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 (which have since been remove), and the
41 "if ... else" trinary operator which can select between two expressions
42 based on a third (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;
114 struct parse_context {
115 struct token_config config;
123 #define container_of(ptr, type, member) ({ \
124 const typeof( ((type *)0)->member ) *__mptr = (ptr); \
125 (type *)( (char *)__mptr - offsetof(type,member) );})
127 #define config2context(_conf) container_of(_conf, struct parse_context, \
130 ###### Parser: reduce
131 struct parse_context *c = config2context(config);
139 #include <sys/mman.h>
158 static char Usage[] =
159 "Usage: oceani --trace --print --noexec --brackets --section=SectionName prog.ocn\n";
160 static const struct option long_options[] = {
161 {"trace", 0, NULL, 't'},
162 {"print", 0, NULL, 'p'},
163 {"noexec", 0, NULL, 'n'},
164 {"brackets", 0, NULL, 'b'},
165 {"section", 1, NULL, 's'},
168 const char *options = "tpnbs";
170 static void pr_err(char *msg) // NOTEST
172 fprintf(stderr, "%s\n", msg); // NOTEST
175 int main(int argc, char *argv[])
180 struct section *s = NULL, *ss;
181 char *section = NULL;
182 struct parse_context context = {
184 .ignored = (1 << TK_mark),
185 .number_chars = ".,_+- ",
190 int doprint=0, dotrace=0, doexec=1, brackets=0;
192 while ((opt = getopt_long(argc, argv, options, long_options, NULL))
195 case 't': dotrace=1; break;
196 case 'p': doprint=1; break;
197 case 'n': doexec=0; break;
198 case 'b': brackets=1; break;
199 case 's': section = optarg; break;
200 default: fprintf(stderr, Usage);
204 if (optind >= argc) {
205 fprintf(stderr, "oceani: no input file given\n");
208 fd = open(argv[optind], O_RDONLY);
210 fprintf(stderr, "oceani: cannot open %s\n", argv[optind]);
213 context.file_name = argv[optind];
214 len = lseek(fd, 0, 2);
215 file = mmap(NULL, len, PROT_READ, MAP_SHARED, fd, 0);
216 s = code_extract(file, file+len, pr_err);
218 fprintf(stderr, "oceani: could not find any code in %s\n",
223 ## context initialization
226 for (ss = s; ss; ss = ss->next) {
227 struct text sec = ss->section;
228 if (sec.len == strlen(section) &&
229 strncmp(sec.txt, section, sec.len) == 0)
233 fprintf(stderr, "oceani: cannot find section %s\n",
240 fprintf(stderr, "oceani: no code found in requested section\n"); // NOTEST
241 goto cleanup; // NOTEST
244 parse_oceani(ss->code, &context.config, dotrace ? stderr : NULL);
246 resolve_consts(&context);
247 prepare_types(&context);
248 if (!context.parse_error && !analyse_funcs(&context)) {
249 fprintf(stderr, "oceani: type error in program - not running.\n");
250 context.parse_error += 1;
258 if (doexec && !context.parse_error)
259 interp_main(&context, argc - optind, argv + optind);
262 struct section *t = s->next;
267 // FIXME parser should pop scope even on error
268 while (context.scope_depth > 0)
272 ## free context types
273 ## free context storage
274 exit(context.parse_error ? 1 : 0);
279 The four requirements of parse, analyse, print, interpret apply to
280 each language element individually so that is how most of the code
283 Three of the four are fairly self explanatory. The one that requires
284 a little explanation is the analysis step.
286 The current language design does not require the types of variables to
287 be declared, but they must still have a single type. Different
288 operations impose different requirements on the variables, for example
289 addition requires both arguments to be numeric, and assignment
290 requires the variable on the left to have the same type as the
291 expression on the right.
293 Analysis involves propagating these type requirements around and
294 consequently setting the type of each variable. If any requirements
295 are violated (e.g. a string is compared with a number) or if a
296 variable needs to have two different types, then an error is raised
297 and the program will not run.
299 If the same variable is declared in both branchs of an 'if/else', or
300 in all cases of a 'switch' then the multiple instances may be merged
301 into just one variable if the variable is referenced after the
302 conditional statement. When this happens, the types must naturally be
303 consistent across all the branches. When the variable is not used
304 outside the if, the variables in the different branches are distinct
305 and can be of different types.
307 Undeclared names may only appear in "use" statements and "case" expressions.
308 These names are given a type of "label" and a unique value.
309 This allows them to fill the role of a name in an enumerated type, which
310 is useful for testing the `switch` statement.
312 As we will see, the condition part of a `while` statement can return
313 either a Boolean or some other type. This requires that the expected
314 type that gets passed around comprises a type and a flag to indicate
315 that `Tbool` is also permitted.
317 As there are, as yet, no distinct types that are compatible, there
318 isn't much subtlety in the analysis. When we have distinct number
319 types, this will become more interesting.
323 When analysis discovers an inconsistency it needs to report an error;
324 just refusing to run the code ensures that the error doesn't cascade,
325 but by itself it isn't very useful. A clear understanding of the sort
326 of error message that are useful will help guide the process of
329 At a simplistic level, the only sort of error that type analysis can
330 report is that the type of some construct doesn't match a contextual
331 requirement. For example, in `4 + "hello"` the addition provides a
332 contextual requirement for numbers, but `"hello"` is not a number. In
333 this particular example no further information is needed as the types
334 are obvious from local information. When a variable is involved that
335 isn't the case. It may be helpful to explain why the variable has a
336 particular type, by indicating the location where the type was set,
337 whether by declaration or usage.
339 Using a recursive-descent analysis we can easily detect a problem at
340 multiple locations. In "`hello:= "there"; 4 + hello`" the addition
341 will detect that one argument is not a number and the usage of `hello`
342 will detect that a number was wanted, but not provided. In this
343 (early) version of the language, we will generate error reports at
344 multiple locations, so the use of `hello` will report an error and
345 explain were the value was set, and the addition will report an error
346 and say why numbers are needed. To be able to report locations for
347 errors, each language element will need to record a file location
348 (line and column) and each variable will need to record the language
349 element where its type was set. For now we will assume that each line
350 of an error message indicates one location in the file, and up to 2
351 types. So we provide a `printf`-like function which takes a format, a
352 location (a `struct exec` which has not yet been introduced), and 2
353 types. "`%1`" reports the first type, "`%2`" reports the second. We
354 will need a function to print the location, once we know how that is
355 stored. e As will be explained later, there are sometimes extra rules for
356 type matching and they might affect error messages, we need to pass those
359 As well as type errors, we sometimes need to report problems with
360 tokens, which might be unexpected or might name a type that has not
361 been defined. For these we have `tok_err()` which reports an error
362 with a given token. Each of the error functions sets the flag in the
363 context so indicate that parsing failed.
367 static void fput_loc(struct exec *loc, FILE *f);
368 static void type_err(struct parse_context *c,
369 char *fmt, struct exec *loc,
370 struct type *t1, enum val_rules rules, struct type *t2);
371 static void tok_err(struct parse_context *c, char *fmt, struct token *t);
373 ###### core functions
375 static void type_err(struct parse_context *c,
376 char *fmt, struct exec *loc,
377 struct type *t1, enum val_rules rules, struct type *t2)
379 fprintf(stderr, "%s:", c->file_name);
380 fput_loc(loc, stderr);
381 for (; *fmt ; fmt++) {
388 case '%': fputc(*fmt, stderr); break; // NOTEST
389 default: fputc('?', stderr); break; // NOTEST
391 type_print(t1, stderr);
394 type_print(t2, stderr);
403 static void tok_err(struct parse_context *c, char *fmt, struct token *t)
405 fprintf(stderr, "%s:%d:%d: %s: %.*s\n", c->file_name, t->line, t->col, fmt,
406 t->txt.len, t->txt.txt);
410 ## Entities: declared and predeclared.
412 There are various "things" that the language and/or the interpreter
413 needs to know about to parse and execute a program. These include
414 types, variables, values, and executable code. These are all lumped
415 together under the term "entities" (calling them "objects" would be
416 confusing) and introduced here. The following section will present the
417 different specific code elements which comprise or manipulate these
422 Executables can be lots of different things. In many cases an
423 executable is just an operation combined with one or two other
424 executables. This allows for expressions and lists etc. Other times an
425 executable is something quite specific like a constant or variable name.
426 So we define a `struct exec` to be a general executable with a type, and
427 a `struct binode` which is a subclass of `exec`, forms a node in a
428 binary tree, and holds an operation. The simplest operation is "List"
429 which can be used to combine several execs together.
431 There will be other subclasses, and to access these we need to be able
432 to `cast` the `exec` into the various other types. The first field in
433 any `struct exec` is the type from the `exec_types` enum.
436 #define cast(structname, pointer) ({ \
437 const typeof( ((struct structname *)0)->type) *__mptr = &(pointer)->type; \
438 if (__mptr && *__mptr != X##structname) abort(); \
439 (struct structname *)( (char *)__mptr);})
441 #define new(structname) ({ \
442 struct structname *__ptr = ((struct structname *)calloc(1,sizeof(struct structname))); \
443 __ptr->type = X##structname; \
444 __ptr->line = -1; __ptr->column = -1; \
447 #define new_pos(structname, token) ({ \
448 struct structname *__ptr = ((struct structname *)calloc(1,sizeof(struct structname))); \
449 __ptr->type = X##structname; \
450 __ptr->line = token.line; __ptr->column = token.col; \
459 enum exec_types type;
469 struct exec *left, *right;
474 static int __fput_loc(struct exec *loc, FILE *f)
478 if (loc->line >= 0) {
479 fprintf(f, "%d:%d: ", loc->line, loc->column);
482 if (loc->type == Xbinode)
483 return __fput_loc(cast(binode,loc)->left, f) ||
484 __fput_loc(cast(binode,loc)->right, f); // NOTEST
487 static void fput_loc(struct exec *loc, FILE *f)
489 if (!__fput_loc(loc, f))
490 fprintf(f, "??:??: "); // NOTEST
493 Each different type of `exec` node needs a number of functions defined,
494 a bit like methods. We must be able to free it, print it, analyse it
495 and execute it. Once we have specific `exec` types we will need to
496 parse them too. Let's take this a bit more slowly.
500 The parser generator requires a `free_foo` function for each struct
501 that stores attributes and they will often be `exec`s and subtypes
502 there-of. So we need `free_exec` which can handle all the subtypes,
503 and we need `free_binode`.
507 static void free_binode(struct binode *b)
516 ###### core functions
517 static void free_exec(struct exec *e)
528 static void free_exec(struct exec *e);
530 ###### free exec cases
531 case Xbinode: free_binode(cast(binode, e)); break;
535 Printing an `exec` requires that we know the current indent level for
536 printing line-oriented components. As will become clear later, we
537 also want to know what sort of bracketing to use.
541 static void do_indent(int i, char *str)
548 ###### core functions
549 static void print_binode(struct binode *b, int indent, int bracket)
553 case List: abort(); // must be handled by parent NOTEST
554 ## print binode cases
558 static void print_exec(struct exec *e, int indent, int bracket)
564 print_binode(cast(binode, e), indent, bracket); break;
569 do_indent(indent, "/* FREE");
570 for (v = e->to_free; v; v = v->next_free) {
571 printf(" %.*s", v->name->name.len, v->name->name.txt);
572 printf("[%d,%d]", v->scope_start, v->scope_end);
573 if (v->frame_pos >= 0)
574 printf("(%d+%d)", v->frame_pos,
575 v->type ? v->type->size:0);
583 static void print_exec(struct exec *e, int indent, int bracket);
587 As discussed, analysis involves propagating type requirements around the
588 program and looking for errors.
590 So `propagate_types` is passed an expected type (being a `struct type`
591 pointer together with some `val_rules` flags) that the `exec` is
592 expected to return, and returns the type that it does return, either of
593 which can be `NULL` signifying "unknown". A `prop_err` flag set is
594 passed by reference. It has `Efail` set when an error is found, and
595 `Eretry` when the type for some element is set via propagation. If
596 any expression cannot be evaluated a compile time, `Eruntime` is set.
597 If the expression can be copied, `Emaycopy` is set.
599 If `Erval` is set, then the value cannot be assigned to because it is
600 a temporary result. If `Erval` is clear but `Econst` is set, then
601 the value can only be assigned once, when the variable is declared.
605 enum val_rules {Rboolok = 1<<0, Rrefok = 1<<1,};
606 enum prop_err {Efail = 1<<0, Eretry = 1<<1, Eruntime = 1<<2,
607 Emaycopy = 1<<3, Erval = 1<<4, Econst = 1<<5};
610 static struct type *propagate_types(struct exec *prog, struct parse_context *c, enum prop_err *perr,
611 struct type *type, enum val_rules rules);
612 ###### core functions
614 static struct type *__propagate_types(struct exec *prog, struct parse_context *c, enum prop_err *perr,
615 enum prop_err *perr_local,
616 struct type *type, enum val_rules rules)
623 switch (prog->type) {
626 struct binode *b = cast(binode, prog);
628 case List: abort(); // NOTEST
629 ## propagate binode cases
633 ## propagate exec cases
638 static struct type *propagate_types(struct exec *prog, struct parse_context *c, enum prop_err *perr,
639 struct type *type, enum val_rules rules)
641 int pre_err = c->parse_error;
642 enum prop_err perr_local = 0;
643 struct type *ret = __propagate_types(prog, c, perr, &perr_local, type, rules);
645 *perr |= perr_local & (Efail | Eretry);
646 if (c->parse_error > pre_err)
653 Interpreting an `exec` doesn't require anything but the `exec`. State
654 is stored in variables and each variable will be directly linked from
655 within the `exec` tree. The exception to this is the `main` function
656 which needs to look at command line arguments. This function will be
657 interpreted separately.
659 Each `exec` can return a value combined with a type in `struct lrval`.
660 The type may be `Tnone` but must be non-NULL. Some `exec`s will return
661 the location of a value, which can be updated, in `lval`. Others will
662 set `lval` to NULL indicating that there is a value of appropriate type
666 static struct value interp_exec(struct parse_context *c, struct exec *e,
667 struct type **typeret);
668 ###### core functions
672 struct value rval, *lval;
675 /* If dest is passed, dtype must give the expected type, and
676 * result can go there, in which case type is returned as NULL.
678 static struct lrval _interp_exec(struct parse_context *c, struct exec *e,
679 struct value *dest, struct type *dtype);
681 static struct value interp_exec(struct parse_context *c, struct exec *e,
682 struct type **typeret)
684 struct lrval ret = _interp_exec(c, e, NULL, NULL);
686 if (!ret.type) abort();
690 dup_value(ret.type, ret.lval, &ret.rval);
694 static struct value *linterp_exec(struct parse_context *c, struct exec *e,
695 struct type **typeret)
697 struct lrval ret = _interp_exec(c, e, NULL, NULL);
699 if (!ret.type) abort();
703 free_value(ret.type, &ret.rval);
707 /* dinterp_exec is used when the destination type is certain and
708 * the value has a place to go.
710 static void dinterp_exec(struct parse_context *c, struct exec *e,
711 struct value *dest, struct type *dtype,
714 struct lrval ret = _interp_exec(c, e, dest, dtype);
718 free_value(dtype, dest);
720 dup_value(dtype, ret.lval, dest);
722 memcpy(dest, &ret.rval, dtype->size);
725 static struct lrval _interp_exec(struct parse_context *c, struct exec *e,
726 struct value *dest, struct type *dtype)
728 /* If the result is copied to dest, ret.type is set to NULL */
730 struct value rv = {}, *lrv = NULL;
733 rvtype = ret.type = Tnone;
743 struct binode *b = cast(binode, e);
744 struct value left, right, *lleft;
745 struct type *ltype, *rtype;
746 ltype = rtype = Tnone;
748 case List: abort(); // NOTEST
749 ## interp binode cases
751 free_value(ltype, &left);
752 free_value(rtype, &right);
762 ## interp exec cleanup
768 Values come in a wide range of types, with more likely to be added.
769 Each type needs to be able to print its own values (for convenience at
770 least) as well as to compare two values, at least for equality and
771 possibly for order. For now, values might need to be duplicated and
772 freed, though eventually such manipulations will be better integrated
775 Rather than requiring every numeric type to support all numeric
776 operations (add, multiply, etc), we allow types to be able to present
777 as one of a few standard types: integer, float, and fraction. The
778 existence of these conversion functions eventually enable types to
779 determine if they are compatible with other types, though such types
780 have not yet been implemented.
782 Named type are stored in a simple linked list. Objects of each type are
783 "values" which are often passed around by value.
785 There are both explicitly named types, and anonymous types. Anonymous
786 cannot be accessed by name, but are used internally and have a name
787 which might be reported in error messages.
794 ## value union fields
802 struct token first_use;
805 void (*init)(struct type *type, struct value *val);
806 int (*prepare_type)(struct parse_context *c, struct type *type, int parse_time);
807 void (*print)(struct type *type, struct value *val, FILE *f);
808 void (*print_type)(struct type *type, FILE *f);
809 int (*cmp_order)(struct type *t1, struct type *t2,
810 struct value *v1, struct value *v2);
811 int (*cmp_eq)(struct type *t1, struct type *t2,
812 struct value *v1, struct value *v2);
813 void (*dup)(struct type *type, struct value *vold, struct value *vnew);
814 int (*test)(struct type *type, struct value *val);
815 void (*free)(struct type *type, struct value *val);
816 void (*free_type)(struct type *t);
817 long long (*to_int)(struct value *v);
818 double (*to_float)(struct value *v);
819 int (*to_mpq)(mpq_t *q, struct value *v);
828 struct type *typelist;
835 static struct type *find_type(struct parse_context *c, struct text s)
837 struct type *t = c->typelist;
839 while (t && (t->anon ||
840 text_cmp(t->name, s) != 0))
845 static struct type *_add_type(struct parse_context *c, struct text s,
846 struct type *proto, int anon)
850 n = calloc(1, sizeof(*n));
857 n->next = c->typelist;
862 static struct type *add_type(struct parse_context *c, struct text s,
865 return _add_type(c, s, proto, 0);
868 static struct type *add_anon_type(struct parse_context *c,
869 struct type *proto, char *name, ...)
875 vasprintf(&t.txt, name, ap);
877 t.len = strlen(t.txt);
878 return _add_type(c, t, proto, 1);
881 static struct type *find_anon_type(struct parse_context *c,
882 struct type *proto, char *name, ...)
884 struct type *t = c->typelist;
889 vasprintf(&nm.txt, name, ap);
891 nm.len = strlen(name);
893 while (t && (!t->anon ||
894 text_cmp(t->name, nm) != 0))
900 return _add_type(c, nm, proto, 1);
903 static void free_type(struct type *t)
905 /* The type is always a reference to something in the
906 * context, so we don't need to free anything.
910 static void free_value(struct type *type, struct value *v)
914 memset(v, 0x5a, type->size);
918 static void type_print(struct type *type, FILE *f)
921 fputs("*unknown*type*", f); // NOTEST
922 else if (type->name.len && !type->anon)
923 fprintf(f, "%.*s", type->name.len, type->name.txt);
924 else if (type->print_type)
925 type->print_type(type, f);
926 else if (type->name.len && type->anon)
927 fprintf(f, "\"%.*s\"", type->name.len, type->name.txt);
929 fputs("*invalid*type*", f); // NOTEST
932 static void val_init(struct type *type, struct value *val)
934 if (type && type->init)
935 type->init(type, val);
938 static void dup_value(struct type *type,
939 struct value *vold, struct value *vnew)
941 if (type && type->dup)
942 type->dup(type, vold, vnew);
945 static int value_cmp(struct type *tl, struct type *tr,
946 struct value *left, struct value *right)
948 if (tl && tl->cmp_order)
949 return tl->cmp_order(tl, tr, left, right);
950 if (tl && tl->cmp_eq)
951 return tl->cmp_eq(tl, tr, left, right);
955 static void print_value(struct type *type, struct value *v, FILE *f)
957 if (type && type->print)
958 type->print(type, v, f);
960 fprintf(f, "*Unknown*"); // NOTEST
963 static void prepare_types(struct parse_context *c)
967 enum { none, some, cannot } progress = none;
972 for (t = c->typelist; t; t = t->next) {
974 tok_err(c, "error: type used but not declared",
976 if (t->size == 0 && t->prepare_type) {
977 if (t->prepare_type(c, t, 1))
979 else if (progress == cannot)
980 tok_err(c, "error: type has recursive definition",
990 progress = cannot; break;
992 progress = none; break;
999 static void free_value(struct type *type, struct value *v);
1000 static int type_compat(struct type *require, struct type *have, enum val_rules rules);
1001 static void type_print(struct type *type, FILE *f);
1002 static void val_init(struct type *type, struct value *v);
1003 static void dup_value(struct type *type,
1004 struct value *vold, struct value *vnew);
1005 static int value_cmp(struct type *tl, struct type *tr,
1006 struct value *left, struct value *right);
1007 static void print_value(struct type *type, struct value *v, FILE *f);
1009 ###### free context types
1011 while (context.typelist) {
1012 struct type *t = context.typelist;
1014 context.typelist = t->next;
1022 Type can be specified for local variables, for fields in a structure,
1023 for formal parameters to functions, and possibly elsewhere. Different
1024 rules may apply in different contexts. As a minimum, a named type may
1025 always be used. Currently the type of a formal parameter can be
1026 different from types in other contexts, so we have a separate grammar
1032 Type -> IDENTIFIER ${
1033 $0 = find_type(c, $ID.txt);
1035 $0 = add_type(c, $ID.txt, NULL);
1036 $0->first_use = $ID;
1041 FormalType -> Type ${ $0 = $<1; }$
1042 ## formal type grammar
1046 Values of the base types can be numbers, which we represent as
1047 multi-precision fractions, strings, Booleans and labels. When
1048 analysing the program we also need to allow for places where no value
1049 is meaningful (type `Tnone`) and where we don't know what type to
1050 expect yet (type is `NULL`).
1052 Values are never shared, they are always copied when used, and freed
1053 when no longer needed.
1055 When propagating type information around the program, we need to
1056 determine if two types are compatible, where type `NULL` is compatible
1057 with anything. There are two special cases with type compatibility,
1058 both related to the Conditional Statement which will be described
1059 later. In some cases a Boolean can be accepted as well as some other
1060 primary type, and in others any type is acceptable except a label (`Vlabel`).
1061 A separate function encoding these cases will simplify some code later.
1063 ###### type functions
1065 int (*compat)(struct type *this, struct type *other, enum val_rules rules);
1067 ###### ast functions
1069 static int type_compat(struct type *require, struct type *have,
1070 enum val_rules rules)
1072 if ((rules & Rboolok) && have == Tbool)
1074 if (!require || !have)
1077 if (require->compat)
1078 return require->compat(require, have, rules);
1080 return require == have;
1085 #include "parse_string.h"
1086 #include "parse_number.h"
1089 myLDLIBS := libnumber.o libstring.o -lgmp
1090 LDLIBS := $(filter-out $(myLDLIBS),$(LDLIBS)) $(myLDLIBS)
1092 ###### type union fields
1093 enum vtype {Vnone, Vstr, Vnum, Vbool, Vlabel} vtype;
1095 ###### value union fields
1101 ###### ast functions
1102 static void _free_value(struct type *type, struct value *v)
1106 switch (type->vtype) {
1108 case Vstr: free(v->str.txt); break;
1109 case Vnum: mpq_clear(v->num); break;
1115 ###### value functions
1117 static void _val_init(struct type *type, struct value *val)
1119 switch(type->vtype) {
1120 case Vnone: // NOTEST
1123 mpq_init(val->num); break;
1125 val->str.txt = malloc(1);
1132 val->label = 0; // NOTEST
1137 static void _dup_value(struct type *type,
1138 struct value *vold, struct value *vnew)
1140 switch (type->vtype) {
1141 case Vnone: // NOTEST
1144 vnew->label = vold->label; // NOTEST
1147 vnew->bool = vold->bool;
1150 mpq_init(vnew->num);
1151 mpq_set(vnew->num, vold->num);
1154 vnew->str.len = vold->str.len;
1155 vnew->str.txt = malloc(vnew->str.len);
1156 memcpy(vnew->str.txt, vold->str.txt, vnew->str.len);
1161 static int _value_cmp(struct type *tl, struct type *tr,
1162 struct value *left, struct value *right)
1167 switch (tl->vtype) {
1168 case Vlabel: cmp = left->label == right->label ? 0 : 1; break;
1169 case Vnum: cmp = mpq_cmp(left->num, right->num); break;
1170 case Vstr: cmp = text_cmp(left->str, right->str); break;
1171 case Vbool: cmp = left->bool - right->bool; break;
1172 case Vnone: cmp = 0; // NOTEST
1177 static void _print_value(struct type *type, struct value *v, FILE *f)
1179 switch (type->vtype) {
1180 case Vnone: // NOTEST
1181 fprintf(f, "*no-value*"); break; // NOTEST
1182 case Vlabel: // NOTEST
1183 fprintf(f, "*label-%d*", v->label); break; // NOTEST
1185 fprintf(f, "%.*s", v->str.len, v->str.txt); break;
1187 fprintf(f, "%s", v->bool ? "True":"False"); break;
1192 mpf_set_q(fl, v->num);
1193 gmp_fprintf(f, "%.10Fg", fl);
1200 static void _free_value(struct type *type, struct value *v);
1202 static int bool_test(struct type *type, struct value *v)
1207 static struct type base_prototype = {
1209 .print = _print_value,
1210 .cmp_order = _value_cmp,
1211 .cmp_eq = _value_cmp,
1213 .free = _free_value,
1216 static struct type *Tbool, *Tstr, *Tnum, *Tnone, *Tlabel;
1218 ###### ast functions
1219 static struct type *add_base_type(struct parse_context *c, char *n,
1220 enum vtype vt, int size)
1222 struct text txt = { n, strlen(n) };
1225 t = add_type(c, txt, &base_prototype);
1228 t->align = size > sizeof(void*) ? sizeof(void*) : size;
1229 if (t->size & (t->align - 1))
1230 t->size = (t->size | (t->align - 1)) + 1; // NOTEST
1234 ###### context initialization
1236 Tbool = add_base_type(&context, "Boolean", Vbool, sizeof(char));
1237 Tbool->test = bool_test;
1238 Tstr = add_base_type(&context, "string", Vstr, sizeof(struct text));
1239 Tnum = add_base_type(&context, "number", Vnum, sizeof(mpq_t));
1240 Tnone = add_base_type(&context, "none", Vnone, 0);
1241 Tlabel = add_base_type(&context, "label", Vlabel, sizeof(void*));
1245 We have already met values as separate objects. When manifest constants
1246 appear in the program text, that must result in an executable which has
1247 a constant value. So the `val` structure embeds a value in an
1260 ###### ast functions
1261 struct val *new_val(struct type *T, struct token tk)
1263 struct val *v = new_pos(val, tk);
1268 ###### declare terminals
1275 $0 = new_val(Tbool, $1);
1279 $0 = new_val(Tbool, $1);
1284 $0 = new_val(Tnum, $1);
1285 if (number_parse($0->val.num, tail, $1.txt) == 0) {
1286 mpq_init($0->val.num);
1287 tok_err(c, "error: unsupported number format", &$NUM);
1289 tok_err(c, "error: unsupported number suffix", &$1);
1293 $0 = new_val(Tstr, $1);
1294 string_parse(&$1, '\\', &$0->val.str, tail);
1296 tok_err(c, "error: unsupported string suffix",
1301 $0 = new_val(Tstr, $1);
1302 string_parse(&$1, '\\', &$0->val.str, tail);
1304 tok_err(c, "error: unsupported string suffix",
1308 ###### print exec cases
1311 struct val *v = cast(val, e);
1312 if (v->vtype == Tstr)
1314 // FIXME how to ensure numbers have same precision.
1315 print_value(v->vtype, &v->val, stdout);
1316 if (v->vtype == Tstr)
1321 ###### propagate exec cases
1324 struct val *val = cast(val, prog);
1325 if (!type_compat(type, val->vtype, rules))
1326 type_err(c, "error: expected %1 found %2",
1327 prog, type, rules, val->vtype);
1332 ###### interp exec cases
1334 rvtype = cast(val, e)->vtype;
1335 dup_value(rvtype, &cast(val, e)->val, &rv);
1338 ###### ast functions
1339 static void free_val(struct val *v)
1342 free_value(v->vtype, &v->val);
1346 ###### free exec cases
1347 case Xval: free_val(cast(val, e)); break;
1349 ###### ast functions
1350 // Move all nodes from 'b' to 'rv', reversing their order.
1351 // In 'b' 'left' is a list, and 'right' is the last node.
1352 // In 'rv', left' is the first node and 'right' is a list.
1353 static struct binode *reorder_bilist(struct binode *b)
1355 struct binode *rv = NULL;
1358 struct exec *t = b->right;
1362 b = cast(binode, b->left);
1372 Labels are a temporary concept until I implement enums. There are an
1373 anonymous enum which is declared by usage. Thet are only allowed in
1374 `use` statements and corresponding `case` entries. They appear as a
1375 period followed by an identifier. All identifiers that are "used" must
1378 For now, we have a global list of labels, and don't check that all "use"
1390 ###### free exec cases
1394 ###### print exec cases
1396 struct label *l = cast(label, e);
1397 printf(".%.*s", l->name.len, l->name.txt);
1403 struct labels *next;
1407 ###### parse context
1408 struct labels *labels;
1410 ###### ast functions
1411 static int label_lookup(struct parse_context *c, struct text name)
1413 struct labels *l, **lp = &c->labels;
1414 while (*lp && text_cmp((*lp)->name, name) < 0)
1416 if (*lp && text_cmp((*lp)->name, name) == 0)
1417 return (*lp)->value;
1418 l = calloc(1, sizeof(*l));
1421 if (c->next_label == 0)
1423 l->value = c->next_label;
1429 ###### free context storage
1430 while (context.labels) {
1431 struct labels *l = context.labels;
1432 context.labels = l->next;
1436 ###### declare terminals
1440 struct label *l = new_pos(label, $ID);
1444 ###### propagate exec cases
1446 struct label *l = cast(label, prog);
1447 l->value = label_lookup(c, l->name);
1448 if (!type_compat(type, Tlabel, rules))
1449 type_err(c, "error: expected %1 found %2",
1450 prog, type, rules, Tlabel);
1454 ###### interp exec cases
1456 struct label *l = cast(label, e);
1457 rv.label = l->value;
1465 Variables are scoped named values. We store the names in a linked list
1466 of "bindings" sorted in lexical order, and use sequential search and
1473 struct binding *next; // in lexical order
1477 This linked list is stored in the parse context so that "reduce"
1478 functions can find or add variables, and so the analysis phase can
1479 ensure that every variable gets a type.
1481 ###### parse context
1483 struct binding *varlist; // In lexical order
1485 ###### ast functions
1487 static struct binding *find_binding(struct parse_context *c, struct text s)
1489 struct binding **l = &c->varlist;
1494 (cmp = text_cmp((*l)->name, s)) < 0)
1498 n = calloc(1, sizeof(*n));
1505 Each name can be linked to multiple variables defined in different
1506 scopes. Each scope starts where the name is declared and continues
1507 until the end of the containing code block. Scopes of a given name
1508 cannot nest, so a declaration while a name is in-scope is an error.
1510 ###### binding fields
1511 struct variable *var;
1515 struct variable *previous;
1517 struct binding *name;
1518 struct exec *where_decl;// where name was declared
1519 struct exec *where_set; // where type was set
1523 When a scope closes, the values of the variables might need to be freed.
1524 This happens in the context of some `struct exec` and each `exec` will
1525 need to know which variables need to be freed when it completes.
1528 struct variable *to_free;
1530 ####### variable fields
1531 struct exec *cleanup_exec;
1532 struct variable *next_free;
1534 ####### interp exec cleanup
1537 for (v = e->to_free; v; v = v->next_free) {
1538 struct value *val = var_value(c, v);
1539 free_value(v->type, val);
1543 ###### ast functions
1544 static void variable_unlink_exec(struct variable *v)
1546 struct variable **vp;
1547 if (!v->cleanup_exec)
1549 for (vp = &v->cleanup_exec->to_free;
1550 *vp; vp = &(*vp)->next_free) {
1554 v->cleanup_exec = NULL;
1559 While the naming seems strange, we include local constants in the
1560 definition of variables. A name declared `var := value` can
1561 subsequently be changed, but a name declared `var ::= value` cannot -
1564 ###### variable fields
1567 Scopes in parallel branches can be partially merged. More
1568 specifically, if a given name is declared in both branches of an
1569 if/else then its scope is a candidate for merging. Similarly if
1570 every branch of an exhaustive switch (e.g. has an "else" clause)
1571 declares a given name, then the scopes from the branches are
1572 candidates for merging.
1574 Note that names declared inside a loop (which is only parallel to
1575 itself) are never visible after the loop. Similarly names defined in
1576 scopes which are not parallel, such as those started by `for` and
1577 `switch`, are never visible after the scope. Only variables defined in
1578 both `then` and `else` (including the implicit then after an `if`, and
1579 excluding `then` used with `for`) and in all `case`s and `else` of a
1580 `switch` or `while` can be visible beyond the `if`/`switch`/`while`.
1582 Labels, which are a bit like variables, follow different rules.
1583 Labels are not explicitly declared, but if an undeclared name appears
1584 in a context where a label is legal, that effectively declares the
1585 name as a label. The declaration remains in force (or in scope) at
1586 least to the end of the immediately containing block and conditionally
1587 in any larger containing block which does not declare the name in some
1588 other way. Importantly, the conditional scope extension happens even
1589 if the label is only used in one parallel branch of a conditional --
1590 when used in one branch it is treated as having been declared in all
1593 Merge candidates are tentatively visible beyond the end of the
1594 branching statement which creates them. If the name is used, the
1595 merge is affirmed and they become a single variable visible at the
1596 outer layer. If not - if it is redeclared first - the merge lapses.
1598 To track scopes we have an extra stack, implemented as a linked list,
1599 which roughly parallels the parse stack and which is used exclusively
1600 for scoping. When a new scope is opened, a new frame is pushed and
1601 the child-count of the parent frame is incremented. This child-count
1602 is used to distinguish between the first of a set of parallel scopes,
1603 in which declared variables must not be in scope, and subsequent
1604 branches, whether they may already be conditionally scoped.
1606 We need a total ordering of scopes so we can easily compare to variables
1607 to see if they are concurrently in scope. To achieve this we record a
1608 `scope_count` which is actually a count of both beginnings and endings
1609 of scopes. Then each variable has a record of the scope count where it
1610 enters scope, and where it leaves.
1612 To push a new frame *before* any code in the frame is parsed, we need a
1613 grammar reduction. This is most easily achieved with a grammar
1614 element which derives the empty string, and creates the new scope when
1615 it is recognised. This can be placed, for example, between a keyword
1616 like "if" and the code following it.
1620 struct scope *parent;
1624 ###### parse context
1627 struct scope *scope_stack;
1629 ###### variable fields
1630 int scope_start, scope_end;
1632 ###### ast functions
1633 static void scope_pop(struct parse_context *c)
1635 struct scope *s = c->scope_stack;
1637 c->scope_stack = s->parent;
1639 c->scope_depth -= 1;
1640 c->scope_count += 1;
1643 static void scope_push(struct parse_context *c)
1645 struct scope *s = calloc(1, sizeof(*s));
1647 c->scope_stack->child_count += 1;
1648 s->parent = c->scope_stack;
1650 c->scope_depth += 1;
1651 c->scope_count += 1;
1657 OpenScope -> ${ scope_push(c); }$
1659 Each variable records a scope depth and is in one of four states:
1661 - "in scope". This is the case between the declaration of the
1662 variable and the end of the containing block, and also between
1663 the usage with affirms a merge and the end of that block.
1665 The scope depth is not greater than the current parse context scope
1666 nest depth. When the block of that depth closes, the state will
1667 change. To achieve this, all "in scope" variables are linked
1668 together as a stack in nesting order.
1670 - "pending". The "in scope" block has closed, but other parallel
1671 scopes are still being processed. So far, every parallel block at
1672 the same level that has closed has declared the name.
1674 The scope depth is the depth of the last parallel block that
1675 enclosed the declaration, and that has closed.
1677 - "conditionally in scope". The "in scope" block and all parallel
1678 scopes have closed, and no further mention of the name has been seen.
1679 This state includes a secondary nest depth (`min_depth`) which records
1680 the outermost scope seen since the variable became conditionally in
1681 scope. If a use of the name is found, the variable becomes "in scope"
1682 and that secondary depth becomes the recorded scope depth. If the
1683 name is declared as a new variable, the old variable becomes "out of
1684 scope" and the recorded scope depth stays unchanged.
1686 - "out of scope". The variable is neither in scope nor conditionally
1687 in scope. It is permanently out of scope now and can be removed from
1688 the "in scope" stack. When a variable becomes out-of-scope it is
1689 moved to a separate list (`out_scope`) of variables which have fully
1690 known scope. This will be used at the end of each function to assign
1691 each variable a place in the stack frame.
1693 ###### variable fields
1694 int depth, min_depth;
1695 enum { OutScope, PendingScope, CondScope, InScope } scope;
1696 struct variable *in_scope;
1698 ###### parse context
1700 struct variable *in_scope;
1701 struct variable *out_scope;
1703 All variables with the same name are linked together using the
1704 'previous' link. Those variable that have been affirmatively merged all
1705 have a 'merged' pointer that points to one primary variable - the most
1706 recently declared instance. When merging variables, we need to also
1707 adjust the 'merged' pointer on any other variables that had previously
1708 been merged with the one that will no longer be primary.
1710 A variable that is no longer the most recent instance of a name may
1711 still have "pending" scope, if it might still be merged with most
1712 recent instance. These variables don't really belong in the
1713 "in_scope" list, but are not immediately removed when a new instance
1714 is found. Instead, they are detected and ignored when considering the
1715 list of in_scope names.
1717 The storage of the value of a variable will be described later. For now
1718 we just need to know that when a variable goes out of scope, it might
1719 need to be freed. For this we need to be able to find it, so assume that
1720 `var_value()` will provide that.
1722 ###### variable fields
1723 struct variable *merged;
1725 ###### ast functions
1727 static void variable_merge(struct variable *primary, struct variable *secondary)
1731 primary = primary->merged;
1733 for (v = primary->previous; v; v=v->previous)
1734 if (v == secondary || v == secondary->merged ||
1735 v->merged == secondary ||
1736 v->merged == secondary->merged) {
1737 v->scope = OutScope;
1738 v->merged = primary;
1739 if (v->scope_start < primary->scope_start)
1740 primary->scope_start = v->scope_start;
1741 if (v->scope_end > primary->scope_end)
1742 primary->scope_end = v->scope_end; // NOTEST
1743 variable_unlink_exec(v);
1747 ###### forward decls
1748 static struct value *var_value(struct parse_context *c, struct variable *v);
1750 ###### free global vars
1752 while (context.varlist) {
1753 struct binding *b = context.varlist;
1754 struct variable *v = b->var;
1755 context.varlist = b->next;
1758 struct variable *next = v->previous;
1760 if (v->global && v->frame_pos >= 0) {
1761 free_value(v->type, var_value(&context, v));
1762 if (v->depth == 0 && v->type->free == function_free)
1763 // This is a function constant
1764 free_exec(v->where_decl);
1771 #### Manipulating Bindings
1773 When a name is conditionally visible, a new declaration discards the old
1774 binding - the condition lapses. Similarly when we reach the end of a
1775 function (outermost non-global scope) any conditional scope must lapse.
1776 Conversely a usage of the name affirms the visibility and extends it to
1777 the end of the containing block - i.e. the block that contains both the
1778 original declaration and the latest usage. This is determined from
1779 `min_depth`. When a conditionally visible variable gets affirmed like
1780 this, it is also merged with other conditionally visible variables with
1783 When we parse a variable declaration we either report an error if the
1784 name is currently bound, or create a new variable at the current nest
1785 depth if the name is unbound or bound to a conditionally scoped or
1786 pending-scope variable. If the previous variable was conditionally
1787 scoped, it and its homonyms becomes out-of-scope.
1789 When we parse a variable reference (including non-declarative assignment
1790 "foo = bar") we report an error if the name is not bound or is bound to
1791 a pending-scope variable; update the scope if the name is bound to a
1792 conditionally scoped variable; or just proceed normally if the named
1793 variable is in scope.
1795 When we exit a scope, any variables bound at this level are either
1796 marked out of scope or pending-scoped, depending on whether the scope
1797 was sequential or parallel. Here a "parallel" scope means the "then"
1798 or "else" part of a conditional, or any "case" or "else" branch of a
1799 switch. Other scopes are "sequential".
1801 When exiting a parallel scope we check if there are any variables that
1802 were previously pending and are still visible. If there are, then
1803 they weren't redeclared in the most recent scope, so they cannot be
1804 merged and must become out-of-scope. If it is not the first of
1805 parallel scopes (based on `child_count`), we check that there was a
1806 previous binding that is still pending-scope. If there isn't, the new
1807 variable must now be out-of-scope.
1809 When exiting a sequential scope that immediately enclosed parallel
1810 scopes, we need to resolve any pending-scope variables. If there was
1811 no `else` clause, and we cannot determine that the `switch` was exhaustive,
1812 we need to mark all pending-scope variable as out-of-scope. Otherwise
1813 all pending-scope variables become conditionally scoped.
1816 enum closetype { CloseSequential, CloseFunction, CloseParallel, CloseElse };
1818 ###### ast functions
1820 static struct variable *var_decl(struct parse_context *c, struct text s)
1822 struct binding *b = find_binding(c, s);
1823 struct variable *v = b->var;
1825 switch (v ? v->scope : OutScope) {
1827 /* Caller will report the error */
1831 v && v->scope == CondScope;
1833 v->scope = OutScope;
1837 v = calloc(1, sizeof(*v));
1838 v->previous = b->var;
1842 v->min_depth = v->depth = c->scope_depth;
1844 v->in_scope = c->in_scope;
1845 v->scope_start = c->scope_count;
1851 static struct variable *var_ref(struct parse_context *c, struct text s)
1853 struct binding *b = find_binding(c, s);
1854 struct variable *v = b->var;
1855 struct variable *v2;
1857 switch (v ? v->scope : OutScope) {
1860 /* Caller will report the error */
1863 /* All CondScope variables of this name need to be merged
1864 * and become InScope
1866 v->depth = v->min_depth;
1868 for (v2 = v->previous;
1869 v2 && v2->scope == CondScope;
1871 variable_merge(v, v2);
1879 static int var_refile(struct parse_context *c, struct variable *v)
1881 /* Variable just went out of scope. Add it to the out_scope
1882 * list, sorted by ->scope_start
1884 struct variable **vp = &c->out_scope;
1885 while ((*vp) && (*vp)->scope_start < v->scope_start)
1886 vp = &(*vp)->in_scope;
1892 static void var_block_close(struct parse_context *c, enum closetype ct,
1895 /* Close off all variables that are in_scope.
1896 * Some variables in c->scope may already be not-in-scope,
1897 * such as when a PendingScope variable is hidden by a new
1898 * variable with the same name.
1899 * So we check for v->name->var != v and drop them.
1900 * If we choose to make a variable OutScope, we drop it
1903 struct variable *v, **vp, *v2;
1906 for (vp = &c->in_scope;
1907 (v = *vp) && v->min_depth > c->scope_depth;
1908 (v->scope == OutScope || v->name->var != v)
1909 ? (*vp = v->in_scope, var_refile(c, v))
1910 : ( vp = &v->in_scope, 0)) {
1911 v->min_depth = c->scope_depth;
1912 if (v->name->var != v)
1913 /* This is still in scope, but we haven't just
1917 v->min_depth = c->scope_depth;
1918 if (v->scope == InScope)
1919 v->scope_end = c->scope_count;
1920 if (v->scope == InScope && e && !v->global) {
1921 /* This variable gets cleaned up when 'e' finishes */
1922 variable_unlink_exec(v);
1923 v->cleanup_exec = e;
1924 v->next_free = e->to_free;
1929 case CloseParallel: /* handle PendingScope */
1933 if (c->scope_stack->child_count == 1)
1934 /* first among parallel branches */
1935 v->scope = PendingScope;
1936 else if (v->previous &&
1937 v->previous->scope == PendingScope)
1938 /* all previous branches used name */
1939 v->scope = PendingScope;
1941 v->scope = OutScope;
1942 if (ct == CloseElse) {
1943 /* All Pending variables with this name
1944 * are now Conditional */
1946 v2 && v2->scope == PendingScope;
1948 v2->scope = CondScope;
1952 /* Not possible as it would require
1953 * parallel scope to be nested immediately
1954 * in a parallel scope, and that never
1958 /* Not possible as we already tested for
1965 if (v->scope == CondScope)
1966 /* Condition cannot continue past end of function */
1969 case CloseSequential:
1972 v->scope = OutScope;
1975 /* There was no 'else', so we can only become
1976 * conditional if we know the cases were exhaustive,
1977 * and that doesn't mean anything yet.
1978 * So only labels become conditional..
1981 v2 && v2->scope == PendingScope;
1983 v2->scope = OutScope;
1986 case OutScope: break;
1995 The value of a variable is store separately from the variable, on an
1996 analogue of a stack frame. There are (currently) two frames that can be
1997 active. A global frame which currently only stores constants, and a
1998 stacked frame which stores local variables. Each variable knows if it
1999 is global or not, and what its index into the frame is.
2001 Values in the global frame are known immediately they are relevant, so
2002 the frame needs to be reallocated as it grows so it can store those
2003 values. The local frame doesn't get values until the interpreted phase
2004 is started, so there is no need to allocate until the size is known.
2006 We initialize the `frame_pos` to an impossible value, so that we can
2007 tell if it was set or not later.
2009 ###### variable fields
2013 ###### variable init
2016 ###### parse context
2018 short global_size, global_alloc;
2020 void *global, *local;
2022 ###### forward decls
2023 static struct value *global_alloc(struct parse_context *c, struct type *t,
2024 struct variable *v, struct value *init);
2026 ###### ast functions
2028 static struct value *var_value(struct parse_context *c, struct variable *v)
2031 if (!c->local || !v->type)
2032 return NULL; // NOTEST
2033 if (v->frame_pos + v->type->size > c->local_size) {
2034 printf("INVALID frame_pos\n"); // NOTEST
2037 return c->local + v->frame_pos;
2039 if (c->global_size > c->global_alloc) {
2040 int old = c->global_alloc;
2041 c->global_alloc = (c->global_size | 1023) + 1024;
2042 c->global = realloc(c->global, c->global_alloc);
2043 memset(c->global + old, 0, c->global_alloc - old);
2045 return c->global + v->frame_pos;
2048 static struct value *global_alloc(struct parse_context *c, struct type *t,
2049 struct variable *v, struct value *init)
2052 struct variable scratch;
2054 if (t->prepare_type)
2055 t->prepare_type(c, t, 1); // NOTEST
2057 if (c->global_size & (t->align - 1))
2058 c->global_size = (c->global_size + t->align) & ~(t->align-1);
2063 v->frame_pos = c->global_size;
2065 c->global_size += v->type->size;
2066 ret = var_value(c, v);
2068 memcpy(ret, init, t->size);
2070 val_init(t, ret); // NOTEST
2074 As global values are found -- struct field initializers, labels etc --
2075 `global_alloc()` is called to record the value in the global frame.
2077 When the program is fully parsed, each function is analysed, we need to
2078 walk the list of variables local to that function and assign them an
2079 offset in the stack frame. For this we have `scope_finalize()`.
2081 We keep the stack from dense by re-using space for between variables
2082 that are not in scope at the same time. The `out_scope` list is sorted
2083 by `scope_start` and as we process a varible, we move it to an FIFO
2084 stack. For each variable we consider, we first discard any from the
2085 stack anything that went out of scope before the new variable came in.
2086 Then we place the new variable just after the one at the top of the
2089 ###### ast functions
2091 static void scope_finalize(struct parse_context *c, struct type *ft)
2093 int size = ft->function.local_size;
2094 struct variable *next = ft->function.scope;
2095 struct variable *done = NULL;
2098 struct variable *v = next;
2099 struct type *t = v->type;
2106 if (v->frame_pos >= 0)
2108 while (done && done->scope_end < v->scope_start)
2109 done = done->in_scope;
2111 pos = done->frame_pos + done->type->size;
2113 pos = ft->function.local_size;
2114 if (pos & (t->align - 1))
2115 pos = (pos + t->align) & ~(t->align-1);
2117 if (size < pos + v->type->size)
2118 size = pos + v->type->size;
2122 c->out_scope = NULL;
2123 ft->function.local_size = size;
2126 ###### free context storage
2127 free(context.global);
2129 #### Variables as executables
2131 Just as we used a `val` to wrap a value into an `exec`, we similarly
2132 need a `var` to wrap a `variable` into an exec. While each `val`
2133 contained a copy of the value, each `var` holds a link to the variable
2134 because it really is the same variable no matter where it appears.
2135 When a variable is used, we need to remember to follow the `->merged`
2136 link to find the primary instance.
2138 When a variable is declared, it may or may not be given an explicit
2139 type. We need to record which so that we can report the parsed code
2148 struct variable *var;
2151 ###### variable fields
2159 VariableDecl -> IDENTIFIER : ${ {
2160 struct variable *v = var_decl(c, $1.txt);
2161 $0 = new_pos(var, $1);
2166 v = var_ref(c, $1.txt);
2168 type_err(c, "error: variable '%v' redeclared",
2170 type_err(c, "info: this is where '%v' was first declared",
2171 v->where_decl, NULL, 0, NULL);
2174 | IDENTIFIER :: ${ {
2175 struct variable *v = var_decl(c, $1.txt);
2176 $0 = new_pos(var, $1);
2182 v = var_ref(c, $1.txt);
2184 type_err(c, "error: variable '%v' redeclared",
2186 type_err(c, "info: this is where '%v' was first declared",
2187 v->where_decl, NULL, 0, NULL);
2190 | IDENTIFIER : Type ${ {
2191 struct variable *v = var_decl(c, $1.txt);
2192 $0 = new_pos(var, $1);
2198 v->explicit_type = 1;
2200 v = var_ref(c, $1.txt);
2202 type_err(c, "error: variable '%v' redeclared",
2204 type_err(c, "info: this is where '%v' was first declared",
2205 v->where_decl, NULL, 0, NULL);
2208 | IDENTIFIER :: Type ${ {
2209 struct variable *v = var_decl(c, $1.txt);
2210 $0 = new_pos(var, $1);
2217 v->explicit_type = 1;
2219 v = var_ref(c, $1.txt);
2221 type_err(c, "error: variable '%v' redeclared",
2223 type_err(c, "info: this is where '%v' was first declared",
2224 v->where_decl, NULL, 0, NULL);
2229 Variable -> IDENTIFIER ${ {
2230 struct variable *v = var_ref(c, $1.txt);
2231 $0 = new_pos(var, $1);
2233 /* This might be a global const or a label
2234 * Allocate a var with impossible type Tnone,
2235 * which will be adjusted when we find out what it is,
2236 * or will trigger an error.
2238 v = var_decl(c, $1.txt);
2245 cast(var, $0)->var = v;
2248 ###### print exec cases
2251 struct var *v = cast(var, e);
2253 struct binding *b = v->var->name;
2254 printf("%.*s", b->name.len, b->name.txt);
2261 if (loc && loc->type == Xvar) {
2262 struct var *v = cast(var, loc);
2264 struct binding *b = v->var->name;
2265 fprintf(stderr, "%.*s", b->name.len, b->name.txt);
2267 fputs("???", stderr); // NOTEST
2269 fputs("NOTVAR", stderr); // NOTEST
2272 ###### propagate exec cases
2276 struct var *var = cast(var, prog);
2277 struct variable *v = var->var;
2279 type_err(c, "%d:BUG: no variable!!", prog, NULL, 0, NULL); // NOTEST
2280 return Tnone; // NOTEST
2283 if (v->type == Tnone && v->where_decl == prog)
2284 type_err(c, "error: variable used but not declared: %v",
2285 prog, NULL, 0, NULL);
2286 if (v->type == NULL) {
2287 if (type && !(*perr & Efail)) {
2289 v->where_set = prog;
2292 } else if (!type_compat(type, v->type, rules)) {
2293 type_err(c, "error: expected %1 but variable '%v' is %2", prog,
2294 type, rules, v->type);
2295 type_err(c, "info: this is where '%v' was set to %1", v->where_set,
2296 v->type, rules, NULL);
2298 if (!v->global || v->frame_pos < 0)
2305 ###### interp exec cases
2308 struct var *var = cast(var, e);
2309 struct variable *v = var->var;
2312 lrv = var_value(c, v);
2317 ###### ast functions
2319 static void free_var(struct var *v)
2324 ###### free exec cases
2325 case Xvar: free_var(cast(var, e)); break;
2330 Now that we have the shape of the interpreter in place we can add some
2331 complex types and connected them in to the data structures and the
2332 different phases of parse, analyse, print, interpret.
2334 Being "complex" the language will naturally have syntax to access
2335 specifics of objects of these types. These will fit into the grammar as
2336 "Terms" which are the things that are combined with various operators to
2337 form an "Expression". Where a Term is formed by some operation on another
2338 Term, the subordinate Term will always come first, so for example a
2339 member of an array will be expressed as the Term for the array followed
2340 by an index in square brackets. The strict rule of using postfix
2341 operations makes precedence irrelevant within terms. To provide a place
2342 to put the grammar for terms of each type, we will start out by
2343 introducing the "Term" grammar production, with contains at least a
2344 simple "Value" (to be explained later).
2346 We also take this opportunity to introduce the "ExpressionsList" which
2347 is a simple comma-separated list of expressions - it may be used in
2350 ###### declare terminals
2355 Term -> Value ${ $0 = $<1; }$
2356 | Variable ${ $0 = $<1; }$
2360 ExpressionList -> ExpressionList , Expression ${
2373 Thus far the complex types we have are arrays and structs.
2377 Arrays can be declared by giving a size and a type, as `[size]type' so
2378 `freq:[26]number` declares `freq` to be an array of 26 numbers. The
2379 size can be either a literal number, or a named constant. Some day an
2380 arbitrary expression will be supported.
2382 As a formal parameter to a function, the array can be declared with a
2383 new variable as the size: `name:[size::number]string`. The `size`
2384 variable is set to the size of the array and must be a constant. As
2385 `number` is the only supported type, it can be left out:
2386 `name:[size::]string`.
2388 Arrays cannot be assigned. When pointers are introduced we will also
2389 introduce array slices which can refer to part or all of an array -
2390 the assignment syntax will create a slice. For now, an array can only
2391 ever be referenced by the name it is declared with. It is likely that
2392 a "`copy`" primitive will eventually be define which can be used to
2393 make a copy of an array with controllable recursive depth.
2395 For now we have two sorts of array, those with fixed size either because
2396 it is given as a literal number or because it is a struct member (which
2397 cannot have a runtime-changing size), and those with a size that is
2398 determined at runtime - local variables with a const size. The former
2399 have their size calculated at parse time, the latter at run time.
2401 For the latter type, the `size` field of the type is the size of a
2402 pointer, and the array is reallocated every time it comes into scope.
2404 We differentiate struct fields with a const size from local variables
2405 with a const size by whether they are prepared at parse time or not.
2407 ###### type union fields
2410 int unspec; // size is unspecified - vsize must be set.
2413 struct variable *vsize;
2414 struct type *member;
2417 ###### value union fields
2418 void *array; // used if not static_size
2420 ###### value functions
2422 static int array_prepare_type(struct parse_context *c, struct type *type,
2425 struct value *vsize;
2427 if (type->array.static_size)
2428 return 1; // NOTEST - guard against reentry
2429 if (type->array.unspec && parse_time)
2430 return 1; // NOTEST - unspec is still incomplete
2431 if (parse_time && type->array.vsize && !type->array.vsize->global)
2432 return 1; // NOTEST - should be impossible
2434 if (type->array.vsize) {
2435 vsize = var_value(c, type->array.vsize);
2437 return 1; // NOTEST - should be impossible
2439 mpz_tdiv_q(q, mpq_numref(vsize->num), mpq_denref(vsize->num));
2440 type->array.size = mpz_get_si(q);
2445 if (type->array.member->size <= 0)
2446 return 0; // NOTEST - error caught before here
2448 type->array.static_size = 1;
2449 type->size = type->array.size * type->array.member->size;
2450 type->align = type->array.member->align;
2455 static void array_init(struct type *type, struct value *val)
2458 void *ptr = val->ptr;
2462 if (!type->array.static_size) {
2463 val->array = calloc(type->array.size,
2464 type->array.member->size);
2467 for (i = 0; i < type->array.size; i++) {
2469 v = (void*)ptr + i * type->array.member->size;
2470 val_init(type->array.member, v);
2474 static void array_free(struct type *type, struct value *val)
2477 void *ptr = val->ptr;
2479 if (!type->array.static_size)
2481 for (i = 0; i < type->array.size; i++) {
2483 v = (void*)ptr + i * type->array.member->size;
2484 free_value(type->array.member, v);
2486 if (!type->array.static_size)
2490 static int array_compat(struct type *require, struct type *have,
2491 enum val_rules rules)
2493 if (have->compat != require->compat)
2495 /* Both are arrays, so we can look at details */
2496 if (!type_compat(require->array.member, have->array.member, 0))
2498 if (have->array.unspec && require->array.unspec &&
2499 have->array.size != require->array.size)
2501 if (have->array.unspec || require->array.unspec)
2503 if (require->array.vsize == NULL && have->array.vsize == NULL)
2504 return require->array.size == have->array.size;
2506 return require->array.vsize == have->array.vsize;
2509 static void array_print_type(struct type *type, FILE *f)
2512 if (type->array.vsize) {
2513 struct binding *b = type->array.vsize->name;
2514 fprintf(f, "%.*s%s]", b->name.len, b->name.txt,
2515 type->array.unspec ? "::" : "");
2516 } else if (type->array.size)
2517 fprintf(f, "%d]", type->array.size);
2520 type_print(type->array.member, f);
2523 static struct type array_prototype = {
2525 .prepare_type = array_prepare_type,
2526 .print_type = array_print_type,
2527 .compat = array_compat,
2529 .size = sizeof(void*),
2530 .align = sizeof(void*),
2533 ###### declare terminals
2538 | [ NUMBER ] Type ${ {
2544 if (number_parse(num, tail, $2.txt) == 0)
2545 tok_err(c, "error: unrecognised number", &$2);
2547 tok_err(c, "error: unsupported number suffix", &$2);
2550 elements = mpz_get_ui(mpq_numref(num));
2551 if (mpz_cmp_ui(mpq_denref(num), 1) != 0) {
2552 tok_err(c, "error: array size must be an integer",
2554 } else if (mpz_cmp_ui(mpq_numref(num), 1UL << 30) >= 0)
2555 tok_err(c, "error: array size is too large",
2560 $0 = t = add_anon_type(c, &array_prototype, "array[%d]", elements );
2561 t->array.size = elements;
2562 t->array.member = $<4;
2563 t->array.vsize = NULL;
2566 | [ IDENTIFIER ] Type ${ {
2567 struct variable *v = var_ref(c, $2.txt);
2570 tok_err(c, "error: name undeclared", &$2);
2571 else if (!v->constant)
2572 tok_err(c, "error: array size must be a constant", &$2);
2574 $0 = add_anon_type(c, &array_prototype, "array[%.*s]", $2.txt.len, $2.txt.txt);
2575 $0->array.member = $<4;
2577 $0->array.vsize = v;
2580 ###### formal type grammar
2583 $0 = add_anon_type(c, &array_prototype, "array[]");
2584 $0->array.member = $<Type;
2586 $0->array.unspec = 1;
2587 $0->array.vsize = NULL;
2595 | Term [ Expression ] ${ {
2596 struct binode *b = new(binode);
2604 struct binode *b = new(binode);
2610 ###### print binode cases
2612 print_exec(b->left, -1, bracket);
2614 print_exec(b->right, -1, bracket);
2619 print_exec(b->left, -1, bracket);
2623 ###### propagate binode cases
2625 /* left must be an array, right must be a number,
2626 * result is the member type of the array
2628 propagate_types(b->right, c, perr_local, Tnum, 0);
2629 t = propagate_types(b->left, c, perr, NULL, 0);
2630 if (!t || t->compat != array_compat) {
2631 type_err(c, "error: %1 cannot be indexed", prog, t, 0, NULL);
2634 if (!type_compat(type, t->array.member, rules)) {
2635 type_err(c, "error: have %1 but need %2", prog,
2636 t->array.member, rules, type);
2638 return t->array.member;
2643 /* left must be an array, result is a number
2645 t = propagate_types(b->left, c, perr, NULL, 0);
2646 if (!t || t->compat != array_compat) {
2647 type_err(c, "error: %1 cannot provide length", prog, t, 0, NULL);
2650 if (!type_compat(type, Tnum, rules))
2651 type_err(c, "error: have %1 but need %2", prog,
2656 ###### interp binode cases
2662 lleft = linterp_exec(c, b->left, <ype);
2663 right = interp_exec(c, b->right, &rtype);
2665 mpz_tdiv_q(q, mpq_numref(right.num), mpq_denref(right.num));
2669 if (ltype->array.static_size)
2672 ptr = *(void**)lleft;
2673 rvtype = ltype->array.member;
2674 if (i >= 0 && i < ltype->array.size)
2675 lrv = ptr + i * rvtype->size;
2677 val_init(ltype->array.member, &rv); // UNSAFE
2682 lleft = linterp_exec(c, b->left, <ype);
2683 mpq_set_ui(rv.num, ltype->array.size, 1);
2691 A `struct` is a data-type that contains one or more other data-types.
2692 It differs from an array in that each member can be of a different
2693 type, and they are accessed by name rather than by number. Thus you
2694 cannot choose an element by calculation, you need to know what you
2697 The language makes no promises about how a given structure will be
2698 stored in memory - it is free to rearrange fields to suit whatever
2699 criteria seems important.
2701 Structs are declared separately from program code - they cannot be
2702 declared in-line in a variable declaration like arrays can. A struct
2703 is given a name and this name is used to identify the type - the name
2704 is not prefixed by the word `struct` as it would be in C.
2706 Structs are only treated as the same if they have the same name.
2707 Simply having the same fields in the same order is not enough. This
2708 might change once we can create structure initializers from a list of
2711 Each component datum is identified much like a variable is declared,
2712 with a name, one or two colons, and a type. The type cannot be omitted
2713 as there is no opportunity to deduce the type from usage. An initial
2714 value can be given following an equals sign, so
2716 ##### Example: a struct type
2722 would declare a type called "complex" which has two number fields,
2723 each initialised to zero.
2725 Struct will need to be declared separately from the code that uses
2726 them, so we will need to be able to print out the declaration of a
2727 struct when reprinting the whole program. So a `print_type_decl` type
2728 function will be needed.
2730 ###### type union fields
2739 } *fields; // This is created when field_list is analysed.
2741 struct fieldlist *prev;
2744 } *field_list; // This is created during parsing
2747 ###### type functions
2748 void (*print_type_decl)(struct type *type, FILE *f);
2749 struct type *(*fieldref)(struct type *t, struct parse_context *c,
2750 struct fieldref *f, struct value **vp);
2752 ###### value functions
2754 static void structure_init(struct type *type, struct value *val)
2758 for (i = 0; i < type->structure.nfields; i++) {
2760 v = (void*) val->ptr + type->structure.fields[i].offset;
2761 if (type->structure.fields[i].init)
2762 dup_value(type->structure.fields[i].type,
2763 type->structure.fields[i].init,
2766 val_init(type->structure.fields[i].type, v);
2770 static void structure_free(struct type *type, struct value *val)
2774 for (i = 0; i < type->structure.nfields; i++) {
2776 v = (void*)val->ptr + type->structure.fields[i].offset;
2777 free_value(type->structure.fields[i].type, v);
2781 static void free_fieldlist(struct fieldlist *f)
2785 free_fieldlist(f->prev);
2790 static void structure_free_type(struct type *t)
2793 for (i = 0; i < t->structure.nfields; i++)
2794 if (t->structure.fields[i].init) {
2795 free_value(t->structure.fields[i].type,
2796 t->structure.fields[i].init);
2798 free(t->structure.fields);
2799 free_fieldlist(t->structure.field_list);
2802 static int structure_prepare_type(struct parse_context *c,
2803 struct type *t, int parse_time)
2806 struct fieldlist *f;
2808 if (!parse_time || t->structure.fields)
2811 for (f = t->structure.field_list; f; f=f->prev) {
2815 if (f->f.type->size <= 0)
2817 if (f->f.type->prepare_type)
2818 f->f.type->prepare_type(c, f->f.type, parse_time);
2820 if (f->init == NULL)
2824 propagate_types(f->init, c, &perr, f->f.type, 0);
2825 } while (perr & Eretry);
2827 c->parse_error += 1; // NOTEST
2830 t->structure.nfields = cnt;
2831 t->structure.fields = calloc(cnt, sizeof(struct field));
2832 f = t->structure.field_list;
2834 int a = f->f.type->align;
2836 t->structure.fields[cnt] = f->f;
2837 if (t->size & (a-1))
2838 t->size = (t->size | (a-1)) + 1;
2839 t->structure.fields[cnt].offset = t->size;
2840 t->size += ((f->f.type->size - 1) | (a-1)) + 1;
2844 if (f->init && !c->parse_error) {
2845 struct value vl = interp_exec(c, f->init, NULL);
2846 t->structure.fields[cnt].init =
2847 global_alloc(c, f->f.type, NULL, &vl);
2855 static int find_struct_index(struct type *type, struct text field)
2858 for (i = 0; i < type->structure.nfields; i++)
2859 if (text_cmp(type->structure.fields[i].name, field) == 0)
2861 return IndexInvalid;
2864 static struct type *structure_fieldref(struct type *t, struct parse_context *c,
2865 struct fieldref *f, struct value **vp)
2867 if (f->index == IndexUnknown) {
2868 f->index = find_struct_index(t, f->name);
2870 type_err(c, "error: cannot find requested field in %1",
2871 f->left, t, 0, NULL);
2876 struct value *v = *vp;
2877 v = (void*)v->ptr + t->structure.fields[f->index].offset;
2880 return t->structure.fields[f->index].type;
2883 static struct type structure_prototype = {
2884 .init = structure_init,
2885 .free = structure_free,
2886 .free_type = structure_free_type,
2887 .print_type_decl = structure_print_type,
2888 .prepare_type = structure_prepare_type,
2889 .fieldref = structure_fieldref,
2902 enum { IndexUnknown = -1, IndexInvalid = -2 };
2904 ###### free exec cases
2906 free_exec(cast(fieldref, e)->left);
2910 ###### declare terminals
2915 | Term . IDENTIFIER ${ {
2916 struct fieldref *fr = new_pos(fieldref, $2);
2919 fr->index = IndexUnknown;
2923 ###### print exec cases
2927 struct fieldref *f = cast(fieldref, e);
2928 print_exec(f->left, -1, bracket);
2929 printf(".%.*s", f->name.len, f->name.txt);
2933 ###### propagate exec cases
2937 struct fieldref *f = cast(fieldref, prog);
2938 struct type *st = propagate_types(f->left, c, perr, NULL, 0);
2940 if (!st || !st->fieldref)
2941 type_err(c, "error: field reference on %1 is not supported",
2942 f->left, st, 0, NULL);
2944 t = st->fieldref(st, c, f, NULL);
2945 if (t && !type_compat(type, t, rules))
2946 type_err(c, "error: have %1 but need %2", prog,
2953 ###### interp exec cases
2956 struct fieldref *f = cast(fieldref, e);
2958 struct value *lleft = linterp_exec(c, f->left, <ype);
2960 rvtype = ltype->fieldref(ltype, c, f, &lrv);
2964 ###### top level grammar
2966 StructName -> IDENTIFIER ${ {
2967 struct type *t = find_type(c, $ID.txt);
2969 if (t && t->size >= 0) {
2970 tok_err(c, "error: type already declared", &$ID);
2971 tok_err(c, "info: this is location of declartion", &t->first_use);
2975 t = add_type(c, $ID.txt, NULL);
2980 DeclareStruct -> struct StructName FieldBlock Newlines ${ {
2981 struct type *t = $<SN;
2982 struct type tmp = *t;
2984 *t = structure_prototype;
2987 t->first_use = tmp.first_use;
2989 t->structure.field_list = $<FB;
2993 FieldBlock -> { IN OptNL FieldLines OUT OptNL } ${ $0 = $<FL; }$
2994 | { SimpleFieldList } ${ $0 = $<SFL; }$
2995 | IN OptNL FieldLines OUT ${ $0 = $<FL; }$
2996 | SimpleFieldList EOL ${ $0 = $<SFL; }$
2998 FieldLines -> SimpleFieldList Newlines ${ $0 = $<SFL; }$
2999 | FieldLines SimpleFieldList Newlines ${ {
3000 struct fieldlist *f = $<SFL;
3011 SimpleFieldList -> Field ${ $0 = $<F; }$
3012 | SimpleFieldList ; Field ${
3016 | SimpleFieldList ; ${
3019 | ERROR ${ tok_err(c, "Syntax error in struct field", &$1); }$
3021 Field -> IDENTIFIER : Type = Expression ${ {
3022 $0 = calloc(1, sizeof(struct fieldlist));
3023 $0->f.name = $ID.txt;
3024 $0->f.type = $<Type;
3028 | IDENTIFIER : Type ${
3029 $0 = calloc(1, sizeof(struct fieldlist));
3030 $0->f.name = $ID.txt;
3031 $0->f.type = $<Type;
3034 ###### forward decls
3035 static void structure_print_type(struct type *t, FILE *f);
3037 ###### value functions
3038 static void structure_print_type(struct type *t, FILE *f)
3042 fprintf(f, "struct %.*s\n", t->name.len, t->name.txt);
3044 for (i = 0; i < t->structure.nfields; i++) {
3045 struct field *fl = t->structure.fields + i;
3046 fprintf(f, " %.*s : ", fl->name.len, fl->name.txt);
3047 type_print(fl->type, f);
3048 if (fl->type->print && fl->init) {
3050 if (fl->type == Tstr)
3052 print_value(fl->type, fl->init, f);
3053 if (fl->type == Tstr)
3060 ###### print type decls
3065 while (target != 0) {
3067 for (t = context.typelist; t ; t=t->next)
3068 if (!t->anon && t->print_type_decl &&
3078 t->print_type_decl(t, stdout);
3086 References, or pointers, are values that refer to another value. They
3087 can only refer to a `struct`, though as a struct can embed anything they
3088 can effectively refer to anything.
3090 References are potentially dangerous as they might refer to some
3091 variable which no longer exists - either because a stack frame
3092 containing it has been discarded or because the value was allocated on
3093 the heap and has now been free. Ocean does not yet provide any
3094 protection against these problems. It will in due course.
3096 With references comes the opportunity and the need to explicitly
3097 allocate values on the "heap" and to free them. We currently provide
3098 fairly basic support for this.
3100 Reference make use of the `@` symbol in various ways. A type that starts
3101 with `@` is a reference to whatever follows. A reference value
3102 followed by an `@` acts as the referred value, though the `@` is often
3103 not needed. Finally, an expression that starts with `@` is a special
3104 reference related expression. Some examples might help.
3106 ##### Example: Reference examples
3113 bar.number = 23; bar.string = "hello"
3124 Obviously this is very contrived. `ref` is a reference to a `foo` which
3125 is initially set to refer to the value stored in `bar` - no extra syntax
3126 is needed to "Take the address of" `bar` - the fact that `ref` is a
3127 reference means that only the address make sense.
3129 When `ref.a` is accessed, that is whatever value is stored in `bar.a`.
3130 The same syntax is used for accessing fields both in structs and in
3131 references to structs. It would be correct to use `ref@.a`, but not
3134 `@new()` creates an object of whatever type is needed for the program
3135 to by type-correct. In future iterations of Ocean, arguments a
3136 constructor will access arguments, so the the syntax now looks like a
3137 function call. `@free` can be assigned any reference that was returned
3138 by `@new()`, and it will be freed. `@nil` is a value of whatever
3139 reference type is appropriate, and is stable and never the address of
3140 anything in the heap or on the stack. A reference can be assigned
3141 `@nil` or compared against that value.
3143 ###### declare terminals
3146 ###### type union fields
3149 struct type *referent;
3152 ###### value union fields
3155 ###### value functions
3157 static void reference_print_type(struct type *t, FILE *f)
3160 type_print(t->reference.referent, f);
3163 static int reference_cmp(struct type *tl, struct type *tr,
3164 struct value *left, struct value *right)
3166 return left->ref == right->ref ? 0 : 1;
3169 static void reference_dup(struct type *t,
3170 struct value *vold, struct value *vnew)
3172 vnew->ref = vold->ref;
3175 static void reference_free(struct type *t, struct value *v)
3177 /* Nothing to do here */
3180 static int reference_compat(struct type *require, struct type *have,
3181 enum val_rules rules)
3184 if (require->reference.referent == have)
3186 if (have->compat != require->compat)
3188 if (have->reference.referent != require->reference.referent)
3193 static int reference_test(struct type *type, struct value *val)
3195 return val->ref != NULL;
3198 static struct type *reference_fieldref(struct type *t, struct parse_context *c,
3199 struct fieldref *f, struct value **vp)
3201 struct type *rt = t->reference.referent;
3206 return rt->fieldref(rt, c, f, vp);
3208 type_err(c, "error: field reference on %1 is not supported",
3209 f->left, rt, 0, NULL);
3213 static struct type reference_prototype = {
3214 .print_type = reference_print_type,
3215 .cmp_eq = reference_cmp,
3216 .dup = reference_dup,
3217 .test = reference_test,
3218 .free = reference_free,
3219 .compat = reference_compat,
3220 .fieldref = reference_fieldref,
3221 .size = sizeof(void*),
3222 .align = sizeof(void*),
3228 struct type *t = find_type(c, $ID.txt);
3230 t = add_type(c, $ID.txt, NULL);
3233 $0 = find_anon_type(c, &reference_prototype, "@%.*s",
3234 $ID.txt.len, $ID.txt.txt);
3235 $0->reference.referent = t;
3238 ###### core functions
3239 static int text_is(struct text t, char *s)
3241 return (strlen(s) == t.len &&
3242 strncmp(s, t.txt, t.len) == 0);
3251 enum ref_func { RefNew, RefFree, RefNil } action;
3252 struct type *reftype;
3256 ###### SimpleStatement Grammar
3258 | @ IDENTIFIER = Expression ${ {
3259 struct ref *r = new_pos(ref, $ID);
3261 if (!text_is($ID.txt, "free"))
3262 tok_err(c, "error: only \"@free\" makes sense here",
3266 r->action = RefFree;
3270 ###### expression grammar
3271 | @ IDENTIFIER ( ) ${
3272 // Only 'new' valid here
3273 if (!text_is($ID.txt, "new")) {
3274 tok_err(c, "error: Only reference function is \"@new()\"",
3277 struct ref *r = new_pos(ref,$ID);
3283 // Only 'nil' valid here
3284 if (!text_is($ID.txt, "nil")) {
3285 tok_err(c, "error: Only reference value is \"@nil\"",
3288 struct ref *r = new_pos(ref,$ID);
3294 ###### print exec cases
3296 struct ref *r = cast(ref, e);
3297 switch (r->action) {
3299 printf("@new()"); break;
3301 printf("@nil"); break;
3303 do_indent(indent, "@free = ");
3304 print_exec(r->right, indent, bracket);
3310 ###### propagate exec cases
3312 struct ref *r = cast(ref, prog);
3313 switch (r->action) {
3315 if (type && type->free != reference_free) {
3316 type_err(c, "error: @new() can only be used with references, not %1",
3317 prog, type, 0, NULL);
3320 if (type && !r->reftype) {
3327 if (type && type->free != reference_free)
3328 type_err(c, "error: @nil can only be used with reference, not %1",
3329 prog, type, 0, NULL);
3330 if (type && !r->reftype) {
3337 t = propagate_types(r->right, c, perr_local, NULL, 0);
3338 if (t && t->free != reference_free)
3339 type_err(c, "error: @free can only be assigned a reference, not %1",
3348 ###### interp exec cases
3350 struct ref *r = cast(ref, e);
3351 switch (r->action) {
3354 rv.ref = calloc(1, r->reftype->reference.referent->size);
3355 rvtype = r->reftype;
3359 rvtype = r->reftype;
3362 rv = interp_exec(c, r->right, &rvtype);
3363 free_value(rvtype->reference.referent, rv.ref);
3371 ###### free exec cases
3373 struct ref *r = cast(ref, e);
3374 free_exec(r->right);
3379 ###### Expressions: dereference
3387 struct binode *b = new(binode);
3393 ###### print binode cases
3395 print_exec(b->left, -1, bracket);
3399 print_exec(b->left, -1, bracket);
3402 ###### propagate binode cases
3404 /* left must be a reference, and we return what it refers to */
3405 /* FIXME how can I pass the expected type down? */
3406 t = propagate_types(b->left, c, perr, NULL, 0);
3408 if (!t || t->free != reference_free)
3409 type_err(c, "error: Cannot dereference %1", b, t, 0, NULL);
3411 return t->reference.referent;
3415 /* left must be lval, we create reference to it */
3416 if (!type || type->free != reference_free)
3417 t = propagate_types(b->left, c, perr, type, 0); // NOTEST impossible
3419 t = propagate_types(b->left, c, perr,
3420 type->reference.referent, 0);
3422 t = find_anon_type(c, &reference_prototype, "@%.*s",
3423 t->name.len, t->name.txt);
3426 ###### interp binode cases
3428 left = interp_exec(c, b->left, <ype);
3430 rvtype = ltype->reference.referent;
3434 rv.ref = linterp_exec(c, b->left, &rvtype);
3435 rvtype = find_anon_type(c, &reference_prototype, "@%.*s",
3436 rvtype->name.len, rvtype->name.txt);
3442 A function is a chunk of code which can be passed parameters and can
3443 return results. Each function has a type which includes the set of
3444 parameters and the return value. As yet these types cannot be declared
3445 separately from the function itself.
3447 The parameters can be specified either in parentheses as a ';' separated
3450 ##### Example: function 1
3452 func main(av:[ac::number]string; env:[envc::number]string)
3455 or as an indented list of one parameter per line (though each line can
3456 be a ';' separated list)
3458 ##### Example: function 2
3461 argv:[argc::number]string
3462 env:[envc::number]string
3466 In the first case a return type can follow the parentheses after a colon,
3467 in the second it is given on a line starting with the word `return`.
3469 ##### Example: functions that return
3471 func add(a:number; b:number): number
3481 Rather than returning a type, the function can specify a set of local
3482 variables to return as a struct. The values of these variables when the
3483 function exits will be provided to the caller. For this the return type
3484 is replaced with a block of result declarations, either in parentheses
3485 or bracketed by `return` and `do`.
3487 ##### Example: functions returning multiple variables
3489 func to_cartesian(rho:number; theta:number):(x:number; y:number)
3502 For constructing the lists we use a `List` binode, which will be
3503 further detailed when Expression Lists are introduced.
3505 ###### type union fields
3508 struct binode *params;
3509 struct type *return_type;
3510 struct variable *scope;
3511 int inline_result; // return value is at start of 'local'
3515 ###### value union fields
3516 struct exec *function;
3518 ###### type functions
3519 void (*check_args)(struct parse_context *c, enum prop_err *perr,
3520 struct type *require, struct exec *args);
3522 ###### value functions
3524 static void function_free(struct type *type, struct value *val)
3526 free_exec(val->function);
3527 val->function = NULL;
3530 static int function_compat(struct type *require, struct type *have,
3531 enum val_rules rules)
3533 // FIXME can I do anything here yet?
3537 static struct exec *take_addr(struct exec *e)
3539 struct binode *rv = new(binode);
3545 static void function_check_args(struct parse_context *c, enum prop_err *perr,
3546 struct type *require, struct exec *args)
3548 /* This should be 'compat', but we don't have a 'tuple' type to
3549 * hold the type of 'args'
3551 struct binode *arg = cast(binode, args);
3552 struct binode *param = require->function.params;
3555 struct var *pv = cast(var, param->left);
3556 struct type *t = pv->var->type, *t2;
3558 type_err(c, "error: insufficient arguments to function.",
3559 args, NULL, 0, NULL);
3563 t2 = propagate_types(arg->left, c, perr, t, Rrefok);
3564 if (t->free == reference_free &&
3565 t->reference.referent == t2 &&
3567 arg->left = take_addr(arg->left);
3568 } else if (!(*perr & Efail) && !type_compat(t2, t, 0)) {
3569 type_err(c, "error: cannot pass rval when reference expected",
3570 arg->left, NULL, 0, NULL);
3572 param = cast(binode, param->right);
3573 arg = cast(binode, arg->right);
3576 type_err(c, "error: too many arguments to function.",
3577 args, NULL, 0, NULL);
3580 static void function_print(struct type *type, struct value *val, FILE *f)
3583 print_exec(val->function, 1, 0);
3586 static void function_print_type_decl(struct type *type, FILE *f)
3590 for (b = type->function.params; b; b = cast(binode, b->right)) {
3591 struct variable *v = cast(var, b->left)->var;
3592 fprintf(f, "%.*s%s", v->name->name.len, v->name->name.txt,
3593 v->constant ? "::" : ":");
3594 type_print(v->type, f);
3599 if (type->function.return_type != Tnone) {
3601 if (type->function.inline_result) {
3603 struct type *t = type->function.return_type;
3605 for (i = 0; i < t->structure.nfields; i++) {
3606 struct field *fl = t->structure.fields + i;
3609 fprintf(f, "%.*s:", fl->name.len, fl->name.txt);
3610 type_print(fl->type, f);
3614 type_print(type->function.return_type, f);
3618 static void function_free_type(struct type *t)
3620 free_exec(t->function.params);
3623 static struct type function_prototype = {
3624 .size = sizeof(void*),
3625 .align = sizeof(void*),
3626 .free = function_free,
3627 .compat = function_compat,
3628 .check_args = function_check_args,
3629 .print = function_print,
3630 .print_type_decl = function_print_type_decl,
3631 .free_type = function_free_type,
3634 ###### declare terminals
3641 FuncName -> IDENTIFIER ${ {
3642 struct variable *v = var_decl(c, $1.txt);
3643 struct var *e = new_pos(var, $1);
3650 v = var_ref(c, $1.txt);
3652 type_err(c, "error: function '%v' redeclared",
3654 type_err(c, "info: this is where '%v' was first declared",
3655 v->where_decl, NULL, 0, NULL);
3661 Args -> ArgsLine NEWLINE ${ $0 = $<AL; }$
3662 | Args ArgsLine NEWLINE ${ {
3663 struct binode *b = $<AL;
3664 struct binode **bp = &b;
3666 bp = (struct binode **)&(*bp)->left;
3671 ArgsLine -> ${ $0 = NULL; }$
3672 | Varlist ${ $0 = $<1; }$
3673 | Varlist ; ${ $0 = $<1; }$
3675 Varlist -> Varlist ; ArgDecl ${
3676 $0 = new_pos(binode, $2);
3689 ArgDecl -> IDENTIFIER : FormalType ${ {
3690 struct variable *v = var_decl(c, $ID.txt);
3691 $0 = new_pos(var, $ID);
3698 ##### Function calls
3700 A function call can appear either as an expression or as a statement.
3701 We use a new 'Funcall' binode type to link the function with a list of
3702 arguments, form with the 'List' nodes.
3704 We have already seen the "Term" which is how a function call can appear
3705 in an expression. To parse a function call into a statement we include
3706 it in the "SimpleStatement Grammar" which will be described later.
3712 | Term ( ExpressionList ) ${ {
3713 struct binode *b = new(binode);
3716 b->right = reorder_bilist($<EL);
3720 struct binode *b = new(binode);
3727 ###### SimpleStatement Grammar
3729 | Term ( ExpressionList ) ${ {
3730 struct binode *b = new(binode);
3733 b->right = reorder_bilist($<EL);
3737 ###### print binode cases
3740 do_indent(indent, "");
3741 print_exec(b->left, -1, bracket);
3743 for (b = cast(binode, b->right); b; b = cast(binode, b->right)) {
3746 print_exec(b->left, -1, bracket);
3756 ###### propagate binode cases
3759 /* Every arg must match formal parameter, and result
3760 * is return type of function
3762 struct binode *args = cast(binode, b->right);
3763 struct var *v = cast(var, b->left);
3765 if (!v->var->type || v->var->type->check_args == NULL) {
3766 type_err(c, "error: attempt to call a non-function.",
3767 prog, NULL, 0, NULL);
3771 v->var->type->check_args(c, perr_local, v->var->type, args);
3772 if (v->var->type->function.inline_result)
3775 return v->var->type->function.return_type;
3778 ###### interp binode cases
3781 struct var *v = cast(var, b->left);
3782 struct type *t = v->var->type;
3783 void *oldlocal = c->local;
3784 int old_size = c->local_size;
3785 void *local = calloc(1, t->function.local_size);
3786 struct value *fbody = var_value(c, v->var);
3787 struct binode *arg = cast(binode, b->right);
3788 struct binode *param = t->function.params;
3791 struct var *pv = cast(var, param->left);
3792 struct type *vtype = NULL;
3793 struct value val = interp_exec(c, arg->left, &vtype);
3795 c->local = local; c->local_size = t->function.local_size;
3796 lval = var_value(c, pv->var);
3797 c->local = oldlocal; c->local_size = old_size;
3798 memcpy(lval, &val, vtype->size);
3799 param = cast(binode, param->right);
3800 arg = cast(binode, arg->right);
3802 c->local = local; c->local_size = t->function.local_size;
3803 if (t->function.inline_result && dtype) {
3804 _interp_exec(c, fbody->function, NULL, NULL);
3805 memcpy(dest, local, dtype->size);
3806 rvtype = ret.type = NULL;
3808 rv = interp_exec(c, fbody->function, &rvtype);
3809 c->local = oldlocal; c->local_size = old_size;
3814 ## Complex executables: statements and expressions
3816 Now that we have types and values and variables and most of the basic
3817 Terms which provide access to these, we can explore the more complex
3818 code that combine all of these to get useful work done. Specifically
3819 statements and expressions.
3821 Expressions are various combinations of Terms. We will use operator
3822 precedence to ensure correct parsing. The simplest Expression is just a
3823 Term - others will follow.
3828 Expression -> Term ${ $0 = $<Term; }$
3829 ## expression grammar
3831 ### Expressions: Conditional
3833 Our first user of the `binode` will be conditional expressions, which
3834 is a bit odd as they actually have three components. That will be
3835 handled by having 2 binodes for each expression. The conditional
3836 expression is the lowest precedence operator which is why we define it
3837 first - to start the precedence list.
3839 Conditional expressions are of the form "value `if` condition `else`
3840 other_value". They associate to the right, so everything to the right
3841 of `else` is part of an else value, while only a higher-precedence to
3842 the left of `if` is the if values. Between `if` and `else` there is no
3843 room for ambiguity, so a full conditional expression is allowed in
3849 ###### declare terminals
3853 ###### expression grammar
3855 | Expression if Expression else Expression $$ifelse ${ {
3856 struct binode *b1 = new(binode);
3857 struct binode *b2 = new(binode);
3867 ###### print binode cases
3870 b2 = cast(binode, b->right);
3871 if (bracket) printf("(");
3872 print_exec(b2->left, -1, bracket);
3874 print_exec(b->left, -1, bracket);
3876 print_exec(b2->right, -1, bracket);
3877 if (bracket) printf(")");
3880 ###### propagate binode cases
3883 /* cond must be Tbool, others must match */
3884 struct binode *b2 = cast(binode, b->right);
3887 propagate_types(b->left, c, perr_local, Tbool, 0);
3888 t = propagate_types(b2->left, c, perr, type, 0);
3889 t2 = propagate_types(b2->right, c, perr, type ?: t, 0);
3893 ###### interp binode cases
3896 struct binode *b2 = cast(binode, b->right);
3897 left = interp_exec(c, b->left, <ype);
3899 rv = interp_exec(c, b2->left, &rvtype);
3901 rv = interp_exec(c, b2->right, &rvtype);
3905 ### Expressions: Boolean
3907 The next class of expressions to use the `binode` will be Boolean
3908 expressions. `and` and `or` are short-circuit operators that don't
3909 evaluate the second expression if not necessary.
3916 ###### declare terminals
3921 ###### expression grammar
3922 | Expression or Expression ${ {
3923 struct binode *b = new(binode);
3929 | Expression and Expression ${ {
3930 struct binode *b = new(binode);
3936 | not Expression ${ {
3937 struct binode *b = new(binode);
3943 ###### print binode cases
3945 if (bracket) printf("(");
3946 print_exec(b->left, -1, bracket);
3948 print_exec(b->right, -1, bracket);
3949 if (bracket) printf(")");
3952 if (bracket) printf("(");
3953 print_exec(b->left, -1, bracket);
3955 print_exec(b->right, -1, bracket);
3956 if (bracket) printf(")");
3959 if (bracket) printf("(");
3961 print_exec(b->right, -1, bracket);
3962 if (bracket) printf(")");
3965 ###### propagate binode cases
3969 /* both must be Tbool, result is Tbool */
3970 propagate_types(b->left, c, perr, Tbool, 0);
3971 propagate_types(b->right, c, perr, Tbool, 0);
3972 if (type && type != Tbool)
3973 type_err(c, "error: %1 operation found where %2 expected", prog,
3978 ###### interp binode cases
3980 rv = interp_exec(c, b->left, &rvtype);
3982 rv = interp_exec(c, b->right, NULL);
3985 rv = interp_exec(c, b->left, &rvtype);
3987 rv = interp_exec(c, b->right, NULL);
3990 rv = interp_exec(c, b->right, &rvtype);
3994 ### Expressions: Comparison
3996 Of slightly higher precedence that Boolean expressions are Comparisons.
3997 A comparison takes arguments of any comparable type, but the two types
4000 To simplify the parsing we introduce an `eop` which can record an
4001 expression operator, and the `CMPop` non-terminal will match one of them.
4008 ###### ast functions
4009 static void free_eop(struct eop *e)
4023 ###### declare terminals
4024 $LEFT < > <= >= == != CMPop
4026 ###### expression grammar
4027 | Expression CMPop Expression ${ {
4028 struct binode *b = new(binode);
4038 CMPop -> < ${ $0.op = Less; }$
4039 | > ${ $0.op = Gtr; }$
4040 | <= ${ $0.op = LessEq; }$
4041 | >= ${ $0.op = GtrEq; }$
4042 | == ${ $0.op = Eql; }$
4043 | != ${ $0.op = NEql; }$
4045 ###### print binode cases
4053 if (bracket) printf("(");
4054 print_exec(b->left, -1, bracket);
4056 case Less: printf(" < "); break;
4057 case LessEq: printf(" <= "); break;
4058 case Gtr: printf(" > "); break;
4059 case GtrEq: printf(" >= "); break;
4060 case Eql: printf(" == "); break;
4061 case NEql: printf(" != "); break;
4062 default: abort(); // NOTEST
4064 print_exec(b->right, -1, bracket);
4065 if (bracket) printf(")");
4068 ###### propagate binode cases
4075 /* Both must match but not be labels, result is Tbool */
4076 t = propagate_types(b->left, c, perr, NULL, 0);
4078 propagate_types(b->right, c, perr, t, 0);
4080 t = propagate_types(b->right, c, perr, NULL, 0); // NOTEST
4082 t = propagate_types(b->left, c, perr, t, 0); // NOTEST
4084 if (!type_compat(type, Tbool, 0))
4085 type_err(c, "error: Comparison returns %1 but %2 expected", prog,
4086 Tbool, rules, type);
4090 ###### interp binode cases
4099 left = interp_exec(c, b->left, <ype);
4100 right = interp_exec(c, b->right, &rtype);
4101 cmp = value_cmp(ltype, rtype, &left, &right);
4104 case Less: rv.bool = cmp < 0; break;
4105 case LessEq: rv.bool = cmp <= 0; break;
4106 case Gtr: rv.bool = cmp > 0; break;
4107 case GtrEq: rv.bool = cmp >= 0; break;
4108 case Eql: rv.bool = cmp == 0; break;
4109 case NEql: rv.bool = cmp != 0; break;
4110 default: rv.bool = 0; break; // NOTEST
4115 ### Expressions: Arithmetic etc.
4117 The remaining expressions with the highest precedence are arithmetic,
4118 string concatenation, string conversion, and testing. String concatenation
4119 (`++`) has the same precedence as multiplication and division, but lower
4122 Testing comes in two forms. A single question mark (`?`) is a uniary
4123 operator which converts come types into Boolean. The general meaning is
4124 "is this a value value" and there will be more uses as the language
4125 develops. A double questionmark (`??`) is a binary operator (Choose),
4126 with same precedence as multiplication, which returns the LHS if it
4127 tests successfully, else returns the RHS.
4129 String conversion is a temporary feature until I get a better type
4130 system. `$` is a prefix operator which expects a string and returns
4133 `+` and `-` are both infix and prefix operations (where they are
4134 absolute value and negation). These have different operator names.
4136 We also have a 'Bracket' operator which records where parentheses were
4137 found. This makes it easy to reproduce these when printing. Possibly I
4138 should only insert brackets were needed for precedence. Putting
4139 parentheses around an expression converts it into a Term,
4145 Absolute, Negate, Test,
4149 ###### declare terminals
4151 $LEFT * / % ++ ?? Top
4155 ###### expression grammar
4156 | Expression Eop Expression ${ {
4157 struct binode *b = new(binode);
4164 | Expression Top Expression ${ {
4165 struct binode *b = new(binode);
4172 | Uop Expression ${ {
4173 struct binode *b = new(binode);
4181 | ( Expression ) ${ {
4182 struct binode *b = new_pos(binode, $1);
4191 Eop -> + ${ $0.op = Plus; }$
4192 | - ${ $0.op = Minus; }$
4194 Uop -> + ${ $0.op = Absolute; }$
4195 | - ${ $0.op = Negate; }$
4196 | $ ${ $0.op = StringConv; }$
4197 | ? ${ $0.op = Test; }$
4199 Top -> * ${ $0.op = Times; }$
4200 | / ${ $0.op = Divide; }$
4201 | % ${ $0.op = Rem; }$
4202 | ++ ${ $0.op = Concat; }$
4203 | ?? ${ $0.op = Choose; }$
4205 ###### print binode cases
4213 if (bracket) printf("(");
4214 print_exec(b->left, indent, bracket);
4216 case Plus: fputs(" + ", stdout); break;
4217 case Minus: fputs(" - ", stdout); break;
4218 case Times: fputs(" * ", stdout); break;
4219 case Divide: fputs(" / ", stdout); break;
4220 case Rem: fputs(" % ", stdout); break;
4221 case Concat: fputs(" ++ ", stdout); break;
4222 case Choose: fputs(" ?? ", stdout); break;
4223 default: abort(); // NOTEST
4225 print_exec(b->right, indent, bracket);
4226 if (bracket) printf(")");
4232 if (bracket) printf("(");
4234 case Absolute: fputs("+", stdout); break;
4235 case Negate: fputs("-", stdout); break;
4236 case StringConv: fputs("$", stdout); break;
4237 case Test: fputs("?", stdout); break;
4238 default: abort(); // NOTEST
4240 print_exec(b->right, indent, bracket);
4241 if (bracket) printf(")");
4244 /* Avoid double brackets... */
4245 if (!bracket) printf("(");
4246 print_exec(b->right, indent, bracket);
4247 if (!bracket) printf(")");
4250 ###### propagate binode cases
4256 /* both must be numbers, result is Tnum */
4259 /* as propagate_types ignores a NULL,
4260 * unary ops fit here too */
4261 propagate_types(b->left, c, perr, Tnum, 0);
4262 propagate_types(b->right, c, perr, Tnum, 0);
4263 if (!type_compat(type, Tnum, 0))
4264 type_err(c, "error: Arithmetic returns %1 but %2 expected", prog,
4270 /* both must be Tstr, result is Tstr */
4271 propagate_types(b->left, c, perr, Tstr, 0);
4272 propagate_types(b->right, c, perr, Tstr, 0);
4273 if (!type_compat(type, Tstr, 0))
4274 type_err(c, "error: Concat returns %1 but %2 expected", prog,
4280 /* op must be string, result is number */
4281 propagate_types(b->left, c, perr, Tstr, 0);
4282 if (!type_compat(type, Tnum, 0))
4284 "error: Can only convert string to number, not %1",
4285 prog, type, 0, NULL);
4290 /* LHS must support ->test, result is Tbool */
4291 t = propagate_types(b->right, c, perr, NULL, 0);
4293 type_err(c, "error: '?' requires a testable value, not %1",
4299 /* LHS and RHS must match and are returned. Must support
4302 t = propagate_types(b->left, c, perr, type, rules);
4303 t = propagate_types(b->right, c, perr, t, rules);
4304 if (t && t->test == NULL)
4305 type_err(c, "error: \"??\" requires a testable value, not %1",
4311 return propagate_types(b->right, c, perr, type, rules);
4313 ###### interp binode cases
4316 rv = interp_exec(c, b->left, &rvtype);
4317 right = interp_exec(c, b->right, &rtype);
4318 mpq_add(rv.num, rv.num, right.num);
4321 rv = interp_exec(c, b->left, &rvtype);
4322 right = interp_exec(c, b->right, &rtype);
4323 mpq_sub(rv.num, rv.num, right.num);
4326 rv = interp_exec(c, b->left, &rvtype);
4327 right = interp_exec(c, b->right, &rtype);
4328 mpq_mul(rv.num, rv.num, right.num);
4331 rv = interp_exec(c, b->left, &rvtype);
4332 right = interp_exec(c, b->right, &rtype);
4333 mpq_div(rv.num, rv.num, right.num);
4338 left = interp_exec(c, b->left, <ype);
4339 right = interp_exec(c, b->right, &rtype);
4340 mpz_init(l); mpz_init(r); mpz_init(rem);
4341 mpz_tdiv_q(l, mpq_numref(left.num), mpq_denref(left.num));
4342 mpz_tdiv_q(r, mpq_numref(right.num), mpq_denref(right.num));
4343 mpz_tdiv_r(rem, l, r);
4344 val_init(Tnum, &rv);
4345 mpq_set_z(rv.num, rem);
4346 mpz_clear(r); mpz_clear(l); mpz_clear(rem);
4351 rv = interp_exec(c, b->right, &rvtype);
4352 mpq_neg(rv.num, rv.num);
4355 rv = interp_exec(c, b->right, &rvtype);
4356 mpq_abs(rv.num, rv.num);
4359 rv = interp_exec(c, b->right, &rvtype);
4362 left = interp_exec(c, b->left, <ype);
4363 right = interp_exec(c, b->right, &rtype);
4365 rv.str = text_join(left.str, right.str);
4368 right = interp_exec(c, b->right, &rvtype);
4372 struct text tx = right.str;
4375 if (tx.txt[0] == '-') {
4380 if (number_parse(rv.num, tail, tx) == 0)
4383 mpq_neg(rv.num, rv.num);
4385 printf("Unsupported suffix: %.*s\n", tx.len, tx.txt);
4389 right = interp_exec(c, b->right, &rtype);
4391 rv.bool = !!rtype->test(rtype, &right);
4394 left = interp_exec(c, b->left, <ype);
4395 if (ltype->test(ltype, &left)) {
4400 rv = interp_exec(c, b->right, &rvtype);
4403 ###### value functions
4405 static struct text text_join(struct text a, struct text b)
4408 rv.len = a.len + b.len;
4409 rv.txt = malloc(rv.len);
4410 memcpy(rv.txt, a.txt, a.len);
4411 memcpy(rv.txt+a.len, b.txt, b.len);
4415 ### Blocks, Statements, and Statement lists.
4417 Now that we have expressions out of the way we need to turn to
4418 statements. There are simple statements and more complex statements.
4419 Simple statements do not contain (syntactic) newlines, complex statements do.
4421 Statements often come in sequences and we have corresponding simple
4422 statement lists and complex statement lists.
4423 The former comprise only simple statements separated by semicolons.
4424 The later comprise complex statements and simple statement lists. They are
4425 separated by newlines. Thus the semicolon is only used to separate
4426 simple statements on the one line. This may be overly restrictive,
4427 but I'm not sure I ever want a complex statement to share a line with
4430 Note that a simple statement list can still use multiple lines if
4431 subsequent lines are indented, so
4433 ###### Example: wrapped simple statement list
4438 is a single simple statement list. This might allow room for
4439 confusion, so I'm not set on it yet.
4441 A simple statement list needs no extra syntax. A complex statement
4442 list has two syntactic forms. It can be enclosed in braces (much like
4443 C blocks), or it can be introduced by an indent and continue until an
4444 unindented newline (much like Python blocks). With this extra syntax
4445 it is referred to as a block.
4447 Note that a block does not have to include any newlines if it only
4448 contains simple statements. So both of:
4450 if condition: a=b; d=f
4452 if condition { a=b; print f }
4456 In either case the list is constructed from a `binode` list with
4457 `Block` as the operator. When parsing the list it is most convenient
4458 to append to the end, so a list is a list and a statement. When using
4459 the list it is more convenient to consider a list to be a statement
4460 and a list. So we need a function to re-order a list.
4461 `reorder_bilist` serves this purpose.
4463 The only stand-alone statement we introduce at this stage is `pass`
4464 which does nothing and is represented as a `NULL` pointer in a `Block`
4465 list. Other stand-alone statements will follow once the infrastructure
4468 As many statements will use binodes, we declare a binode pointer 'b' in
4469 the common header for all reductions to use.
4471 ###### Parser: reduce
4482 Block -> { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
4483 | { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
4484 | SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
4485 | SimpleStatements EOL ${ $0 = reorder_bilist($<SS);
4487 | IN OptNL Statementlist OUT ${ $0 = $<Sl; }$
4489 OpenBlock -> OpenScope { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
4490 | OpenScope { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
4491 | OpenScope SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
4492 | OpenScope SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
4493 | IN OpenScope OptNL Statementlist OUT ${ $0 = $<Sl; }$
4495 UseBlock -> { IN OpenScope OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
4496 | { OpenScope SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
4497 | IN OpenScope OptNL Statementlist OUT ${ $0 = $<Sl; }$
4499 ColonBlock -> { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
4500 | { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
4501 | : SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
4502 | : SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
4503 | : IN OptNL Statementlist OUT ${ $0 = $<Sl; }$
4505 Statementlist -> ComplexStatements ${ $0 = reorder_bilist($<CS); }$
4507 ComplexStatements -> ComplexStatements ComplexStatement ${
4509 $0 = $<1; // NOTEST - impossible
4517 | ComplexStatement ${
4519 $0 = NULL; // NOTEST - impossible
4529 ComplexStatement -> SimpleStatements Newlines ${
4530 $0 = reorder_bilist($<SS);
4532 | SimpleStatements ; Newlines ${
4533 $0 = reorder_bilist($<SS);
4535 ## ComplexStatement Grammar
4538 SimpleStatements -> SimpleStatements ; SimpleStatement ${
4544 | SimpleStatement ${
4553 SimpleStatement -> pass ${ $0 = NULL; }$
4554 | ERROR ${ tok_err(c, "Syntax error in statement", &$1); }$
4555 ## SimpleStatement Grammar
4557 ###### print binode cases
4559 // block, one per line
4560 if (b->left == NULL)
4561 do_indent(indent, "pass\n");
4563 print_exec(b->left, indent, bracket);
4565 print_exec(b->right, indent, bracket);
4568 ###### propagate binode cases
4571 /* If any statement returns something other than Tnone
4572 * or Tbool then all such must return same type.
4573 * As each statement may be Tnone or something else,
4574 * we must always pass NULL (unknown) down, otherwise an incorrect
4575 * error might occur. We never return Tnone unless it is
4580 for (e = b; e; e = cast(binode, e->right)) {
4581 *perr |= *perr_local;
4583 t = propagate_types(e->left, c, perr_local, NULL, rules);
4584 if ((rules & Rboolok) && (t == Tbool || t == Tnone))
4586 if (t == Tnone && e->right)
4587 /* Only the final statement *must* return a value
4595 type_err(c, "error: expected %1, found %2",
4596 e->left, type, rules, t);
4602 ###### interp binode cases
4604 while (rvtype == Tnone &&
4607 rv = interp_exec(c, b->left, &rvtype);
4608 b = cast(binode, b->right);
4612 ### The Print statement
4614 `print` is a simple statement that takes a comma-separated list of
4615 expressions and prints the values separated by spaces and terminated
4616 by a newline. No control of formatting is possible.
4618 `print` uses `ExpressionList` to collect the expressions and stores them
4619 on the left side of a `Print` binode unlessthere is a trailing comma
4620 when the list is stored on the `right` side and no trailing newline is
4626 ##### declare terminals
4629 ###### SimpleStatement Grammar
4631 | print ExpressionList ${
4632 $0 = b = new_pos(binode, $1);
4635 b->left = reorder_bilist($<EL);
4637 | print ExpressionList , ${ {
4638 $0 = b = new_pos(binode, $1);
4640 b->right = reorder_bilist($<EL);
4644 $0 = b = new_pos(binode, $1);
4650 ###### print binode cases
4653 do_indent(indent, "print");
4654 b2 = cast(binode, b->left ?: b->right);
4657 print_exec(b2->left, -1, bracket);
4660 b2 = cast(binode, b2->right);
4668 ###### propagate binode cases
4671 /* don't care but all must be consistent */
4673 b = cast(binode, b->left);
4675 b = cast(binode, b->right);
4677 propagate_types(b->left, c, perr_local, NULL, 0);
4678 b = cast(binode, b->right);
4682 ###### interp binode cases
4686 struct binode *b2 = cast(binode, b->left);
4688 b2 = cast(binode, b->right);
4689 for (; b2; b2 = cast(binode, b2->right)) {
4690 left = interp_exec(c, b2->left, <ype);
4691 print_value(ltype, &left, stdout);
4692 free_value(ltype, &left);
4696 if (b->right == NULL)
4702 ###### Assignment statement
4704 An assignment will assign a value to a variable, providing it hasn't
4705 been declared as a constant. The analysis phase ensures that the type
4706 will be correct so the interpreter just needs to perform the
4707 calculation. There is a form of assignment which declares a new
4708 variable as well as assigning a value. If a name is used before
4709 it is declared, it is assumed to be a global constant which are allowed to
4710 be declared at any time.
4716 ###### declare terminals
4719 ###### SimpleStatement Grammar
4720 | Term = Expression ${
4721 $0 = b= new(binode);
4726 | VariableDecl = Expression ${
4727 $0 = b= new(binode);
4734 if ($1->var->where_set == NULL) {
4736 "Variable declared with no type or value: %v",
4740 $0 = b = new(binode);
4747 ###### print binode cases
4750 do_indent(indent, "");
4751 print_exec(b->left, -1, bracket);
4753 print_exec(b->right, -1, bracket);
4760 struct variable *v = cast(var, b->left)->var;
4761 do_indent(indent, "");
4762 print_exec(b->left, -1, bracket);
4763 if (cast(var, b->left)->var->constant) {
4765 if (v->explicit_type) {
4766 type_print(v->type, stdout);
4771 if (v->explicit_type) {
4772 type_print(v->type, stdout);
4778 print_exec(b->right, -1, bracket);
4785 ###### propagate binode cases
4789 /* Both must match, or left may be ref and right an lval
4790 * Type must support 'dup',
4791 * For Assign, left must not be constant.
4794 *perr &= ~(Erval | Econst);
4795 t = propagate_types(b->left, c, perr, NULL, 0);
4800 struct type *t2 = propagate_types(b->right, c, perr_local,
4802 if (!t2 || t2 == t || (*perr_local & Efail))
4803 ; // No more effort needed
4804 else if (t->free == reference_free &&
4805 t->reference.referent == t2 &&
4806 !(*perr_local & Erval))
4807 b->right = take_addr(b->right);
4808 else if (t->free == reference_free &&
4809 t->reference.referent == t2 &&
4810 (*perr_local & Erval))
4811 type_err(c, "error: Cannot assign an rval to a reference.",
4814 t = propagate_types(b->right, c, perr_local, NULL, 0);
4816 propagate_types(b->left, c, perr, t, 0);
4819 type_err(c, "error: cannot assign to an rval", b,
4821 else if (b->op == Assign && (*perr & Econst)) {
4822 type_err(c, "error: Cannot assign to a constant: %v",
4823 b->left, NULL, 0, NULL);
4824 if (b->left->type == Xvar) {
4825 struct var *var = cast(var, b->left);
4826 struct variable *v = var->var;
4827 type_err(c, "info: name was defined as a constant here",
4828 v->where_decl, NULL, 0, NULL);
4831 if (t && t->dup == NULL && !(*perr_local & Emaycopy))
4832 type_err(c, "error: cannot assign value of type %1", b, t, 0, NULL);
4833 if (b->left->type == Xvar && (*perr_local & Efail))
4834 type_err(c, "info: variable '%v' was set as %1 here.",
4835 cast(var, b->left)->var->where_set, t, rules, NULL);
4840 ###### interp binode cases
4843 lleft = linterp_exec(c, b->left, <ype);
4845 dinterp_exec(c, b->right, lleft, ltype, 1);
4851 struct variable *v = cast(var, b->left)->var;
4854 val = var_value(c, v);
4855 if (v->type->prepare_type)
4856 v->type->prepare_type(c, v->type, 0);
4858 val_init(v->type, val);
4860 dinterp_exec(c, b->right, val, v->type, 0);
4864 ### The `use` statement
4866 The `use` statement is the last "simple" statement. It is needed when a
4867 statement block can return a value. This includes the body of a
4868 function which has a return type, and the "condition" code blocks in
4869 `if`, `while`, and `switch` statements.
4874 ###### declare terminals
4877 ###### SimpleStatement Grammar
4879 $0 = b = new_pos(binode, $1);
4884 ###### print binode cases
4887 do_indent(indent, "use ");
4888 print_exec(b->right, -1, bracket);
4893 ###### propagate binode cases
4896 /* result matches value */
4897 return propagate_types(b->right, c, perr, type, 0);
4899 ###### interp binode cases
4902 rv = interp_exec(c, b->right, &rvtype);
4905 ### The Conditional Statement
4907 This is the biggy and currently the only complex statement. This
4908 subsumes `if`, `while`, `do/while`, `switch`, and some parts of `for`.
4909 It is comprised of a number of parts, all of which are optional though
4910 set combinations apply. Each part is (usually) a key word (`then` is
4911 sometimes optional) followed by either an expression or a code block,
4912 except the `casepart` which is a "key word and an expression" followed
4913 by a code block. The code-block option is valid for all parts and,
4914 where an expression is also allowed, the code block can use the `use`
4915 statement to report a value. If the code block does not report a value
4916 the effect is similar to reporting `True`.
4918 The `else` and `case` parts, as well as `then` when combined with
4919 `if`, can contain a `use` statement which will apply to some
4920 containing conditional statement. `for` parts, `do` parts and `then`
4921 parts used with `for` can never contain a `use`, except in some
4922 subordinate conditional statement.
4924 If there is a `forpart`, it is executed first, only once.
4925 If there is a `dopart`, then it is executed repeatedly providing
4926 always that the `condpart` or `cond`, if present, does not return a non-True
4927 value. `condpart` can fail to return any value if it simply executes
4928 to completion. This is treated the same as returning `True`.
4930 If there is a `thenpart` it will be executed whenever the `condpart`
4931 or `cond` returns True (or does not return any value), but this will happen
4932 *after* `dopart` (when present).
4934 If `elsepart` is present it will be executed at most once when the
4935 condition returns `False` or some value that isn't `True` and isn't
4936 matched by any `casepart`. If there are any `casepart`s, they will be
4937 executed when the condition returns a matching value.
4939 The particular sorts of values allowed in case parts has not yet been
4940 determined in the language design, so nothing is prohibited.
4942 The various blocks in this complex statement potentially provide scope
4943 for variables as described earlier. Each such block must include the
4944 "OpenScope" nonterminal before parsing the block, and must call
4945 `var_block_close()` when closing the block.
4947 The code following "`if`", "`switch`" and "`for`" does not get its own
4948 scope, but is in a scope covering the whole statement, so names
4949 declared there cannot be redeclared elsewhere. Similarly the
4950 condition following "`while`" is in a scope the covers the body
4951 ("`do`" part) of the loop, and which does not allow conditional scope
4952 extension. Code following "`then`" (both looping and non-looping),
4953 "`else`" and "`case`" each get their own local scope.
4955 The type requirements on the code block in a `whilepart` are quite
4956 unusal. It is allowed to return a value of some identifiable type, in
4957 which case the loop aborts and an appropriate `casepart` is run, or it
4958 can return a Boolean, in which case the loop either continues to the
4959 `dopart` (on `True`) or aborts and runs the `elsepart` (on `False`).
4960 This is different both from the `ifpart` code block which is expected to
4961 return a Boolean, or the `switchpart` code block which is expected to
4962 return the same type as the casepart values. The correct analysis of
4963 the type of the `whilepart` code block is the reason for the
4964 `Rboolok` flag which is passed to `propagate_types()`.
4966 The `cond_statement` cannot fit into a `binode` so a new `exec` is
4967 defined. As there are two scopes which cover multiple parts - one for
4968 the whole statement and one for "while" and "do" - and as we will use
4969 the 'struct exec' to track scopes, we actually need two new types of
4970 exec. One is a `binode` for the looping part, the rest is the
4971 `cond_statement`. The `cond_statement` will use an auxilliary `struct
4972 casepart` to track a list of case parts.
4983 struct exec *action;
4984 struct casepart *next;
4986 struct cond_statement {
4988 struct exec *forpart, *condpart, *thenpart, *elsepart;
4989 struct binode *looppart;
4990 struct casepart *casepart;
4993 ###### ast functions
4995 static void free_casepart(struct casepart *cp)
4999 free_exec(cp->value);
5000 free_exec(cp->action);
5007 static void free_cond_statement(struct cond_statement *s)
5011 free_exec(s->forpart);
5012 free_exec(s->condpart);
5013 free_exec(s->looppart);
5014 free_exec(s->thenpart);
5015 free_exec(s->elsepart);
5016 free_casepart(s->casepart);
5020 ###### free exec cases
5021 case Xcond_statement: free_cond_statement(cast(cond_statement, e)); break;
5023 ###### ComplexStatement Grammar
5024 | CondStatement ${ $0 = $<1; }$
5026 ###### declare terminals
5027 $TERM for then while do
5034 // A CondStatement must end with EOL, as does CondSuffix and
5036 // ForPart, ThenPart, SwitchPart, CasePart are non-empty and
5037 // may or may not end with EOL
5038 // WhilePart and IfPart include an appropriate Suffix
5040 // ForPart, SwitchPart, and IfPart open scopes, o we have to close
5041 // them. WhilePart opens and closes its own scope.
5042 CondStatement -> ForPart OptNL ThenPart OptNL WhilePart CondSuffix ${
5045 $0->thenpart = $<TP;
5046 $0->looppart = $<WP;
5047 var_block_close(c, CloseSequential, $0);
5049 | ForPart OptNL WhilePart CondSuffix ${
5052 $0->looppart = $<WP;
5053 var_block_close(c, CloseSequential, $0);
5055 | WhilePart CondSuffix ${
5057 $0->looppart = $<WP;
5059 | SwitchPart OptNL CasePart CondSuffix ${
5061 $0->condpart = $<SP;
5062 $CP->next = $0->casepart;
5063 $0->casepart = $<CP;
5064 var_block_close(c, CloseSequential, $0);
5066 | SwitchPart : IN OptNL CasePart CondSuffix OUT Newlines ${
5068 $0->condpart = $<SP;
5069 $CP->next = $0->casepart;
5070 $0->casepart = $<CP;
5071 var_block_close(c, CloseSequential, $0);
5073 | IfPart IfSuffix ${
5075 $0->condpart = $IP.condpart; $IP.condpart = NULL;
5076 $0->thenpart = $IP.thenpart; $IP.thenpart = NULL;
5077 // This is where we close an "if" statement
5078 var_block_close(c, CloseSequential, $0);
5081 CondSuffix -> IfSuffix ${
5084 | Newlines CasePart CondSuffix ${
5086 $CP->next = $0->casepart;
5087 $0->casepart = $<CP;
5089 | CasePart CondSuffix ${
5091 $CP->next = $0->casepart;
5092 $0->casepart = $<CP;
5095 IfSuffix -> Newlines ${ $0 = new(cond_statement); }$
5096 | Newlines ElsePart ${ $0 = $<EP; }$
5097 | ElsePart ${$0 = $<EP; }$
5099 ElsePart -> else OpenBlock Newlines ${
5100 $0 = new(cond_statement);
5101 $0->elsepart = $<OB;
5102 var_block_close(c, CloseElse, $0->elsepart);
5104 | else OpenScope CondStatement ${
5105 $0 = new(cond_statement);
5106 $0->elsepart = $<CS;
5107 var_block_close(c, CloseElse, $0->elsepart);
5111 CasePart -> case Expression OpenScope ColonBlock ${
5112 $0 = calloc(1,sizeof(struct casepart));
5115 var_block_close(c, CloseParallel, $0->action);
5119 // These scopes are closed in CondStatement
5120 ForPart -> for OpenBlock ${
5124 ThenPart -> then OpenBlock ${
5126 var_block_close(c, CloseSequential, $0);
5130 // This scope is closed in CondStatement
5131 WhilePart -> while UseBlock OptNL do OpenBlock ${
5136 var_block_close(c, CloseSequential, $0->right);
5137 var_block_close(c, CloseSequential, $0);
5139 | while OpenScope Expression OpenScope ColonBlock ${
5144 var_block_close(c, CloseSequential, $0->right);
5145 var_block_close(c, CloseSequential, $0);
5149 IfPart -> if UseBlock OptNL then OpenBlock ${
5152 var_block_close(c, CloseParallel, $0.thenpart);
5154 | if OpenScope Expression OpenScope ColonBlock ${
5157 var_block_close(c, CloseParallel, $0.thenpart);
5159 | if OpenScope Expression OpenScope OptNL then Block ${
5162 var_block_close(c, CloseParallel, $0.thenpart);
5166 // This scope is closed in CondStatement
5167 SwitchPart -> switch OpenScope Expression ${
5170 | switch UseBlock ${
5174 ###### print binode cases
5176 if (b->left && b->left->type == Xbinode &&
5177 cast(binode, b->left)->op == Block) {
5179 do_indent(indent, "while {\n");
5181 do_indent(indent, "while\n");
5182 print_exec(b->left, indent+1, bracket);
5184 do_indent(indent, "} do {\n");
5186 do_indent(indent, "do\n");
5187 print_exec(b->right, indent+1, bracket);
5189 do_indent(indent, "}\n");
5191 do_indent(indent, "while ");
5192 print_exec(b->left, 0, bracket);
5197 print_exec(b->right, indent+1, bracket);
5199 do_indent(indent, "}\n");
5203 ###### print exec cases
5205 case Xcond_statement:
5207 struct cond_statement *cs = cast(cond_statement, e);
5208 struct casepart *cp;
5210 do_indent(indent, "for");
5211 if (bracket) printf(" {\n"); else printf("\n");
5212 print_exec(cs->forpart, indent+1, bracket);
5215 do_indent(indent, "} then {\n");
5217 do_indent(indent, "then\n");
5218 print_exec(cs->thenpart, indent+1, bracket);
5220 if (bracket) do_indent(indent, "}\n");
5223 print_exec(cs->looppart, indent, bracket);
5227 do_indent(indent, "switch");
5229 do_indent(indent, "if");
5230 if (cs->condpart && cs->condpart->type == Xbinode &&
5231 cast(binode, cs->condpart)->op == Block) {
5236 print_exec(cs->condpart, indent+1, bracket);
5238 do_indent(indent, "}\n");
5240 do_indent(indent, "then\n");
5241 print_exec(cs->thenpart, indent+1, bracket);
5245 print_exec(cs->condpart, 0, bracket);
5251 print_exec(cs->thenpart, indent+1, bracket);
5253 do_indent(indent, "}\n");
5258 for (cp = cs->casepart; cp; cp = cp->next) {
5259 do_indent(indent, "case ");
5260 print_exec(cp->value, -1, 0);
5265 print_exec(cp->action, indent+1, bracket);
5267 do_indent(indent, "}\n");
5270 do_indent(indent, "else");
5275 print_exec(cs->elsepart, indent+1, bracket);
5277 do_indent(indent, "}\n");
5282 ###### propagate binode cases
5284 propagate_types(b->right, c, perr_local, Tnone, 0);
5285 return propagate_types(b->left, c, perr, type, rules);
5287 ###### propagate exec cases
5288 case Xcond_statement:
5290 // forpart and looppart->right must return Tnone
5291 // thenpart must return Tnone if there is a loopart,
5292 // otherwise it is like elsepart.
5294 // be bool if there is no casepart
5295 // match casepart->values if there is a switchpart
5296 // either be bool or match casepart->value if there
5298 // elsepart and casepart->action must match the return type
5299 // expected of this statement.
5300 struct cond_statement *cs = cast(cond_statement, prog);
5301 struct casepart *cp;
5303 t = propagate_types(cs->forpart, c, perr, Tnone, 0);
5306 t = propagate_types(cs->thenpart, c, perr, Tnone, 0);
5308 if (cs->casepart == NULL) {
5309 propagate_types(cs->condpart, c, perr, Tbool, 0);
5310 propagate_types(cs->looppart, c, perr, Tbool, 0);
5312 /* Condpart must match case values, with bool permitted */
5314 for (cp = cs->casepart;
5315 cp && !t; cp = cp->next)
5316 t = propagate_types(cp->value, c, perr, NULL, 0);
5317 if (!t && cs->condpart)
5318 t = propagate_types(cs->condpart, c, perr, NULL, Rboolok); // NOTEST
5319 if (!t && cs->looppart)
5320 t = propagate_types(cs->looppart, c, perr, NULL, Rboolok); // NOTEST
5321 // Now we have a type (I hope) push it down
5323 for (cp = cs->casepart; cp; cp = cp->next)
5324 propagate_types(cp->value, c, perr, t, 0);
5325 propagate_types(cs->condpart, c, perr, t, Rboolok);
5326 propagate_types(cs->looppart, c, perr, t, Rboolok);
5329 // (if)then, else, and case parts must return expected type.
5330 if (!cs->looppart && !type)
5331 type = propagate_types(cs->thenpart, c, perr, NULL, rules);
5333 type = propagate_types(cs->elsepart, c, perr, NULL, rules);
5334 for (cp = cs->casepart;
5336 cp = cp->next) // NOTEST
5337 type = propagate_types(cp->action, c, perr, NULL, rules); // NOTEST
5340 propagate_types(cs->thenpart, c, perr, type, rules);
5341 propagate_types(cs->elsepart, c, perr, type, rules);
5342 for (cp = cs->casepart; cp ; cp = cp->next)
5343 propagate_types(cp->action, c, perr, type, rules);
5349 ###### interp binode cases
5351 // This just performs one iterration of the loop
5352 rv = interp_exec(c, b->left, &rvtype);
5353 if (rvtype == Tnone ||
5354 (rvtype == Tbool && rv.bool != 0))
5355 // rvtype is Tnone or Tbool, doesn't need to be freed
5356 interp_exec(c, b->right, NULL);
5359 ###### interp exec cases
5360 case Xcond_statement:
5362 struct value v, cnd;
5363 struct type *vtype, *cndtype;
5364 struct casepart *cp;
5365 struct cond_statement *cs = cast(cond_statement, e);
5368 interp_exec(c, cs->forpart, NULL);
5370 while ((cnd = interp_exec(c, cs->looppart, &cndtype)),
5371 cndtype == Tnone || (cndtype == Tbool && cnd.bool != 0))
5372 interp_exec(c, cs->thenpart, NULL);
5374 cnd = interp_exec(c, cs->condpart, &cndtype);
5375 if ((cndtype == Tnone ||
5376 (cndtype == Tbool && cnd.bool != 0))) {
5377 // cnd is Tnone or Tbool, doesn't need to be freed
5378 rv = interp_exec(c, cs->thenpart, &rvtype);
5379 // skip else (and cases)
5383 for (cp = cs->casepart; cp; cp = cp->next) {
5384 v = interp_exec(c, cp->value, &vtype);
5385 if (value_cmp(cndtype, vtype, &v, &cnd) == 0) {
5386 free_value(vtype, &v);
5387 free_value(cndtype, &cnd);
5388 rv = interp_exec(c, cp->action, &rvtype);
5391 free_value(vtype, &v);
5393 free_value(cndtype, &cnd);
5395 rv = interp_exec(c, cs->elsepart, &rvtype);
5402 ### Top level structure
5404 All the language elements so far can be used in various places. Now
5405 it is time to clarify what those places are.
5407 At the top level of a file there will be a number of declarations.
5408 Many of the things that can be declared haven't been described yet,
5409 such as functions, procedures, imports, and probably more.
5410 For now there are two sorts of things that can appear at the top
5411 level. They are predefined constants, `struct` types, and the `main`
5412 function. While the syntax will allow the `main` function to appear
5413 multiple times, that will trigger an error if it is actually attempted.
5415 The various declarations do not return anything. They store the
5416 various declarations in the parse context.
5418 ###### Parser: grammar
5421 Ocean -> OptNL DeclarationList
5423 ## declare terminals
5431 DeclarationList -> Declaration
5432 | DeclarationList Declaration
5434 Declaration -> ERROR Newlines ${
5435 tok_err(c, // NOTEST
5436 "error: unhandled parse error", &$1);
5442 ## top level grammar
5446 ### The `const` section
5448 As well as being defined in with the code that uses them, constants can
5449 be declared at the top level. These have full-file scope, so they are
5450 always `InScope`, even before(!) they have been declared. The value of
5451 a top level constant can be given as an expression, and this is
5452 evaluated after parsing and before execution.
5454 A function call can be used to evaluate a constant, but it will not have
5455 access to any program state, once such statement becomes meaningful.
5456 e.g. arguments and filesystem will not be visible.
5458 Constants are defined in a section that starts with the reserved word
5459 `const` and then has a block with a list of assignment statements.
5460 For syntactic consistency, these must use the double-colon syntax to
5461 make it clear that they are constants. Type can also be given: if
5462 not, the type will be determined during analysis, as with other
5465 ###### parse context
5466 struct binode *constlist;
5468 ###### top level grammar
5472 DeclareConstant -> const { IN OptNL ConstList OUT OptNL } Newlines
5473 | const { SimpleConstList } Newlines
5474 | const IN OptNL ConstList OUT Newlines
5475 | const SimpleConstList Newlines
5477 ConstList -> ConstList SimpleConstLine
5480 SimpleConstList -> SimpleConstList ; Const
5484 SimpleConstLine -> SimpleConstList Newlines
5485 | ERROR Newlines ${ tok_err(c, "Syntax error in constant", &$1); }$
5488 CType -> Type ${ $0 = $<1; }$
5492 Const -> IDENTIFIER :: CType = Expression ${ {
5494 struct binode *bl, *bv;
5495 struct var *var = new_pos(var, $ID);
5497 v = var_decl(c, $ID.txt);
5499 v->where_decl = var;
5505 v = var_ref(c, $1.txt);
5506 if (v->type == Tnone) {
5507 v->where_decl = var;
5513 tok_err(c, "error: name already declared", &$1);
5514 type_err(c, "info: this is where '%v' was first declared",
5515 v->where_decl, NULL, 0, NULL);
5527 bl->left = c->constlist;
5532 ###### core functions
5533 static void resolve_consts(struct parse_context *c)
5537 enum { none, some, cannot } progress = none;
5539 c->constlist = reorder_bilist(c->constlist);
5542 for (b = cast(binode, c->constlist); b;
5543 b = cast(binode, b->right)) {
5545 struct binode *vb = cast(binode, b->left);
5546 struct var *v = cast(var, vb->left);
5547 if (v->var->frame_pos >= 0)
5551 propagate_types(vb->right, c, &perr,
5553 } while (perr & Eretry);
5555 c->parse_error += 1;
5556 else if (!(perr & Eruntime)) {
5558 struct value res = interp_exec(
5559 c, vb->right, &v->var->type);
5560 global_alloc(c, v->var->type, v->var, &res);
5562 if (progress == cannot)
5563 type_err(c, "error: const %v cannot be resolved.",
5573 progress = cannot; break;
5575 progress = none; break;
5580 ###### print const decls
5585 for (b = cast(binode, context.constlist); b;
5586 b = cast(binode, b->right)) {
5587 struct binode *vb = cast(binode, b->left);
5588 struct var *vr = cast(var, vb->left);
5589 struct variable *v = vr->var;
5595 printf(" %.*s :: ", v->name->name.len, v->name->name.txt);
5596 type_print(v->type, stdout);
5598 print_exec(vb->right, -1, 0);
5603 ###### free const decls
5604 free_binode(context.constlist);
5606 ### Function declarations
5608 The code in an Ocean program is all stored in function declarations.
5609 One of the functions must be named `main` and it must accept an array of
5610 strings as a parameter - the command line arguments.
5612 As this is the top level, several things are handled a bit differently.
5613 The function is not interpreted by `interp_exec` as that isn't passed
5614 the argument list which the program requires. Similarly type analysis
5615 is a bit more interesting at this level.
5617 ###### ast functions
5619 static struct type *handle_results(struct parse_context *c,
5620 struct binode *results)
5622 /* Create a 'struct' type from the results list, which
5623 * is a list for 'struct var'
5625 struct type *t = add_anon_type(c, &structure_prototype,
5630 for (b = results; b; b = cast(binode, b->right))
5632 t->structure.nfields = cnt;
5633 t->structure.fields = calloc(cnt, sizeof(struct field));
5635 for (b = results; b; b = cast(binode, b->right)) {
5636 struct var *v = cast(var, b->left);
5637 struct field *f = &t->structure.fields[cnt++];
5638 int a = v->var->type->align;
5639 f->name = v->var->name->name;
5640 f->type = v->var->type;
5642 f->offset = t->size;
5643 v->var->frame_pos = f->offset;
5644 t->size += ((f->type->size - 1) | (a-1)) + 1;
5647 variable_unlink_exec(v->var);
5649 free_binode(results);
5653 static struct variable *declare_function(struct parse_context *c,
5654 struct variable *name,
5655 struct binode *args,
5657 struct binode *results,
5661 struct value fn = {.function = code};
5663 var_block_close(c, CloseFunction, code);
5664 t = add_anon_type(c, &function_prototype,
5665 "func %.*s", name->name->name.len,
5666 name->name->name.txt);
5668 t->function.params = reorder_bilist(args);
5670 ret = handle_results(c, reorder_bilist(results));
5671 t->function.inline_result = 1;
5672 t->function.local_size = ret->size;
5674 t->function.return_type = ret;
5675 global_alloc(c, t, name, &fn);
5676 name->type->function.scope = c->out_scope;
5681 var_block_close(c, CloseFunction, NULL);
5683 c->out_scope = NULL;
5687 ###### declare terminals
5690 ###### top level grammar
5693 DeclareFunction -> func FuncName ( OpenScope ArgsLine ) Block Newlines ${
5694 $0 = declare_function(c, $<FN, $<Ar, Tnone, NULL, $<Bl);
5696 | func FuncName IN OpenScope Args OUT OptNL do Block Newlines ${
5697 $0 = declare_function(c, $<FN, $<Ar, Tnone, NULL, $<Bl);
5699 | func FuncName NEWLINE OpenScope OptNL do Block Newlines ${
5700 $0 = declare_function(c, $<FN, NULL, Tnone, NULL, $<Bl);
5702 | func FuncName ( OpenScope ArgsLine ) : Type Block Newlines ${
5703 $0 = declare_function(c, $<FN, $<Ar, $<Ty, NULL, $<Bl);
5705 | func FuncName ( OpenScope ArgsLine ) : ( ArgsLine ) Block Newlines ${
5706 $0 = declare_function(c, $<FN, $<AL, NULL, $<AL2, $<Bl);
5708 | func FuncName IN OpenScope Args OUT OptNL return Type Newlines do Block Newlines ${
5709 $0 = declare_function(c, $<FN, $<Ar, $<Ty, NULL, $<Bl);
5711 | func FuncName NEWLINE OpenScope return Type Newlines do Block Newlines ${
5712 $0 = declare_function(c, $<FN, NULL, $<Ty, NULL, $<Bl);
5714 | func FuncName IN OpenScope Args OUT OptNL return IN Args OUT OptNL do Block Newlines ${
5715 $0 = declare_function(c, $<FN, $<Ar, NULL, $<Ar2, $<Bl);
5717 | func FuncName NEWLINE OpenScope return IN Args OUT OptNL do Block Newlines ${
5718 $0 = declare_function(c, $<FN, NULL, NULL, $<Ar, $<Bl);
5721 ###### print func decls
5726 while (target != 0) {
5728 for (v = context.in_scope; v; v=v->in_scope)
5729 if (v->depth == 0 && v->type && v->type->check_args) {
5738 struct value *val = var_value(&context, v);
5739 printf("func %.*s", v->name->name.len, v->name->name.txt);
5740 v->type->print_type_decl(v->type, stdout);
5743 print_exec(val->function, 1, brackets);
5746 print_value(v->type, val, stdout);
5748 printf("/* frame size %d */\n", v->type->function.local_size);
5754 ###### core functions
5756 static int analyse_funcs(struct parse_context *c)
5760 for (v = c->in_scope; v; v = v->in_scope) {
5764 if (v->depth != 0 || !v->type || !v->type->check_args)
5766 ret = v->type->function.inline_result ?
5767 Tnone : v->type->function.return_type;
5768 val = var_value(c, v);
5771 propagate_types(val->function, c, &perr, ret, 0);
5772 } while (!(perr & Efail) && (perr & Eretry));
5773 if (!(perr & Efail))
5774 /* Make sure everything is still consistent */
5775 propagate_types(val->function, c, &perr, ret, 0);
5778 if (!v->type->function.inline_result &&
5779 !v->type->function.return_type->dup) {
5780 type_err(c, "error: function cannot return value of type %1",
5781 v->where_decl, v->type->function.return_type, 0, NULL);
5784 scope_finalize(c, v->type);
5789 static int analyse_main(struct type *type, struct parse_context *c)
5791 struct binode *bp = type->function.params;
5795 struct type *argv_type;
5797 argv_type = add_anon_type(c, &array_prototype, "argv");
5798 argv_type->array.member = Tstr;
5799 argv_type->array.unspec = 1;
5801 for (b = bp; b; b = cast(binode, b->right)) {
5805 propagate_types(b->left, c, &perr, argv_type, 0);
5807 default: /* invalid */ // NOTEST
5808 propagate_types(b->left, c, &perr, Tnone, 0); // NOTEST
5811 c->parse_error += 1;
5814 return !c->parse_error;
5817 static void interp_main(struct parse_context *c, int argc, char **argv)
5819 struct value *progp = NULL;
5820 struct text main_name = { "main", 4 };
5821 struct variable *mainv;
5827 mainv = var_ref(c, main_name);
5829 progp = var_value(c, mainv);
5830 if (!progp || !progp->function) {
5831 fprintf(stderr, "oceani: no main function found.\n");
5832 c->parse_error += 1;
5835 if (!analyse_main(mainv->type, c)) {
5836 fprintf(stderr, "oceani: main has wrong type.\n");
5837 c->parse_error += 1;
5840 al = mainv->type->function.params;
5842 c->local_size = mainv->type->function.local_size;
5843 c->local = calloc(1, c->local_size);
5845 struct var *v = cast(var, al->left);
5846 struct value *vl = var_value(c, v->var);
5854 t->array.size = argc;
5855 t->prepare_type(c, t, 0);
5856 array_init(v->var->type, vl);
5857 for (i = 0; i < argc; i++) {
5858 struct value *vl2 = vl->array + i * v->var->type->array.member->size;
5860 arg.str.txt = argv[i];
5861 arg.str.len = strlen(argv[i]);
5862 free_value(Tstr, vl2);
5863 dup_value(Tstr, &arg, vl2);
5867 al = cast(binode, al->right);
5869 v = interp_exec(c, progp->function, &vtype);
5870 free_value(vtype, &v);
5875 ###### ast functions
5876 void free_variable(struct variable *v)
5880 ## And now to test it out.
5882 Having a language requires having a "hello world" program. I'll
5883 provide a little more than that: a program that prints "Hello world"
5884 finds the GCD of two numbers, prints the first few elements of
5885 Fibonacci, performs a binary search for a number, and a few other
5886 things which will likely grow as the languages grows.
5888 ###### File: oceani.mk
5891 @echo "===== DEMO ====="
5892 ./oceani --section "demo: hello" oceani.mdc 55 33
5898 four ::= 2 + 2 ; five ::= 10/2
5899 const pie ::= "I like Pie";
5900 cake ::= "The cake is"
5908 func main(argv:[]string)
5909 print "Hello World, what lovely oceans you have!"
5910 print "Are there", five, "?"
5911 print pi, pie, "but", cake
5913 A := $argv[1]; B := $argv[2]
5915 /* When a variable is defined in both branches of an 'if',
5916 * and used afterwards, the variables are merged.
5922 print "Is", A, "bigger than", B,"? ", bigger
5923 /* If a variable is not used after the 'if', no
5924 * merge happens, so types can be different
5927 double:string = "yes"
5928 print A, "is more than twice", B, "?", double
5931 print "double", B, "is", double
5942 print "GCD of", A, "and", B,"is", a
5944 print a, "is not positive, cannot calculate GCD"
5946 print b, "is not positive, cannot calculate GCD"
5951 print "Fibonacci:", f1,f2,
5952 then togo = togo - 1
5960 /* Binary search... */
5965 mid := (lo + hi) / 2
5978 print "Yay, I found", target
5980 print "Closest I found was", lo
5985 // "middle square" PRNG. Not particularly good, but one my
5986 // Dad taught me - the first one I ever heard of.
5987 for i:=1; then i = i + 1; while i < size:
5988 n := list[i-1] * list[i-1]
5989 list[i] = (n / 100) % 10 000
5991 print "Before sort:",
5992 for i:=0; then i = i + 1; while i < size:
5996 for i := 1; then i=i+1; while i < size:
5997 for j:=i-1; then j=j-1; while j >= 0:
5998 if list[j] > list[j+1]:
6002 print " After sort:",
6003 for i:=0; then i = i + 1; while i < size:
6007 if 1 == 2 then print "yes"; else print "no"
6011 bob.alive = (bob.name == "Hello")
6012 print "bob", "is" if bob.alive else "isn't", "alive"