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, "%.10Fg", 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 // FIXME how to ensure numbers have same precision.
1243 print_value(v->vtype, &v->val, stdout);
1244 if (v->vtype == Tstr)
1249 ###### propagate exec cases
1252 struct val *val = cast(val, prog);
1253 if (!type_compat(type, val->vtype, rules))
1254 type_err(c, "error: expected %1%r found %2",
1255 prog, type, rules, val->vtype);
1259 ###### interp exec cases
1261 rvtype = cast(val, e)->vtype;
1262 dup_value(rvtype, &cast(val, e)->val, &rv);
1265 ###### ast functions
1266 static void free_val(struct val *v)
1269 free_value(v->vtype, &v->val);
1273 ###### free exec cases
1274 case Xval: free_val(cast(val, e)); break;
1276 ###### ast functions
1277 // Move all nodes from 'b' to 'rv', reversing their order.
1278 // In 'b' 'left' is a list, and 'right' is the last node.
1279 // In 'rv', left' is the first node and 'right' is a list.
1280 static struct binode *reorder_bilist(struct binode *b)
1282 struct binode *rv = NULL;
1285 struct exec *t = b->right;
1289 b = cast(binode, b->left);
1299 Variables are scoped named values. We store the names in a linked list
1300 of "bindings" sorted in lexical order, and use sequential search and
1307 struct binding *next; // in lexical order
1311 This linked list is stored in the parse context so that "reduce"
1312 functions can find or add variables, and so the analysis phase can
1313 ensure that every variable gets a type.
1315 ###### parse context
1317 struct binding *varlist; // In lexical order
1319 ###### ast functions
1321 static struct binding *find_binding(struct parse_context *c, struct text s)
1323 struct binding **l = &c->varlist;
1328 (cmp = text_cmp((*l)->name, s)) < 0)
1332 n = calloc(1, sizeof(*n));
1339 Each name can be linked to multiple variables defined in different
1340 scopes. Each scope starts where the name is declared and continues
1341 until the end of the containing code block. Scopes of a given name
1342 cannot nest, so a declaration while a name is in-scope is an error.
1344 ###### binding fields
1345 struct variable *var;
1349 struct variable *previous;
1351 struct binding *name;
1352 struct exec *where_decl;// where name was declared
1353 struct exec *where_set; // where type was set
1357 When a scope closes, the values of the variables might need to be freed.
1358 This happens in the context of some `struct exec` and each `exec` will
1359 need to know which variables need to be freed when it completes.
1362 struct variable *to_free;
1364 ####### variable fields
1365 struct exec *cleanup_exec;
1366 struct variable *next_free;
1368 ####### interp exec cleanup
1371 for (v = e->to_free; v; v = v->next_free) {
1372 struct value *val = var_value(c, v);
1373 free_value(v->type, val);
1377 ###### ast functions
1378 static void variable_unlink_exec(struct variable *v)
1380 struct variable **vp;
1381 if (!v->cleanup_exec)
1383 for (vp = &v->cleanup_exec->to_free;
1384 *vp; vp = &(*vp)->next_free) {
1388 v->cleanup_exec = NULL;
1393 While the naming seems strange, we include local constants in the
1394 definition of variables. A name declared `var := value` can
1395 subsequently be changed, but a name declared `var ::= value` cannot -
1398 ###### variable fields
1401 Scopes in parallel branches can be partially merged. More
1402 specifically, if a given name is declared in both branches of an
1403 if/else then its scope is a candidate for merging. Similarly if
1404 every branch of an exhaustive switch (e.g. has an "else" clause)
1405 declares a given name, then the scopes from the branches are
1406 candidates for merging.
1408 Note that names declared inside a loop (which is only parallel to
1409 itself) are never visible after the loop. Similarly names defined in
1410 scopes which are not parallel, such as those started by `for` and
1411 `switch`, are never visible after the scope. Only variables defined in
1412 both `then` and `else` (including the implicit then after an `if`, and
1413 excluding `then` used with `for`) and in all `case`s and `else` of a
1414 `switch` or `while` can be visible beyond the `if`/`switch`/`while`.
1416 Labels, which are a bit like variables, follow different rules.
1417 Labels are not explicitly declared, but if an undeclared name appears
1418 in a context where a label is legal, that effectively declares the
1419 name as a label. The declaration remains in force (or in scope) at
1420 least to the end of the immediately containing block and conditionally
1421 in any larger containing block which does not declare the name in some
1422 other way. Importantly, the conditional scope extension happens even
1423 if the label is only used in one parallel branch of a conditional --
1424 when used in one branch it is treated as having been declared in all
1427 Merge candidates are tentatively visible beyond the end of the
1428 branching statement which creates them. If the name is used, the
1429 merge is affirmed and they become a single variable visible at the
1430 outer layer. If not - if it is redeclared first - the merge lapses.
1432 To track scopes we have an extra stack, implemented as a linked list,
1433 which roughly parallels the parse stack and which is used exclusively
1434 for scoping. When a new scope is opened, a new frame is pushed and
1435 the child-count of the parent frame is incremented. This child-count
1436 is used to distinguish between the first of a set of parallel scopes,
1437 in which declared variables must not be in scope, and subsequent
1438 branches, whether they may already be conditionally scoped.
1440 We need a total ordering of scopes so we can easily compare to variables
1441 to see if they are concurrently in scope. To achieve this we record a
1442 `scope_count` which is actually a count of both beginnings and endings
1443 of scopes. Then each variable has a record of the scope count where it
1444 enters scope, and where it leaves.
1446 To push a new frame *before* any code in the frame is parsed, we need a
1447 grammar reduction. This is most easily achieved with a grammar
1448 element which derives the empty string, and creates the new scope when
1449 it is recognised. This can be placed, for example, between a keyword
1450 like "if" and the code following it.
1454 struct scope *parent;
1458 ###### parse context
1461 struct scope *scope_stack;
1463 ###### variable fields
1464 int scope_start, scope_end;
1466 ###### ast functions
1467 static void scope_pop(struct parse_context *c)
1469 struct scope *s = c->scope_stack;
1471 c->scope_stack = s->parent;
1473 c->scope_depth -= 1;
1474 c->scope_count += 1;
1477 static void scope_push(struct parse_context *c)
1479 struct scope *s = calloc(1, sizeof(*s));
1481 c->scope_stack->child_count += 1;
1482 s->parent = c->scope_stack;
1484 c->scope_depth += 1;
1485 c->scope_count += 1;
1491 OpenScope -> ${ scope_push(c); }$
1493 Each variable records a scope depth and is in one of four states:
1495 - "in scope". This is the case between the declaration of the
1496 variable and the end of the containing block, and also between
1497 the usage with affirms a merge and the end of that block.
1499 The scope depth is not greater than the current parse context scope
1500 nest depth. When the block of that depth closes, the state will
1501 change. To achieve this, all "in scope" variables are linked
1502 together as a stack in nesting order.
1504 - "pending". The "in scope" block has closed, but other parallel
1505 scopes are still being processed. So far, every parallel block at
1506 the same level that has closed has declared the name.
1508 The scope depth is the depth of the last parallel block that
1509 enclosed the declaration, and that has closed.
1511 - "conditionally in scope". The "in scope" block and all parallel
1512 scopes have closed, and no further mention of the name has been seen.
1513 This state includes a secondary nest depth (`min_depth`) which records
1514 the outermost scope seen since the variable became conditionally in
1515 scope. If a use of the name is found, the variable becomes "in scope"
1516 and that secondary depth becomes the recorded scope depth. If the
1517 name is declared as a new variable, the old variable becomes "out of
1518 scope" and the recorded scope depth stays unchanged.
1520 - "out of scope". The variable is neither in scope nor conditionally
1521 in scope. It is permanently out of scope now and can be removed from
1522 the "in scope" stack. When a variable becomes out-of-scope it is
1523 moved to a separate list (`out_scope`) of variables which have fully
1524 known scope. This will be used at the end of each function to assign
1525 each variable a place in the stack frame.
1527 ###### variable fields
1528 int depth, min_depth;
1529 enum { OutScope, PendingScope, CondScope, InScope } scope;
1530 struct variable *in_scope;
1532 ###### parse context
1534 struct variable *in_scope;
1535 struct variable *out_scope;
1537 All variables with the same name are linked together using the
1538 'previous' link. Those variable that have been affirmatively merged all
1539 have a 'merged' pointer that points to one primary variable - the most
1540 recently declared instance. When merging variables, we need to also
1541 adjust the 'merged' pointer on any other variables that had previously
1542 been merged with the one that will no longer be primary.
1544 A variable that is no longer the most recent instance of a name may
1545 still have "pending" scope, if it might still be merged with most
1546 recent instance. These variables don't really belong in the
1547 "in_scope" list, but are not immediately removed when a new instance
1548 is found. Instead, they are detected and ignored when considering the
1549 list of in_scope names.
1551 The storage of the value of a variable will be described later. For now
1552 we just need to know that when a variable goes out of scope, it might
1553 need to be freed. For this we need to be able to find it, so assume that
1554 `var_value()` will provide that.
1556 ###### variable fields
1557 struct variable *merged;
1559 ###### ast functions
1561 static void variable_merge(struct variable *primary, struct variable *secondary)
1565 primary = primary->merged;
1567 for (v = primary->previous; v; v=v->previous)
1568 if (v == secondary || v == secondary->merged ||
1569 v->merged == secondary ||
1570 v->merged == secondary->merged) {
1571 v->scope = OutScope;
1572 v->merged = primary;
1573 if (v->scope_start < primary->scope_start)
1574 primary->scope_start = v->scope_start;
1575 if (v->scope_end > primary->scope_end)
1576 primary->scope_end = v->scope_end; // NOTEST
1577 variable_unlink_exec(v);
1581 ###### forward decls
1582 static struct value *var_value(struct parse_context *c, struct variable *v);
1584 ###### free global vars
1586 while (context.varlist) {
1587 struct binding *b = context.varlist;
1588 struct variable *v = b->var;
1589 context.varlist = b->next;
1592 struct variable *next = v->previous;
1594 if (v->global && v->frame_pos >= 0) {
1595 free_value(v->type, var_value(&context, v));
1596 if (v->depth == 0 && v->type->free == function_free)
1597 // This is a function constant
1598 free_exec(v->where_decl);
1605 #### Manipulating Bindings
1607 When a name is conditionally visible, a new declaration discards the old
1608 binding - the condition lapses. Similarly when we reach the end of a
1609 function (outermost non-global scope) any conditional scope must lapse.
1610 Conversely a usage of the name affirms the visibility and extends it to
1611 the end of the containing block - i.e. the block that contains both the
1612 original declaration and the latest usage. This is determined from
1613 `min_depth`. When a conditionally visible variable gets affirmed like
1614 this, it is also merged with other conditionally visible variables with
1617 When we parse a variable declaration we either report an error if the
1618 name is currently bound, or create a new variable at the current nest
1619 depth if the name is unbound or bound to a conditionally scoped or
1620 pending-scope variable. If the previous variable was conditionally
1621 scoped, it and its homonyms becomes out-of-scope.
1623 When we parse a variable reference (including non-declarative assignment
1624 "foo = bar") we report an error if the name is not bound or is bound to
1625 a pending-scope variable; update the scope if the name is bound to a
1626 conditionally scoped variable; or just proceed normally if the named
1627 variable is in scope.
1629 When we exit a scope, any variables bound at this level are either
1630 marked out of scope or pending-scoped, depending on whether the scope
1631 was sequential or parallel. Here a "parallel" scope means the "then"
1632 or "else" part of a conditional, or any "case" or "else" branch of a
1633 switch. Other scopes are "sequential".
1635 When exiting a parallel scope we check if there are any variables that
1636 were previously pending and are still visible. If there are, then
1637 they weren't redeclared in the most recent scope, so they cannot be
1638 merged and must become out-of-scope. If it is not the first of
1639 parallel scopes (based on `child_count`), we check that there was a
1640 previous binding that is still pending-scope. If there isn't, the new
1641 variable must now be out-of-scope.
1643 When exiting a sequential scope that immediately enclosed parallel
1644 scopes, we need to resolve any pending-scope variables. If there was
1645 no `else` clause, and we cannot determine that the `switch` was exhaustive,
1646 we need to mark all pending-scope variable as out-of-scope. Otherwise
1647 all pending-scope variables become conditionally scoped.
1650 enum closetype { CloseSequential, CloseFunction, CloseParallel, CloseElse };
1652 ###### ast functions
1654 static struct variable *var_decl(struct parse_context *c, struct text s)
1656 struct binding *b = find_binding(c, s);
1657 struct variable *v = b->var;
1659 switch (v ? v->scope : OutScope) {
1661 /* Caller will report the error */
1665 v && v->scope == CondScope;
1667 v->scope = OutScope;
1671 v = calloc(1, sizeof(*v));
1672 v->previous = b->var;
1676 v->min_depth = v->depth = c->scope_depth;
1678 v->in_scope = c->in_scope;
1679 v->scope_start = c->scope_count;
1685 static struct variable *var_ref(struct parse_context *c, struct text s)
1687 struct binding *b = find_binding(c, s);
1688 struct variable *v = b->var;
1689 struct variable *v2;
1691 switch (v ? v->scope : OutScope) {
1694 /* Caller will report the error */
1697 /* All CondScope variables of this name need to be merged
1698 * and become InScope
1700 v->depth = v->min_depth;
1702 for (v2 = v->previous;
1703 v2 && v2->scope == CondScope;
1705 variable_merge(v, v2);
1713 static int var_refile(struct parse_context *c, struct variable *v)
1715 /* Variable just went out of scope. Add it to the out_scope
1716 * list, sorted by ->scope_start
1718 struct variable **vp = &c->out_scope;
1719 while ((*vp) && (*vp)->scope_start < v->scope_start)
1720 vp = &(*vp)->in_scope;
1726 static void var_block_close(struct parse_context *c, enum closetype ct,
1729 /* Close off all variables that are in_scope.
1730 * Some variables in c->scope may already be not-in-scope,
1731 * such as when a PendingScope variable is hidden by a new
1732 * variable with the same name.
1733 * So we check for v->name->var != v and drop them.
1734 * If we choose to make a variable OutScope, we drop it
1737 struct variable *v, **vp, *v2;
1740 for (vp = &c->in_scope;
1741 (v = *vp) && v->min_depth > c->scope_depth;
1742 (v->scope == OutScope || v->name->var != v)
1743 ? (*vp = v->in_scope, var_refile(c, v))
1744 : ( vp = &v->in_scope, 0)) {
1745 v->min_depth = c->scope_depth;
1746 if (v->name->var != v)
1747 /* This is still in scope, but we haven't just
1751 v->min_depth = c->scope_depth;
1752 if (v->scope == InScope)
1753 v->scope_end = c->scope_count;
1754 if (v->scope == InScope && e && !v->global) {
1755 /* This variable gets cleaned up when 'e' finishes */
1756 variable_unlink_exec(v);
1757 v->cleanup_exec = e;
1758 v->next_free = e->to_free;
1763 case CloseParallel: /* handle PendingScope */
1767 if (c->scope_stack->child_count == 1)
1768 /* first among parallel branches */
1769 v->scope = PendingScope;
1770 else if (v->previous &&
1771 v->previous->scope == PendingScope)
1772 /* all previous branches used name */
1773 v->scope = PendingScope;
1774 else if (v->type == Tlabel)
1775 /* Labels remain pending even when not used */
1776 v->scope = PendingScope; // UNTESTED
1778 v->scope = OutScope;
1779 if (ct == CloseElse) {
1780 /* All Pending variables with this name
1781 * are now Conditional */
1783 v2 && v2->scope == PendingScope;
1785 v2->scope = CondScope;
1789 /* Not possible as it would require
1790 * parallel scope to be nested immediately
1791 * in a parallel scope, and that never
1795 /* Not possible as we already tested for
1802 if (v->scope == CondScope)
1803 /* Condition cannot continue past end of function */
1806 case CloseSequential:
1807 if (v->type == Tlabel)
1808 v->scope = PendingScope;
1811 v->scope = OutScope;
1814 /* There was no 'else', so we can only become
1815 * conditional if we know the cases were exhaustive,
1816 * and that doesn't mean anything yet.
1817 * So only labels become conditional..
1820 v2 && v2->scope == PendingScope;
1822 if (v2->type == Tlabel)
1823 v2->scope = CondScope;
1825 v2->scope = OutScope;
1828 case OutScope: break;
1837 The value of a variable is store separately from the variable, on an
1838 analogue of a stack frame. There are (currently) two frames that can be
1839 active. A global frame which currently only stores constants, and a
1840 stacked frame which stores local variables. Each variable knows if it
1841 is global or not, and what its index into the frame is.
1843 Values in the global frame are known immediately they are relevant, so
1844 the frame needs to be reallocated as it grows so it can store those
1845 values. The local frame doesn't get values until the interpreted phase
1846 is started, so there is no need to allocate until the size is known.
1848 We initialize the `frame_pos` to an impossible value, so that we can
1849 tell if it was set or not later.
1851 ###### variable fields
1855 ###### variable init
1858 ###### parse context
1860 short global_size, global_alloc;
1862 void *global, *local;
1864 ###### forward decls
1865 static struct value *global_alloc(struct parse_context *c, struct type *t,
1866 struct variable *v, struct value *init);
1868 ###### ast functions
1870 static struct value *var_value(struct parse_context *c, struct variable *v)
1873 if (!c->local || !v->type)
1875 if (v->frame_pos + v->type->size > c->local_size) {
1876 printf("INVALID frame_pos\n"); // NOTEST
1879 return c->local + v->frame_pos;
1881 if (c->global_size > c->global_alloc) {
1882 int old = c->global_alloc;
1883 c->global_alloc = (c->global_size | 1023) + 1024;
1884 c->global = realloc(c->global, c->global_alloc);
1885 memset(c->global + old, 0, c->global_alloc - old);
1887 return c->global + v->frame_pos;
1890 static struct value *global_alloc(struct parse_context *c, struct type *t,
1891 struct variable *v, struct value *init)
1894 struct variable scratch;
1896 if (t->prepare_type)
1897 t->prepare_type(c, t, 1); // NOTEST
1899 if (c->global_size & (t->align - 1))
1900 c->global_size = (c->global_size + t->align) & ~(t->align-1); // NOTEST
1905 v->frame_pos = c->global_size;
1907 c->global_size += v->type->size;
1908 ret = var_value(c, v);
1910 memcpy(ret, init, t->size);
1916 As global values are found -- struct field initializers, labels etc --
1917 `global_alloc()` is called to record the value in the global frame.
1919 When the program is fully parsed, each function is analysed, we need to
1920 walk the list of variables local to that function and assign them an
1921 offset in the stack frame. For this we have `scope_finalize()`.
1923 We keep the stack from dense by re-using space for between variables
1924 that are not in scope at the same time. The `out_scope` list is sorted
1925 by `scope_start` and as we process a varible, we move it to an FIFO
1926 stack. For each variable we consider, we first discard any from the
1927 stack anything that went out of scope before the new variable came in.
1928 Then we place the new variable just after the one at the top of the
1931 ###### ast functions
1933 static void scope_finalize(struct parse_context *c, struct type *ft)
1935 int size = ft->function.local_size;
1936 struct variable *next = ft->function.scope;
1937 struct variable *done = NULL;
1940 struct variable *v = next;
1941 struct type *t = v->type;
1948 if (v->frame_pos >= 0)
1950 while (done && done->scope_end < v->scope_start)
1951 done = done->in_scope;
1953 pos = done->frame_pos + done->type->size;
1955 pos = ft->function.local_size;
1956 if (pos & (t->align - 1))
1957 pos = (pos + t->align) & ~(t->align-1);
1959 if (size < pos + v->type->size)
1960 size = pos + v->type->size;
1964 c->out_scope = NULL;
1965 ft->function.local_size = size;
1968 ###### free context storage
1969 free(context.global);
1971 #### Variables as executables
1973 Just as we used a `val` to wrap a value into an `exec`, we similarly
1974 need a `var` to wrap a `variable` into an exec. While each `val`
1975 contained a copy of the value, each `var` holds a link to the variable
1976 because it really is the same variable no matter where it appears.
1977 When a variable is used, we need to remember to follow the `->merged`
1978 link to find the primary instance.
1980 When a variable is declared, it may or may not be given an explicit
1981 type. We need to record which so that we can report the parsed code
1990 struct variable *var;
1993 ###### variable fields
2001 VariableDecl -> IDENTIFIER : ${ {
2002 struct variable *v = var_decl(c, $1.txt);
2003 $0 = new_pos(var, $1);
2008 v = var_ref(c, $1.txt);
2010 type_err(c, "error: variable '%v' redeclared",
2012 type_err(c, "info: this is where '%v' was first declared",
2013 v->where_decl, NULL, 0, NULL);
2016 | IDENTIFIER :: ${ {
2017 struct variable *v = var_decl(c, $1.txt);
2018 $0 = new_pos(var, $1);
2024 v = var_ref(c, $1.txt);
2026 type_err(c, "error: variable '%v' redeclared",
2028 type_err(c, "info: this is where '%v' was first declared",
2029 v->where_decl, NULL, 0, NULL);
2032 | IDENTIFIER : Type ${ {
2033 struct variable *v = var_decl(c, $1.txt);
2034 $0 = new_pos(var, $1);
2040 v->explicit_type = 1;
2042 v = var_ref(c, $1.txt);
2044 type_err(c, "error: variable '%v' redeclared",
2046 type_err(c, "info: this is where '%v' was first declared",
2047 v->where_decl, NULL, 0, NULL);
2050 | IDENTIFIER :: Type ${ {
2051 struct variable *v = var_decl(c, $1.txt);
2052 $0 = new_pos(var, $1);
2059 v->explicit_type = 1;
2061 v = var_ref(c, $1.txt);
2063 type_err(c, "error: variable '%v' redeclared",
2065 type_err(c, "info: this is where '%v' was first declared",
2066 v->where_decl, NULL, 0, NULL);
2071 Variable -> IDENTIFIER ${ {
2072 struct variable *v = var_ref(c, $1.txt);
2073 $0 = new_pos(var, $1);
2075 /* This might be a label - allocate a var just in case */
2076 v = var_decl(c, $1.txt);
2083 cast(var, $0)->var = v;
2086 ###### print exec cases
2089 struct var *v = cast(var, e);
2091 struct binding *b = v->var->name;
2092 printf("%.*s", b->name.len, b->name.txt);
2099 if (loc && loc->type == Xvar) {
2100 struct var *v = cast(var, loc);
2102 struct binding *b = v->var->name;
2103 fprintf(stderr, "%.*s", b->name.len, b->name.txt);
2105 fputs("???", stderr); // NOTEST
2107 fputs("NOTVAR", stderr);
2110 ###### propagate exec cases
2114 struct var *var = cast(var, prog);
2115 struct variable *v = var->var;
2117 type_err(c, "%d:BUG: no variable!!", prog, NULL, 0, NULL); // NOTEST
2118 return Tnone; // NOTEST
2121 if (v->constant && (rules & Rnoconstant)) {
2122 type_err(c, "error: Cannot assign to a constant: %v",
2123 prog, NULL, 0, NULL);
2124 type_err(c, "info: name was defined as a constant here",
2125 v->where_decl, NULL, 0, NULL);
2128 if (v->type == Tnone && v->where_decl == prog)
2129 type_err(c, "error: variable used but not declared: %v",
2130 prog, NULL, 0, NULL);
2131 if (v->type == NULL) {
2132 if (type && *ok != 0) {
2134 v->where_set = prog;
2139 if (!type_compat(type, v->type, rules)) {
2140 type_err(c, "error: expected %1%r but variable '%v' is %2", prog,
2141 type, rules, v->type);
2142 type_err(c, "info: this is where '%v' was set to %1", v->where_set,
2143 v->type, rules, NULL);
2150 ###### interp exec cases
2153 struct var *var = cast(var, e);
2154 struct variable *v = var->var;
2157 lrv = var_value(c, v);
2162 ###### ast functions
2164 static void free_var(struct var *v)
2169 ###### free exec cases
2170 case Xvar: free_var(cast(var, e)); break;
2175 Now that we have the shape of the interpreter in place we can add some
2176 complex types and connected them in to the data structures and the
2177 different phases of parse, analyse, print, interpret.
2179 Being "complex" the language will naturally have syntax to access
2180 specifics of objects of these types. These will fit into the grammar as
2181 "Terms" which are the things that are combined with various operators to
2182 form "Expression". Where a Term is formed by some operation on another
2183 Term, the subordinate Term will always come first, so for example a
2184 member of an array will be expressed as the Term for the array followed
2185 by an index in square brackets. The strict rule of using postfix
2186 operations makes precedence irrelevant within terms. To provide a place
2187 to put the grammar for each terms of each type, we will start out by
2188 introducing the "Term" grammar production, with contains at least a
2189 simple "Value" (to be explained later).
2193 Term -> Value ${ $0 = $<1; }$
2194 | Variable ${ $0 = $<1; }$
2197 Thus far the complex types we have are arrays and structs.
2201 Arrays can be declared by giving a size and a type, as `[size]type' so
2202 `freq:[26]number` declares `freq` to be an array of 26 numbers. The
2203 size can be either a literal number, or a named constant. Some day an
2204 arbitrary expression will be supported.
2206 As a formal parameter to a function, the array can be declared with a
2207 new variable as the size: `name:[size::number]string`. The `size`
2208 variable is set to the size of the array and must be a constant. As
2209 `number` is the only supported type, it can be left out:
2210 `name:[size::]string`.
2212 Arrays cannot be assigned. When pointers are introduced we will also
2213 introduce array slices which can refer to part or all of an array -
2214 the assignment syntax will create a slice. For now, an array can only
2215 ever be referenced by the name it is declared with. It is likely that
2216 a "`copy`" primitive will eventually be define which can be used to
2217 make a copy of an array with controllable recursive depth.
2219 For now we have two sorts of array, those with fixed size either because
2220 it is given as a literal number or because it is a struct member (which
2221 cannot have a runtime-changing size), and those with a size that is
2222 determined at runtime - local variables with a const size. The former
2223 have their size calculated at parse time, the latter at run time.
2225 For the latter type, the `size` field of the type is the size of a
2226 pointer, and the array is reallocated every time it comes into scope.
2228 We differentiate struct fields with a const size from local variables
2229 with a const size by whether they are prepared at parse time or not.
2231 ###### type union fields
2234 int unspec; // size is unspecified - vsize must be set.
2237 struct variable *vsize;
2238 struct type *member;
2241 ###### value union fields
2242 void *array; // used if not static_size
2244 ###### value functions
2246 static void array_prepare_type(struct parse_context *c, struct type *type,
2249 struct value *vsize;
2251 if (type->array.static_size)
2253 if (type->array.unspec && parse_time)
2256 if (type->array.vsize) {
2257 vsize = var_value(c, type->array.vsize);
2261 mpz_tdiv_q(q, mpq_numref(vsize->num), mpq_denref(vsize->num));
2262 type->array.size = mpz_get_si(q);
2266 if (parse_time && type->array.member->size) {
2267 type->array.static_size = 1;
2268 type->size = type->array.size * type->array.member->size;
2269 type->align = type->array.member->align;
2273 static void array_init(struct type *type, struct value *val)
2276 void *ptr = val->ptr;
2280 if (!type->array.static_size) {
2281 val->array = calloc(type->array.size,
2282 type->array.member->size);
2285 for (i = 0; i < type->array.size; i++) {
2287 v = (void*)ptr + i * type->array.member->size;
2288 val_init(type->array.member, v);
2292 static void array_free(struct type *type, struct value *val)
2295 void *ptr = val->ptr;
2297 if (!type->array.static_size)
2299 for (i = 0; i < type->array.size; i++) {
2301 v = (void*)ptr + i * type->array.member->size;
2302 free_value(type->array.member, v);
2304 if (!type->array.static_size)
2308 static int array_compat(struct type *require, struct type *have)
2310 if (have->compat != require->compat)
2312 /* Both are arrays, so we can look at details */
2313 if (!type_compat(require->array.member, have->array.member, 0))
2315 if (have->array.unspec && require->array.unspec) {
2316 if (have->array.vsize && require->array.vsize &&
2317 have->array.vsize != require->array.vsize) // UNTESTED
2318 /* sizes might not be the same */
2319 return 0; // UNTESTED
2322 if (have->array.unspec || require->array.unspec)
2323 return 1; // UNTESTED
2324 if (require->array.vsize == NULL && have->array.vsize == NULL)
2325 return require->array.size == have->array.size;
2327 return require->array.vsize == have->array.vsize; // UNTESTED
2330 static void array_print_type(struct type *type, FILE *f)
2333 if (type->array.vsize) {
2334 struct binding *b = type->array.vsize->name;
2335 fprintf(f, "%.*s%s]", b->name.len, b->name.txt,
2336 type->array.unspec ? "::" : "");
2337 } else if (type->array.size)
2338 fprintf(f, "%d]", type->array.size);
2341 type_print(type->array.member, f);
2344 static struct type array_prototype = {
2346 .prepare_type = array_prepare_type,
2347 .print_type = array_print_type,
2348 .compat = array_compat,
2350 .size = sizeof(void*),
2351 .align = sizeof(void*),
2354 ###### declare terminals
2359 | [ NUMBER ] Type ${ {
2365 if (number_parse(num, tail, $2.txt) == 0)
2366 tok_err(c, "error: unrecognised number", &$2);
2368 tok_err(c, "error: unsupported number suffix", &$2);
2371 elements = mpz_get_ui(mpq_numref(num));
2372 if (mpz_cmp_ui(mpq_denref(num), 1) != 0) {
2373 tok_err(c, "error: array size must be an integer",
2375 } else if (mpz_cmp_ui(mpq_numref(num), 1UL << 30) >= 0)
2376 tok_err(c, "error: array size is too large",
2381 $0 = t = add_anon_type(c, &array_prototype, "array[%d]", elements );
2382 t->array.size = elements;
2383 t->array.member = $<4;
2384 t->array.vsize = NULL;
2387 | [ IDENTIFIER ] Type ${ {
2388 struct variable *v = var_ref(c, $2.txt);
2391 tok_err(c, "error: name undeclared", &$2);
2392 else if (!v->constant)
2393 tok_err(c, "error: array size must be a constant", &$2);
2395 $0 = add_anon_type(c, &array_prototype, "array[%.*s]", $2.txt.len, $2.txt.txt);
2396 $0->array.member = $<4;
2398 $0->array.vsize = v;
2403 OptType -> Type ${ $0 = $<1; }$
2406 ###### formal type grammar
2408 | [ IDENTIFIER :: OptType ] Type ${ {
2409 struct variable *v = var_decl(c, $ID.txt);
2415 $0 = add_anon_type(c, &array_prototype, "array[var]");
2416 $0->array.member = $<6;
2418 $0->array.unspec = 1;
2419 $0->array.vsize = v;
2427 | Term [ Expression ] ${ {
2428 struct binode *b = new(binode);
2435 ###### print binode cases
2437 print_exec(b->left, -1, bracket);
2439 print_exec(b->right, -1, bracket);
2443 ###### propagate binode cases
2445 /* left must be an array, right must be a number,
2446 * result is the member type of the array
2448 propagate_types(b->right, c, ok, Tnum, 0);
2449 t = propagate_types(b->left, c, ok, NULL, rules & Rnoconstant);
2450 if (!t || t->compat != array_compat) {
2451 type_err(c, "error: %1 cannot be indexed", prog, t, 0, NULL);
2454 if (!type_compat(type, t->array.member, rules)) {
2455 type_err(c, "error: have %1 but need %2", prog,
2456 t->array.member, rules, type);
2458 return t->array.member;
2462 ###### interp binode cases
2468 lleft = linterp_exec(c, b->left, <ype);
2469 right = interp_exec(c, b->right, &rtype);
2471 mpz_tdiv_q(q, mpq_numref(right.num), mpq_denref(right.num));
2475 if (ltype->array.static_size)
2478 ptr = *(void**)lleft;
2479 rvtype = ltype->array.member;
2480 if (i >= 0 && i < ltype->array.size)
2481 lrv = ptr + i * rvtype->size;
2483 val_init(ltype->array.member, &rv); // UNSAFE
2490 A `struct` is a data-type that contains one or more other data-types.
2491 It differs from an array in that each member can be of a different
2492 type, and they are accessed by name rather than by number. Thus you
2493 cannot choose an element by calculation, you need to know what you
2496 The language makes no promises about how a given structure will be
2497 stored in memory - it is free to rearrange fields to suit whatever
2498 criteria seems important.
2500 Structs are declared separately from program code - they cannot be
2501 declared in-line in a variable declaration like arrays can. A struct
2502 is given a name and this name is used to identify the type - the name
2503 is not prefixed by the word `struct` as it would be in C.
2505 Structs are only treated as the same if they have the same name.
2506 Simply having the same fields in the same order is not enough. This
2507 might change once we can create structure initializers from a list of
2510 Each component datum is identified much like a variable is declared,
2511 with a name, one or two colons, and a type. The type cannot be omitted
2512 as there is no opportunity to deduce the type from usage. An initial
2513 value can be given following an equals sign, so
2515 ##### Example: a struct type
2521 would declare a type called "complex" which has two number fields,
2522 each initialised to zero.
2524 Struct will need to be declared separately from the code that uses
2525 them, so we will need to be able to print out the declaration of a
2526 struct when reprinting the whole program. So a `print_type_decl` type
2527 function will be needed.
2529 ###### type union fields
2538 } *fields; // This is created when field_list is analysed.
2540 struct fieldlist *prev;
2543 } *field_list; // This is created during parsing
2546 ###### type functions
2547 void (*print_type_decl)(struct type *type, FILE *f);
2549 ###### value functions
2551 static void structure_init(struct type *type, struct value *val)
2555 for (i = 0; i < type->structure.nfields; i++) {
2557 v = (void*) val->ptr + type->structure.fields[i].offset;
2558 if (type->structure.fields[i].init)
2559 dup_value(type->structure.fields[i].type,
2560 type->structure.fields[i].init,
2563 val_init(type->structure.fields[i].type, v);
2567 static void structure_free(struct type *type, struct value *val)
2571 for (i = 0; i < type->structure.nfields; i++) {
2573 v = (void*)val->ptr + type->structure.fields[i].offset;
2574 free_value(type->structure.fields[i].type, v);
2578 static void free_fieldlist(struct fieldlist *f)
2582 free_fieldlist(f->prev);
2587 static void structure_free_type(struct type *t)
2590 for (i = 0; i < t->structure.nfields; i++)
2591 if (t->structure.fields[i].init) {
2592 free_value(t->structure.fields[i].type,
2593 t->structure.fields[i].init);
2595 free(t->structure.fields);
2596 free_fieldlist(t->structure.field_list);
2599 static void structure_prepare_type(struct parse_context *c,
2600 struct type *t, int parse_time)
2603 struct fieldlist *f;
2605 if (!parse_time || t->structure.fields)
2608 for (f = t->structure.field_list; f; f=f->prev) {
2612 if (f->f.type->prepare_type)
2613 f->f.type->prepare_type(c, f->f.type, 1);
2614 if (f->init == NULL)
2618 propagate_types(f->init, c, &ok, f->f.type, 0);
2621 c->parse_error = 1; // NOTEST
2624 t->structure.nfields = cnt;
2625 t->structure.fields = calloc(cnt, sizeof(struct field));
2626 f = t->structure.field_list;
2628 int a = f->f.type->align;
2630 t->structure.fields[cnt] = f->f;
2631 if (t->size & (a-1))
2632 t->size = (t->size | (a-1)) + 1;
2633 t->structure.fields[cnt].offset = t->size;
2634 t->size += ((f->f.type->size - 1) | (a-1)) + 1;
2638 if (f->init && !c->parse_error) {
2639 struct value vl = interp_exec(c, f->init, NULL);
2640 t->structure.fields[cnt].init =
2641 global_alloc(c, f->f.type, NULL, &vl);
2648 static struct type structure_prototype = {
2649 .init = structure_init,
2650 .free = structure_free,
2651 .free_type = structure_free_type,
2652 .print_type_decl = structure_print_type,
2653 .prepare_type = structure_prepare_type,
2667 ###### free exec cases
2669 free_exec(cast(fieldref, e)->left);
2673 ###### declare terminals
2678 | Term . IDENTIFIER ${ {
2679 struct fieldref *fr = new_pos(fieldref, $2);
2686 ###### print exec cases
2690 struct fieldref *f = cast(fieldref, e);
2691 print_exec(f->left, -1, bracket);
2692 printf(".%.*s", f->name.len, f->name.txt);
2696 ###### ast functions
2697 static int find_struct_index(struct type *type, struct text field)
2700 for (i = 0; i < type->structure.nfields; i++)
2701 if (text_cmp(type->structure.fields[i].name, field) == 0)
2706 ###### propagate exec cases
2710 struct fieldref *f = cast(fieldref, prog);
2711 struct type *st = propagate_types(f->left, c, ok, NULL, 0);
2714 type_err(c, "error: unknown type for field access", f->left, // UNTESTED
2716 else if (st->init != structure_init)
2717 type_err(c, "error: field reference attempted on %1, not a struct",
2718 f->left, st, 0, NULL);
2719 else if (f->index == -2) {
2720 f->index = find_struct_index(st, f->name);
2722 type_err(c, "error: cannot find requested field in %1",
2723 f->left, st, 0, NULL);
2725 if (f->index >= 0) {
2726 struct type *ft = st->structure.fields[f->index].type;
2727 if (!type_compat(type, ft, rules))
2728 type_err(c, "error: have %1 but need %2", prog,
2735 ###### interp exec cases
2738 struct fieldref *f = cast(fieldref, e);
2740 struct value *lleft = linterp_exec(c, f->left, <ype);
2741 lrv = (void*)lleft->ptr + ltype->structure.fields[f->index].offset;
2742 rvtype = ltype->structure.fields[f->index].type;
2746 ###### top level grammar
2747 DeclareStruct -> struct IDENTIFIER FieldBlock Newlines ${ {
2749 add_type(c, $2.txt, &structure_prototype);
2750 t->structure.field_list = $<FB;
2754 FieldBlock -> { IN OptNL FieldLines OUT OptNL } ${ $0 = $<FL; }$
2755 | { SimpleFieldList } ${ $0 = $<SFL; }$
2756 | IN OptNL FieldLines OUT ${ $0 = $<FL; }$
2757 | SimpleFieldList EOL ${ $0 = $<SFL; }$
2759 FieldLines -> SimpleFieldList Newlines ${ $0 = $<SFL; }$
2760 | FieldLines SimpleFieldList Newlines ${
2765 SimpleFieldList -> Field ${ $0 = $<F; }$
2766 | SimpleFieldList ; Field ${
2770 | SimpleFieldList ; ${
2773 | ERROR ${ tok_err(c, "Syntax error in struct field", &$1); }$
2775 Field -> IDENTIFIER : Type = Expression ${ {
2776 $0 = calloc(1, sizeof(struct fieldlist));
2777 $0->f.name = $ID.txt;
2778 $0->f.type = $<Type;
2782 | IDENTIFIER : Type ${
2783 $0 = calloc(1, sizeof(struct fieldlist));
2784 $0->f.name = $ID.txt;
2785 $0->f.type = $<Type;
2788 ###### forward decls
2789 static void structure_print_type(struct type *t, FILE *f);
2791 ###### value functions
2792 static void structure_print_type(struct type *t, FILE *f)
2796 fprintf(f, "struct %.*s\n", t->name.len, t->name.txt);
2798 for (i = 0; i < t->structure.nfields; i++) {
2799 struct field *fl = t->structure.fields + i;
2800 fprintf(f, " %.*s : ", fl->name.len, fl->name.txt);
2801 type_print(fl->type, f);
2802 if (fl->type->print && fl->init) {
2804 if (fl->type == Tstr)
2805 fprintf(f, "\""); // UNTESTED
2806 print_value(fl->type, fl->init, f);
2807 if (fl->type == Tstr)
2808 fprintf(f, "\""); // UNTESTED
2814 ###### print type decls
2819 while (target != 0) {
2821 for (t = context.typelist; t ; t=t->next)
2822 if (!t->anon && t->print_type_decl &&
2832 t->print_type_decl(t, stdout);
2840 A function is a chunk of code which can be passed parameters and can
2841 return results. Each function has a type which includes the set of
2842 parameters and the return value. As yet these types cannot be declared
2843 separately from the function itself.
2845 The parameters can be specified either in parentheses as a ';' separated
2848 ##### Example: function 1
2850 func main(av:[ac::number]string; env:[envc::number]string)
2853 or as an indented list of one parameter per line (though each line can
2854 be a ';' separated list)
2856 ##### Example: function 2
2859 argv:[argc::number]string
2860 env:[envc::number]string
2864 In the first case a return type can follow the parentheses after a colon,
2865 in the second it is given on a line starting with the word `return`.
2867 ##### Example: functions that return
2869 func add(a:number; b:number): number
2879 Rather than returning a type, the function can specify a set of local
2880 variables to return as a struct. The values of these variables when the
2881 function exits will be provided to the caller. For this the return type
2882 is replaced with a block of result declarations, either in parentheses
2883 or bracketed by `return` and `do`.
2885 ##### Example: functions returning multiple variables
2887 func to_cartesian(rho:number; theta:number):(x:number; y:number)
2900 For constructing the lists we use a `List` binode, which will be
2901 further detailed when Expression Lists are introduced.
2903 ###### type union fields
2906 struct binode *params;
2907 struct type *return_type;
2908 struct variable *scope;
2909 int inline_result; // return value is at start of 'local'
2913 ###### value union fields
2914 struct exec *function;
2916 ###### type functions
2917 void (*check_args)(struct parse_context *c, int *ok,
2918 struct type *require, struct exec *args);
2920 ###### value functions
2922 static void function_free(struct type *type, struct value *val)
2924 free_exec(val->function);
2925 val->function = NULL;
2928 static int function_compat(struct type *require, struct type *have)
2930 // FIXME can I do anything here yet?
2934 static void function_check_args(struct parse_context *c, int *ok,
2935 struct type *require, struct exec *args)
2937 /* This should be 'compat', but we don't have a 'tuple' type to
2938 * hold the type of 'args'
2940 struct binode *arg = cast(binode, args);
2941 struct binode *param = require->function.params;
2944 struct var *pv = cast(var, param->left);
2946 type_err(c, "error: insufficient arguments to function.",
2947 args, NULL, 0, NULL);
2951 propagate_types(arg->left, c, ok, pv->var->type, 0);
2952 param = cast(binode, param->right);
2953 arg = cast(binode, arg->right);
2956 type_err(c, "error: too many arguments to function.",
2957 args, NULL, 0, NULL);
2960 static void function_print(struct type *type, struct value *val, FILE *f)
2962 print_exec(val->function, 1, 0);
2965 static void function_print_type_decl(struct type *type, FILE *f)
2969 for (b = type->function.params; b; b = cast(binode, b->right)) {
2970 struct variable *v = cast(var, b->left)->var;
2971 fprintf(f, "%.*s%s", v->name->name.len, v->name->name.txt,
2972 v->constant ? "::" : ":");
2973 type_print(v->type, f);
2978 if (type->function.return_type != Tnone) {
2980 if (type->function.inline_result) {
2982 struct type *t = type->function.return_type;
2984 for (i = 0; i < t->structure.nfields; i++) {
2985 struct field *fl = t->structure.fields + i;
2988 fprintf(f, "%.*s:", fl->name.len, fl->name.txt);
2989 type_print(fl->type, f);
2993 type_print(type->function.return_type, f);
2998 static void function_free_type(struct type *t)
3000 free_exec(t->function.params);
3003 static struct type function_prototype = {
3004 .size = sizeof(void*),
3005 .align = sizeof(void*),
3006 .free = function_free,
3007 .compat = function_compat,
3008 .check_args = function_check_args,
3009 .print = function_print,
3010 .print_type_decl = function_print_type_decl,
3011 .free_type = function_free_type,
3014 ###### declare terminals
3024 FuncName -> IDENTIFIER ${ {
3025 struct variable *v = var_decl(c, $1.txt);
3026 struct var *e = new_pos(var, $1);
3032 v = var_ref(c, $1.txt);
3034 type_err(c, "error: function '%v' redeclared",
3036 type_err(c, "info: this is where '%v' was first declared",
3037 v->where_decl, NULL, 0, NULL);
3043 Args -> ArgsLine NEWLINE ${ $0 = $<AL; }$
3044 | Args ArgsLine NEWLINE ${ {
3045 struct binode *b = $<AL;
3046 struct binode **bp = &b;
3048 bp = (struct binode **)&(*bp)->left;
3053 ArgsLine -> ${ $0 = NULL; }$
3054 | Varlist ${ $0 = $<1; }$
3055 | Varlist ; ${ $0 = $<1; }$
3057 Varlist -> Varlist ; ArgDecl ${
3071 ArgDecl -> IDENTIFIER : FormalType ${ {
3072 struct variable *v = var_decl(c, $1.txt);
3078 ##### Function calls
3080 A function call can appear either as an expression or as a statement.
3081 We use a new 'Funcall' binode type to link the function with a list of
3082 arguments, form with the 'List' nodes.
3084 We have already seen the "Term" which is how a function call can appear
3085 in an expression. To parse a function call into a statement we include
3086 it in the "SimpleStatement Grammar" which will be described later.
3092 | Term ( ExpressionList ) ${ {
3093 struct binode *b = new(binode);
3096 b->right = reorder_bilist($<EL);
3100 struct binode *b = new(binode);
3107 ###### SimpleStatement Grammar
3109 | Term ( ExpressionList ) ${ {
3110 struct binode *b = new(binode);
3113 b->right = reorder_bilist($<EL);
3117 ###### print binode cases
3120 do_indent(indent, "");
3121 print_exec(b->left, -1, bracket);
3123 for (b = cast(binode, b->right); b; b = cast(binode, b->right)) {
3126 print_exec(b->left, -1, bracket);
3136 ###### propagate binode cases
3139 /* Every arg must match formal parameter, and result
3140 * is return type of function
3142 struct binode *args = cast(binode, b->right);
3143 struct var *v = cast(var, b->left);
3145 if (!v->var->type || v->var->type->check_args == NULL) {
3146 type_err(c, "error: attempt to call a non-function.",
3147 prog, NULL, 0, NULL);
3150 v->var->type->check_args(c, ok, v->var->type, args);
3151 return v->var->type->function.return_type;
3154 ###### interp binode cases
3157 struct var *v = cast(var, b->left);
3158 struct type *t = v->var->type;
3159 void *oldlocal = c->local;
3160 int old_size = c->local_size;
3161 void *local = calloc(1, t->function.local_size);
3162 struct value *fbody = var_value(c, v->var);
3163 struct binode *arg = cast(binode, b->right);
3164 struct binode *param = t->function.params;
3167 struct var *pv = cast(var, param->left);
3168 struct type *vtype = NULL;
3169 struct value val = interp_exec(c, arg->left, &vtype);
3171 c->local = local; c->local_size = t->function.local_size;
3172 lval = var_value(c, pv->var);
3173 c->local = oldlocal; c->local_size = old_size;
3174 memcpy(lval, &val, vtype->size);
3175 param = cast(binode, param->right);
3176 arg = cast(binode, arg->right);
3178 c->local = local; c->local_size = t->function.local_size;
3179 if (t->function.inline_result && dtype) {
3180 _interp_exec(c, fbody->function, NULL, NULL);
3181 memcpy(dest, local, dtype->size);
3182 rvtype = ret.type = NULL;
3184 rv = interp_exec(c, fbody->function, &rvtype);
3185 c->local = oldlocal; c->local_size = old_size;
3190 ## Complex executables: statements and expressions
3192 Now that we have types and values and variables and most of the basic
3193 Terms which provide access to these, we can explore the more complex
3194 code that combine all of these to get useful work done. Specifically
3195 statements and expressions.
3197 Expressions are various combinations of Terms. We will use operator
3198 precedence to ensure correct parsing. The simplest Expression is just a
3199 Term - others will follow.
3204 Expression -> Term ${ $0 = $<Term; }$
3205 ## expression grammar
3207 ### Expressions: Conditional
3209 Our first user of the `binode` will be conditional expressions, which
3210 is a bit odd as they actually have three components. That will be
3211 handled by having 2 binodes for each expression. The conditional
3212 expression is the lowest precedence operator which is why we define it
3213 first - to start the precedence list.
3215 Conditional expressions are of the form "value `if` condition `else`
3216 other_value". They associate to the right, so everything to the right
3217 of `else` is part of an else value, while only a higher-precedence to
3218 the left of `if` is the if values. Between `if` and `else` there is no
3219 room for ambiguity, so a full conditional expression is allowed in
3225 ###### declare terminals
3229 ###### expression grammar
3231 | Expression if Expression else Expression $$ifelse ${ {
3232 struct binode *b1 = new(binode);
3233 struct binode *b2 = new(binode);
3243 ###### print binode cases
3246 b2 = cast(binode, b->right);
3247 if (bracket) printf("(");
3248 print_exec(b2->left, -1, bracket);
3250 print_exec(b->left, -1, bracket);
3252 print_exec(b2->right, -1, bracket);
3253 if (bracket) printf(")");
3256 ###### propagate binode cases
3259 /* cond must be Tbool, others must match */
3260 struct binode *b2 = cast(binode, b->right);
3263 propagate_types(b->left, c, ok, Tbool, 0);
3264 t = propagate_types(b2->left, c, ok, type, Rnolabel);
3265 t2 = propagate_types(b2->right, c, ok, type ?: t, Rnolabel);
3269 ###### interp binode cases
3272 struct binode *b2 = cast(binode, b->right);
3273 left = interp_exec(c, b->left, <ype);
3275 rv = interp_exec(c, b2->left, &rvtype); // UNTESTED
3277 rv = interp_exec(c, b2->right, &rvtype);
3283 We take a brief detour, now that we have expressions, to describe lists
3284 of expressions. These will be needed for function parameters and
3285 possibly other situations. They seem generic enough to introduce here
3286 to be used elsewhere.
3288 And ExpressionList will use the `List` type of `binode`, building up at
3289 the end. And place where they are used will probably call
3290 `reorder_bilist()` to get a more normal first/next arrangement.
3292 ###### declare terminals
3295 `List` execs have no implicit semantics, so they are never propagated or
3296 interpreted. The can be printed as a comma separate list, which is how
3297 they are parsed. Note they are also used for function formal parameter
3298 lists. In that case a separate function is used to print them.
3300 ###### print binode cases
3304 print_exec(b->left, -1, bracket);
3307 b = cast(binode, b->right);
3311 ###### propagate binode cases
3312 case List: abort(); // NOTEST
3313 ###### interp binode cases
3314 case List: abort(); // NOTEST
3319 ExpressionList -> ExpressionList , Expression ${
3332 ### Expressions: Boolean
3334 The next class of expressions to use the `binode` will be Boolean
3335 expressions. "`and then`" and "`or else`" are similar to `and` and `or`
3336 have same corresponding precendence. The difference is that they don't
3337 evaluate the second expression if not necessary.
3346 ###### declare terminals
3351 ###### expression grammar
3352 | Expression or Expression ${ {
3353 struct binode *b = new(binode);
3359 | Expression or else Expression ${ {
3360 struct binode *b = new(binode);
3367 | Expression and Expression ${ {
3368 struct binode *b = new(binode);
3374 | Expression and then Expression ${ {
3375 struct binode *b = new(binode);
3382 | not Expression ${ {
3383 struct binode *b = new(binode);
3389 ###### print binode cases
3391 if (bracket) printf("(");
3392 print_exec(b->left, -1, bracket);
3394 print_exec(b->right, -1, bracket);
3395 if (bracket) printf(")");
3398 if (bracket) printf("(");
3399 print_exec(b->left, -1, bracket);
3400 printf(" and then ");
3401 print_exec(b->right, -1, bracket);
3402 if (bracket) printf(")");
3405 if (bracket) printf("(");
3406 print_exec(b->left, -1, bracket);
3408 print_exec(b->right, -1, bracket);
3409 if (bracket) printf(")");
3412 if (bracket) printf("(");
3413 print_exec(b->left, -1, bracket);
3414 printf(" or else ");
3415 print_exec(b->right, -1, bracket);
3416 if (bracket) printf(")");
3419 if (bracket) printf("(");
3421 print_exec(b->right, -1, bracket);
3422 if (bracket) printf(")");
3425 ###### propagate binode cases
3431 /* both must be Tbool, result is Tbool */
3432 propagate_types(b->left, c, ok, Tbool, 0);
3433 propagate_types(b->right, c, ok, Tbool, 0);
3434 if (type && type != Tbool)
3435 type_err(c, "error: %1 operation found where %2 expected", prog,
3439 ###### interp binode cases
3441 rv = interp_exec(c, b->left, &rvtype);
3442 right = interp_exec(c, b->right, &rtype);
3443 rv.bool = rv.bool && right.bool;
3446 rv = interp_exec(c, b->left, &rvtype);
3448 rv = interp_exec(c, b->right, NULL);
3451 rv = interp_exec(c, b->left, &rvtype);
3452 right = interp_exec(c, b->right, &rtype);
3453 rv.bool = rv.bool || right.bool;
3456 rv = interp_exec(c, b->left, &rvtype);
3458 rv = interp_exec(c, b->right, NULL);
3461 rv = interp_exec(c, b->right, &rvtype);
3465 ### Expressions: Comparison
3467 Of slightly higher precedence that Boolean expressions are Comparisons.
3468 A comparison takes arguments of any comparable type, but the two types
3471 To simplify the parsing we introduce an `eop` which can record an
3472 expression operator, and the `CMPop` non-terminal will match one of them.
3479 ###### ast functions
3480 static void free_eop(struct eop *e)
3494 ###### declare terminals
3495 $LEFT < > <= >= == != CMPop
3497 ###### expression grammar
3498 | Expression CMPop Expression ${ {
3499 struct binode *b = new(binode);
3509 CMPop -> < ${ $0.op = Less; }$
3510 | > ${ $0.op = Gtr; }$
3511 | <= ${ $0.op = LessEq; }$
3512 | >= ${ $0.op = GtrEq; }$
3513 | == ${ $0.op = Eql; }$
3514 | != ${ $0.op = NEql; }$
3516 ###### print binode cases
3524 if (bracket) printf("(");
3525 print_exec(b->left, -1, bracket);
3527 case Less: printf(" < "); break;
3528 case LessEq: printf(" <= "); break;
3529 case Gtr: printf(" > "); break;
3530 case GtrEq: printf(" >= "); break;
3531 case Eql: printf(" == "); break;
3532 case NEql: printf(" != "); break;
3533 default: abort(); // NOTEST
3535 print_exec(b->right, -1, bracket);
3536 if (bracket) printf(")");
3539 ###### propagate binode cases
3546 /* Both must match but not be labels, result is Tbool */
3547 t = propagate_types(b->left, c, ok, NULL, Rnolabel);
3549 propagate_types(b->right, c, ok, t, 0);
3551 t = propagate_types(b->right, c, ok, NULL, Rnolabel); // UNTESTED
3553 t = propagate_types(b->left, c, ok, t, 0); // UNTESTED
3555 if (!type_compat(type, Tbool, 0))
3556 type_err(c, "error: Comparison returns %1 but %2 expected", prog,
3557 Tbool, rules, type);
3560 ###### interp binode cases
3569 left = interp_exec(c, b->left, <ype);
3570 right = interp_exec(c, b->right, &rtype);
3571 cmp = value_cmp(ltype, rtype, &left, &right);
3574 case Less: rv.bool = cmp < 0; break;
3575 case LessEq: rv.bool = cmp <= 0; break;
3576 case Gtr: rv.bool = cmp > 0; break;
3577 case GtrEq: rv.bool = cmp >= 0; break;
3578 case Eql: rv.bool = cmp == 0; break;
3579 case NEql: rv.bool = cmp != 0; break;
3580 default: rv.bool = 0; break; // NOTEST
3585 ### Expressions: Arithmetic etc.
3587 The remaining expressions with the highest precedence are arithmetic,
3588 string concatenation, and string conversion. String concatenation
3589 (`++`) has the same precedence as multiplication and division, but lower
3592 String conversion is a temporary feature until I get a better type
3593 system. `$` is a prefix operator which expects a string and returns
3596 `+` and `-` are both infix and prefix operations (where they are
3597 absolute value and negation). These have different operator names.
3599 We also have a 'Bracket' operator which records where parentheses were
3600 found. This makes it easy to reproduce these when printing. Possibly I
3601 should only insert brackets were needed for precedence. Putting
3602 parentheses around an expression converts it into a Term,
3612 ###### declare terminals
3618 ###### expression grammar
3619 | Expression Eop Expression ${ {
3620 struct binode *b = new(binode);
3627 | Expression Top Expression ${ {
3628 struct binode *b = new(binode);
3635 | Uop Expression ${ {
3636 struct binode *b = new(binode);
3644 | ( Expression ) ${ {
3645 struct binode *b = new_pos(binode, $1);
3654 Eop -> + ${ $0.op = Plus; }$
3655 | - ${ $0.op = Minus; }$
3657 Uop -> + ${ $0.op = Absolute; }$
3658 | - ${ $0.op = Negate; }$
3659 | $ ${ $0.op = StringConv; }$
3661 Top -> * ${ $0.op = Times; }$
3662 | / ${ $0.op = Divide; }$
3663 | % ${ $0.op = Rem; }$
3664 | ++ ${ $0.op = Concat; }$
3666 ###### print binode cases
3673 if (bracket) printf("(");
3674 print_exec(b->left, indent, bracket);
3676 case Plus: fputs(" + ", stdout); break;
3677 case Minus: fputs(" - ", stdout); break;
3678 case Times: fputs(" * ", stdout); break;
3679 case Divide: fputs(" / ", stdout); break;
3680 case Rem: fputs(" % ", stdout); break;
3681 case Concat: fputs(" ++ ", stdout); break;
3682 default: abort(); // NOTEST
3684 print_exec(b->right, indent, bracket);
3685 if (bracket) printf(")");
3690 if (bracket) printf("(");
3692 case Absolute: fputs("+", stdout); break;
3693 case Negate: fputs("-", stdout); break;
3694 case StringConv: fputs("$", stdout); break;
3695 default: abort(); // NOTEST
3697 print_exec(b->right, indent, bracket);
3698 if (bracket) printf(")");
3702 print_exec(b->right, indent, bracket);
3706 ###### propagate binode cases
3712 /* both must be numbers, result is Tnum */
3715 /* as propagate_types ignores a NULL,
3716 * unary ops fit here too */
3717 propagate_types(b->left, c, ok, Tnum, 0);
3718 propagate_types(b->right, c, ok, Tnum, 0);
3719 if (!type_compat(type, Tnum, 0))
3720 type_err(c, "error: Arithmetic returns %1 but %2 expected", prog,
3725 /* both must be Tstr, result is Tstr */
3726 propagate_types(b->left, c, ok, Tstr, 0);
3727 propagate_types(b->right, c, ok, Tstr, 0);
3728 if (!type_compat(type, Tstr, 0))
3729 type_err(c, "error: Concat returns %1 but %2 expected", prog,
3734 /* op must be string, result is number */
3735 propagate_types(b->left, c, ok, Tstr, 0);
3736 if (!type_compat(type, Tnum, 0))
3737 type_err(c, // UNTESTED
3738 "error: Can only convert string to number, not %1",
3739 prog, type, 0, NULL);
3743 return propagate_types(b->right, c, ok, type, 0);
3745 ###### interp binode cases
3748 rv = interp_exec(c, b->left, &rvtype);
3749 right = interp_exec(c, b->right, &rtype);
3750 mpq_add(rv.num, rv.num, right.num);
3753 rv = interp_exec(c, b->left, &rvtype);
3754 right = interp_exec(c, b->right, &rtype);
3755 mpq_sub(rv.num, rv.num, right.num);
3758 rv = interp_exec(c, b->left, &rvtype);
3759 right = interp_exec(c, b->right, &rtype);
3760 mpq_mul(rv.num, rv.num, right.num);
3763 rv = interp_exec(c, b->left, &rvtype);
3764 right = interp_exec(c, b->right, &rtype);
3765 mpq_div(rv.num, rv.num, right.num);
3770 left = interp_exec(c, b->left, <ype);
3771 right = interp_exec(c, b->right, &rtype);
3772 mpz_init(l); mpz_init(r); mpz_init(rem);
3773 mpz_tdiv_q(l, mpq_numref(left.num), mpq_denref(left.num));
3774 mpz_tdiv_q(r, mpq_numref(right.num), mpq_denref(right.num));
3775 mpz_tdiv_r(rem, l, r);
3776 val_init(Tnum, &rv);
3777 mpq_set_z(rv.num, rem);
3778 mpz_clear(r); mpz_clear(l); mpz_clear(rem);
3783 rv = interp_exec(c, b->right, &rvtype);
3784 mpq_neg(rv.num, rv.num);
3787 rv = interp_exec(c, b->right, &rvtype);
3788 mpq_abs(rv.num, rv.num);
3791 rv = interp_exec(c, b->right, &rvtype);
3794 left = interp_exec(c, b->left, <ype);
3795 right = interp_exec(c, b->right, &rtype);
3797 rv.str = text_join(left.str, right.str);
3800 right = interp_exec(c, b->right, &rvtype);
3804 struct text tx = right.str;
3807 if (tx.txt[0] == '-') {
3808 neg = 1; // UNTESTED
3809 tx.txt++; // UNTESTED
3810 tx.len--; // UNTESTED
3812 if (number_parse(rv.num, tail, tx) == 0)
3813 mpq_init(rv.num); // UNTESTED
3815 mpq_neg(rv.num, rv.num); // UNTESTED
3817 printf("Unsupported suffix: %.*s\n", tx.len, tx.txt); // UNTESTED
3821 ###### value functions
3823 static struct text text_join(struct text a, struct text b)
3826 rv.len = a.len + b.len;
3827 rv.txt = malloc(rv.len);
3828 memcpy(rv.txt, a.txt, a.len);
3829 memcpy(rv.txt+a.len, b.txt, b.len);
3833 ### Blocks, Statements, and Statement lists.
3835 Now that we have expressions out of the way we need to turn to
3836 statements. There are simple statements and more complex statements.
3837 Simple statements do not contain (syntactic) newlines, complex statements do.
3839 Statements often come in sequences and we have corresponding simple
3840 statement lists and complex statement lists.
3841 The former comprise only simple statements separated by semicolons.
3842 The later comprise complex statements and simple statement lists. They are
3843 separated by newlines. Thus the semicolon is only used to separate
3844 simple statements on the one line. This may be overly restrictive,
3845 but I'm not sure I ever want a complex statement to share a line with
3848 Note that a simple statement list can still use multiple lines if
3849 subsequent lines are indented, so
3851 ###### Example: wrapped simple statement list
3856 is a single simple statement list. This might allow room for
3857 confusion, so I'm not set on it yet.
3859 A simple statement list needs no extra syntax. A complex statement
3860 list has two syntactic forms. It can be enclosed in braces (much like
3861 C blocks), or it can be introduced by an indent and continue until an
3862 unindented newline (much like Python blocks). With this extra syntax
3863 it is referred to as a block.
3865 Note that a block does not have to include any newlines if it only
3866 contains simple statements. So both of:
3868 if condition: a=b; d=f
3870 if condition { a=b; print f }
3874 In either case the list is constructed from a `binode` list with
3875 `Block` as the operator. When parsing the list it is most convenient
3876 to append to the end, so a list is a list and a statement. When using
3877 the list it is more convenient to consider a list to be a statement
3878 and a list. So we need a function to re-order a list.
3879 `reorder_bilist` serves this purpose.
3881 The only stand-alone statement we introduce at this stage is `pass`
3882 which does nothing and is represented as a `NULL` pointer in a `Block`
3883 list. Other stand-alone statements will follow once the infrastructure
3886 As many statements will use binodes, we declare a binode pointer 'b' in
3887 the common header for all reductions to use.
3889 ###### Parser: reduce
3900 Block -> { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3901 | { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3902 | SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
3903 | SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
3904 | IN OptNL Statementlist OUT ${ $0 = $<Sl; }$
3906 OpenBlock -> OpenScope { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3907 | OpenScope { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3908 | OpenScope SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
3909 | OpenScope SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
3910 | IN OpenScope OptNL Statementlist OUT ${ $0 = $<Sl; }$
3912 UseBlock -> { OpenScope IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3913 | { OpenScope SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3914 | IN OpenScope OptNL Statementlist OUT ${ $0 = $<Sl; }$
3916 ColonBlock -> { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3917 | { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3918 | : SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
3919 | : SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
3920 | : IN OptNL Statementlist OUT ${ $0 = $<Sl; }$
3922 Statementlist -> ComplexStatements ${ $0 = reorder_bilist($<CS); }$
3924 ComplexStatements -> ComplexStatements ComplexStatement ${
3934 | ComplexStatement ${
3946 ComplexStatement -> SimpleStatements Newlines ${
3947 $0 = reorder_bilist($<SS);
3949 | SimpleStatements ; Newlines ${
3950 $0 = reorder_bilist($<SS);
3952 ## ComplexStatement Grammar
3955 SimpleStatements -> SimpleStatements ; SimpleStatement ${
3961 | SimpleStatement ${
3970 SimpleStatement -> pass ${ $0 = NULL; }$
3971 | ERROR ${ tok_err(c, "Syntax error in statement", &$1); }$
3972 ## SimpleStatement Grammar
3974 ###### print binode cases
3978 if (b->left == NULL) // UNTESTED
3979 printf("pass"); // UNTESTED
3981 print_exec(b->left, indent, bracket); // UNTESTED
3982 if (b->right) { // UNTESTED
3983 printf("; "); // UNTESTED
3984 print_exec(b->right, indent, bracket); // UNTESTED
3987 // block, one per line
3988 if (b->left == NULL)
3989 do_indent(indent, "pass\n");
3991 print_exec(b->left, indent, bracket);
3993 print_exec(b->right, indent, bracket);
3997 ###### propagate binode cases
4000 /* If any statement returns something other than Tnone
4001 * or Tbool then all such must return same type.
4002 * As each statement may be Tnone or something else,
4003 * we must always pass NULL (unknown) down, otherwise an incorrect
4004 * error might occur. We never return Tnone unless it is
4009 for (e = b; e; e = cast(binode, e->right)) {
4010 t = propagate_types(e->left, c, ok, NULL, rules);
4011 if ((rules & Rboolok) && (t == Tbool || t == Tnone))
4013 if (t == Tnone && e->right)
4014 /* Only the final statement *must* return a value
4022 type_err(c, "error: expected %1%r, found %2",
4023 e->left, type, rules, t);
4029 ###### interp binode cases
4031 while (rvtype == Tnone &&
4034 rv = interp_exec(c, b->left, &rvtype);
4035 b = cast(binode, b->right);
4039 ### The Print statement
4041 `print` is a simple statement that takes a comma-separated list of
4042 expressions and prints the values separated by spaces and terminated
4043 by a newline. No control of formatting is possible.
4045 `print` uses `ExpressionList` to collect the expressions and stores them
4046 on the left side of a `Print` binode unlessthere is a trailing comma
4047 when the list is stored on the `right` side and no trailing newline is
4053 ##### declare terminals
4056 ###### SimpleStatement Grammar
4058 | print ExpressionList ${
4059 $0 = b = new(binode);
4062 b->left = reorder_bilist($<EL);
4064 | print ExpressionList , ${ {
4065 $0 = b = new(binode);
4067 b->right = reorder_bilist($<EL);
4071 $0 = b = new(binode);
4077 ###### print binode cases
4080 do_indent(indent, "print");
4082 print_exec(b->right, -1, bracket);
4085 print_exec(b->left, -1, bracket);
4090 ###### propagate binode cases
4093 /* don't care but all must be consistent */
4095 b = cast(binode, b->left);
4097 b = cast(binode, b->right);
4099 propagate_types(b->left, c, ok, NULL, Rnolabel);
4100 b = cast(binode, b->right);
4104 ###### interp binode cases
4108 struct binode *b2 = cast(binode, b->left);
4110 b2 = cast(binode, b->right);
4111 for (; b2; b2 = cast(binode, b2->right)) {
4112 left = interp_exec(c, b2->left, <ype);
4113 print_value(ltype, &left, stdout);
4114 free_value(ltype, &left);
4118 if (b->right == NULL)
4124 ###### Assignment statement
4126 An assignment will assign a value to a variable, providing it hasn't
4127 been declared as a constant. The analysis phase ensures that the type
4128 will be correct so the interpreter just needs to perform the
4129 calculation. There is a form of assignment which declares a new
4130 variable as well as assigning a value. If a name is assigned before
4131 it is declared, and error will be raised as the name is created as
4132 `Tlabel` and it is illegal to assign to such names.
4138 ###### declare terminals
4141 ###### SimpleStatement Grammar
4142 | Term = Expression ${
4143 $0 = b= new(binode);
4148 | VariableDecl = Expression ${
4149 $0 = b= new(binode);
4156 if ($1->var->where_set == NULL) {
4158 "Variable declared with no type or value: %v",
4162 $0 = b = new(binode);
4169 ###### print binode cases
4172 do_indent(indent, "");
4173 print_exec(b->left, indent, bracket);
4175 print_exec(b->right, indent, bracket);
4182 struct variable *v = cast(var, b->left)->var;
4183 do_indent(indent, "");
4184 print_exec(b->left, indent, bracket);
4185 if (cast(var, b->left)->var->constant) {
4187 if (v->explicit_type) {
4188 type_print(v->type, stdout);
4193 if (v->explicit_type) {
4194 type_print(v->type, stdout);
4200 print_exec(b->right, indent, bracket);
4207 ###### propagate binode cases
4211 /* Both must match and not be labels,
4212 * Type must support 'dup',
4213 * For Assign, left must not be constant.
4216 t = propagate_types(b->left, c, ok, NULL,
4217 Rnolabel | (b->op == Assign ? Rnoconstant : 0));
4222 if (propagate_types(b->right, c, ok, t, 0) != t)
4223 if (b->left->type == Xvar)
4224 type_err(c, "info: variable '%v' was set as %1 here.",
4225 cast(var, b->left)->var->where_set, t, rules, NULL);
4227 t = propagate_types(b->right, c, ok, NULL, Rnolabel);
4229 propagate_types(b->left, c, ok, t,
4230 (b->op == Assign ? Rnoconstant : 0));
4232 if (t && t->dup == NULL && t->name.txt[0] != ' ') // HACK
4233 type_err(c, "error: cannot assign value of type %1", b, t, 0, NULL);
4238 ###### interp binode cases
4241 lleft = linterp_exec(c, b->left, <ype);
4243 dinterp_exec(c, b->right, lleft, ltype, 1);
4249 struct variable *v = cast(var, b->left)->var;
4252 val = var_value(c, v);
4253 if (v->type->prepare_type)
4254 v->type->prepare_type(c, v->type, 0);
4256 dinterp_exec(c, b->right, val, v->type, 0);
4258 val_init(v->type, val);
4262 ### The `use` statement
4264 The `use` statement is the last "simple" statement. It is needed when a
4265 statement block can return a value. This includes the body of a
4266 function which has a return type, and the "condition" code blocks in
4267 `if`, `while`, and `switch` statements.
4272 ###### declare terminals
4275 ###### SimpleStatement Grammar
4277 $0 = b = new_pos(binode, $1);
4280 if (b->right->type == Xvar) {
4281 struct var *v = cast(var, b->right);
4282 if (v->var->type == Tnone) {
4283 /* Convert this to a label */
4286 v->var->type = Tlabel;
4287 val = global_alloc(c, Tlabel, v->var, NULL);
4293 ###### print binode cases
4296 do_indent(indent, "use ");
4297 print_exec(b->right, -1, bracket);
4302 ###### propagate binode cases
4305 /* result matches value */
4306 return propagate_types(b->right, c, ok, type, 0);
4308 ###### interp binode cases
4311 rv = interp_exec(c, b->right, &rvtype);
4314 ### The Conditional Statement
4316 This is the biggy and currently the only complex statement. This
4317 subsumes `if`, `while`, `do/while`, `switch`, and some parts of `for`.
4318 It is comprised of a number of parts, all of which are optional though
4319 set combinations apply. Each part is (usually) a key word (`then` is
4320 sometimes optional) followed by either an expression or a code block,
4321 except the `casepart` which is a "key word and an expression" followed
4322 by a code block. The code-block option is valid for all parts and,
4323 where an expression is also allowed, the code block can use the `use`
4324 statement to report a value. If the code block does not report a value
4325 the effect is similar to reporting `True`.
4327 The `else` and `case` parts, as well as `then` when combined with
4328 `if`, can contain a `use` statement which will apply to some
4329 containing conditional statement. `for` parts, `do` parts and `then`
4330 parts used with `for` can never contain a `use`, except in some
4331 subordinate conditional statement.
4333 If there is a `forpart`, it is executed first, only once.
4334 If there is a `dopart`, then it is executed repeatedly providing
4335 always that the `condpart` or `cond`, if present, does not return a non-True
4336 value. `condpart` can fail to return any value if it simply executes
4337 to completion. This is treated the same as returning `True`.
4339 If there is a `thenpart` it will be executed whenever the `condpart`
4340 or `cond` returns True (or does not return any value), but this will happen
4341 *after* `dopart` (when present).
4343 If `elsepart` is present it will be executed at most once when the
4344 condition returns `False` or some value that isn't `True` and isn't
4345 matched by any `casepart`. If there are any `casepart`s, they will be
4346 executed when the condition returns a matching value.
4348 The particular sorts of values allowed in case parts has not yet been
4349 determined in the language design, so nothing is prohibited.
4351 The various blocks in this complex statement potentially provide scope
4352 for variables as described earlier. Each such block must include the
4353 "OpenScope" nonterminal before parsing the block, and must call
4354 `var_block_close()` when closing the block.
4356 The code following "`if`", "`switch`" and "`for`" does not get its own
4357 scope, but is in a scope covering the whole statement, so names
4358 declared there cannot be redeclared elsewhere. Similarly the
4359 condition following "`while`" is in a scope the covers the body
4360 ("`do`" part) of the loop, and which does not allow conditional scope
4361 extension. Code following "`then`" (both looping and non-looping),
4362 "`else`" and "`case`" each get their own local scope.
4364 The type requirements on the code block in a `whilepart` are quite
4365 unusal. It is allowed to return a value of some identifiable type, in
4366 which case the loop aborts and an appropriate `casepart` is run, or it
4367 can return a Boolean, in which case the loop either continues to the
4368 `dopart` (on `True`) or aborts and runs the `elsepart` (on `False`).
4369 This is different both from the `ifpart` code block which is expected to
4370 return a Boolean, or the `switchpart` code block which is expected to
4371 return the same type as the casepart values. The correct analysis of
4372 the type of the `whilepart` code block is the reason for the
4373 `Rboolok` flag which is passed to `propagate_types()`.
4375 The `cond_statement` cannot fit into a `binode` so a new `exec` is
4376 defined. As there are two scopes which cover multiple parts - one for
4377 the whole statement and one for "while" and "do" - and as we will use
4378 the 'struct exec' to track scopes, we actually need two new types of
4379 exec. One is a `binode` for the looping part, the rest is the
4380 `cond_statement`. The `cond_statement` will use an auxilliary `struct
4381 casepart` to track a list of case parts.
4392 struct exec *action;
4393 struct casepart *next;
4395 struct cond_statement {
4397 struct exec *forpart, *condpart, *thenpart, *elsepart;
4398 struct binode *looppart;
4399 struct casepart *casepart;
4402 ###### ast functions
4404 static void free_casepart(struct casepart *cp)
4408 free_exec(cp->value);
4409 free_exec(cp->action);
4416 static void free_cond_statement(struct cond_statement *s)
4420 free_exec(s->forpart);
4421 free_exec(s->condpart);
4422 free_exec(s->looppart);
4423 free_exec(s->thenpart);
4424 free_exec(s->elsepart);
4425 free_casepart(s->casepart);
4429 ###### free exec cases
4430 case Xcond_statement: free_cond_statement(cast(cond_statement, e)); break;
4432 ###### ComplexStatement Grammar
4433 | CondStatement ${ $0 = $<1; }$
4435 ###### declare terminals
4436 $TERM for then while do
4443 // A CondStatement must end with EOL, as does CondSuffix and
4445 // ForPart, ThenPart, SwitchPart, CasePart are non-empty and
4446 // may or may not end with EOL
4447 // WhilePart and IfPart include an appropriate Suffix
4449 // ForPart, SwitchPart, and IfPart open scopes, o we have to close
4450 // them. WhilePart opens and closes its own scope.
4451 CondStatement -> ForPart OptNL ThenPart OptNL WhilePart CondSuffix ${
4454 $0->thenpart = $<TP;
4455 $0->looppart = $<WP;
4456 var_block_close(c, CloseSequential, $0);
4458 | ForPart OptNL WhilePart CondSuffix ${
4461 $0->looppart = $<WP;
4462 var_block_close(c, CloseSequential, $0);
4464 | WhilePart CondSuffix ${
4466 $0->looppart = $<WP;
4468 | SwitchPart OptNL CasePart CondSuffix ${
4470 $0->condpart = $<SP;
4471 $CP->next = $0->casepart;
4472 $0->casepart = $<CP;
4473 var_block_close(c, CloseSequential, $0);
4475 | SwitchPart : IN OptNL CasePart CondSuffix OUT Newlines ${
4477 $0->condpart = $<SP;
4478 $CP->next = $0->casepart;
4479 $0->casepart = $<CP;
4480 var_block_close(c, CloseSequential, $0);
4482 | IfPart IfSuffix ${
4484 $0->condpart = $IP.condpart; $IP.condpart = NULL;
4485 $0->thenpart = $IP.thenpart; $IP.thenpart = NULL;
4486 // This is where we close an "if" statement
4487 var_block_close(c, CloseSequential, $0);
4490 CondSuffix -> IfSuffix ${
4493 | Newlines CasePart CondSuffix ${
4495 $CP->next = $0->casepart;
4496 $0->casepart = $<CP;
4498 | CasePart CondSuffix ${
4500 $CP->next = $0->casepart;
4501 $0->casepart = $<CP;
4504 IfSuffix -> Newlines ${ $0 = new(cond_statement); }$
4505 | Newlines ElsePart ${ $0 = $<EP; }$
4506 | ElsePart ${$0 = $<EP; }$
4508 ElsePart -> else OpenBlock Newlines ${
4509 $0 = new(cond_statement);
4510 $0->elsepart = $<OB;
4511 var_block_close(c, CloseElse, $0->elsepart);
4513 | else OpenScope CondStatement ${
4514 $0 = new(cond_statement);
4515 $0->elsepart = $<CS;
4516 var_block_close(c, CloseElse, $0->elsepart);
4520 CasePart -> case Expression OpenScope ColonBlock ${
4521 $0 = calloc(1,sizeof(struct casepart));
4524 var_block_close(c, CloseParallel, $0->action);
4528 // These scopes are closed in CondStatement
4529 ForPart -> for OpenBlock ${
4533 ThenPart -> then OpenBlock ${
4535 var_block_close(c, CloseSequential, $0);
4539 // This scope is closed in CondStatement
4540 WhilePart -> while UseBlock OptNL do OpenBlock ${
4545 var_block_close(c, CloseSequential, $0->right);
4546 var_block_close(c, CloseSequential, $0);
4548 | while OpenScope Expression OpenScope ColonBlock ${
4553 var_block_close(c, CloseSequential, $0->right);
4554 var_block_close(c, CloseSequential, $0);
4558 IfPart -> if UseBlock OptNL then OpenBlock ${
4561 var_block_close(c, CloseParallel, $0.thenpart);
4563 | if OpenScope Expression OpenScope ColonBlock ${
4566 var_block_close(c, CloseParallel, $0.thenpart);
4568 | if OpenScope Expression OpenScope OptNL then Block ${
4571 var_block_close(c, CloseParallel, $0.thenpart);
4575 // This scope is closed in CondStatement
4576 SwitchPart -> switch OpenScope Expression ${
4579 | switch UseBlock ${
4583 ###### print binode cases
4585 if (b->left && b->left->type == Xbinode &&
4586 cast(binode, b->left)->op == Block) {
4588 do_indent(indent, "while {\n");
4590 do_indent(indent, "while\n");
4591 print_exec(b->left, indent+1, bracket);
4593 do_indent(indent, "} do {\n");
4595 do_indent(indent, "do\n");
4596 print_exec(b->right, indent+1, bracket);
4598 do_indent(indent, "}\n");
4600 do_indent(indent, "while ");
4601 print_exec(b->left, 0, bracket);
4606 print_exec(b->right, indent+1, bracket);
4608 do_indent(indent, "}\n");
4612 ###### print exec cases
4614 case Xcond_statement:
4616 struct cond_statement *cs = cast(cond_statement, e);
4617 struct casepart *cp;
4619 do_indent(indent, "for");
4620 if (bracket) printf(" {\n"); else printf("\n");
4621 print_exec(cs->forpart, indent+1, bracket);
4624 do_indent(indent, "} then {\n");
4626 do_indent(indent, "then\n");
4627 print_exec(cs->thenpart, indent+1, bracket);
4629 if (bracket) do_indent(indent, "}\n");
4632 print_exec(cs->looppart, indent, bracket);
4636 do_indent(indent, "switch");
4638 do_indent(indent, "if");
4639 if (cs->condpart && cs->condpart->type == Xbinode &&
4640 cast(binode, cs->condpart)->op == Block) {
4645 print_exec(cs->condpart, indent+1, bracket);
4647 do_indent(indent, "}\n");
4649 do_indent(indent, "then\n");
4650 print_exec(cs->thenpart, indent+1, bracket);
4654 print_exec(cs->condpart, 0, bracket);
4660 print_exec(cs->thenpart, indent+1, bracket);
4662 do_indent(indent, "}\n");
4667 for (cp = cs->casepart; cp; cp = cp->next) {
4668 do_indent(indent, "case ");
4669 print_exec(cp->value, -1, 0);
4674 print_exec(cp->action, indent+1, bracket);
4676 do_indent(indent, "}\n");
4679 do_indent(indent, "else");
4684 print_exec(cs->elsepart, indent+1, bracket);
4686 do_indent(indent, "}\n");
4691 ###### propagate binode cases
4693 t = propagate_types(b->right, c, ok, Tnone, 0);
4694 if (!type_compat(Tnone, t, 0))
4695 *ok = 0; // UNTESTED
4696 return propagate_types(b->left, c, ok, type, rules);
4698 ###### propagate exec cases
4699 case Xcond_statement:
4701 // forpart and looppart->right must return Tnone
4702 // thenpart must return Tnone if there is a loopart,
4703 // otherwise it is like elsepart.
4705 // be bool if there is no casepart
4706 // match casepart->values if there is a switchpart
4707 // either be bool or match casepart->value if there
4709 // elsepart and casepart->action must match the return type
4710 // expected of this statement.
4711 struct cond_statement *cs = cast(cond_statement, prog);
4712 struct casepart *cp;
4714 t = propagate_types(cs->forpart, c, ok, Tnone, 0);
4715 if (!type_compat(Tnone, t, 0))
4716 *ok = 0; // UNTESTED
4719 t = propagate_types(cs->thenpart, c, ok, Tnone, 0);
4720 if (!type_compat(Tnone, t, 0))
4721 *ok = 0; // UNTESTED
4723 if (cs->casepart == NULL) {
4724 propagate_types(cs->condpart, c, ok, Tbool, 0);
4725 propagate_types(cs->looppart, c, ok, Tbool, 0);
4727 /* Condpart must match case values, with bool permitted */
4729 for (cp = cs->casepart;
4730 cp && !t; cp = cp->next)
4731 t = propagate_types(cp->value, c, ok, NULL, 0);
4732 if (!t && cs->condpart)
4733 t = propagate_types(cs->condpart, c, ok, NULL, Rboolok); // UNTESTED
4734 if (!t && cs->looppart)
4735 t = propagate_types(cs->looppart, c, ok, NULL, Rboolok); // UNTESTED
4736 // Now we have a type (I hope) push it down
4738 for (cp = cs->casepart; cp; cp = cp->next)
4739 propagate_types(cp->value, c, ok, t, 0);
4740 propagate_types(cs->condpart, c, ok, t, Rboolok);
4741 propagate_types(cs->looppart, c, ok, t, Rboolok);
4744 // (if)then, else, and case parts must return expected type.
4745 if (!cs->looppart && !type)
4746 type = propagate_types(cs->thenpart, c, ok, NULL, rules);
4748 type = propagate_types(cs->elsepart, c, ok, NULL, rules);
4749 for (cp = cs->casepart;
4751 cp = cp->next) // UNTESTED
4752 type = propagate_types(cp->action, c, ok, NULL, rules); // UNTESTED
4755 propagate_types(cs->thenpart, c, ok, type, rules);
4756 propagate_types(cs->elsepart, c, ok, type, rules);
4757 for (cp = cs->casepart; cp ; cp = cp->next)
4758 propagate_types(cp->action, c, ok, type, rules);
4764 ###### interp binode cases
4766 // This just performs one iterration of the loop
4767 rv = interp_exec(c, b->left, &rvtype);
4768 if (rvtype == Tnone ||
4769 (rvtype == Tbool && rv.bool != 0))
4770 // rvtype is Tnone or Tbool, doesn't need to be freed
4771 interp_exec(c, b->right, NULL);
4774 ###### interp exec cases
4775 case Xcond_statement:
4777 struct value v, cnd;
4778 struct type *vtype, *cndtype;
4779 struct casepart *cp;
4780 struct cond_statement *cs = cast(cond_statement, e);
4783 interp_exec(c, cs->forpart, NULL);
4785 while ((cnd = interp_exec(c, cs->looppart, &cndtype)),
4786 cndtype == Tnone || (cndtype == Tbool && cnd.bool != 0))
4787 interp_exec(c, cs->thenpart, NULL);
4789 cnd = interp_exec(c, cs->condpart, &cndtype);
4790 if ((cndtype == Tnone ||
4791 (cndtype == Tbool && cnd.bool != 0))) {
4792 // cnd is Tnone or Tbool, doesn't need to be freed
4793 rv = interp_exec(c, cs->thenpart, &rvtype);
4794 // skip else (and cases)
4798 for (cp = cs->casepart; cp; cp = cp->next) {
4799 v = interp_exec(c, cp->value, &vtype);
4800 if (value_cmp(cndtype, vtype, &v, &cnd) == 0) {
4801 free_value(vtype, &v);
4802 free_value(cndtype, &cnd);
4803 rv = interp_exec(c, cp->action, &rvtype);
4806 free_value(vtype, &v);
4808 free_value(cndtype, &cnd);
4810 rv = interp_exec(c, cs->elsepart, &rvtype);
4817 ### Top level structure
4819 All the language elements so far can be used in various places. Now
4820 it is time to clarify what those places are.
4822 At the top level of a file there will be a number of declarations.
4823 Many of the things that can be declared haven't been described yet,
4824 such as functions, procedures, imports, and probably more.
4825 For now there are two sorts of things that can appear at the top
4826 level. They are predefined constants, `struct` types, and the `main`
4827 function. While the syntax will allow the `main` function to appear
4828 multiple times, that will trigger an error if it is actually attempted.
4830 The various declarations do not return anything. They store the
4831 various declarations in the parse context.
4833 ###### Parser: grammar
4836 Ocean -> OptNL DeclarationList
4838 ## declare terminals
4846 DeclarationList -> Declaration
4847 | DeclarationList Declaration
4849 Declaration -> ERROR Newlines ${
4850 tok_err(c, // UNTESTED
4851 "error: unhandled parse error", &$1);
4857 ## top level grammar
4861 ### The `const` section
4863 As well as being defined in with the code that uses them, constants can
4864 be declared at the top level. These have full-file scope, so they are
4865 always `InScope`, even before(!) they have been declared. The value of
4866 a top level constant can be given as an expression, and this is
4867 evaluated after parsing and before execution.
4869 A function call can be used to evaluate a constant, but it will not have
4870 access to any program state, once such statement becomes meaningful.
4871 e.g. arguments and filesystem will not be visible.
4873 Constants are defined in a section that starts with the reserved word
4874 `const` and then has a block with a list of assignment statements.
4875 For syntactic consistency, these must use the double-colon syntax to
4876 make it clear that they are constants. Type can also be given: if
4877 not, the type will be determined during analysis, as with other
4880 ###### parse context
4881 struct binode *constlist;
4883 ###### top level grammar
4887 DeclareConstant -> const { IN OptNL ConstList OUT OptNL } Newlines
4888 | const { SimpleConstList } Newlines
4889 | const IN OptNL ConstList OUT Newlines
4890 | const SimpleConstList Newlines
4892 ConstList -> ConstList SimpleConstLine
4895 SimpleConstList -> SimpleConstList ; Const
4899 SimpleConstLine -> SimpleConstList Newlines
4900 | ERROR Newlines ${ tok_err(c, "Syntax error in constant", &$1); }$
4903 CType -> Type ${ $0 = $<1; }$
4907 Const -> IDENTIFIER :: CType = Expression ${ {
4909 struct binode *bl, *bv;
4910 struct var *var = new_pos(var, $ID);
4912 v = var_decl(c, $ID.txt);
4914 v->where_decl = var;
4920 v = var_ref(c, $1.txt);
4921 tok_err(c, "error: name already declared", &$1);
4922 type_err(c, "info: this is where '%v' was first declared",
4923 v->where_decl, NULL, 0, NULL);
4934 bl->left = c->constlist;
4939 ###### core functions
4940 static void resolve_consts(struct parse_context *c)
4943 c->constlist = reorder_bilist(c->constlist);
4944 for (b = cast(binode, c->constlist); b;
4945 b = cast(binode, b->right)) {
4947 struct binode *vb = cast(binode, b->left);
4948 struct var *v = cast(var, vb->left);
4951 propagate_types(vb->right, c, &ok,
4957 struct value res = interp_exec(
4958 c, vb->right, &v->var->type);
4959 global_alloc(c, v->var->type, v->var, &res);
4964 ###### print const decls
4969 for (b = cast(binode, context.constlist); b;
4970 b = cast(binode, b->right)) {
4971 struct binode *vb = cast(binode, b->left);
4972 struct var *vr = cast(var, vb->left);
4973 struct variable *v = vr->var;
4979 printf(" %.*s :: ", v->name->name.len, v->name->name.txt);
4980 type_print(v->type, stdout);
4982 print_exec(vb->right, -1, 0);
4987 ###### free const decls
4988 free_binode(context.constlist);
4990 ### Function declarations
4992 The code in an Ocean program is all stored in function declarations.
4993 One of the functions must be named `main` and it must accept an array of
4994 strings as a parameter - the command line arguments.
4996 As this is the top level, several things are handled a bit differently.
4997 The function is not interpreted by `interp_exec` as that isn't passed
4998 the argument list which the program requires. Similarly type analysis
4999 is a bit more interesting at this level.
5001 ###### ast functions
5003 static struct type *handle_results(struct parse_context *c,
5004 struct binode *results)
5006 /* Create a 'struct' type from the results list, which
5007 * is a list for 'struct var'
5009 struct type *t = add_anon_type(c, &structure_prototype,
5010 " function result");
5014 for (b = results; b; b = cast(binode, b->right))
5016 t->structure.nfields = cnt;
5017 t->structure.fields = calloc(cnt, sizeof(struct field));
5019 for (b = results; b; b = cast(binode, b->right)) {
5020 struct var *v = cast(var, b->left);
5021 struct field *f = &t->structure.fields[cnt++];
5022 int a = v->var->type->align;
5023 f->name = v->var->name->name;
5024 f->type = v->var->type;
5026 f->offset = t->size;
5027 v->var->frame_pos = f->offset;
5028 t->size += ((f->type->size - 1) | (a-1)) + 1;
5031 variable_unlink_exec(v->var);
5033 free_binode(results);
5037 static struct variable *declare_function(struct parse_context *c,
5038 struct variable *name,
5039 struct binode *args,
5041 struct binode *results,
5045 struct value fn = {.function = code};
5047 var_block_close(c, CloseFunction, code);
5048 t = add_anon_type(c, &function_prototype,
5049 "func %.*s", name->name->name.len,
5050 name->name->name.txt);
5052 t->function.params = reorder_bilist(args);
5054 ret = handle_results(c, reorder_bilist(results));
5055 t->function.inline_result = 1;
5056 t->function.local_size = ret->size;
5058 t->function.return_type = ret;
5059 global_alloc(c, t, name, &fn);
5060 name->type->function.scope = c->out_scope;
5065 var_block_close(c, CloseFunction, NULL);
5067 c->out_scope = NULL;
5071 ###### declare terminals
5074 ###### top level grammar
5077 DeclareFunction -> func FuncName ( OpenScope ArgsLine ) Block Newlines ${
5078 $0 = declare_function(c, $<FN, $<Ar, Tnone, NULL, $<Bl);
5080 | func FuncName IN OpenScope Args OUT OptNL do Block Newlines ${
5081 $0 = declare_function(c, $<FN, $<Ar, Tnone, NULL, $<Bl);
5083 | func FuncName NEWLINE OpenScope OptNL do Block Newlines ${
5084 $0 = declare_function(c, $<FN, NULL, Tnone, NULL, $<Bl);
5086 | func FuncName ( OpenScope ArgsLine ) : Type Block Newlines ${
5087 $0 = declare_function(c, $<FN, $<Ar, $<Ty, NULL, $<Bl);
5089 | func FuncName ( OpenScope ArgsLine ) : ( ArgsLine ) Block Newlines ${
5090 $0 = declare_function(c, $<FN, $<AL, NULL, $<AL2, $<Bl);
5092 | func FuncName IN OpenScope Args OUT OptNL return Type Newlines do Block Newlines ${
5093 $0 = declare_function(c, $<FN, $<Ar, $<Ty, NULL, $<Bl);
5095 | func FuncName NEWLINE OpenScope return Type Newlines do Block Newlines ${
5096 $0 = declare_function(c, $<FN, NULL, $<Ty, NULL, $<Bl);
5098 | func FuncName IN OpenScope Args OUT OptNL return IN Args OUT OptNL do Block Newlines ${
5099 $0 = declare_function(c, $<FN, $<Ar, NULL, $<Ar2, $<Bl);
5101 | func FuncName NEWLINE OpenScope return IN Args OUT OptNL do Block Newlines ${
5102 $0 = declare_function(c, $<FN, NULL, NULL, $<Ar, $<Bl);
5105 ###### print func decls
5110 while (target != 0) {
5112 for (v = context.in_scope; v; v=v->in_scope)
5113 if (v->depth == 0 && v->type && v->type->check_args) {
5122 struct value *val = var_value(&context, v);
5123 printf("func %.*s", v->name->name.len, v->name->name.txt);
5124 v->type->print_type_decl(v->type, stdout);
5126 print_exec(val->function, 0, brackets);
5128 print_value(v->type, val, stdout);
5129 printf("/* frame size %d */\n", v->type->function.local_size);
5135 ###### core functions
5137 static int analyse_funcs(struct parse_context *c)
5141 for (v = c->in_scope; v; v = v->in_scope) {
5145 if (v->depth != 0 || !v->type || !v->type->check_args)
5147 ret = v->type->function.inline_result ?
5148 Tnone : v->type->function.return_type;
5149 val = var_value(c, v);
5152 propagate_types(val->function, c, &ok, ret, 0);
5155 /* Make sure everything is still consistent */
5156 propagate_types(val->function, c, &ok, ret, 0);
5159 if (!v->type->function.inline_result &&
5160 !v->type->function.return_type->dup) {
5161 type_err(c, "error: function cannot return value of type %1",
5162 v->where_decl, v->type->function.return_type, 0, NULL);
5165 scope_finalize(c, v->type);
5170 static int analyse_main(struct type *type, struct parse_context *c)
5172 struct binode *bp = type->function.params;
5176 struct type *argv_type;
5178 argv_type = add_anon_type(c, &array_prototype, "argv");
5179 argv_type->array.member = Tstr;
5180 argv_type->array.unspec = 1;
5182 for (b = bp; b; b = cast(binode, b->right)) {
5186 propagate_types(b->left, c, &ok, argv_type, 0);
5188 default: /* invalid */ // NOTEST
5189 propagate_types(b->left, c, &ok, Tnone, 0); // NOTEST
5195 return !c->parse_error;
5198 static void interp_main(struct parse_context *c, int argc, char **argv)
5200 struct value *progp = NULL;
5201 struct text main_name = { "main", 4 };
5202 struct variable *mainv;
5208 mainv = var_ref(c, main_name);
5210 progp = var_value(c, mainv);
5211 if (!progp || !progp->function) {
5212 fprintf(stderr, "oceani: no main function found.\n");
5216 if (!analyse_main(mainv->type, c)) {
5217 fprintf(stderr, "oceani: main has wrong type.\n");
5221 al = mainv->type->function.params;
5223 c->local_size = mainv->type->function.local_size;
5224 c->local = calloc(1, c->local_size);
5226 struct var *v = cast(var, al->left);
5227 struct value *vl = var_value(c, v->var);
5237 mpq_set_ui(argcq, argc, 1);
5238 memcpy(var_value(c, t->array.vsize), &argcq, sizeof(argcq));
5239 t->prepare_type(c, t, 0);
5240 array_init(v->var->type, vl);
5241 for (i = 0; i < argc; i++) {
5242 struct value *vl2 = vl->array + i * v->var->type->array.member->size;
5244 arg.str.txt = argv[i];
5245 arg.str.len = strlen(argv[i]);
5246 free_value(Tstr, vl2);
5247 dup_value(Tstr, &arg, vl2);
5251 al = cast(binode, al->right);
5253 v = interp_exec(c, progp->function, &vtype);
5254 free_value(vtype, &v);
5259 ###### ast functions
5260 void free_variable(struct variable *v)
5264 ## And now to test it out.
5266 Having a language requires having a "hello world" program. I'll
5267 provide a little more than that: a program that prints "Hello world"
5268 finds the GCD of two numbers, prints the first few elements of
5269 Fibonacci, performs a binary search for a number, and a few other
5270 things which will likely grow as the languages grows.
5272 ###### File: oceani.mk
5275 @echo "===== DEMO ====="
5276 ./oceani --section "demo: hello" oceani.mdc 55 33
5282 four ::= 2 + 2 ; five ::= 10/2
5283 const pie ::= "I like Pie";
5284 cake ::= "The cake is"
5292 func main(argv:[argc::]string)
5293 print "Hello World, what lovely oceans you have!"
5294 print "Are there", five, "?"
5295 print pi, pie, "but", cake
5297 A := $argv[1]; B := $argv[2]
5299 /* When a variable is defined in both branches of an 'if',
5300 * and used afterwards, the variables are merged.
5306 print "Is", A, "bigger than", B,"? ", bigger
5307 /* If a variable is not used after the 'if', no
5308 * merge happens, so types can be different
5311 double:string = "yes"
5312 print A, "is more than twice", B, "?", double
5315 print "double", B, "is", double
5320 if a > 0 and then b > 0:
5326 print "GCD of", A, "and", B,"is", a
5328 print a, "is not positive, cannot calculate GCD"
5330 print b, "is not positive, cannot calculate GCD"
5335 print "Fibonacci:", f1,f2,
5336 then togo = togo - 1
5344 /* Binary search... */
5349 mid := (lo + hi) / 2
5362 print "Yay, I found", target
5364 print "Closest I found was", lo
5369 // "middle square" PRNG. Not particularly good, but one my
5370 // Dad taught me - the first one I ever heard of.
5371 for i:=1; then i = i + 1; while i < size:
5372 n := list[i-1] * list[i-1]
5373 list[i] = (n / 100) % 10 000
5375 print "Before sort:",
5376 for i:=0; then i = i + 1; while i < size:
5380 for i := 1; then i=i+1; while i < size:
5381 for j:=i-1; then j=j-1; while j >= 0:
5382 if list[j] > list[j+1]:
5386 print " After sort:",
5387 for i:=0; then i = i + 1; while i < size:
5391 if 1 == 2 then print "yes"; else print "no"
5395 bob.alive = (bob.name == "Hello")
5396 print "bob", "is" if bob.alive else "isn't", "alive"