = new_pos(binode, $1);
4279 if (b->right->type == Xvar) {
4280 struct var *v = cast(var, b->right);
4281 if (v->var->type == Tnone) {
4282 /* Convert this to a label */
4285 v->var->type = Tlabel;
4286 val = global_alloc(c, Tlabel, v->var, NULL);
4292 ###### print binode cases
4295 do_indent(indent, "use ");
4296 print_exec(b->right, -1, bracket);
4301 ###### propagate binode cases
4304 /* result matches value */
4305 return propagate_types(b->right, c, ok, type, 0);
4307 ###### interp binode cases
4310 rv = interp_exec(c, b->right, &rvtype);
4313 ### The Conditional Statement
4315 This is the biggy and currently the only complex statement. This
4316 subsumes `if`, `while`, `do/while`, `switch`, and some parts of `for`.
4317 It is comprised of a number of parts, all of which are optional though
4318 set combinations apply. Each part is (usually) a key word (`then` is
4319 sometimes optional) followed by either an expression or a code block,
4320 except the `casepart` which is a "key word and an expression" followed
4321 by a code block. The code-block option is valid for all parts and,
4322 where an expression is also allowed, the code block can use the `use`
4323 statement to report a value. If the code block does not report a value
4324 the effect is similar to reporting `True`.
4326 The `else` and `case` parts, as well as `then` when combined with
4327 `if`, can contain a `use` statement which will apply to some
4328 containing conditional statement. `for` parts, `do` parts and `then`
4329 parts used with `for` can never contain a `use`, except in some
4330 subordinate conditional statement.
4332 If there is a `forpart`, it is executed first, only once.
4333 If there is a `dopart`, then it is executed repeatedly providing
4334 always that the `condpart` or `cond`, if present, does not return a non-True
4335 value. `condpart` can fail to return any value if it simply executes
4336 to completion. This is treated the same as returning `True`.
4338 If there is a `thenpart` it will be executed whenever the `condpart`
4339 or `cond` returns True (or does not return any value), but this will happen
4340 *after* `dopart` (when present).
4342 If `elsepart` is present it will be executed at most once when the
4343 condition returns `False` or some value that isn't `True` and isn't
4344 matched by any `casepart`. If there are any `casepart`s, they will be
4345 executed when the condition returns a matching value.
4347 The particular sorts of values allowed in case parts has not yet been
4348 determined in the language design, so nothing is prohibited.
4350 The various blocks in this complex statement potentially provide scope
4351 for variables as described earlier. Each such block must include the
4352 "OpenScope" nonterminal before parsing the block, and must call
4353 `var_block_close()` when closing the block.
4355 The code following "`if`", "`switch`" and "`for`" does not get its own
4356 scope, but is in a scope covering the whole statement, so names
4357 declared there cannot be redeclared elsewhere. Similarly the
4358 condition following "`while`" is in a scope the covers the body
4359 ("`do`" part) of the loop, and which does not allow conditional scope
4360 extension. Code following "`then`" (both looping and non-looping),
4361 "`else`" and "`case`" each get their own local scope.
4363 The type requirements on the code block in a `whilepart` are quite
4364 unusal. It is allowed to return a value of some identifiable type, in
4365 which case the loop aborts and an appropriate `casepart` is run, or it
4366 can return a Boolean, in which case the loop either continues to the
4367 `dopart` (on `True`) or aborts and runs the `elsepart` (on `False`).
4368 This is different both from the `ifpart` code block which is expected to
4369 return a Boolean, or the `switchpart` code block which is expected to
4370 return the same type as the casepart values. The correct analysis of
4371 the type of the `whilepart` code block is the reason for the
4372 `Rboolok` flag which is passed to `propagate_types()`.
4374 The `cond_statement` cannot fit into a `binode` so a new `exec` is
4375 defined. As there are two scopes which cover multiple parts - one for
4376 the whole statement and one for "while" and "do" - and as we will use
4377 the 'struct exec' to track scopes, we actually need two new types of
4378 exec. One is a `binode` for the looping part, the rest is the
4379 `cond_statement`. The `cond_statement` will use an auxilliary `struct
4380 casepart` to track a list of case parts.
4391 struct exec *action;
4392 struct casepart *next;
4394 struct cond_statement {
4396 struct exec *forpart, *condpart, *thenpart, *elsepart;
4397 struct binode *looppart;
4398 struct casepart *casepart;
4401 ###### ast functions
4403 static void free_casepart(struct casepart *cp)
4407 free_exec(cp->value);
4408 free_exec(cp->action);
4415 static void free_cond_statement(struct cond_statement *s)
4419 free_exec(s->forpart);
4420 free_exec(s->condpart);
4421 free_exec(s->looppart);
4422 free_exec(s->thenpart);
4423 free_exec(s->elsepart);
4424 free_casepart(s->casepart);
4428 ###### free exec cases
4429 case Xcond_statement: free_cond_statement(cast(cond_statement, e)); break;
4431 ###### ComplexStatement Grammar
4432 | CondStatement ${ $0 = $<1; }$
4434 ###### declare terminals
4435 $TERM for then while do
4442 // A CondStatement must end with EOL, as does CondSuffix and
4444 // ForPart, ThenPart, SwitchPart, CasePart are non-empty and
4445 // may or may not end with EOL
4446 // WhilePart and IfPart include an appropriate Suffix
4448 // ForPart, SwitchPart, and IfPart open scopes, o we have to close
4449 // them. WhilePart opens and closes its own scope.
4450 CondStatement -> ForPart OptNL ThenPart OptNL WhilePart CondSuffix ${
4453 $0->thenpart = $<TP;
4454 $0->looppart = $<WP;
4455 var_block_close(c, CloseSequential, $0);
4457 | ForPart OptNL WhilePart CondSuffix ${
4460 $0->looppart = $<WP;
4461 var_block_close(c, CloseSequential, $0);
4463 | WhilePart CondSuffix ${
4465 $0->looppart = $<WP;
4467 | SwitchPart OptNL CasePart CondSuffix ${
4469 $0->condpart = $<SP;
4470 $CP->next = $0->casepart;
4471 $0->casepart = $<CP;
4472 var_block_close(c, CloseSequential, $0);
4474 | SwitchPart : IN OptNL CasePart CondSuffix OUT Newlines ${
4476 $0->condpart = $<SP;
4477 $CP->next = $0->casepart;
4478 $0->casepart = $<CP;
4479 var_block_close(c, CloseSequential, $0);
4481 | IfPart IfSuffix ${
4483 $0->condpart = $IP.condpart; $IP.condpart = NULL;
4484 $0->thenpart = $IP.thenpart; $IP.thenpart = NULL;
4485 // This is where we close an "if" statement
4486 var_block_close(c, CloseSequential, $0);
4489 CondSuffix -> IfSuffix ${
4492 | Newlines CasePart CondSuffix ${
4494 $CP->next = $0->casepart;
4495 $0->casepart = $<CP;
4497 | CasePart CondSuffix ${
4499 $CP->next = $0->casepart;
4500 $0->casepart = $<CP;
4503 IfSuffix -> Newlines ${ $0 = new(cond_statement); }$
4504 | Newlines ElsePart ${ $0 = $<EP; }$
4505 | ElsePart ${$0 = $<EP; }$
4507 ElsePart -> else OpenBlock Newlines ${
4508 $0 = new(cond_statement);
4509 $0->elsepart = $<OB;
4510 var_block_close(c, CloseElse, $0->elsepart);
4512 | else OpenScope CondStatement ${
4513 $0 = new(cond_statement);
4514 $0->elsepart = $<CS;
4515 var_block_close(c, CloseElse, $0->elsepart);
4519 CasePart -> case Expression OpenScope ColonBlock ${
4520 $0 = calloc(1,sizeof(struct casepart));
4523 var_block_close(c, CloseParallel, $0->action);
4527 // These scopes are closed in CondStatement
4528 ForPart -> for OpenBlock ${
4532 ThenPart -> then OpenBlock ${
4534 var_block_close(c, CloseSequential, $0);
4538 // This scope is closed in CondStatement
4539 WhilePart -> while UseBlock OptNL do OpenBlock ${
4544 var_block_close(c, CloseSequential, $0->right);
4545 var_block_close(c, CloseSequential, $0);
4547 | while OpenScope Expression OpenScope ColonBlock ${
4552 var_block_close(c, CloseSequential, $0->right);
4553 var_block_close(c, CloseSequential, $0);
4557 IfPart -> if UseBlock OptNL then OpenBlock ${
4560 var_block_close(c, CloseParallel, $0.thenpart);
4562 | if OpenScope Expression OpenScope ColonBlock ${
4565 var_block_close(c, CloseParallel, $0.thenpart);
4567 | if OpenScope Expression OpenScope OptNL then Block ${
4570 var_block_close(c, CloseParallel, $0.thenpart);
4574 // This scope is closed in CondStatement
4575 SwitchPart -> switch OpenScope Expression ${
4578 | switch UseBlock ${
4582 ###### print binode cases
4584 if (b->left && b->left->type == Xbinode &&
4585 cast(binode, b->left)->op == Block) {
4587 do_indent(indent, "while {\n");
4589 do_indent(indent, "while\n");
4590 print_exec(b->left, indent+1, bracket);
4592 do_indent(indent, "} do {\n");
4594 do_indent(indent, "do\n");
4595 print_exec(b->right, indent+1, bracket);
4597 do_indent(indent, "}\n");
4599 do_indent(indent, "while ");
4600 print_exec(b->left, 0, bracket);
4605 print_exec(b->right, indent+1, bracket);
4607 do_indent(indent, "}\n");
4611 ###### print exec cases
4613 case Xcond_statement:
4615 struct cond_statement *cs = cast(cond_statement, e);
4616 struct casepart *cp;
4618 do_indent(indent, "for");
4619 if (bracket) printf(" {\n"); else printf("\n");
4620 print_exec(cs->forpart, indent+1, bracket);
4623 do_indent(indent, "} then {\n");
4625 do_indent(indent, "then\n");
4626 print_exec(cs->thenpart, indent+1, bracket);
4628 if (bracket) do_indent(indent, "}\n");
4631 print_exec(cs->looppart, indent, bracket);
4635 do_indent(indent, "switch");
4637 do_indent(indent, "if");
4638 if (cs->condpart && cs->condpart->type == Xbinode &&
4639 cast(binode, cs->condpart)->op == Block) {
4644 print_exec(cs->condpart, indent+1, bracket);
4646 do_indent(indent, "}\n");
4648 do_indent(indent, "then\n");
4649 print_exec(cs->thenpart, indent+1, bracket);
4653 print_exec(cs->condpart, 0, bracket);
4659 print_exec(cs->thenpart, indent+1, bracket);
4661 do_indent(indent, "}\n");
4666 for (cp = cs->casepart; cp; cp = cp->next) {
4667 do_indent(indent, "case ");
4668 print_exec(cp->value, -1, 0);
4673 print_exec(cp->action, indent+1, bracket);
4675 do_indent(indent, "}\n");
4678 do_indent(indent, "else");
4683 print_exec(cs->elsepart, indent+1, bracket);
4685 do_indent(indent, "}\n");
4690 ###### propagate binode cases
4692 t = propagate_types(b->right, c, ok, Tnone, 0);
4693 if (!type_compat(Tnone, t, 0))
4694 *ok = 0; // UNTESTED
4695 return propagate_types(b->left, c, ok, type, rules);
4697 ###### propagate exec cases
4698 case Xcond_statement:
4700 // forpart and looppart->right must return Tnone
4701 // thenpart must return Tnone if there is a loopart,
4702 // otherwise it is like elsepart.
4704 // be bool if there is no casepart
4705 // match casepart->values if there is a switchpart
4706 // either be bool or match casepart->value if there
4708 // elsepart and casepart->action must match the return type
4709 // expected of this statement.
4710 struct cond_statement *cs = cast(cond_statement, prog);
4711 struct casepart *cp;
4713 t = propagate_types(cs->forpart, c, ok, Tnone, 0);
4714 if (!type_compat(Tnone, t, 0))
4715 *ok = 0; // UNTESTED
4718 t = propagate_types(cs->thenpart, c, ok, Tnone, 0);
4719 if (!type_compat(Tnone, t, 0))
4720 *ok = 0; // UNTESTED
4722 if (cs->casepart == NULL) {
4723 propagate_types(cs->condpart, c, ok, Tbool, 0);
4724 propagate_types(cs->looppart, c, ok, Tbool, 0);
4726 /* Condpart must match case values, with bool permitted */
4728 for (cp = cs->casepart;
4729 cp && !t; cp = cp->next)
4730 t = propagate_types(cp->value, c, ok, NULL, 0);
4731 if (!t && cs->condpart)
4732 t = propagate_types(cs->condpart, c, ok, NULL, Rboolok); // UNTESTED
4733 if (!t && cs->looppart)
4734 t = propagate_types(cs->looppart, c, ok, NULL, Rboolok); // UNTESTED
4735 // Now we have a type (I hope) push it down
4737 for (cp = cs->casepart; cp; cp = cp->next)
4738 propagate_types(cp->value, c, ok, t, 0);
4739 propagate_types(cs->condpart, c, ok, t, Rboolok);
4740 propagate_types(cs->looppart, c, ok, t, Rboolok);
4743 // (if)then, else, and case parts must return expected type.
4744 if (!cs->looppart && !type)
4745 type = propagate_types(cs->thenpart, c, ok, NULL, rules);
4747 type = propagate_types(cs->elsepart, c, ok, NULL, rules);
4748 for (cp = cs->casepart;
4750 cp = cp->next) // UNTESTED
4751 type = propagate_types(cp->action, c, ok, NULL, rules); // UNTESTED
4754 propagate_types(cs->thenpart, c, ok, type, rules);
4755 propagate_types(cs->elsepart, c, ok, type, rules);
4756 for (cp = cs->casepart; cp ; cp = cp->next)
4757 propagate_types(cp->action, c, ok, type, rules);
4763 ###### interp binode cases
4765 // This just performs one iterration of the loop
4766 rv = interp_exec(c, b->left, &rvtype);
4767 if (rvtype == Tnone ||
4768 (rvtype == Tbool && rv.bool != 0))
4769 // rvtype is Tnone or Tbool, doesn't need to be freed
4770 interp_exec(c, b->right, NULL);
4773 ###### interp exec cases
4774 case Xcond_statement:
4776 struct value v, cnd;
4777 struct type *vtype, *cndtype;
4778 struct casepart *cp;
4779 struct cond_statement *cs = cast(cond_statement, e);
4782 interp_exec(c, cs->forpart, NULL);
4784 while ((cnd = interp_exec(c, cs->looppart, &cndtype)),
4785 cndtype == Tnone || (cndtype == Tbool && cnd.bool != 0))
4786 interp_exec(c, cs->thenpart, NULL);
4788 cnd = interp_exec(c, cs->condpart, &cndtype);
4789 if ((cndtype == Tnone ||
4790 (cndtype == Tbool && cnd.bool != 0))) {
4791 // cnd is Tnone or Tbool, doesn't need to be freed
4792 rv = interp_exec(c, cs->thenpart, &rvtype);
4793 // skip else (and cases)
4797 for (cp = cs->casepart; cp; cp = cp->next) {
4798 v = interp_exec(c, cp->value, &vtype);
4799 if (value_cmp(cndtype, vtype, &v, &cnd) == 0) {
4800 free_value(vtype, &v);
4801 free_value(cndtype, &cnd);
4802 rv = interp_exec(c, cp->action, &rvtype);
4805 free_value(vtype, &v);
4807 free_value(cndtype, &cnd);
4809 rv = interp_exec(c, cs->elsepart, &rvtype);
4816 ### Top level structure
4818 All the language elements so far can be used in various places. Now
4819 it is time to clarify what those places are.
4821 At the top level of a file there will be a number of declarations.
4822 Many of the things that can be declared haven't been described yet,
4823 such as functions, procedures, imports, and probably more.
4824 For now there are two sorts of things that can appear at the top
4825 level. They are predefined constants, `struct` types, and the `main`
4826 function. While the syntax will allow the `main` function to appear
4827 multiple times, that will trigger an error if it is actually attempted.
4829 The various declarations do not return anything. They store the
4830 various declarations in the parse context.
4832 ###### Parser: grammar
4835 Ocean -> OptNL DeclarationList
4837 ## declare terminals
4845 DeclarationList -> Declaration
4846 | DeclarationList Declaration
4848 Declaration -> ERROR Newlines ${
4849 tok_err(c, // UNTESTED
4850 "error: unhandled parse error", &$1);
4856 ## top level grammar
4860 ### The `const` section
4862 As well as being defined in with the code that uses them, constants can
4863 be declared at the top level. These have full-file scope, so they are
4864 always `InScope`, even before(!) they have been declared. The value of
4865 a top level constant can be given as an expression, and this is
4866 evaluated after parsing and before execution.
4868 A function call can be used to evaluate a constant, but it will not have
4869 access to any program state, once such statement becomes meaningful.
4870 e.g. arguments and filesystem will not be visible.
4872 Constants are defined in a section that starts with the reserved word
4873 `const` and then has a block with a list of assignment statements.
4874 For syntactic consistency, these must use the double-colon syntax to
4875 make it clear that they are constants. Type can also be given: if
4876 not, the type will be determined during analysis, as with other
4879 ###### parse context
4880 struct binode *constlist;
4882 ###### top level grammar
4886 DeclareConstant -> const { IN OptNL ConstList OUT OptNL } Newlines
4887 | const { SimpleConstList } Newlines
4888 | const IN OptNL ConstList OUT Newlines
4889 | const SimpleConstList Newlines
4891 ConstList -> ConstList SimpleConstLine
4894 SimpleConstList -> SimpleConstList ; Const
4898 SimpleConstLine -> SimpleConstList Newlines
4899 | ERROR Newlines ${ tok_err(c, "Syntax error in constant", &$1); }$
4902 CType -> Type ${ $0 = $<1; }$
4906 Const -> IDENTIFIER :: CType = Expression ${ {
4908 struct binode *bl, *bv;
4909 struct var *var = new_pos(var, $ID);
4911 v = var_decl(c, $ID.txt);
4913 v->where_decl = var;
4919 v = var_ref(c, $1.txt);
4920 tok_err(c, "error: name already declared", &$1);
4921 type_err(c, "info: this is where '%v' was first declared",
4922 v->where_decl, NULL, 0, NULL);
4933 bl->left = c->constlist;
4938 ###### core functions
4939 static void resolve_consts(struct parse_context *c)
4942 c->constlist = reorder_bilist(c->constlist);
4943 for (b = cast(binode, c->constlist); b;
4944 b = cast(binode, b->right)) {
4946 struct binode *vb = cast(binode, b->left);
4947 struct var *v = cast(var, vb->left);
4950 propagate_types(vb->right, c, &ok,
4956 struct value res = interp_exec(
4957 c, vb->right, &v->var->type);
4958 global_alloc(c, v->var->type, v->var, &res);
4963 ###### print const decls
4968 for (b = cast(binode, context.constlist); b;
4969 b = cast(binode, b->right)) {
4970 struct binode *vb = cast(binode, b->left);
4971 struct var *vr = cast(var, vb->left);
4972 struct variable *v = vr->var;
4978 printf(" %.*s :: ", v->name->name.len, v->name->name.txt);
4979 type_print(v->type, stdout);
4981 print_exec(vb->right, -1, 0);
4986 ###### free const decls
4987 free_binode(context.constlist);
4989 ### Function declarations
4991 The code in an Ocean program is all stored in function declarations.
4992 One of the functions must be named `main` and it must accept an array of
4993 strings as a parameter - the command line arguments.
4995 As this is the top level, several things are handled a bit differently.
4996 The function is not interpreted by `interp_exec` as that isn't passed
4997 the argument list which the program requires. Similarly type analysis
4998 is a bit more interesting at this level.
5000 ###### ast functions
5002 static struct type *handle_results(struct parse_context *c,
5003 struct binode *results)
5005 /* Create a 'struct' type from the results list, which
5006 * is a list for 'struct var'
5008 struct type *t = add_anon_type(c, &structure_prototype,
5009 " function result");
5013 for (b = results; b; b = cast(binode, b->right))
5015 t->structure.nfields = cnt;
5016 t->structure.fields = calloc(cnt, sizeof(struct field));
5018 for (b = results; b; b = cast(binode, b->right)) {
5019 struct var *v = cast(var, b->left);
5020 struct field *f = &t->structure.fields[cnt++];
5021 int a = v->var->type->align;
5022 f->name = v->var->name->name;
5023 f->type = v->var->type;
5025 f->offset = t->size;
5026 v->var->frame_pos = f->offset;
5027 t->size += ((f->type->size - 1) | (a-1)) + 1;
5030 variable_unlink_exec(v->var);
5032 free_binode(results);
5036 static struct variable *declare_function(struct parse_context *c,
5037 struct variable *name,
5038 struct binode *args,
5040 struct binode *results,
5044 struct value fn = {.function = code};
5046 var_block_close(c, CloseFunction, code);
5047 t = add_anon_type(c, &function_prototype,
5048 "func %.*s", name->name->name.len,
5049 name->name->name.txt);
5051 t->function.params = reorder_bilist(args);
5053 ret = handle_results(c, reorder_bilist(results));
5054 t->function.inline_result = 1;
5055 t->function.local_size = ret->size;
5057 t->function.return_type = ret;
5058 global_alloc(c, t, name, &fn);
5059 name->type->function.scope = c->out_scope;
5064 var_block_close(c, CloseFunction, NULL);
5066 c->out_scope = NULL;
5070 ###### declare terminals
5073 ###### top level grammar
5076 DeclareFunction -> func FuncName ( OpenScope ArgsLine ) Block Newlines ${
5077 $0 = declare_function(c, $<FN, $<Ar, Tnone, NULL, $<Bl);
5079 | func FuncName IN OpenScope Args OUT OptNL do Block Newlines ${
5080 $0 = declare_function(c, $<FN, $<Ar, Tnone, NULL, $<Bl);
5082 | func FuncName NEWLINE OpenScope OptNL do Block Newlines ${
5083 $0 = declare_function(c, $<FN, NULL, Tnone, NULL, $<Bl);
5085 | func FuncName ( OpenScope ArgsLine ) : Type Block Newlines ${
5086 $0 = declare_function(c, $<FN, $<Ar, $<Ty, NULL, $<Bl);
5088 | func FuncName ( OpenScope ArgsLine ) : ( ArgsLine ) Block Newlines ${
5089 $0 = declare_function(c, $<FN, $<AL, NULL, $<AL2, $<Bl);
5091 | func FuncName IN OpenScope Args OUT OptNL return Type Newlines do Block Newlines ${
5092 $0 = declare_function(c, $<FN, $<Ar, $<Ty, NULL, $<Bl);
5094 | func FuncName NEWLINE OpenScope return Type Newlines do Block Newlines ${
5095 $0 = declare_function(c, $<FN, NULL, $<Ty, NULL, $<Bl);
5097 | func FuncName IN OpenScope Args OUT OptNL return IN Args OUT OptNL do Block Newlines ${
5098 $0 = declare_function(c, $<FN, $<Ar, NULL, $<Ar2, $<Bl);
5100 | func FuncName NEWLINE OpenScope return IN Args OUT OptNL do Block Newlines ${
5101 $0 = declare_function(c, $<FN, NULL, NULL, $<Ar, $<Bl);
5104 ###### print func decls
5109 while (target != 0) {
5111 for (v = context.in_scope; v; v=v->in_scope)
5112 if (v->depth == 0 && v->type && v->type->check_args) {
5121 struct value *val = var_value(&context, v);
5122 printf("func %.*s", v->name->name.len, v->name->name.txt);
5123 v->type->print_type_decl(v->type, stdout);
5125 print_exec(val->function, 0, brackets);
5127 print_value(v->type, val, stdout);
5128 printf("/* frame size %d */\n", v->type->function.local_size);
5134 ###### core functions
5136 static int analyse_funcs(struct parse_context *c)
5140 for (v = c->in_scope; v; v = v->in_scope) {
5144 if (v->depth != 0 || !v->type || !v->type->check_args)
5146 ret = v->type->function.inline_result ?
5147 Tnone : v->type->function.return_type;
5148 val = var_value(c, v);
5151 propagate_types(val->function, c, &ok, ret, 0);
5154 /* Make sure everything is still consistent */
5155 propagate_types(val->function, c, &ok, ret, 0);
5158 if (!v->type->function.inline_result &&
5159 !v->type->function.return_type->dup) {
5160 type_err(c, "error: function cannot return value of type %1",
5161 v->where_decl, v->type->function.return_type, 0, NULL);
5164 scope_finalize(c, v->type);
5169 static int analyse_main(struct type *type, struct parse_context *c)
5171 struct binode *bp = type->function.params;
5175 struct type *argv_type;
5177 argv_type = add_anon_type(c, &array_prototype, "argv");
5178 argv_type->array.member = Tstr;
5179 argv_type->array.unspec = 1;
5181 for (b = bp; b; b = cast(binode, b->right)) {
5185 propagate_types(b->left, c, &ok, argv_type, 0);
5187 default: /* invalid */ // NOTEST
5188 propagate_types(b->left, c, &ok, Tnone, 0); // NOTEST
5194 return !c->parse_error;
5197 static void interp_main(struct parse_context *c, int argc, char **argv)
5199 struct value *progp = NULL;
5200 struct text main_name = { "main", 4 };
5201 struct variable *mainv;
5207 mainv = var_ref(c, main_name);
5209 progp = var_value(c, mainv);
5210 if (!progp || !progp->function) {
5211 fprintf(stderr, "oceani: no main function found.\n");
5215 if (!analyse_main(mainv->type, c)) {
5216 fprintf(stderr, "oceani: main has wrong type.\n");
5220 al = mainv->type->function.params;
5222 c->local_size = mainv->type->function.local_size;
5223 c->local = calloc(1, c->local_size);
5225 struct var *v = cast(var, al->left);
5226 struct value *vl = var_value(c, v->var);
5236 mpq_set_ui(argcq, argc, 1);
5237 memcpy(var_value(c, t->array.vsize), &argcq, sizeof(argcq));
5238 t->prepare_type(c, t, 0);
5239 array_init(v->var->type, vl);
5240 for (i = 0; i < argc; i++) {
5241 struct value *vl2 = vl->array + i * v->var->type->array.member->size;
5243 arg.str.txt = argv[i];
5244 arg.str.len = strlen(argv[i]);
5245 free_value(Tstr, vl2);
5246 dup_value(Tstr, &arg, vl2);
5250 al = cast(binode, al->right);
5252 v = interp_exec(c, progp->function, &vtype);
5253 free_value(vtype, &v);
5258 ###### ast functions
5259 void free_variable(struct variable *v)
5263 ## And now to test it out.
5265 Having a language requires having a "hello world" program. I'll
5266 provide a little more than that: a program that prints "Hello world"
5267 finds the GCD of two numbers, prints the first few elements of
5268 Fibonacci, performs a binary search for a number, and a few other
5269 things which will likely grow as the languages grows.
5271 ###### File: oceani.mk
5274 @echo "===== DEMO ====="
5275 ./oceani --section "demo: hello" oceani.mdc 55 33
5281 four ::= 2 + 2 ; five ::= 10/2
5282 const pie ::= "I like Pie";
5283 cake ::= "The cake is"
5291 func main(argv:[argc::]string)
5292 print "Hello World, what lovely oceans you have!"
5293 print "Are there", five, "?"
5294 print pi, pie, "but", cake
5296 A := $argv[1]; B := $argv[2]
5298 /* When a variable is defined in both branches of an 'if',
5299 * and used afterwards, the variables are merged.
5305 print "Is", A, "bigger than", B,"? ", bigger
5306 /* If a variable is not used after the 'if', no
5307 * merge happens, so types can be different
5310 double:string = "yes"
5311 print A, "is more than twice", B, "?", double
5314 print "double", B, "is", double
5319 if a > 0 and then b > 0:
5325 print "GCD of", A, "and", B,"is", a
5327 print a, "is not positive, cannot calculate GCD"
5329 print b, "is not positive, cannot calculate GCD"
5334 print "Fibonacci:", f1,f2,
5335 then togo = togo - 1
5343 /* Binary search... */
5348 mid := (lo + hi) / 2
5361 print "Yay, I found", target
5363 print "Closest I found was", lo
5368 // "middle square" PRNG. Not particularly good, but one my
5369 // Dad taught me - the first one I ever heard of.
5370 for i:=1; then i = i + 1; while i < size:
5371 n := list[i-1] * list[i-1]
5372 list[i] = (n / 100) % 10 000
5374 print "Before sort:",
5375 for i:=0; then i = i + 1; while i < size:
5379 for i := 1; then i=i+1; while i < size:
5380 for j:=i-1; then j=j-1; while j >= 0:
5381 if list[j] > list[j+1]:
5385 print " After sort:",
5386 for i:=0; then i = i + 1; while i < size:
5390 if 1 == 2 then print "yes"; else print "no"
5394 bob.alive = (bob.name == "Hello")
5395 print "bob", "is" if bob.alive else "isn't", "alive"