1 # Ocean Interpreter - Jamison Creek version
3 Ocean is intended to be a compiled language, so this interpreter is
4 not targeted at being the final product. It is, rather, an intermediate
5 stage and fills that role in two distinct ways.
7 Firstly, it exists as a platform to experiment with the early language
8 design. An interpreter is easy to write and easy to get working, so
9 the barrier for entry is lower if I aim to start with an interpreter.
11 Secondly, the plan for the Ocean compiler is to write it in the
12 [Ocean language](http://ocean-lang.org). To achieve this we naturally
13 need some sort of boot-strap process and this interpreter - written in
14 portable C - will fill that role. It will be used to bootstrap the
17 Two features that are not needed to fill either of these roles are
18 performance and completeness. The interpreter only needs to be fast
19 enough to run small test programs and occasionally to run the compiler
20 on itself. It only needs to be complete enough to test aspects of the
21 design which are developed before the compiler is working, and to run
22 the compiler on itself. Any features not used by the compiler when
23 compiling itself are superfluous. They may be included anyway, but
26 Nonetheless, the interpreter should end up being reasonably complete,
27 and any performance bottlenecks which appear and are easily fixed, will
32 This third version of the interpreter exists to test out some initial
33 ideas relating to types. Particularly it adds arrays (indexed from
34 zero) and simple structures. Basic control flow and variable scoping
35 are already fairly well established, as are basic numerical and
38 Some operators that have only recently been added, and so have not
39 generated all that much experience yet are "and then" and "or else" as
40 short-circuit Boolean operators, and the "if ... else" trinary
41 operator which can select between two expressions based on a third
42 (which appears syntactically in the middle).
44 The "func" clause currently only allows a "main" function to be
45 declared. That will be extended when proper function support is added.
47 An element that is present purely to make a usable language, and
48 without any expectation that they will remain, is the "print" statement
49 which performs simple output.
51 The current scalar types are "number", "Boolean", and "string".
52 Boolean will likely stay in its current form, the other two might, but
53 could just as easily be changed.
57 Versions of the interpreter which obviously do not support a complete
58 language will be named after creeks and streams. This one is Jamison
61 Once we have something reasonably resembling a complete language, the
62 names of rivers will be used.
63 Early versions of the compiler will be named after seas. Major
64 releases of the compiler will be named after oceans. Hopefully I will
65 be finished once I get to the Pacific Ocean release.
69 As well as parsing and executing a program, the interpreter can print
70 out the program from the parsed internal structure. This is useful
71 for validating the parsing.
72 So the main requirements of the interpreter are:
74 - Parse the program, possibly with tracing,
75 - Analyse the parsed program to ensure consistency,
77 - Execute the "main" function in the program, if no parsing or
78 consistency errors were found.
80 This is all performed by a single C program extracted with
83 There will be two formats for printing the program: a default and one
84 that uses bracketing. So a `--bracket` command line option is needed
85 for that. Normally the first code section found is used, however an
86 alternate section can be requested so that a file (such as this one)
87 can contain multiple programs. This is effected with the `--section`
90 This code must be compiled with `-fplan9-extensions` so that anonymous
91 structures can be used.
93 ###### File: oceani.mk
95 myCFLAGS := -Wall -g -fplan9-extensions
96 CFLAGS := $(filter-out $(myCFLAGS),$(CFLAGS)) $(myCFLAGS)
97 myLDLIBS:= libparser.o libscanner.o libmdcode.o -licuuc
98 LDLIBS := $(filter-out $(myLDLIBS),$(LDLIBS)) $(myLDLIBS)
100 all :: $(LDLIBS) oceani
101 oceani.c oceani.h : oceani.mdc parsergen
102 ./parsergen -o oceani --LALR --tag Parser oceani.mdc
103 oceani.mk: oceani.mdc md2c
106 oceani: oceani.o $(LDLIBS)
107 $(CC) $(CFLAGS) -o oceani oceani.o $(LDLIBS)
109 ###### Parser: header
111 struct parse_context;
113 struct parse_context {
114 struct token_config config;
122 #define container_of(ptr, type, member) ({ \
123 const typeof( ((type *)0)->member ) *__mptr = (ptr); \
124 (type *)( (char *)__mptr - offsetof(type,member) );})
126 #define config2context(_conf) container_of(_conf, struct parse_context, \
129 ###### Parser: reduce
130 struct parse_context *c = config2context(config);
138 #include <sys/mman.h>
157 static char Usage[] =
158 "Usage: oceani --trace --print --noexec --brackets --section=SectionName prog.ocn\n";
159 static const struct option long_options[] = {
160 {"trace", 0, NULL, 't'},
161 {"print", 0, NULL, 'p'},
162 {"noexec", 0, NULL, 'n'},
163 {"brackets", 0, NULL, 'b'},
164 {"section", 1, NULL, 's'},
167 const char *options = "tpnbs";
169 static void pr_err(char *msg) // NOTEST
171 fprintf(stderr, "%s\n", msg); // NOTEST
174 int main(int argc, char *argv[])
179 struct section *s = NULL, *ss;
180 char *section = NULL;
181 struct parse_context context = {
183 .ignored = (1 << TK_mark),
184 .number_chars = ".,_+- ",
189 int doprint=0, dotrace=0, doexec=1, brackets=0;
191 while ((opt = getopt_long(argc, argv, options, long_options, NULL))
194 case 't': dotrace=1; break;
195 case 'p': doprint=1; break;
196 case 'n': doexec=0; break;
197 case 'b': brackets=1; break;
198 case 's': section = optarg; break;
199 default: fprintf(stderr, Usage);
203 if (optind >= argc) {
204 fprintf(stderr, "oceani: no input file given\n");
207 fd = open(argv[optind], O_RDONLY);
209 fprintf(stderr, "oceani: cannot open %s\n", argv[optind]);
212 context.file_name = argv[optind];
213 len = lseek(fd, 0, 2);
214 file = mmap(NULL, len, PROT_READ, MAP_SHARED, fd, 0);
215 s = code_extract(file, file+len, pr_err);
217 fprintf(stderr, "oceani: could not find any code in %s\n",
222 ## context initialization
225 for (ss = s; ss; ss = ss->next) {
226 struct text sec = ss->section;
227 if (sec.len == strlen(section) &&
228 strncmp(sec.txt, section, sec.len) == 0)
232 fprintf(stderr, "oceani: cannot find section %s\n",
239 fprintf(stderr, "oceani: no code found in requested section\n"); // NOTEST
240 goto cleanup; // NOTEST
243 parse_oceani(ss->code, &context.config, dotrace ? stderr : NULL);
245 resolve_consts(&context);
246 prepare_types(&context);
247 if (!context.parse_error && !analyse_funcs(&context)) {
248 fprintf(stderr, "oceani: type error in program - not running.\n");
249 context.parse_error += 1;
257 if (doexec && !context.parse_error)
258 interp_main(&context, argc - optind, argv + optind);
261 struct section *t = s->next;
266 // FIXME parser should pop scope even on error
267 while (context.scope_depth > 0)
271 ## free context types
272 ## free context storage
273 exit(context.parse_error ? 1 : 0);
278 The four requirements of parse, analyse, print, interpret apply to
279 each language element individually so that is how most of the code
282 Three of the four are fairly self explanatory. The one that requires
283 a little explanation is the analysis step.
285 The current language design does not require the types of variables to
286 be declared, but they must still have a single type. Different
287 operations impose different requirements on the variables, for example
288 addition requires both arguments to be numeric, and assignment
289 requires the variable on the left to have the same type as the
290 expression on the right.
292 Analysis involves propagating these type requirements around and
293 consequently setting the type of each variable. If any requirements
294 are violated (e.g. a string is compared with a number) or if a
295 variable needs to have two different types, then an error is raised
296 and the program will not run.
298 If the same variable is declared in both branchs of an 'if/else', or
299 in all cases of a 'switch' then the multiple instances may be merged
300 into just one variable if the variable is referenced after the
301 conditional statement. When this happens, the types must naturally be
302 consistent across all the branches. When the variable is not used
303 outside the if, the variables in the different branches are distinct
304 and can be of different types.
306 Undeclared names may only appear in "use" statements and "case" expressions.
307 These names are given a type of "label" and a unique value.
308 This allows them to fill the role of a name in an enumerated type, which
309 is useful for testing the `switch` statement.
311 As we will see, the condition part of a `while` statement can return
312 either a Boolean or some other type. This requires that the expected
313 type that gets passed around comprises a type and a flag to indicate
314 that `Tbool` is also permitted.
316 As there are, as yet, no distinct types that are compatible, there
317 isn't much subtlety in the analysis. When we have distinct number
318 types, this will become more interesting.
322 When analysis discovers an inconsistency it needs to report an error;
323 just refusing to run the code ensures that the error doesn't cascade,
324 but by itself it isn't very useful. A clear understanding of the sort
325 of error message that are useful will help guide the process of
328 At a simplistic level, the only sort of error that type analysis can
329 report is that the type of some construct doesn't match a contextual
330 requirement. For example, in `4 + "hello"` the addition provides a
331 contextual requirement for numbers, but `"hello"` is not a number. In
332 this particular example no further information is needed as the types
333 are obvious from local information. When a variable is involved that
334 isn't the case. It may be helpful to explain why the variable has a
335 particular type, by indicating the location where the type was set,
336 whether by declaration or usage.
338 Using a recursive-descent analysis we can easily detect a problem at
339 multiple locations. In "`hello:= "there"; 4 + hello`" the addition
340 will detect that one argument is not a number and the usage of `hello`
341 will detect that a number was wanted, but not provided. In this
342 (early) version of the language, we will generate error reports at
343 multiple locations, so the use of `hello` will report an error and
344 explain were the value was set, and the addition will report an error
345 and say why numbers are needed. To be able to report locations for
346 errors, each language element will need to record a file location
347 (line and column) and each variable will need to record the language
348 element where its type was set. For now we will assume that each line
349 of an error message indicates one location in the file, and up to 2
350 types. So we provide a `printf`-like function which takes a format, a
351 location (a `struct exec` which has not yet been introduced), and 2
352 types. "`%1`" reports the first type, "`%2`" reports the second. We
353 will need a function to print the location, once we know how that is
354 stored. e As will be explained later, there are sometimes extra rules for
355 type matching and they might affect error messages, we need to pass those
358 As well as type errors, we sometimes need to report problems with
359 tokens, which might be unexpected or might name a type that has not
360 been defined. For these we have `tok_err()` which reports an error
361 with a given token. Each of the error functions sets the flag in the
362 context so indicate that parsing failed.
366 static void fput_loc(struct exec *loc, FILE *f);
367 static void type_err(struct parse_context *c,
368 char *fmt, struct exec *loc,
369 struct type *t1, int rules, struct type *t2);
370 static void tok_err(struct parse_context *c, char *fmt, struct token *t);
372 ###### core functions
374 static void type_err(struct parse_context *c,
375 char *fmt, struct exec *loc,
376 struct type *t1, int rules, struct type *t2)
378 fprintf(stderr, "%s:", c->file_name);
379 fput_loc(loc, stderr);
380 for (; *fmt ; fmt++) {
387 case '%': fputc(*fmt, stderr); break; // NOTEST
388 default: fputc('?', stderr); break; // NOTEST
390 type_print(t1, stderr);
393 type_print(t2, stderr);
402 static void tok_err(struct parse_context *c, char *fmt, struct token *t)
404 fprintf(stderr, "%s:%d:%d: %s: %.*s\n", c->file_name, t->line, t->col, fmt,
405 t->txt.len, t->txt.txt);
409 ## Entities: declared and predeclared.
411 There are various "things" that the language and/or the interpreter
412 needs to know about to parse and execute a program. These include
413 types, variables, values, and executable code. These are all lumped
414 together under the term "entities" (calling them "objects" would be
415 confusing) and introduced here. The following section will present the
416 different specific code elements which comprise or manipulate these
421 Executables can be lots of different things. In many cases an
422 executable is just an operation combined with one or two other
423 executables. This allows for expressions and lists etc. Other times an
424 executable is something quite specific like a constant or variable name.
425 So we define a `struct exec` to be a general executable with a type, and
426 a `struct binode` which is a subclass of `exec`, forms a node in a
427 binary tree, and holds an operation. There will be other subclasses,
428 and to access these we need to be able to `cast` the `exec` into the
429 various other types. The first field in any `struct exec` is the type
430 from the `exec_types` enum.
433 #define cast(structname, pointer) ({ \
434 const typeof( ((struct structname *)0)->type) *__mptr = &(pointer)->type; \
435 if (__mptr && *__mptr != X##structname) abort(); \
436 (struct structname *)( (char *)__mptr);})
438 #define new(structname) ({ \
439 struct structname *__ptr = ((struct structname *)calloc(1,sizeof(struct structname))); \
440 __ptr->type = X##structname; \
441 __ptr->line = -1; __ptr->column = -1; \
444 #define new_pos(structname, token) ({ \
445 struct structname *__ptr = ((struct structname *)calloc(1,sizeof(struct structname))); \
446 __ptr->type = X##structname; \
447 __ptr->line = token.line; __ptr->column = token.col; \
456 enum exec_types type;
465 struct exec *left, *right;
470 static int __fput_loc(struct exec *loc, FILE *f)
474 if (loc->line >= 0) {
475 fprintf(f, "%d:%d: ", loc->line, loc->column);
478 if (loc->type == Xbinode)
479 return __fput_loc(cast(binode,loc)->left, f) ||
480 __fput_loc(cast(binode,loc)->right, f); // NOTEST
483 static void fput_loc(struct exec *loc, FILE *f)
485 if (!__fput_loc(loc, f))
486 fprintf(f, "??:??: ");
489 Each different type of `exec` node needs a number of functions defined,
490 a bit like methods. We must be able to free it, print it, analyse it
491 and execute it. Once we have specific `exec` types we will need to
492 parse them too. Let's take this a bit more slowly.
496 The parser generator requires a `free_foo` function for each struct
497 that stores attributes and they will often be `exec`s and subtypes
498 there-of. So we need `free_exec` which can handle all the subtypes,
499 and we need `free_binode`.
503 static void free_binode(struct binode *b)
512 ###### core functions
513 static void free_exec(struct exec *e)
524 static void free_exec(struct exec *e);
526 ###### free exec cases
527 case Xbinode: free_binode(cast(binode, e)); break;
531 Printing an `exec` requires that we know the current indent level for
532 printing line-oriented components. As will become clear later, we
533 also want to know what sort of bracketing to use.
537 static void do_indent(int i, char *str)
544 ###### core functions
545 static void print_binode(struct binode *b, int indent, int bracket)
549 ## print binode cases
553 static void print_exec(struct exec *e, int indent, int bracket)
559 print_binode(cast(binode, e), indent, bracket); break;
564 do_indent(indent, "/* FREE");
565 for (v = e->to_free; v; v = v->next_free) {
566 printf(" %.*s", v->name->name.len, v->name->name.txt);
567 printf("[%d,%d]", v->scope_start, v->scope_end);
568 if (v->frame_pos >= 0)
569 printf("(%d+%d)", v->frame_pos,
570 v->type ? v->type->size:0);
578 static void print_exec(struct exec *e, int indent, int bracket);
582 As discussed, analysis involves propagating type requirements around the
583 program and looking for errors.
585 So `propagate_types` is passed an expected type (being a `struct type`
586 pointer together with some `val_rules` flags) that the `exec` is
587 expected to return, and returns the type that it does return, either of
588 which can be `NULL` signifying "unknown". A `prop_err` flag set is
589 passed by reference. It has `Efail` set when an error is found, and
590 `Eretry` when the type for some element is set via propagation. If
591 any expression cannot be evaluated immediately, `Enoconst` is set.
592 If the expression can be copied, `Emaycopy` is set.
594 If it remains unchanged at `0`, then no more propagation is needed.
598 enum val_rules {Rnolabel = 1<<0, Rboolok = 1<<1, Rnoconstant = 1<<2};
599 enum prop_err {Efail = 1<<0, Eretry = 1<<1, Enoconst = 1<<2,
604 if (rules & Rnolabel)
605 fputs(" (labels not permitted)", stderr);
609 static struct type *propagate_types(struct exec *prog, struct parse_context *c, enum prop_err *perr,
610 struct type *type, int rules);
611 ###### core functions
613 static struct type *__propagate_types(struct exec *prog, struct parse_context *c, enum prop_err *perr,
614 struct type *type, int rules)
621 switch (prog->type) {
624 struct binode *b = cast(binode, prog);
626 ## propagate binode cases
630 ## propagate exec cases
635 static struct type *propagate_types(struct exec *prog, struct parse_context *c, enum prop_err *perr,
636 struct type *type, int rules)
638 int pre_err = c->parse_error;
639 struct type *ret = __propagate_types(prog, c, perr, type, rules);
641 if (c->parse_error > pre_err)
648 Interpreting an `exec` doesn't require anything but the `exec`. State
649 is stored in variables and each variable will be directly linked from
650 within the `exec` tree. The exception to this is the `main` function
651 which needs to look at command line arguments. This function will be
652 interpreted separately.
654 Each `exec` can return a value combined with a type in `struct lrval`.
655 The type may be `Tnone` but must be non-NULL. Some `exec`s will return
656 the location of a value, which can be updated, in `lval`. Others will
657 set `lval` to NULL indicating that there is a value of appropriate type
661 static struct value interp_exec(struct parse_context *c, struct exec *e,
662 struct type **typeret);
663 ###### core functions
667 struct value rval, *lval;
670 /* If dest is passed, dtype must give the expected type, and
671 * result can go there, in which case type is returned as NULL.
673 static struct lrval _interp_exec(struct parse_context *c, struct exec *e,
674 struct value *dest, struct type *dtype);
676 static struct value interp_exec(struct parse_context *c, struct exec *e,
677 struct type **typeret)
679 struct lrval ret = _interp_exec(c, e, NULL, NULL);
681 if (!ret.type) abort();
685 dup_value(ret.type, ret.lval, &ret.rval);
689 static struct value *linterp_exec(struct parse_context *c, struct exec *e,
690 struct type **typeret)
692 struct lrval ret = _interp_exec(c, e, NULL, NULL);
694 if (!ret.type) abort();
698 free_value(ret.type, &ret.rval);
702 /* dinterp_exec is used when the destination type is certain and
703 * the value has a place to go.
705 static void dinterp_exec(struct parse_context *c, struct exec *e,
706 struct value *dest, struct type *dtype,
709 struct lrval ret = _interp_exec(c, e, dest, dtype);
713 free_value(dtype, dest);
715 dup_value(dtype, ret.lval, dest);
717 memcpy(dest, &ret.rval, dtype->size);
720 static struct lrval _interp_exec(struct parse_context *c, struct exec *e,
721 struct value *dest, struct type *dtype)
723 /* If the result is copied to dest, ret.type is set to NULL */
725 struct value rv = {}, *lrv = NULL;
728 rvtype = ret.type = Tnone;
738 struct binode *b = cast(binode, e);
739 struct value left, right, *lleft;
740 struct type *ltype, *rtype;
741 ltype = rtype = Tnone;
743 ## interp binode cases
745 free_value(ltype, &left);
746 free_value(rtype, &right);
756 ## interp exec cleanup
762 Values come in a wide range of types, with more likely to be added.
763 Each type needs to be able to print its own values (for convenience at
764 least) as well as to compare two values, at least for equality and
765 possibly for order. For now, values might need to be duplicated and
766 freed, though eventually such manipulations will be better integrated
769 Rather than requiring every numeric type to support all numeric
770 operations (add, multiply, etc), we allow types to be able to present
771 as one of a few standard types: integer, float, and fraction. The
772 existence of these conversion functions eventually enable types to
773 determine if they are compatible with other types, though such types
774 have not yet been implemented.
776 Named type are stored in a simple linked list. Objects of each type are
777 "values" which are often passed around by value.
779 There are both explicitly named types, and anonymous types. Anonymous
780 cannot be accessed by name, but are used internally and have a name
781 which might be reported in error messages.
788 ## value union fields
795 struct token first_use;
798 void (*init)(struct type *type, struct value *val);
799 int (*prepare_type)(struct parse_context *c, struct type *type, int parse_time);
800 void (*print)(struct type *type, struct value *val, FILE *f);
801 void (*print_type)(struct type *type, FILE *f);
802 int (*cmp_order)(struct type *t1, struct type *t2,
803 struct value *v1, struct value *v2);
804 int (*cmp_eq)(struct type *t1, struct type *t2,
805 struct value *v1, struct value *v2);
806 void (*dup)(struct type *type, struct value *vold, struct value *vnew);
807 int (*test)(struct type *type, struct value *val);
808 void (*free)(struct type *type, struct value *val);
809 void (*free_type)(struct type *t);
810 long long (*to_int)(struct value *v);
811 double (*to_float)(struct value *v);
812 int (*to_mpq)(mpq_t *q, struct value *v);
821 struct type *typelist;
828 static struct type *find_type(struct parse_context *c, struct text s)
830 struct type *t = c->typelist;
832 while (t && (t->anon ||
833 text_cmp(t->name, s) != 0))
838 static struct type *_add_type(struct parse_context *c, struct text s,
839 struct type *proto, int anon)
843 n = calloc(1, sizeof(*n));
850 n->next = c->typelist;
855 static struct type *add_type(struct parse_context *c, struct text s,
858 return _add_type(c, s, proto, 0);
861 static struct type *add_anon_type(struct parse_context *c,
862 struct type *proto, char *name, ...)
868 vasprintf(&t.txt, name, ap);
870 t.len = strlen(name);
871 return _add_type(c, t, proto, 1);
874 static void free_type(struct type *t)
876 /* The type is always a reference to something in the
877 * context, so we don't need to free anything.
881 static void free_value(struct type *type, struct value *v)
885 memset(v, 0x5a, type->size);
889 static void type_print(struct type *type, FILE *f)
892 fputs("*unknown*type*", f); // NOTEST
893 else if (type->name.len && !type->anon)
894 fprintf(f, "%.*s", type->name.len, type->name.txt);
895 else if (type->print_type)
896 type->print_type(type, f);
898 fputs("*invalid*type*", f);
901 static void val_init(struct type *type, struct value *val)
903 if (type && type->init)
904 type->init(type, val);
907 static void dup_value(struct type *type,
908 struct value *vold, struct value *vnew)
910 if (type && type->dup)
911 type->dup(type, vold, vnew);
914 static int value_cmp(struct type *tl, struct type *tr,
915 struct value *left, struct value *right)
917 if (tl && tl->cmp_order)
918 return tl->cmp_order(tl, tr, left, right);
919 if (tl && tl->cmp_eq) // NOTEST
920 return tl->cmp_eq(tl, tr, left, right); // NOTEST
924 static void print_value(struct type *type, struct value *v, FILE *f)
926 if (type && type->print)
927 type->print(type, v, f);
929 fprintf(f, "*Unknown*"); // NOTEST
932 static void prepare_types(struct parse_context *c)
936 enum { none, some, cannot } progress = none;
941 for (t = c->typelist; t; t = t->next) {
943 tok_err(c, "error: type used but not declared",
945 if (t->size == 0 && t->prepare_type) {
946 if (t->prepare_type(c, t, 1))
948 else if (progress == cannot)
949 tok_err(c, "error: type has recursive definition",
959 progress = cannot; break;
961 progress = none; break;
968 static void free_value(struct type *type, struct value *v);
969 static int type_compat(struct type *require, struct type *have, int rules);
970 static void type_print(struct type *type, FILE *f);
971 static void val_init(struct type *type, struct value *v);
972 static void dup_value(struct type *type,
973 struct value *vold, struct value *vnew);
974 static int value_cmp(struct type *tl, struct type *tr,
975 struct value *left, struct value *right);
976 static void print_value(struct type *type, struct value *v, FILE *f);
978 ###### free context types
980 while (context.typelist) {
981 struct type *t = context.typelist;
983 context.typelist = t->next;
991 Type can be specified for local variables, for fields in a structure,
992 for formal parameters to functions, and possibly elsewhere. Different
993 rules may apply in different contexts. As a minimum, a named type may
994 always be used. Currently the type of a formal parameter can be
995 different from types in other contexts, so we have a separate grammar
1001 Type -> IDENTIFIER ${
1002 $0 = find_type(c, $ID.txt);
1004 $0 = add_type(c, $ID.txt, NULL);
1005 $0->first_use = $ID;
1010 FormalType -> Type ${ $0 = $<1; }$
1011 ## formal type grammar
1015 Values of the base types can be numbers, which we represent as
1016 multi-precision fractions, strings, Booleans and labels. When
1017 analysing the program we also need to allow for places where no value
1018 is meaningful (type `Tnone`) and where we don't know what type to
1019 expect yet (type is `NULL`).
1021 Values are never shared, they are always copied when used, and freed
1022 when no longer needed.
1024 When propagating type information around the program, we need to
1025 determine if two types are compatible, where type `NULL` is compatible
1026 with anything. There are two special cases with type compatibility,
1027 both related to the Conditional Statement which will be described
1028 later. In some cases a Boolean can be accepted as well as some other
1029 primary type, and in others any type is acceptable except a label (`Vlabel`).
1030 A separate function encoding these cases will simplify some code later.
1032 ###### type functions
1034 int (*compat)(struct type *this, struct type *other);
1036 ###### ast functions
1038 static int type_compat(struct type *require, struct type *have, int rules)
1040 if ((rules & Rboolok) && have == Tbool)
1042 if ((rules & Rnolabel) && have == Tlabel)
1044 if (!require || !have)
1047 if (require->compat)
1048 return require->compat(require, have);
1050 return require == have;
1055 #include "parse_string.h"
1056 #include "parse_number.h"
1059 myLDLIBS := libnumber.o libstring.o -lgmp
1060 LDLIBS := $(filter-out $(myLDLIBS),$(LDLIBS)) $(myLDLIBS)
1062 ###### type union fields
1063 enum vtype {Vnone, Vstr, Vnum, Vbool, Vlabel} vtype;
1065 ###### value union fields
1071 ###### ast functions
1072 static void _free_value(struct type *type, struct value *v)
1076 switch (type->vtype) {
1078 case Vstr: free(v->str.txt); break;
1079 case Vnum: mpq_clear(v->num); break;
1085 ###### value functions
1087 static void _val_init(struct type *type, struct value *val)
1089 switch(type->vtype) {
1090 case Vnone: // NOTEST
1093 mpq_init(val->num); break;
1095 val->str.txt = malloc(1);
1107 static void _dup_value(struct type *type,
1108 struct value *vold, struct value *vnew)
1110 switch (type->vtype) {
1111 case Vnone: // NOTEST
1114 vnew->label = vold->label;
1117 vnew->bool = vold->bool;
1120 mpq_init(vnew->num);
1121 mpq_set(vnew->num, vold->num);
1124 vnew->str.len = vold->str.len;
1125 vnew->str.txt = malloc(vnew->str.len);
1126 memcpy(vnew->str.txt, vold->str.txt, vnew->str.len);
1131 static int _value_cmp(struct type *tl, struct type *tr,
1132 struct value *left, struct value *right)
1136 return tl - tr; // NOTEST
1137 switch (tl->vtype) {
1138 case Vlabel: cmp = left->label == right->label ? 0 : 1; break;
1139 case Vnum: cmp = mpq_cmp(left->num, right->num); break;
1140 case Vstr: cmp = text_cmp(left->str, right->str); break;
1141 case Vbool: cmp = left->bool - right->bool; break;
1142 case Vnone: cmp = 0; // NOTEST
1147 static void _print_value(struct type *type, struct value *v, FILE *f)
1149 switch (type->vtype) {
1150 case Vnone: // NOTEST
1151 fprintf(f, "*no-value*"); break; // NOTEST
1152 case Vlabel: // NOTEST
1153 fprintf(f, "*label-%p*", v->label); break; // NOTEST
1155 fprintf(f, "%.*s", v->str.len, v->str.txt); break;
1157 fprintf(f, "%s", v->bool ? "True":"False"); break;
1162 mpf_set_q(fl, v->num);
1163 gmp_fprintf(f, "%.10Fg", fl);
1170 static void _free_value(struct type *type, struct value *v);
1172 static int bool_test(struct type *type, struct value *v)
1177 static struct type base_prototype = {
1179 .print = _print_value,
1180 .cmp_order = _value_cmp,
1181 .cmp_eq = _value_cmp,
1183 .free = _free_value,
1186 static struct type *Tbool, *Tstr, *Tnum, *Tnone, *Tlabel;
1188 ###### ast functions
1189 static struct type *add_base_type(struct parse_context *c, char *n,
1190 enum vtype vt, int size)
1192 struct text txt = { n, strlen(n) };
1195 t = add_type(c, txt, &base_prototype);
1198 t->align = size > sizeof(void*) ? sizeof(void*) : size;
1199 if (t->size & (t->align - 1))
1200 t->size = (t->size | (t->align - 1)) + 1; // NOTEST
1204 ###### context initialization
1206 Tbool = add_base_type(&context, "Boolean", Vbool, sizeof(char));
1207 Tbool->test = bool_test;
1208 Tstr = add_base_type(&context, "string", Vstr, sizeof(struct text));
1209 Tnum = add_base_type(&context, "number", Vnum, sizeof(mpq_t));
1210 Tnone = add_base_type(&context, "none", Vnone, 0);
1211 Tlabel = add_base_type(&context, "label", Vlabel, sizeof(void*));
1215 We have already met values as separate objects. When manifest constants
1216 appear in the program text, that must result in an executable which has
1217 a constant value. So the `val` structure embeds a value in an
1230 ###### ast functions
1231 struct val *new_val(struct type *T, struct token tk)
1233 struct val *v = new_pos(val, tk);
1244 $0 = new_val(Tbool, $1);
1248 $0 = new_val(Tbool, $1);
1253 $0 = new_val(Tnum, $1);
1254 if (number_parse($0->val.num, tail, $1.txt) == 0)
1255 mpq_init($0->val.num); // UNTESTED
1257 tok_err(c, "error: unsupported number suffix",
1262 $0 = new_val(Tstr, $1);
1263 string_parse(&$1, '\\', &$0->val.str, tail);
1265 tok_err(c, "error: unsupported string suffix",
1270 $0 = new_val(Tstr, $1);
1271 string_parse(&$1, '\\', &$0->val.str, tail);
1273 tok_err(c, "error: unsupported string suffix",
1277 ###### print exec cases
1280 struct val *v = cast(val, e);
1281 if (v->vtype == Tstr)
1283 // FIXME how to ensure numbers have same precision.
1284 print_value(v->vtype, &v->val, stdout);
1285 if (v->vtype == Tstr)
1290 ###### propagate exec cases
1293 struct val *val = cast(val, prog);
1294 if (!type_compat(type, val->vtype, rules))
1295 type_err(c, "error: expected %1%r found %2",
1296 prog, type, rules, val->vtype);
1300 ###### interp exec cases
1302 rvtype = cast(val, e)->vtype;
1303 dup_value(rvtype, &cast(val, e)->val, &rv);
1306 ###### ast functions
1307 static void free_val(struct val *v)
1310 free_value(v->vtype, &v->val);
1314 ###### free exec cases
1315 case Xval: free_val(cast(val, e)); break;
1317 ###### ast functions
1318 // Move all nodes from 'b' to 'rv', reversing their order.
1319 // In 'b' 'left' is a list, and 'right' is the last node.
1320 // In 'rv', left' is the first node and 'right' is a list.
1321 static struct binode *reorder_bilist(struct binode *b)
1323 struct binode *rv = NULL;
1326 struct exec *t = b->right;
1330 b = cast(binode, b->left);
1340 Variables are scoped named values. We store the names in a linked list
1341 of "bindings" sorted in lexical order, and use sequential search and
1348 struct binding *next; // in lexical order
1352 This linked list is stored in the parse context so that "reduce"
1353 functions can find or add variables, and so the analysis phase can
1354 ensure that every variable gets a type.
1356 ###### parse context
1358 struct binding *varlist; // In lexical order
1360 ###### ast functions
1362 static struct binding *find_binding(struct parse_context *c, struct text s)
1364 struct binding **l = &c->varlist;
1369 (cmp = text_cmp((*l)->name, s)) < 0)
1373 n = calloc(1, sizeof(*n));
1380 Each name can be linked to multiple variables defined in different
1381 scopes. Each scope starts where the name is declared and continues
1382 until the end of the containing code block. Scopes of a given name
1383 cannot nest, so a declaration while a name is in-scope is an error.
1385 ###### binding fields
1386 struct variable *var;
1390 struct variable *previous;
1392 struct binding *name;
1393 struct exec *where_decl;// where name was declared
1394 struct exec *where_set; // where type was set
1398 When a scope closes, the values of the variables might need to be freed.
1399 This happens in the context of some `struct exec` and each `exec` will
1400 need to know which variables need to be freed when it completes.
1403 struct variable *to_free;
1405 ####### variable fields
1406 struct exec *cleanup_exec;
1407 struct variable *next_free;
1409 ####### interp exec cleanup
1412 for (v = e->to_free; v; v = v->next_free) {
1413 struct value *val = var_value(c, v);
1414 free_value(v->type, val);
1418 ###### ast functions
1419 static void variable_unlink_exec(struct variable *v)
1421 struct variable **vp;
1422 if (!v->cleanup_exec)
1424 for (vp = &v->cleanup_exec->to_free;
1425 *vp; vp = &(*vp)->next_free) {
1429 v->cleanup_exec = NULL;
1434 While the naming seems strange, we include local constants in the
1435 definition of variables. A name declared `var := value` can
1436 subsequently be changed, but a name declared `var ::= value` cannot -
1439 ###### variable fields
1442 Scopes in parallel branches can be partially merged. More
1443 specifically, if a given name is declared in both branches of an
1444 if/else then its scope is a candidate for merging. Similarly if
1445 every branch of an exhaustive switch (e.g. has an "else" clause)
1446 declares a given name, then the scopes from the branches are
1447 candidates for merging.
1449 Note that names declared inside a loop (which is only parallel to
1450 itself) are never visible after the loop. Similarly names defined in
1451 scopes which are not parallel, such as those started by `for` and
1452 `switch`, are never visible after the scope. Only variables defined in
1453 both `then` and `else` (including the implicit then after an `if`, and
1454 excluding `then` used with `for`) and in all `case`s and `else` of a
1455 `switch` or `while` can be visible beyond the `if`/`switch`/`while`.
1457 Labels, which are a bit like variables, follow different rules.
1458 Labels are not explicitly declared, but if an undeclared name appears
1459 in a context where a label is legal, that effectively declares the
1460 name as a label. The declaration remains in force (or in scope) at
1461 least to the end of the immediately containing block and conditionally
1462 in any larger containing block which does not declare the name in some
1463 other way. Importantly, the conditional scope extension happens even
1464 if the label is only used in one parallel branch of a conditional --
1465 when used in one branch it is treated as having been declared in all
1468 Merge candidates are tentatively visible beyond the end of the
1469 branching statement which creates them. If the name is used, the
1470 merge is affirmed and they become a single variable visible at the
1471 outer layer. If not - if it is redeclared first - the merge lapses.
1473 To track scopes we have an extra stack, implemented as a linked list,
1474 which roughly parallels the parse stack and which is used exclusively
1475 for scoping. When a new scope is opened, a new frame is pushed and
1476 the child-count of the parent frame is incremented. This child-count
1477 is used to distinguish between the first of a set of parallel scopes,
1478 in which declared variables must not be in scope, and subsequent
1479 branches, whether they may already be conditionally scoped.
1481 We need a total ordering of scopes so we can easily compare to variables
1482 to see if they are concurrently in scope. To achieve this we record a
1483 `scope_count` which is actually a count of both beginnings and endings
1484 of scopes. Then each variable has a record of the scope count where it
1485 enters scope, and where it leaves.
1487 To push a new frame *before* any code in the frame is parsed, we need a
1488 grammar reduction. This is most easily achieved with a grammar
1489 element which derives the empty string, and creates the new scope when
1490 it is recognised. This can be placed, for example, between a keyword
1491 like "if" and the code following it.
1495 struct scope *parent;
1499 ###### parse context
1502 struct scope *scope_stack;
1504 ###### variable fields
1505 int scope_start, scope_end;
1507 ###### ast functions
1508 static void scope_pop(struct parse_context *c)
1510 struct scope *s = c->scope_stack;
1512 c->scope_stack = s->parent;
1514 c->scope_depth -= 1;
1515 c->scope_count += 1;
1518 static void scope_push(struct parse_context *c)
1520 struct scope *s = calloc(1, sizeof(*s));
1522 c->scope_stack->child_count += 1;
1523 s->parent = c->scope_stack;
1525 c->scope_depth += 1;
1526 c->scope_count += 1;
1532 OpenScope -> ${ scope_push(c); }$
1534 Each variable records a scope depth and is in one of four states:
1536 - "in scope". This is the case between the declaration of the
1537 variable and the end of the containing block, and also between
1538 the usage with affirms a merge and the end of that block.
1540 The scope depth is not greater than the current parse context scope
1541 nest depth. When the block of that depth closes, the state will
1542 change. To achieve this, all "in scope" variables are linked
1543 together as a stack in nesting order.
1545 - "pending". The "in scope" block has closed, but other parallel
1546 scopes are still being processed. So far, every parallel block at
1547 the same level that has closed has declared the name.
1549 The scope depth is the depth of the last parallel block that
1550 enclosed the declaration, and that has closed.
1552 - "conditionally in scope". The "in scope" block and all parallel
1553 scopes have closed, and no further mention of the name has been seen.
1554 This state includes a secondary nest depth (`min_depth`) which records
1555 the outermost scope seen since the variable became conditionally in
1556 scope. If a use of the name is found, the variable becomes "in scope"
1557 and that secondary depth becomes the recorded scope depth. If the
1558 name is declared as a new variable, the old variable becomes "out of
1559 scope" and the recorded scope depth stays unchanged.
1561 - "out of scope". The variable is neither in scope nor conditionally
1562 in scope. It is permanently out of scope now and can be removed from
1563 the "in scope" stack. When a variable becomes out-of-scope it is
1564 moved to a separate list (`out_scope`) of variables which have fully
1565 known scope. This will be used at the end of each function to assign
1566 each variable a place in the stack frame.
1568 ###### variable fields
1569 int depth, min_depth;
1570 enum { OutScope, PendingScope, CondScope, InScope } scope;
1571 struct variable *in_scope;
1573 ###### parse context
1575 struct variable *in_scope;
1576 struct variable *out_scope;
1578 All variables with the same name are linked together using the
1579 'previous' link. Those variable that have been affirmatively merged all
1580 have a 'merged' pointer that points to one primary variable - the most
1581 recently declared instance. When merging variables, we need to also
1582 adjust the 'merged' pointer on any other variables that had previously
1583 been merged with the one that will no longer be primary.
1585 A variable that is no longer the most recent instance of a name may
1586 still have "pending" scope, if it might still be merged with most
1587 recent instance. These variables don't really belong in the
1588 "in_scope" list, but are not immediately removed when a new instance
1589 is found. Instead, they are detected and ignored when considering the
1590 list of in_scope names.
1592 The storage of the value of a variable will be described later. For now
1593 we just need to know that when a variable goes out of scope, it might
1594 need to be freed. For this we need to be able to find it, so assume that
1595 `var_value()` will provide that.
1597 ###### variable fields
1598 struct variable *merged;
1600 ###### ast functions
1602 static void variable_merge(struct variable *primary, struct variable *secondary)
1606 primary = primary->merged;
1608 for (v = primary->previous; v; v=v->previous)
1609 if (v == secondary || v == secondary->merged ||
1610 v->merged == secondary ||
1611 v->merged == secondary->merged) {
1612 v->scope = OutScope;
1613 v->merged = primary;
1614 if (v->scope_start < primary->scope_start)
1615 primary->scope_start = v->scope_start;
1616 if (v->scope_end > primary->scope_end)
1617 primary->scope_end = v->scope_end; // NOTEST
1618 variable_unlink_exec(v);
1622 ###### forward decls
1623 static struct value *var_value(struct parse_context *c, struct variable *v);
1625 ###### free global vars
1627 while (context.varlist) {
1628 struct binding *b = context.varlist;
1629 struct variable *v = b->var;
1630 context.varlist = b->next;
1633 struct variable *next = v->previous;
1635 if (v->global && v->frame_pos >= 0) {
1636 free_value(v->type, var_value(&context, v));
1637 if (v->depth == 0 && v->type->free == function_free)
1638 // This is a function constant
1639 free_exec(v->where_decl);
1646 #### Manipulating Bindings
1648 When a name is conditionally visible, a new declaration discards the old
1649 binding - the condition lapses. Similarly when we reach the end of a
1650 function (outermost non-global scope) any conditional scope must lapse.
1651 Conversely a usage of the name affirms the visibility and extends it to
1652 the end of the containing block - i.e. the block that contains both the
1653 original declaration and the latest usage. This is determined from
1654 `min_depth`. When a conditionally visible variable gets affirmed like
1655 this, it is also merged with other conditionally visible variables with
1658 When we parse a variable declaration we either report an error if the
1659 name is currently bound, or create a new variable at the current nest
1660 depth if the name is unbound or bound to a conditionally scoped or
1661 pending-scope variable. If the previous variable was conditionally
1662 scoped, it and its homonyms becomes out-of-scope.
1664 When we parse a variable reference (including non-declarative assignment
1665 "foo = bar") we report an error if the name is not bound or is bound to
1666 a pending-scope variable; update the scope if the name is bound to a
1667 conditionally scoped variable; or just proceed normally if the named
1668 variable is in scope.
1670 When we exit a scope, any variables bound at this level are either
1671 marked out of scope or pending-scoped, depending on whether the scope
1672 was sequential or parallel. Here a "parallel" scope means the "then"
1673 or "else" part of a conditional, or any "case" or "else" branch of a
1674 switch. Other scopes are "sequential".
1676 When exiting a parallel scope we check if there are any variables that
1677 were previously pending and are still visible. If there are, then
1678 they weren't redeclared in the most recent scope, so they cannot be
1679 merged and must become out-of-scope. If it is not the first of
1680 parallel scopes (based on `child_count`), we check that there was a
1681 previous binding that is still pending-scope. If there isn't, the new
1682 variable must now be out-of-scope.
1684 When exiting a sequential scope that immediately enclosed parallel
1685 scopes, we need to resolve any pending-scope variables. If there was
1686 no `else` clause, and we cannot determine that the `switch` was exhaustive,
1687 we need to mark all pending-scope variable as out-of-scope. Otherwise
1688 all pending-scope variables become conditionally scoped.
1691 enum closetype { CloseSequential, CloseFunction, CloseParallel, CloseElse };
1693 ###### ast functions
1695 static struct variable *var_decl(struct parse_context *c, struct text s)
1697 struct binding *b = find_binding(c, s);
1698 struct variable *v = b->var;
1700 switch (v ? v->scope : OutScope) {
1702 /* Caller will report the error */
1706 v && v->scope == CondScope;
1708 v->scope = OutScope;
1712 v = calloc(1, sizeof(*v));
1713 v->previous = b->var;
1717 v->min_depth = v->depth = c->scope_depth;
1719 v->in_scope = c->in_scope;
1720 v->scope_start = c->scope_count;
1726 static struct variable *var_ref(struct parse_context *c, struct text s)
1728 struct binding *b = find_binding(c, s);
1729 struct variable *v = b->var;
1730 struct variable *v2;
1732 switch (v ? v->scope : OutScope) {
1735 /* Caller will report the error */
1738 /* All CondScope variables of this name need to be merged
1739 * and become InScope
1741 v->depth = v->min_depth;
1743 for (v2 = v->previous;
1744 v2 && v2->scope == CondScope;
1746 variable_merge(v, v2);
1754 static int var_refile(struct parse_context *c, struct variable *v)
1756 /* Variable just went out of scope. Add it to the out_scope
1757 * list, sorted by ->scope_start
1759 struct variable **vp = &c->out_scope;
1760 while ((*vp) && (*vp)->scope_start < v->scope_start)
1761 vp = &(*vp)->in_scope;
1767 static void var_block_close(struct parse_context *c, enum closetype ct,
1770 /* Close off all variables that are in_scope.
1771 * Some variables in c->scope may already be not-in-scope,
1772 * such as when a PendingScope variable is hidden by a new
1773 * variable with the same name.
1774 * So we check for v->name->var != v and drop them.
1775 * If we choose to make a variable OutScope, we drop it
1778 struct variable *v, **vp, *v2;
1781 for (vp = &c->in_scope;
1782 (v = *vp) && v->min_depth > c->scope_depth;
1783 (v->scope == OutScope || v->name->var != v)
1784 ? (*vp = v->in_scope, var_refile(c, v))
1785 : ( vp = &v->in_scope, 0)) {
1786 v->min_depth = c->scope_depth;
1787 if (v->name->var != v)
1788 /* This is still in scope, but we haven't just
1792 v->min_depth = c->scope_depth;
1793 if (v->scope == InScope)
1794 v->scope_end = c->scope_count;
1795 if (v->scope == InScope && e && !v->global) {
1796 /* This variable gets cleaned up when 'e' finishes */
1797 variable_unlink_exec(v);
1798 v->cleanup_exec = e;
1799 v->next_free = e->to_free;
1804 case CloseParallel: /* handle PendingScope */
1808 if (c->scope_stack->child_count == 1)
1809 /* first among parallel branches */
1810 v->scope = PendingScope;
1811 else if (v->previous &&
1812 v->previous->scope == PendingScope)
1813 /* all previous branches used name */
1814 v->scope = PendingScope;
1815 else if (v->type == Tlabel)
1816 /* Labels remain pending even when not used */
1817 v->scope = PendingScope; // UNTESTED
1819 v->scope = OutScope;
1820 if (ct == CloseElse) {
1821 /* All Pending variables with this name
1822 * are now Conditional */
1824 v2 && v2->scope == PendingScope;
1826 v2->scope = CondScope;
1830 /* Not possible as it would require
1831 * parallel scope to be nested immediately
1832 * in a parallel scope, and that never
1836 /* Not possible as we already tested for
1843 if (v->scope == CondScope)
1844 /* Condition cannot continue past end of function */
1847 case CloseSequential:
1848 if (v->type == Tlabel)
1849 v->scope = PendingScope;
1852 v->scope = OutScope;
1855 /* There was no 'else', so we can only become
1856 * conditional if we know the cases were exhaustive,
1857 * and that doesn't mean anything yet.
1858 * So only labels become conditional..
1861 v2 && v2->scope == PendingScope;
1863 if (v2->type == Tlabel)
1864 v2->scope = CondScope;
1866 v2->scope = OutScope;
1869 case OutScope: break;
1878 The value of a variable is store separately from the variable, on an
1879 analogue of a stack frame. There are (currently) two frames that can be
1880 active. A global frame which currently only stores constants, and a
1881 stacked frame which stores local variables. Each variable knows if it
1882 is global or not, and what its index into the frame is.
1884 Values in the global frame are known immediately they are relevant, so
1885 the frame needs to be reallocated as it grows so it can store those
1886 values. The local frame doesn't get values until the interpreted phase
1887 is started, so there is no need to allocate until the size is known.
1889 We initialize the `frame_pos` to an impossible value, so that we can
1890 tell if it was set or not later.
1892 ###### variable fields
1896 ###### variable init
1899 ###### parse context
1901 short global_size, global_alloc;
1903 void *global, *local;
1905 ###### forward decls
1906 static struct value *global_alloc(struct parse_context *c, struct type *t,
1907 struct variable *v, struct value *init);
1909 ###### ast functions
1911 static struct value *var_value(struct parse_context *c, struct variable *v)
1914 if (!c->local || !v->type)
1915 return NULL; // UNTESTED
1916 if (v->frame_pos + v->type->size > c->local_size) {
1917 printf("INVALID frame_pos\n"); // NOTEST
1920 return c->local + v->frame_pos;
1922 if (c->global_size > c->global_alloc) {
1923 int old = c->global_alloc;
1924 c->global_alloc = (c->global_size | 1023) + 1024;
1925 c->global = realloc(c->global, c->global_alloc);
1926 memset(c->global + old, 0, c->global_alloc - old);
1928 return c->global + v->frame_pos;
1931 static struct value *global_alloc(struct parse_context *c, struct type *t,
1932 struct variable *v, struct value *init)
1935 struct variable scratch;
1937 if (t->prepare_type)
1938 t->prepare_type(c, t, 1); // NOTEST
1940 if (c->global_size & (t->align - 1))
1941 c->global_size = (c->global_size + t->align) & ~(t->align-1); // NOTEST
1946 v->frame_pos = c->global_size;
1948 c->global_size += v->type->size;
1949 ret = var_value(c, v);
1951 memcpy(ret, init, t->size);
1957 As global values are found -- struct field initializers, labels etc --
1958 `global_alloc()` is called to record the value in the global frame.
1960 When the program is fully parsed, each function is analysed, we need to
1961 walk the list of variables local to that function and assign them an
1962 offset in the stack frame. For this we have `scope_finalize()`.
1964 We keep the stack from dense by re-using space for between variables
1965 that are not in scope at the same time. The `out_scope` list is sorted
1966 by `scope_start` and as we process a varible, we move it to an FIFO
1967 stack. For each variable we consider, we first discard any from the
1968 stack anything that went out of scope before the new variable came in.
1969 Then we place the new variable just after the one at the top of the
1972 ###### ast functions
1974 static void scope_finalize(struct parse_context *c, struct type *ft)
1976 int size = ft->function.local_size;
1977 struct variable *next = ft->function.scope;
1978 struct variable *done = NULL;
1981 struct variable *v = next;
1982 struct type *t = v->type;
1989 if (v->frame_pos >= 0)
1991 while (done && done->scope_end < v->scope_start)
1992 done = done->in_scope;
1994 pos = done->frame_pos + done->type->size;
1996 pos = ft->function.local_size;
1997 if (pos & (t->align - 1))
1998 pos = (pos + t->align) & ~(t->align-1);
2000 if (size < pos + v->type->size)
2001 size = pos + v->type->size;
2005 c->out_scope = NULL;
2006 ft->function.local_size = size;
2009 ###### free context storage
2010 free(context.global);
2012 #### Variables as executables
2014 Just as we used a `val` to wrap a value into an `exec`, we similarly
2015 need a `var` to wrap a `variable` into an exec. While each `val`
2016 contained a copy of the value, each `var` holds a link to the variable
2017 because it really is the same variable no matter where it appears.
2018 When a variable is used, we need to remember to follow the `->merged`
2019 link to find the primary instance.
2021 When a variable is declared, it may or may not be given an explicit
2022 type. We need to record which so that we can report the parsed code
2031 struct variable *var;
2034 ###### variable fields
2042 VariableDecl -> IDENTIFIER : ${ {
2043 struct variable *v = var_decl(c, $1.txt);
2044 $0 = new_pos(var, $1);
2049 v = var_ref(c, $1.txt);
2051 type_err(c, "error: variable '%v' redeclared",
2053 type_err(c, "info: this is where '%v' was first declared",
2054 v->where_decl, NULL, 0, NULL);
2057 | IDENTIFIER :: ${ {
2058 struct variable *v = var_decl(c, $1.txt);
2059 $0 = new_pos(var, $1);
2065 v = var_ref(c, $1.txt);
2067 type_err(c, "error: variable '%v' redeclared",
2069 type_err(c, "info: this is where '%v' was first declared",
2070 v->where_decl, NULL, 0, NULL);
2073 | IDENTIFIER : Type ${ {
2074 struct variable *v = var_decl(c, $1.txt);
2075 $0 = new_pos(var, $1);
2081 v->explicit_type = 1;
2083 v = var_ref(c, $1.txt);
2085 type_err(c, "error: variable '%v' redeclared",
2087 type_err(c, "info: this is where '%v' was first declared",
2088 v->where_decl, NULL, 0, NULL);
2091 | IDENTIFIER :: Type ${ {
2092 struct variable *v = var_decl(c, $1.txt);
2093 $0 = new_pos(var, $1);
2100 v->explicit_type = 1;
2102 v = var_ref(c, $1.txt);
2104 type_err(c, "error: variable '%v' redeclared",
2106 type_err(c, "info: this is where '%v' was first declared",
2107 v->where_decl, NULL, 0, NULL);
2112 Variable -> IDENTIFIER ${ {
2113 struct variable *v = var_ref(c, $1.txt);
2114 $0 = new_pos(var, $1);
2116 /* This might be a global const or a label
2117 * Allocate a var with impossible type Tnone,
2118 * which will be adjusted when we find out what it is,
2119 * or will trigger an error.
2121 v = var_decl(c, $1.txt);
2128 cast(var, $0)->var = v;
2131 ###### print exec cases
2134 struct var *v = cast(var, e);
2136 struct binding *b = v->var->name;
2137 printf("%.*s", b->name.len, b->name.txt);
2144 if (loc && loc->type == Xvar) {
2145 struct var *v = cast(var, loc);
2147 struct binding *b = v->var->name;
2148 fprintf(stderr, "%.*s", b->name.len, b->name.txt);
2150 fputs("???", stderr); // NOTEST
2152 fputs("NOTVAR", stderr);
2155 ###### propagate exec cases
2159 struct var *var = cast(var, prog);
2160 struct variable *v = var->var;
2162 type_err(c, "%d:BUG: no variable!!", prog, NULL, 0, NULL); // NOTEST
2163 return Tnone; // NOTEST
2166 if (v->constant && (rules & Rnoconstant)) {
2167 type_err(c, "error: Cannot assign to a constant: %v",
2168 prog, NULL, 0, NULL);
2169 type_err(c, "info: name was defined as a constant here",
2170 v->where_decl, NULL, 0, NULL);
2173 if (v->type == Tnone && v->where_decl == prog)
2174 type_err(c, "error: variable used but not declared: %v",
2175 prog, NULL, 0, NULL);
2176 if (v->type == NULL) {
2177 if (type && !(*perr & Efail)) {
2179 v->where_set = prog;
2182 } else if (!type_compat(type, v->type, rules)) {
2183 type_err(c, "error: expected %1%r but variable '%v' is %2", prog,
2184 type, rules, v->type);
2185 type_err(c, "info: this is where '%v' was set to %1", v->where_set,
2186 v->type, rules, NULL);
2188 if (!v->global || v->frame_pos < 0)
2195 ###### interp exec cases
2198 struct var *var = cast(var, e);
2199 struct variable *v = var->var;
2202 lrv = var_value(c, v);
2207 ###### ast functions
2209 static void free_var(struct var *v)
2214 ###### free exec cases
2215 case Xvar: free_var(cast(var, e)); break;
2220 Now that we have the shape of the interpreter in place we can add some
2221 complex types and connected them in to the data structures and the
2222 different phases of parse, analyse, print, interpret.
2224 Being "complex" the language will naturally have syntax to access
2225 specifics of objects of these types. These will fit into the grammar as
2226 "Terms" which are the things that are combined with various operators to
2227 form "Expression". Where a Term is formed by some operation on another
2228 Term, the subordinate Term will always come first, so for example a
2229 member of an array will be expressed as the Term for the array followed
2230 by an index in square brackets. The strict rule of using postfix
2231 operations makes precedence irrelevant within terms. To provide a place
2232 to put the grammar for each terms of each type, we will start out by
2233 introducing the "Term" grammar production, with contains at least a
2234 simple "Value" (to be explained later).
2238 Term -> Value ${ $0 = $<1; }$
2239 | Variable ${ $0 = $<1; }$
2242 Thus far the complex types we have are arrays and structs.
2246 Arrays can be declared by giving a size and a type, as `[size]type' so
2247 `freq:[26]number` declares `freq` to be an array of 26 numbers. The
2248 size can be either a literal number, or a named constant. Some day an
2249 arbitrary expression will be supported.
2251 As a formal parameter to a function, the array can be declared with a
2252 new variable as the size: `name:[size::number]string`. The `size`
2253 variable is set to the size of the array and must be a constant. As
2254 `number` is the only supported type, it can be left out:
2255 `name:[size::]string`.
2257 Arrays cannot be assigned. When pointers are introduced we will also
2258 introduce array slices which can refer to part or all of an array -
2259 the assignment syntax will create a slice. For now, an array can only
2260 ever be referenced by the name it is declared with. It is likely that
2261 a "`copy`" primitive will eventually be define which can be used to
2262 make a copy of an array with controllable recursive depth.
2264 For now we have two sorts of array, those with fixed size either because
2265 it is given as a literal number or because it is a struct member (which
2266 cannot have a runtime-changing size), and those with a size that is
2267 determined at runtime - local variables with a const size. The former
2268 have their size calculated at parse time, the latter at run time.
2270 For the latter type, the `size` field of the type is the size of a
2271 pointer, and the array is reallocated every time it comes into scope.
2273 We differentiate struct fields with a const size from local variables
2274 with a const size by whether they are prepared at parse time or not.
2276 ###### type union fields
2279 int unspec; // size is unspecified - vsize must be set.
2282 struct variable *vsize;
2283 struct type *member;
2286 ###### value union fields
2287 void *array; // used if not static_size
2289 ###### value functions
2291 static int array_prepare_type(struct parse_context *c, struct type *type,
2294 struct value *vsize;
2296 if (type->array.static_size)
2297 return 1; // UNTESTED
2298 if (type->array.unspec && parse_time)
2299 return 1; // UNTESTED
2300 if (parse_time && type->array.vsize && !type->array.vsize->global)
2301 return 1; // UNTESTED
2303 if (type->array.vsize) {
2304 vsize = var_value(c, type->array.vsize);
2306 return 1; // UNTESTED
2308 mpz_tdiv_q(q, mpq_numref(vsize->num), mpq_denref(vsize->num));
2309 type->array.size = mpz_get_si(q);
2314 if (type->array.member->size <= 0)
2315 return 0; // UNTESTED
2317 type->array.static_size = 1;
2318 type->size = type->array.size * type->array.member->size;
2319 type->align = type->array.member->align;
2324 static void array_init(struct type *type, struct value *val)
2327 void *ptr = val->ptr;
2331 if (!type->array.static_size) {
2332 val->array = calloc(type->array.size,
2333 type->array.member->size);
2336 for (i = 0; i < type->array.size; i++) {
2338 v = (void*)ptr + i * type->array.member->size;
2339 val_init(type->array.member, v);
2343 static void array_free(struct type *type, struct value *val)
2346 void *ptr = val->ptr;
2348 if (!type->array.static_size)
2350 for (i = 0; i < type->array.size; i++) {
2352 v = (void*)ptr + i * type->array.member->size;
2353 free_value(type->array.member, v);
2355 if (!type->array.static_size)
2359 static int array_compat(struct type *require, struct type *have)
2361 if (have->compat != require->compat)
2363 /* Both are arrays, so we can look at details */
2364 if (!type_compat(require->array.member, have->array.member, 0))
2366 if (have->array.unspec && require->array.unspec) {
2367 if (have->array.vsize && require->array.vsize &&
2368 have->array.vsize != require->array.vsize) // UNTESTED
2369 /* sizes might not be the same */
2370 return 0; // UNTESTED
2373 if (have->array.unspec || require->array.unspec)
2374 return 1; // UNTESTED
2375 if (require->array.vsize == NULL && have->array.vsize == NULL)
2376 return require->array.size == have->array.size;
2378 return require->array.vsize == have->array.vsize; // UNTESTED
2381 static void array_print_type(struct type *type, FILE *f)
2384 if (type->array.vsize) {
2385 struct binding *b = type->array.vsize->name;
2386 fprintf(f, "%.*s%s]", b->name.len, b->name.txt,
2387 type->array.unspec ? "::" : "");
2388 } else if (type->array.size)
2389 fprintf(f, "%d]", type->array.size);
2392 type_print(type->array.member, f);
2395 static struct type array_prototype = {
2397 .prepare_type = array_prepare_type,
2398 .print_type = array_print_type,
2399 .compat = array_compat,
2401 .size = sizeof(void*),
2402 .align = sizeof(void*),
2405 ###### declare terminals
2410 | [ NUMBER ] Type ${ {
2416 if (number_parse(num, tail, $2.txt) == 0)
2417 tok_err(c, "error: unrecognised number", &$2);
2419 tok_err(c, "error: unsupported number suffix", &$2);
2422 elements = mpz_get_ui(mpq_numref(num));
2423 if (mpz_cmp_ui(mpq_denref(num), 1) != 0) {
2424 tok_err(c, "error: array size must be an integer",
2426 } else if (mpz_cmp_ui(mpq_numref(num), 1UL << 30) >= 0)
2427 tok_err(c, "error: array size is too large",
2432 $0 = t = add_anon_type(c, &array_prototype, "array[%d]", elements );
2433 t->array.size = elements;
2434 t->array.member = $<4;
2435 t->array.vsize = NULL;
2438 | [ IDENTIFIER ] Type ${ {
2439 struct variable *v = var_ref(c, $2.txt);
2442 tok_err(c, "error: name undeclared", &$2);
2443 else if (!v->constant)
2444 tok_err(c, "error: array size must be a constant", &$2);
2446 $0 = add_anon_type(c, &array_prototype, "array[%.*s]", $2.txt.len, $2.txt.txt);
2447 $0->array.member = $<4;
2449 $0->array.vsize = v;
2454 OptType -> Type ${ $0 = $<1; }$
2457 ###### formal type grammar
2459 | [ IDENTIFIER :: OptType ] Type ${ {
2460 struct variable *v = var_decl(c, $ID.txt);
2466 $0 = add_anon_type(c, &array_prototype, "array[var]");
2467 $0->array.member = $<6;
2469 $0->array.unspec = 1;
2470 $0->array.vsize = v;
2478 | Term [ Expression ] ${ {
2479 struct binode *b = new(binode);
2486 ###### print binode cases
2488 print_exec(b->left, -1, bracket);
2490 print_exec(b->right, -1, bracket);
2494 ###### propagate binode cases
2496 /* left must be an array, right must be a number,
2497 * result is the member type of the array
2499 propagate_types(b->right, c, perr, Tnum, 0);
2500 t = propagate_types(b->left, c, perr, NULL, rules & Rnoconstant);
2501 if (!t || t->compat != array_compat) {
2502 type_err(c, "error: %1 cannot be indexed", prog, t, 0, NULL);
2505 if (!type_compat(type, t->array.member, rules)) {
2506 type_err(c, "error: have %1 but need %2", prog,
2507 t->array.member, rules, type);
2509 return t->array.member;
2513 ###### interp binode cases
2519 lleft = linterp_exec(c, b->left, <ype);
2520 right = interp_exec(c, b->right, &rtype);
2522 mpz_tdiv_q(q, mpq_numref(right.num), mpq_denref(right.num));
2526 if (ltype->array.static_size)
2529 ptr = *(void**)lleft;
2530 rvtype = ltype->array.member;
2531 if (i >= 0 && i < ltype->array.size)
2532 lrv = ptr + i * rvtype->size;
2534 val_init(ltype->array.member, &rv); // UNSAFE
2541 A `struct` is a data-type that contains one or more other data-types.
2542 It differs from an array in that each member can be of a different
2543 type, and they are accessed by name rather than by number. Thus you
2544 cannot choose an element by calculation, you need to know what you
2547 The language makes no promises about how a given structure will be
2548 stored in memory - it is free to rearrange fields to suit whatever
2549 criteria seems important.
2551 Structs are declared separately from program code - they cannot be
2552 declared in-line in a variable declaration like arrays can. A struct
2553 is given a name and this name is used to identify the type - the name
2554 is not prefixed by the word `struct` as it would be in C.
2556 Structs are only treated as the same if they have the same name.
2557 Simply having the same fields in the same order is not enough. This
2558 might change once we can create structure initializers from a list of
2561 Each component datum is identified much like a variable is declared,
2562 with a name, one or two colons, and a type. The type cannot be omitted
2563 as there is no opportunity to deduce the type from usage. An initial
2564 value can be given following an equals sign, so
2566 ##### Example: a struct type
2572 would declare a type called "complex" which has two number fields,
2573 each initialised to zero.
2575 Struct will need to be declared separately from the code that uses
2576 them, so we will need to be able to print out the declaration of a
2577 struct when reprinting the whole program. So a `print_type_decl` type
2578 function will be needed.
2580 ###### type union fields
2589 } *fields; // This is created when field_list is analysed.
2591 struct fieldlist *prev;
2594 } *field_list; // This is created during parsing
2597 ###### type functions
2598 void (*print_type_decl)(struct type *type, FILE *f);
2600 ###### value functions
2602 static void structure_init(struct type *type, struct value *val)
2606 for (i = 0; i < type->structure.nfields; i++) {
2608 v = (void*) val->ptr + type->structure.fields[i].offset;
2609 if (type->structure.fields[i].init)
2610 dup_value(type->structure.fields[i].type,
2611 type->structure.fields[i].init,
2614 val_init(type->structure.fields[i].type, v);
2618 static void structure_free(struct type *type, struct value *val)
2622 for (i = 0; i < type->structure.nfields; i++) {
2624 v = (void*)val->ptr + type->structure.fields[i].offset;
2625 free_value(type->structure.fields[i].type, v);
2629 static void free_fieldlist(struct fieldlist *f)
2633 free_fieldlist(f->prev);
2638 static void structure_free_type(struct type *t)
2641 for (i = 0; i < t->structure.nfields; i++)
2642 if (t->structure.fields[i].init) {
2643 free_value(t->structure.fields[i].type,
2644 t->structure.fields[i].init);
2646 free(t->structure.fields);
2647 free_fieldlist(t->structure.field_list);
2650 static int structure_prepare_type(struct parse_context *c,
2651 struct type *t, int parse_time)
2654 struct fieldlist *f;
2656 if (!parse_time || t->structure.fields)
2659 for (f = t->structure.field_list; f; f=f->prev) {
2663 if (f->f.type->size <= 0)
2665 if (f->f.type->prepare_type)
2666 f->f.type->prepare_type(c, f->f.type, parse_time);
2668 if (f->init == NULL)
2672 propagate_types(f->init, c, &perr, f->f.type, 0);
2673 } while (perr & Eretry);
2675 c->parse_error += 1; // NOTEST
2678 t->structure.nfields = cnt;
2679 t->structure.fields = calloc(cnt, sizeof(struct field));
2680 f = t->structure.field_list;
2682 int a = f->f.type->align;
2684 t->structure.fields[cnt] = f->f;
2685 if (t->size & (a-1))
2686 t->size = (t->size | (a-1)) + 1;
2687 t->structure.fields[cnt].offset = t->size;
2688 t->size += ((f->f.type->size - 1) | (a-1)) + 1;
2692 if (f->init && !c->parse_error) {
2693 struct value vl = interp_exec(c, f->init, NULL);
2694 t->structure.fields[cnt].init =
2695 global_alloc(c, f->f.type, NULL, &vl);
2703 static struct type structure_prototype = {
2704 .init = structure_init,
2705 .free = structure_free,
2706 .free_type = structure_free_type,
2707 .print_type_decl = structure_print_type,
2708 .prepare_type = structure_prepare_type,
2722 ###### free exec cases
2724 free_exec(cast(fieldref, e)->left);
2728 ###### declare terminals
2733 | Term . IDENTIFIER ${ {
2734 struct fieldref *fr = new_pos(fieldref, $2);
2741 ###### print exec cases
2745 struct fieldref *f = cast(fieldref, e);
2746 print_exec(f->left, -1, bracket);
2747 printf(".%.*s", f->name.len, f->name.txt);
2751 ###### ast functions
2752 static int find_struct_index(struct type *type, struct text field)
2755 for (i = 0; i < type->structure.nfields; i++)
2756 if (text_cmp(type->structure.fields[i].name, field) == 0)
2761 ###### propagate exec cases
2765 struct fieldref *f = cast(fieldref, prog);
2766 struct type *st = propagate_types(f->left, c, perr, NULL, 0);
2769 type_err(c, "error: unknown type for field access", f->left, // UNTESTED
2771 else if (st->init != structure_init)
2772 type_err(c, "error: field reference attempted on %1, not a struct",
2773 f->left, st, 0, NULL);
2774 else if (f->index == -2) {
2775 f->index = find_struct_index(st, f->name);
2777 type_err(c, "error: cannot find requested field in %1",
2778 f->left, st, 0, NULL);
2780 if (f->index >= 0) {
2781 struct type *ft = st->structure.fields[f->index].type;
2782 if (!type_compat(type, ft, rules))
2783 type_err(c, "error: have %1 but need %2", prog,
2790 ###### interp exec cases
2793 struct fieldref *f = cast(fieldref, e);
2795 struct value *lleft = linterp_exec(c, f->left, <ype);
2796 lrv = (void*)lleft->ptr + ltype->structure.fields[f->index].offset;
2797 rvtype = ltype->structure.fields[f->index].type;
2801 ###### top level grammar
2802 DeclareStruct -> struct IDENTIFIER FieldBlock Newlines ${ {
2804 t = find_type(c, $ID.txt);
2806 t = add_type(c, $ID.txt, &structure_prototype);
2807 else if (t->size >= 0) {
2808 tok_err(c, "error: type already declared", &$ID);
2809 tok_err(c, "info: this is location of declartion", &t->first_use);
2810 /* Create a new one - duplicate */
2811 t = add_type(c, $ID.txt, &structure_prototype);
2813 struct type tmp = *t;
2814 *t = structure_prototype;
2818 t->structure.field_list = $<FB;
2823 FieldBlock -> { IN OptNL FieldLines OUT OptNL } ${ $0 = $<FL; }$
2824 | { SimpleFieldList } ${ $0 = $<SFL; }$
2825 | IN OptNL FieldLines OUT ${ $0 = $<FL; }$
2826 | SimpleFieldList EOL ${ $0 = $<SFL; }$
2828 FieldLines -> SimpleFieldList Newlines ${ $0 = $<SFL; }$
2829 | FieldLines SimpleFieldList Newlines ${
2834 SimpleFieldList -> Field ${ $0 = $<F; }$
2835 | SimpleFieldList ; Field ${
2839 | SimpleFieldList ; ${
2842 | ERROR ${ tok_err(c, "Syntax error in struct field", &$1); }$
2844 Field -> IDENTIFIER : Type = Expression ${ {
2845 $0 = calloc(1, sizeof(struct fieldlist));
2846 $0->f.name = $ID.txt;
2847 $0->f.type = $<Type;
2851 | IDENTIFIER : Type ${
2852 $0 = calloc(1, sizeof(struct fieldlist));
2853 $0->f.name = $ID.txt;
2854 $0->f.type = $<Type;
2857 ###### forward decls
2858 static void structure_print_type(struct type *t, FILE *f);
2860 ###### value functions
2861 static void structure_print_type(struct type *t, FILE *f)
2865 fprintf(f, "struct %.*s\n", t->name.len, t->name.txt);
2867 for (i = 0; i < t->structure.nfields; i++) {
2868 struct field *fl = t->structure.fields + i;
2869 fprintf(f, " %.*s : ", fl->name.len, fl->name.txt);
2870 type_print(fl->type, f);
2871 if (fl->type->print && fl->init) {
2873 if (fl->type == Tstr)
2874 fprintf(f, "\""); // UNTESTED
2875 print_value(fl->type, fl->init, f);
2876 if (fl->type == Tstr)
2877 fprintf(f, "\""); // UNTESTED
2883 ###### print type decls
2888 while (target != 0) {
2890 for (t = context.typelist; t ; t=t->next)
2891 if (!t->anon && t->print_type_decl &&
2901 t->print_type_decl(t, stdout);
2909 A function is a chunk of code which can be passed parameters and can
2910 return results. Each function has a type which includes the set of
2911 parameters and the return value. As yet these types cannot be declared
2912 separately from the function itself.
2914 The parameters can be specified either in parentheses as a ';' separated
2917 ##### Example: function 1
2919 func main(av:[ac::number]string; env:[envc::number]string)
2922 or as an indented list of one parameter per line (though each line can
2923 be a ';' separated list)
2925 ##### Example: function 2
2928 argv:[argc::number]string
2929 env:[envc::number]string
2933 In the first case a return type can follow the parentheses after a colon,
2934 in the second it is given on a line starting with the word `return`.
2936 ##### Example: functions that return
2938 func add(a:number; b:number): number
2948 Rather than returning a type, the function can specify a set of local
2949 variables to return as a struct. The values of these variables when the
2950 function exits will be provided to the caller. For this the return type
2951 is replaced with a block of result declarations, either in parentheses
2952 or bracketed by `return` and `do`.
2954 ##### Example: functions returning multiple variables
2956 func to_cartesian(rho:number; theta:number):(x:number; y:number)
2969 For constructing the lists we use a `List` binode, which will be
2970 further detailed when Expression Lists are introduced.
2972 ###### type union fields
2975 struct binode *params;
2976 struct type *return_type;
2977 struct variable *scope;
2978 int inline_result; // return value is at start of 'local'
2982 ###### value union fields
2983 struct exec *function;
2985 ###### type functions
2986 void (*check_args)(struct parse_context *c, enum prop_err *perr,
2987 struct type *require, struct exec *args);
2989 ###### value functions
2991 static void function_free(struct type *type, struct value *val)
2993 free_exec(val->function);
2994 val->function = NULL;
2997 static int function_compat(struct type *require, struct type *have)
2999 // FIXME can I do anything here yet?
3003 static void function_check_args(struct parse_context *c, enum prop_err *perr,
3004 struct type *require, struct exec *args)
3006 /* This should be 'compat', but we don't have a 'tuple' type to
3007 * hold the type of 'args'
3009 struct binode *arg = cast(binode, args);
3010 struct binode *param = require->function.params;
3013 struct var *pv = cast(var, param->left);
3015 type_err(c, "error: insufficient arguments to function.",
3016 args, NULL, 0, NULL);
3020 propagate_types(arg->left, c, perr, pv->var->type, 0);
3021 param = cast(binode, param->right);
3022 arg = cast(binode, arg->right);
3025 type_err(c, "error: too many arguments to function.",
3026 args, NULL, 0, NULL);
3029 static void function_print(struct type *type, struct value *val, FILE *f)
3031 print_exec(val->function, 1, 0);
3034 static void function_print_type_decl(struct type *type, FILE *f)
3038 for (b = type->function.params; b; b = cast(binode, b->right)) {
3039 struct variable *v = cast(var, b->left)->var;
3040 fprintf(f, "%.*s%s", v->name->name.len, v->name->name.txt,
3041 v->constant ? "::" : ":");
3042 type_print(v->type, f);
3047 if (type->function.return_type != Tnone) {
3049 if (type->function.inline_result) {
3051 struct type *t = type->function.return_type;
3053 for (i = 0; i < t->structure.nfields; i++) {
3054 struct field *fl = t->structure.fields + i;
3057 fprintf(f, "%.*s:", fl->name.len, fl->name.txt);
3058 type_print(fl->type, f);
3062 type_print(type->function.return_type, f);
3067 static void function_free_type(struct type *t)
3069 free_exec(t->function.params);
3072 static struct type function_prototype = {
3073 .size = sizeof(void*),
3074 .align = sizeof(void*),
3075 .free = function_free,
3076 .compat = function_compat,
3077 .check_args = function_check_args,
3078 .print = function_print,
3079 .print_type_decl = function_print_type_decl,
3080 .free_type = function_free_type,
3083 ###### declare terminals
3093 FuncName -> IDENTIFIER ${ {
3094 struct variable *v = var_decl(c, $1.txt);
3095 struct var *e = new_pos(var, $1);
3101 v = var_ref(c, $1.txt);
3103 type_err(c, "error: function '%v' redeclared",
3105 type_err(c, "info: this is where '%v' was first declared",
3106 v->where_decl, NULL, 0, NULL);
3112 Args -> ArgsLine NEWLINE ${ $0 = $<AL; }$
3113 | Args ArgsLine NEWLINE ${ {
3114 struct binode *b = $<AL;
3115 struct binode **bp = &b;
3117 bp = (struct binode **)&(*bp)->left;
3122 ArgsLine -> ${ $0 = NULL; }$
3123 | Varlist ${ $0 = $<1; }$
3124 | Varlist ; ${ $0 = $<1; }$
3126 Varlist -> Varlist ; ArgDecl ${
3140 ArgDecl -> IDENTIFIER : FormalType ${ {
3141 struct variable *v = var_decl(c, $1.txt);
3147 ##### Function calls
3149 A function call can appear either as an expression or as a statement.
3150 We use a new 'Funcall' binode type to link the function with a list of
3151 arguments, form with the 'List' nodes.
3153 We have already seen the "Term" which is how a function call can appear
3154 in an expression. To parse a function call into a statement we include
3155 it in the "SimpleStatement Grammar" which will be described later.
3161 | Term ( ExpressionList ) ${ {
3162 struct binode *b = new(binode);
3165 b->right = reorder_bilist($<EL);
3169 struct binode *b = new(binode);
3176 ###### SimpleStatement Grammar
3178 | Term ( ExpressionList ) ${ {
3179 struct binode *b = new(binode);
3182 b->right = reorder_bilist($<EL);
3186 ###### print binode cases
3189 do_indent(indent, "");
3190 print_exec(b->left, -1, bracket);
3192 for (b = cast(binode, b->right); b; b = cast(binode, b->right)) {
3195 print_exec(b->left, -1, bracket);
3205 ###### propagate binode cases
3208 /* Every arg must match formal parameter, and result
3209 * is return type of function
3211 struct binode *args = cast(binode, b->right);
3212 struct var *v = cast(var, b->left);
3214 if (!v->var->type || v->var->type->check_args == NULL) {
3215 type_err(c, "error: attempt to call a non-function.",
3216 prog, NULL, 0, NULL);
3220 v->var->type->check_args(c, perr, v->var->type, args);
3221 if (v->var->type->function.inline_result)
3223 return v->var->type->function.return_type;
3226 ###### interp binode cases
3229 struct var *v = cast(var, b->left);
3230 struct type *t = v->var->type;
3231 void *oldlocal = c->local;
3232 int old_size = c->local_size;
3233 void *local = calloc(1, t->function.local_size);
3234 struct value *fbody = var_value(c, v->var);
3235 struct binode *arg = cast(binode, b->right);
3236 struct binode *param = t->function.params;
3239 struct var *pv = cast(var, param->left);
3240 struct type *vtype = NULL;
3241 struct value val = interp_exec(c, arg->left, &vtype);
3243 c->local = local; c->local_size = t->function.local_size;
3244 lval = var_value(c, pv->var);
3245 c->local = oldlocal; c->local_size = old_size;
3246 memcpy(lval, &val, vtype->size);
3247 param = cast(binode, param->right);
3248 arg = cast(binode, arg->right);
3250 c->local = local; c->local_size = t->function.local_size;
3251 if (t->function.inline_result && dtype) {
3252 _interp_exec(c, fbody->function, NULL, NULL);
3253 memcpy(dest, local, dtype->size);
3254 rvtype = ret.type = NULL;
3256 rv = interp_exec(c, fbody->function, &rvtype);
3257 c->local = oldlocal; c->local_size = old_size;
3262 ## Complex executables: statements and expressions
3264 Now that we have types and values and variables and most of the basic
3265 Terms which provide access to these, we can explore the more complex
3266 code that combine all of these to get useful work done. Specifically
3267 statements and expressions.
3269 Expressions are various combinations of Terms. We will use operator
3270 precedence to ensure correct parsing. The simplest Expression is just a
3271 Term - others will follow.
3276 Expression -> Term ${ $0 = $<Term; }$
3277 ## expression grammar
3279 ### Expressions: Conditional
3281 Our first user of the `binode` will be conditional expressions, which
3282 is a bit odd as they actually have three components. That will be
3283 handled by having 2 binodes for each expression. The conditional
3284 expression is the lowest precedence operator which is why we define it
3285 first - to start the precedence list.
3287 Conditional expressions are of the form "value `if` condition `else`
3288 other_value". They associate to the right, so everything to the right
3289 of `else` is part of an else value, while only a higher-precedence to
3290 the left of `if` is the if values. Between `if` and `else` there is no
3291 room for ambiguity, so a full conditional expression is allowed in
3297 ###### declare terminals
3301 ###### expression grammar
3303 | Expression if Expression else Expression $$ifelse ${ {
3304 struct binode *b1 = new(binode);
3305 struct binode *b2 = new(binode);
3315 ###### print binode cases
3318 b2 = cast(binode, b->right);
3319 if (bracket) printf("(");
3320 print_exec(b2->left, -1, bracket);
3322 print_exec(b->left, -1, bracket);
3324 print_exec(b2->right, -1, bracket);
3325 if (bracket) printf(")");
3328 ###### propagate binode cases
3331 /* cond must be Tbool, others must match */
3332 struct binode *b2 = cast(binode, b->right);
3335 propagate_types(b->left, c, perr, Tbool, 0);
3336 t = propagate_types(b2->left, c, perr, type, Rnolabel);
3337 t2 = propagate_types(b2->right, c, perr, type ?: t, Rnolabel);
3341 ###### interp binode cases
3344 struct binode *b2 = cast(binode, b->right);
3345 left = interp_exec(c, b->left, <ype);
3347 rv = interp_exec(c, b2->left, &rvtype); // UNTESTED
3349 rv = interp_exec(c, b2->right, &rvtype);
3355 We take a brief detour, now that we have expressions, to describe lists
3356 of expressions. These will be needed for function parameters and
3357 possibly other situations. They seem generic enough to introduce here
3358 to be used elsewhere.
3360 And ExpressionList will use the `List` type of `binode`, building up at
3361 the end. And place where they are used will probably call
3362 `reorder_bilist()` to get a more normal first/next arrangement.
3364 ###### declare terminals
3367 `List` execs have no implicit semantics, so they are never propagated or
3368 interpreted. The can be printed as a comma separate list, which is how
3369 they are parsed. Note they are also used for function formal parameter
3370 lists. In that case a separate function is used to print them.
3372 ###### print binode cases
3376 print_exec(b->left, -1, bracket);
3379 b = cast(binode, b->right);
3383 ###### propagate binode cases
3384 case List: abort(); // NOTEST
3385 ###### interp binode cases
3386 case List: abort(); // NOTEST
3391 ExpressionList -> ExpressionList , Expression ${
3404 ### Expressions: Boolean
3406 The next class of expressions to use the `binode` will be Boolean
3407 expressions. "`and then`" and "`or else`" are similar to `and` and `or`
3408 have same corresponding precendence. The difference is that they don't
3409 evaluate the second expression if not necessary.
3418 ###### declare terminals
3423 ###### expression grammar
3424 | Expression or Expression ${ {
3425 struct binode *b = new(binode);
3431 | Expression or else Expression ${ {
3432 struct binode *b = new(binode);
3439 | Expression and Expression ${ {
3440 struct binode *b = new(binode);
3446 | Expression and then Expression ${ {
3447 struct binode *b = new(binode);
3454 | not Expression ${ {
3455 struct binode *b = new(binode);
3461 ###### print binode cases
3463 if (bracket) printf("(");
3464 print_exec(b->left, -1, bracket);
3466 print_exec(b->right, -1, bracket);
3467 if (bracket) printf(")");
3470 if (bracket) printf("(");
3471 print_exec(b->left, -1, bracket);
3472 printf(" and then ");
3473 print_exec(b->right, -1, bracket);
3474 if (bracket) printf(")");
3477 if (bracket) printf("(");
3478 print_exec(b->left, -1, bracket);
3480 print_exec(b->right, -1, bracket);
3481 if (bracket) printf(")");
3484 if (bracket) printf("(");
3485 print_exec(b->left, -1, bracket);
3486 printf(" or else ");
3487 print_exec(b->right, -1, bracket);
3488 if (bracket) printf(")");
3491 if (bracket) printf("(");
3493 print_exec(b->right, -1, bracket);
3494 if (bracket) printf(")");
3497 ###### propagate binode cases
3503 /* both must be Tbool, result is Tbool */
3504 propagate_types(b->left, c, perr, Tbool, 0);
3505 propagate_types(b->right, c, perr, Tbool, 0);
3506 if (type && type != Tbool)
3507 type_err(c, "error: %1 operation found where %2 expected", prog,
3511 ###### interp binode cases
3513 rv = interp_exec(c, b->left, &rvtype);
3514 right = interp_exec(c, b->right, &rtype);
3515 rv.bool = rv.bool && right.bool;
3518 rv = interp_exec(c, b->left, &rvtype);
3520 rv = interp_exec(c, b->right, NULL);
3523 rv = interp_exec(c, b->left, &rvtype);
3524 right = interp_exec(c, b->right, &rtype);
3525 rv.bool = rv.bool || right.bool;
3528 rv = interp_exec(c, b->left, &rvtype);
3530 rv = interp_exec(c, b->right, NULL);
3533 rv = interp_exec(c, b->right, &rvtype);
3537 ### Expressions: Comparison
3539 Of slightly higher precedence that Boolean expressions are Comparisons.
3540 A comparison takes arguments of any comparable type, but the two types
3543 To simplify the parsing we introduce an `eop` which can record an
3544 expression operator, and the `CMPop` non-terminal will match one of them.
3551 ###### ast functions
3552 static void free_eop(struct eop *e)
3566 ###### declare terminals
3567 $LEFT < > <= >= == != CMPop
3569 ###### expression grammar
3570 | Expression CMPop Expression ${ {
3571 struct binode *b = new(binode);
3581 CMPop -> < ${ $0.op = Less; }$
3582 | > ${ $0.op = Gtr; }$
3583 | <= ${ $0.op = LessEq; }$
3584 | >= ${ $0.op = GtrEq; }$
3585 | == ${ $0.op = Eql; }$
3586 | != ${ $0.op = NEql; }$
3588 ###### print binode cases
3596 if (bracket) printf("(");
3597 print_exec(b->left, -1, bracket);
3599 case Less: printf(" < "); break;
3600 case LessEq: printf(" <= "); break;
3601 case Gtr: printf(" > "); break;
3602 case GtrEq: printf(" >= "); break;
3603 case Eql: printf(" == "); break;
3604 case NEql: printf(" != "); break;
3605 default: abort(); // NOTEST
3607 print_exec(b->right, -1, bracket);
3608 if (bracket) printf(")");
3611 ###### propagate binode cases
3618 /* Both must match but not be labels, result is Tbool */
3619 t = propagate_types(b->left, c, perr, NULL, Rnolabel);
3621 propagate_types(b->right, c, perr, t, 0);
3623 t = propagate_types(b->right, c, perr, NULL, Rnolabel); // UNTESTED
3625 t = propagate_types(b->left, c, perr, t, 0); // UNTESTED
3627 if (!type_compat(type, Tbool, 0))
3628 type_err(c, "error: Comparison returns %1 but %2 expected", prog,
3629 Tbool, rules, type);
3632 ###### interp binode cases
3641 left = interp_exec(c, b->left, <ype);
3642 right = interp_exec(c, b->right, &rtype);
3643 cmp = value_cmp(ltype, rtype, &left, &right);
3646 case Less: rv.bool = cmp < 0; break;
3647 case LessEq: rv.bool = cmp <= 0; break;
3648 case Gtr: rv.bool = cmp > 0; break;
3649 case GtrEq: rv.bool = cmp >= 0; break;
3650 case Eql: rv.bool = cmp == 0; break;
3651 case NEql: rv.bool = cmp != 0; break;
3652 default: rv.bool = 0; break; // NOTEST
3657 ### Expressions: Arithmetic etc.
3659 The remaining expressions with the highest precedence are arithmetic,
3660 string concatenation, string conversion, and testing. String concatenation
3661 (`++`) has the same precedence as multiplication and division, but lower
3664 Testing comes in two forms. A single question mark (`?`) is a uniary
3665 operator which converts come types into Boolean. The general meaning is
3666 "is this a value value" and there will be more uses as the language
3667 develops. A double questionmark (`??`) is a binary operator (Choose),
3668 with same precedence as multiplication, which returns the LHS if it
3669 tests successfully, else returns the RHS.
3671 String conversion is a temporary feature until I get a better type
3672 system. `$` is a prefix operator which expects a string and returns
3675 `+` and `-` are both infix and prefix operations (where they are
3676 absolute value and negation). These have different operator names.
3678 We also have a 'Bracket' operator which records where parentheses were
3679 found. This makes it easy to reproduce these when printing. Possibly I
3680 should only insert brackets were needed for precedence. Putting
3681 parentheses around an expression converts it into a Term,
3687 Absolute, Negate, Test,
3691 ###### declare terminals
3693 $LEFT * / % ++ ?? Top
3697 ###### expression grammar
3698 | Expression Eop Expression ${ {
3699 struct binode *b = new(binode);
3706 | Expression Top Expression ${ {
3707 struct binode *b = new(binode);
3714 | Uop Expression ${ {
3715 struct binode *b = new(binode);
3723 | ( Expression ) ${ {
3724 struct binode *b = new_pos(binode, $1);
3733 Eop -> + ${ $0.op = Plus; }$
3734 | - ${ $0.op = Minus; }$
3736 Uop -> + ${ $0.op = Absolute; }$
3737 | - ${ $0.op = Negate; }$
3738 | $ ${ $0.op = StringConv; }$
3739 | ? ${ $0.op = Test; }$
3741 Top -> * ${ $0.op = Times; }$
3742 | / ${ $0.op = Divide; }$
3743 | % ${ $0.op = Rem; }$
3744 | ++ ${ $0.op = Concat; }$
3745 | ?? ${ $0.op = Choose; }$
3747 ###### print binode cases
3755 if (bracket) printf("(");
3756 print_exec(b->left, indent, bracket);
3758 case Plus: fputs(" + ", stdout); break;
3759 case Minus: fputs(" - ", stdout); break;
3760 case Times: fputs(" * ", stdout); break;
3761 case Divide: fputs(" / ", stdout); break;
3762 case Rem: fputs(" % ", stdout); break;
3763 case Concat: fputs(" ++ ", stdout); break;
3764 case Choose: fputs(" ?? ", stdout); break;
3765 default: abort(); // NOTEST
3767 print_exec(b->right, indent, bracket);
3768 if (bracket) printf(")");
3774 if (bracket) printf("(");
3776 case Absolute: fputs("+", stdout); break;
3777 case Negate: fputs("-", stdout); break;
3778 case StringConv: fputs("$", stdout); break;
3779 case Test: fputs("?", stdout); break;
3780 default: abort(); // NOTEST
3782 print_exec(b->right, indent, bracket);
3783 if (bracket) printf(")");
3787 print_exec(b->right, indent, bracket);
3791 ###### propagate binode cases
3797 /* both must be numbers, result is Tnum */
3800 /* as propagate_types ignores a NULL,
3801 * unary ops fit here too */
3802 propagate_types(b->left, c, perr, Tnum, 0);
3803 propagate_types(b->right, c, perr, Tnum, 0);
3804 if (!type_compat(type, Tnum, 0))
3805 type_err(c, "error: Arithmetic returns %1 but %2 expected", prog,
3810 /* both must be Tstr, result is Tstr */
3811 propagate_types(b->left, c, perr, Tstr, 0);
3812 propagate_types(b->right, c, perr, Tstr, 0);
3813 if (!type_compat(type, Tstr, 0))
3814 type_err(c, "error: Concat returns %1 but %2 expected", prog,
3819 /* op must be string, result is number */
3820 propagate_types(b->left, c, perr, Tstr, 0);
3821 if (!type_compat(type, Tnum, 0))
3822 type_err(c, // UNTESTED
3823 "error: Can only convert string to number, not %1",
3824 prog, type, 0, NULL);
3828 /* LHS must support ->test, result is Tbool */
3829 t = propagate_types(b->right, c, perr, NULL, 0);
3831 type_err(c, "error: '?' requires a testable value, not %1",
3836 /* LHS and RHS must match and are returned. Must support
3839 t = propagate_types(b->left, c, perr, type, rules);
3840 t = propagate_types(b->right, c, perr, t, rules);
3841 if (t && t->test == NULL)
3842 type_err(c, "error: \"??\" requires a testable value, not %1",
3847 return propagate_types(b->right, c, perr, type, 0);
3849 ###### interp binode cases
3852 rv = interp_exec(c, b->left, &rvtype);
3853 right = interp_exec(c, b->right, &rtype);
3854 mpq_add(rv.num, rv.num, right.num);
3857 rv = interp_exec(c, b->left, &rvtype);
3858 right = interp_exec(c, b->right, &rtype);
3859 mpq_sub(rv.num, rv.num, right.num);
3862 rv = interp_exec(c, b->left, &rvtype);
3863 right = interp_exec(c, b->right, &rtype);
3864 mpq_mul(rv.num, rv.num, right.num);
3867 rv = interp_exec(c, b->left, &rvtype);
3868 right = interp_exec(c, b->right, &rtype);
3869 mpq_div(rv.num, rv.num, right.num);
3874 left = interp_exec(c, b->left, <ype);
3875 right = interp_exec(c, b->right, &rtype);
3876 mpz_init(l); mpz_init(r); mpz_init(rem);
3877 mpz_tdiv_q(l, mpq_numref(left.num), mpq_denref(left.num));
3878 mpz_tdiv_q(r, mpq_numref(right.num), mpq_denref(right.num));
3879 mpz_tdiv_r(rem, l, r);
3880 val_init(Tnum, &rv);
3881 mpq_set_z(rv.num, rem);
3882 mpz_clear(r); mpz_clear(l); mpz_clear(rem);
3887 rv = interp_exec(c, b->right, &rvtype);
3888 mpq_neg(rv.num, rv.num);
3891 rv = interp_exec(c, b->right, &rvtype);
3892 mpq_abs(rv.num, rv.num);
3895 rv = interp_exec(c, b->right, &rvtype);
3898 left = interp_exec(c, b->left, <ype);
3899 right = interp_exec(c, b->right, &rtype);
3901 rv.str = text_join(left.str, right.str);
3904 right = interp_exec(c, b->right, &rvtype);
3908 struct text tx = right.str;
3911 if (tx.txt[0] == '-') {
3912 neg = 1; // UNTESTED
3913 tx.txt++; // UNTESTED
3914 tx.len--; // UNTESTED
3916 if (number_parse(rv.num, tail, tx) == 0)
3917 mpq_init(rv.num); // UNTESTED
3919 mpq_neg(rv.num, rv.num); // UNTESTED
3921 printf("Unsupported suffix: %.*s\n", tx.len, tx.txt); // UNTESTED
3925 right = interp_exec(c, b->right, &rtype);
3927 rv.bool = !!rtype->test(rtype, &right);
3930 left = interp_exec(c, b->left, <ype);
3931 if (ltype->test(ltype, &left)) {
3936 rv = interp_exec(c, b->right, &rvtype);
3939 ###### value functions
3941 static struct text text_join(struct text a, struct text b)
3944 rv.len = a.len + b.len;
3945 rv.txt = malloc(rv.len);
3946 memcpy(rv.txt, a.txt, a.len);
3947 memcpy(rv.txt+a.len, b.txt, b.len);
3951 ### Blocks, Statements, and Statement lists.
3953 Now that we have expressions out of the way we need to turn to
3954 statements. There are simple statements and more complex statements.
3955 Simple statements do not contain (syntactic) newlines, complex statements do.
3957 Statements often come in sequences and we have corresponding simple
3958 statement lists and complex statement lists.
3959 The former comprise only simple statements separated by semicolons.
3960 The later comprise complex statements and simple statement lists. They are
3961 separated by newlines. Thus the semicolon is only used to separate
3962 simple statements on the one line. This may be overly restrictive,
3963 but I'm not sure I ever want a complex statement to share a line with
3966 Note that a simple statement list can still use multiple lines if
3967 subsequent lines are indented, so
3969 ###### Example: wrapped simple statement list
3974 is a single simple statement list. This might allow room for
3975 confusion, so I'm not set on it yet.
3977 A simple statement list needs no extra syntax. A complex statement
3978 list has two syntactic forms. It can be enclosed in braces (much like
3979 C blocks), or it can be introduced by an indent and continue until an
3980 unindented newline (much like Python blocks). With this extra syntax
3981 it is referred to as a block.
3983 Note that a block does not have to include any newlines if it only
3984 contains simple statements. So both of:
3986 if condition: a=b; d=f
3988 if condition { a=b; print f }
3992 In either case the list is constructed from a `binode` list with
3993 `Block` as the operator. When parsing the list it is most convenient
3994 to append to the end, so a list is a list and a statement. When using
3995 the list it is more convenient to consider a list to be a statement
3996 and a list. So we need a function to re-order a list.
3997 `reorder_bilist` serves this purpose.
3999 The only stand-alone statement we introduce at this stage is `pass`
4000 which does nothing and is represented as a `NULL` pointer in a `Block`
4001 list. Other stand-alone statements will follow once the infrastructure
4004 As many statements will use binodes, we declare a binode pointer 'b' in
4005 the common header for all reductions to use.
4007 ###### Parser: reduce
4018 Block -> { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
4019 | { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
4020 | SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
4021 | SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
4022 | IN OptNL Statementlist OUT ${ $0 = $<Sl; }$
4024 OpenBlock -> OpenScope { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
4025 | OpenScope { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
4026 | OpenScope SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
4027 | OpenScope SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
4028 | IN OpenScope OptNL Statementlist OUT ${ $0 = $<Sl; }$
4030 UseBlock -> { OpenScope IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
4031 | { OpenScope SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
4032 | IN OpenScope OptNL Statementlist OUT ${ $0 = $<Sl; }$
4034 ColonBlock -> { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
4035 | { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
4036 | : SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
4037 | : SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
4038 | : IN OptNL Statementlist OUT ${ $0 = $<Sl; }$
4040 Statementlist -> ComplexStatements ${ $0 = reorder_bilist($<CS); }$
4042 ComplexStatements -> ComplexStatements ComplexStatement ${
4052 | ComplexStatement ${
4064 ComplexStatement -> SimpleStatements Newlines ${
4065 $0 = reorder_bilist($<SS);
4067 | SimpleStatements ; Newlines ${
4068 $0 = reorder_bilist($<SS);
4070 ## ComplexStatement Grammar
4073 SimpleStatements -> SimpleStatements ; SimpleStatement ${
4079 | SimpleStatement ${
4088 SimpleStatement -> pass ${ $0 = NULL; }$
4089 | ERROR ${ tok_err(c, "Syntax error in statement", &$1); }$
4090 ## SimpleStatement Grammar
4092 ###### print binode cases
4096 if (b->left == NULL) // UNTESTED
4097 printf("pass"); // UNTESTED
4099 print_exec(b->left, indent, bracket); // UNTESTED
4100 if (b->right) { // UNTESTED
4101 printf("; "); // UNTESTED
4102 print_exec(b->right, indent, bracket); // UNTESTED
4105 // block, one per line
4106 if (b->left == NULL)
4107 do_indent(indent, "pass\n");
4109 print_exec(b->left, indent, bracket);
4111 print_exec(b->right, indent, bracket);
4115 ###### propagate binode cases
4118 /* If any statement returns something other than Tnone
4119 * or Tbool then all such must return same type.
4120 * As each statement may be Tnone or something else,
4121 * we must always pass NULL (unknown) down, otherwise an incorrect
4122 * error might occur. We never return Tnone unless it is
4127 for (e = b; e; e = cast(binode, e->right)) {
4128 t = propagate_types(e->left, c, perr, NULL, rules);
4129 if ((rules & Rboolok) && (t == Tbool || t == Tnone))
4131 if (t == Tnone && e->right)
4132 /* Only the final statement *must* return a value
4140 type_err(c, "error: expected %1%r, found %2",
4141 e->left, type, rules, t);
4147 ###### interp binode cases
4149 while (rvtype == Tnone &&
4152 rv = interp_exec(c, b->left, &rvtype);
4153 b = cast(binode, b->right);
4157 ### The Print statement
4159 `print` is a simple statement that takes a comma-separated list of
4160 expressions and prints the values separated by spaces and terminated
4161 by a newline. No control of formatting is possible.
4163 `print` uses `ExpressionList` to collect the expressions and stores them
4164 on the left side of a `Print` binode unlessthere is a trailing comma
4165 when the list is stored on the `right` side and no trailing newline is
4171 ##### declare terminals
4174 ###### SimpleStatement Grammar
4176 | print ExpressionList ${
4177 $0 = b = new(binode);
4180 b->left = reorder_bilist($<EL);
4182 | print ExpressionList , ${ {
4183 $0 = b = new(binode);
4185 b->right = reorder_bilist($<EL);
4189 $0 = b = new(binode);
4195 ###### print binode cases
4198 do_indent(indent, "print");
4200 print_exec(b->right, -1, bracket);
4203 print_exec(b->left, -1, bracket);
4208 ###### propagate binode cases
4211 /* don't care but all must be consistent */
4213 b = cast(binode, b->left);
4215 b = cast(binode, b->right);
4217 propagate_types(b->left, c, perr, NULL, Rnolabel);
4218 b = cast(binode, b->right);
4222 ###### interp binode cases
4226 struct binode *b2 = cast(binode, b->left);
4228 b2 = cast(binode, b->right);
4229 for (; b2; b2 = cast(binode, b2->right)) {
4230 left = interp_exec(c, b2->left, <ype);
4231 print_value(ltype, &left, stdout);
4232 free_value(ltype, &left);
4236 if (b->right == NULL)
4242 ###### Assignment statement
4244 An assignment will assign a value to a variable, providing it hasn't
4245 been declared as a constant. The analysis phase ensures that the type
4246 will be correct so the interpreter just needs to perform the
4247 calculation. There is a form of assignment which declares a new
4248 variable as well as assigning a value. If a name is assigned before
4249 it is declared, and error will be raised as the name is created as
4250 `Tlabel` and it is illegal to assign to such names.
4256 ###### declare terminals
4259 ###### SimpleStatement Grammar
4260 | Term = Expression ${
4261 $0 = b= new(binode);
4266 | VariableDecl = Expression ${
4267 $0 = b= new(binode);
4274 if ($1->var->where_set == NULL) {
4276 "Variable declared with no type or value: %v",
4280 $0 = b = new(binode);
4287 ###### print binode cases
4290 do_indent(indent, "");
4291 print_exec(b->left, indent, bracket);
4293 print_exec(b->right, indent, bracket);
4300 struct variable *v = cast(var, b->left)->var;
4301 do_indent(indent, "");
4302 print_exec(b->left, indent, bracket);
4303 if (cast(var, b->left)->var->constant) {
4305 if (v->explicit_type) {
4306 type_print(v->type, stdout);
4311 if (v->explicit_type) {
4312 type_print(v->type, stdout);
4318 print_exec(b->right, indent, bracket);
4325 ###### propagate binode cases
4329 /* Both must match and not be labels,
4330 * Type must support 'dup',
4331 * For Assign, left must not be constant.
4334 t = propagate_types(b->left, c, perr, NULL,
4335 Rnolabel | (b->op == Assign ? Rnoconstant : 0));
4340 if (propagate_types(b->right, c, perr, t, 0) != t)
4341 if (b->left->type == Xvar)
4342 type_err(c, "info: variable '%v' was set as %1 here.",
4343 cast(var, b->left)->var->where_set, t, rules, NULL);
4345 t = propagate_types(b->right, c, perr, NULL, Rnolabel);
4347 propagate_types(b->left, c, perr, t,
4348 (b->op == Assign ? Rnoconstant : 0));
4350 if (t && t->dup == NULL && !(*perr & Emaycopy))
4351 type_err(c, "error: cannot assign value of type %1", b, t, 0, NULL);
4356 ###### interp binode cases
4359 lleft = linterp_exec(c, b->left, <ype);
4361 dinterp_exec(c, b->right, lleft, ltype, 1);
4367 struct variable *v = cast(var, b->left)->var;
4370 val = var_value(c, v);
4371 if (v->type->prepare_type)
4372 v->type->prepare_type(c, v->type, 0);
4374 dinterp_exec(c, b->right, val, v->type, 0);
4376 val_init(v->type, val);
4380 ### The `use` statement
4382 The `use` statement is the last "simple" statement. It is needed when a
4383 statement block can return a value. This includes the body of a
4384 function which has a return type, and the "condition" code blocks in
4385 `if`, `while`, and `switch` statements.
4390 ###### declare terminals
4393 ###### SimpleStatement Grammar
4395 $0 = b = new_pos(binode, $1);
4398 if (b->right->type == Xvar) {
4399 struct var *v = cast(var, b->right);
4400 if (v->var->type == Tnone) {
4401 /* Convert this to a label */
4404 v->var->type = Tlabel;
4405 val = global_alloc(c, Tlabel, v->var, NULL);
4411 ###### print binode cases
4414 do_indent(indent, "use ");
4415 print_exec(b->right, -1, bracket);
4420 ###### propagate binode cases
4423 /* result matches value */
4424 return propagate_types(b->right, c, perr, type, 0);
4426 ###### interp binode cases
4429 rv = interp_exec(c, b->right, &rvtype);
4432 ### The Conditional Statement
4434 This is the biggy and currently the only complex statement. This
4435 subsumes `if`, `while`, `do/while`, `switch`, and some parts of `for`.
4436 It is comprised of a number of parts, all of which are optional though
4437 set combinations apply. Each part is (usually) a key word (`then` is
4438 sometimes optional) followed by either an expression or a code block,
4439 except the `casepart` which is a "key word and an expression" followed
4440 by a code block. The code-block option is valid for all parts and,
4441 where an expression is also allowed, the code block can use the `use`
4442 statement to report a value. If the code block does not report a value
4443 the effect is similar to reporting `True`.
4445 The `else` and `case` parts, as well as `then` when combined with
4446 `if`, can contain a `use` statement which will apply to some
4447 containing conditional statement. `for` parts, `do` parts and `then`
4448 parts used with `for` can never contain a `use`, except in some
4449 subordinate conditional statement.
4451 If there is a `forpart`, it is executed first, only once.
4452 If there is a `dopart`, then it is executed repeatedly providing
4453 always that the `condpart` or `cond`, if present, does not return a non-True
4454 value. `condpart` can fail to return any value if it simply executes
4455 to completion. This is treated the same as returning `True`.
4457 If there is a `thenpart` it will be executed whenever the `condpart`
4458 or `cond` returns True (or does not return any value), but this will happen
4459 *after* `dopart` (when present).
4461 If `elsepart` is present it will be executed at most once when the
4462 condition returns `False` or some value that isn't `True` and isn't
4463 matched by any `casepart`. If there are any `casepart`s, they will be
4464 executed when the condition returns a matching value.
4466 The particular sorts of values allowed in case parts has not yet been
4467 determined in the language design, so nothing is prohibited.
4469 The various blocks in this complex statement potentially provide scope
4470 for variables as described earlier. Each such block must include the
4471 "OpenScope" nonterminal before parsing the block, and must call
4472 `var_block_close()` when closing the block.
4474 The code following "`if`", "`switch`" and "`for`" does not get its own
4475 scope, but is in a scope covering the whole statement, so names
4476 declared there cannot be redeclared elsewhere. Similarly the
4477 condition following "`while`" is in a scope the covers the body
4478 ("`do`" part) of the loop, and which does not allow conditional scope
4479 extension. Code following "`then`" (both looping and non-looping),
4480 "`else`" and "`case`" each get their own local scope.
4482 The type requirements on the code block in a `whilepart` are quite
4483 unusal. It is allowed to return a value of some identifiable type, in
4484 which case the loop aborts and an appropriate `casepart` is run, or it
4485 can return a Boolean, in which case the loop either continues to the
4486 `dopart` (on `True`) or aborts and runs the `elsepart` (on `False`).
4487 This is different both from the `ifpart` code block which is expected to
4488 return a Boolean, or the `switchpart` code block which is expected to
4489 return the same type as the casepart values. The correct analysis of
4490 the type of the `whilepart` code block is the reason for the
4491 `Rboolok` flag which is passed to `propagate_types()`.
4493 The `cond_statement` cannot fit into a `binode` so a new `exec` is
4494 defined. As there are two scopes which cover multiple parts - one for
4495 the whole statement and one for "while" and "do" - and as we will use
4496 the 'struct exec' to track scopes, we actually need two new types of
4497 exec. One is a `binode` for the looping part, the rest is the
4498 `cond_statement`. The `cond_statement` will use an auxilliary `struct
4499 casepart` to track a list of case parts.
4510 struct exec *action;
4511 struct casepart *next;
4513 struct cond_statement {
4515 struct exec *forpart, *condpart, *thenpart, *elsepart;
4516 struct binode *looppart;
4517 struct casepart *casepart;
4520 ###### ast functions
4522 static void free_casepart(struct casepart *cp)
4526 free_exec(cp->value);
4527 free_exec(cp->action);
4534 static void free_cond_statement(struct cond_statement *s)
4538 free_exec(s->forpart);
4539 free_exec(s->condpart);
4540 free_exec(s->looppart);
4541 free_exec(s->thenpart);
4542 free_exec(s->elsepart);
4543 free_casepart(s->casepart);
4547 ###### free exec cases
4548 case Xcond_statement: free_cond_statement(cast(cond_statement, e)); break;
4550 ###### ComplexStatement Grammar
4551 | CondStatement ${ $0 = $<1; }$
4553 ###### declare terminals
4554 $TERM for then while do
4561 // A CondStatement must end with EOL, as does CondSuffix and
4563 // ForPart, ThenPart, SwitchPart, CasePart are non-empty and
4564 // may or may not end with EOL
4565 // WhilePart and IfPart include an appropriate Suffix
4567 // ForPart, SwitchPart, and IfPart open scopes, o we have to close
4568 // them. WhilePart opens and closes its own scope.
4569 CondStatement -> ForPart OptNL ThenPart OptNL WhilePart CondSuffix ${
4572 $0->thenpart = $<TP;
4573 $0->looppart = $<WP;
4574 var_block_close(c, CloseSequential, $0);
4576 | ForPart OptNL WhilePart CondSuffix ${
4579 $0->looppart = $<WP;
4580 var_block_close(c, CloseSequential, $0);
4582 | WhilePart CondSuffix ${
4584 $0->looppart = $<WP;
4586 | SwitchPart OptNL CasePart CondSuffix ${
4588 $0->condpart = $<SP;
4589 $CP->next = $0->casepart;
4590 $0->casepart = $<CP;
4591 var_block_close(c, CloseSequential, $0);
4593 | SwitchPart : IN OptNL CasePart CondSuffix OUT Newlines ${
4595 $0->condpart = $<SP;
4596 $CP->next = $0->casepart;
4597 $0->casepart = $<CP;
4598 var_block_close(c, CloseSequential, $0);
4600 | IfPart IfSuffix ${
4602 $0->condpart = $IP.condpart; $IP.condpart = NULL;
4603 $0->thenpart = $IP.thenpart; $IP.thenpart = NULL;
4604 // This is where we close an "if" statement
4605 var_block_close(c, CloseSequential, $0);
4608 CondSuffix -> IfSuffix ${
4611 | Newlines CasePart CondSuffix ${
4613 $CP->next = $0->casepart;
4614 $0->casepart = $<CP;
4616 | CasePart CondSuffix ${
4618 $CP->next = $0->casepart;
4619 $0->casepart = $<CP;
4622 IfSuffix -> Newlines ${ $0 = new(cond_statement); }$
4623 | Newlines ElsePart ${ $0 = $<EP; }$
4624 | ElsePart ${$0 = $<EP; }$
4626 ElsePart -> else OpenBlock Newlines ${
4627 $0 = new(cond_statement);
4628 $0->elsepart = $<OB;
4629 var_block_close(c, CloseElse, $0->elsepart);
4631 | else OpenScope CondStatement ${
4632 $0 = new(cond_statement);
4633 $0->elsepart = $<CS;
4634 var_block_close(c, CloseElse, $0->elsepart);
4638 CasePart -> case Expression OpenScope ColonBlock ${
4639 $0 = calloc(1,sizeof(struct casepart));
4642 var_block_close(c, CloseParallel, $0->action);
4646 // These scopes are closed in CondStatement
4647 ForPart -> for OpenBlock ${
4651 ThenPart -> then OpenBlock ${
4653 var_block_close(c, CloseSequential, $0);
4657 // This scope is closed in CondStatement
4658 WhilePart -> while UseBlock OptNL do OpenBlock ${
4663 var_block_close(c, CloseSequential, $0->right);
4664 var_block_close(c, CloseSequential, $0);
4666 | while OpenScope Expression OpenScope ColonBlock ${
4671 var_block_close(c, CloseSequential, $0->right);
4672 var_block_close(c, CloseSequential, $0);
4676 IfPart -> if UseBlock OptNL then OpenBlock ${
4679 var_block_close(c, CloseParallel, $0.thenpart);
4681 | if OpenScope Expression OpenScope ColonBlock ${
4684 var_block_close(c, CloseParallel, $0.thenpart);
4686 | if OpenScope Expression OpenScope OptNL then Block ${
4689 var_block_close(c, CloseParallel, $0.thenpart);
4693 // This scope is closed in CondStatement
4694 SwitchPart -> switch OpenScope Expression ${
4697 | switch UseBlock ${
4701 ###### print binode cases
4703 if (b->left && b->left->type == Xbinode &&
4704 cast(binode, b->left)->op == Block) {
4706 do_indent(indent, "while {\n");
4708 do_indent(indent, "while\n");
4709 print_exec(b->left, indent+1, bracket);
4711 do_indent(indent, "} do {\n");
4713 do_indent(indent, "do\n");
4714 print_exec(b->right, indent+1, bracket);
4716 do_indent(indent, "}\n");
4718 do_indent(indent, "while ");
4719 print_exec(b->left, 0, bracket);
4724 print_exec(b->right, indent+1, bracket);
4726 do_indent(indent, "}\n");
4730 ###### print exec cases
4732 case Xcond_statement:
4734 struct cond_statement *cs = cast(cond_statement, e);
4735 struct casepart *cp;
4737 do_indent(indent, "for");
4738 if (bracket) printf(" {\n"); else printf("\n");
4739 print_exec(cs->forpart, indent+1, bracket);
4742 do_indent(indent, "} then {\n");
4744 do_indent(indent, "then\n");
4745 print_exec(cs->thenpart, indent+1, bracket);
4747 if (bracket) do_indent(indent, "}\n");
4750 print_exec(cs->looppart, indent, bracket);
4754 do_indent(indent, "switch");
4756 do_indent(indent, "if");
4757 if (cs->condpart && cs->condpart->type == Xbinode &&
4758 cast(binode, cs->condpart)->op == Block) {
4763 print_exec(cs->condpart, indent+1, bracket);
4765 do_indent(indent, "}\n");
4767 do_indent(indent, "then\n");
4768 print_exec(cs->thenpart, indent+1, bracket);
4772 print_exec(cs->condpart, 0, bracket);
4778 print_exec(cs->thenpart, indent+1, bracket);
4780 do_indent(indent, "}\n");
4785 for (cp = cs->casepart; cp; cp = cp->next) {
4786 do_indent(indent, "case ");
4787 print_exec(cp->value, -1, 0);
4792 print_exec(cp->action, indent+1, bracket);
4794 do_indent(indent, "}\n");
4797 do_indent(indent, "else");
4802 print_exec(cs->elsepart, indent+1, bracket);
4804 do_indent(indent, "}\n");
4809 ###### propagate binode cases
4811 t = propagate_types(b->right, c, perr, Tnone, 0);
4812 if (!type_compat(Tnone, t, 0))
4813 *perr |= Efail; // UNTESTED
4814 return propagate_types(b->left, c, perr, type, rules);
4816 ###### propagate exec cases
4817 case Xcond_statement:
4819 // forpart and looppart->right must return Tnone
4820 // thenpart must return Tnone if there is a loopart,
4821 // otherwise it is like elsepart.
4823 // be bool if there is no casepart
4824 // match casepart->values if there is a switchpart
4825 // either be bool or match casepart->value if there
4827 // elsepart and casepart->action must match the return type
4828 // expected of this statement.
4829 struct cond_statement *cs = cast(cond_statement, prog);
4830 struct casepart *cp;
4832 t = propagate_types(cs->forpart, c, perr, Tnone, 0);
4833 if (!type_compat(Tnone, t, 0))
4834 *perr |= Efail; // UNTESTED
4837 t = propagate_types(cs->thenpart, c, perr, Tnone, 0);
4838 if (!type_compat(Tnone, t, 0))
4839 *perr |= Efail; // UNTESTED
4841 if (cs->casepart == NULL) {
4842 propagate_types(cs->condpart, c, perr, Tbool, 0);
4843 propagate_types(cs->looppart, c, perr, Tbool, 0);
4845 /* Condpart must match case values, with bool permitted */
4847 for (cp = cs->casepart;
4848 cp && !t; cp = cp->next)
4849 t = propagate_types(cp->value, c, perr, NULL, 0);
4850 if (!t && cs->condpart)
4851 t = propagate_types(cs->condpart, c, perr, NULL, Rboolok); // UNTESTED
4852 if (!t && cs->looppart)
4853 t = propagate_types(cs->looppart, c, perr, NULL, Rboolok); // UNTESTED
4854 // Now we have a type (I hope) push it down
4856 for (cp = cs->casepart; cp; cp = cp->next)
4857 propagate_types(cp->value, c, perr, t, 0);
4858 propagate_types(cs->condpart, c, perr, t, Rboolok);
4859 propagate_types(cs->looppart, c, perr, t, Rboolok);
4862 // (if)then, else, and case parts must return expected type.
4863 if (!cs->looppart && !type)
4864 type = propagate_types(cs->thenpart, c, perr, NULL, rules);
4866 type = propagate_types(cs->elsepart, c, perr, NULL, rules);
4867 for (cp = cs->casepart;
4869 cp = cp->next) // UNTESTED
4870 type = propagate_types(cp->action, c, perr, NULL, rules); // UNTESTED
4873 propagate_types(cs->thenpart, c, perr, type, rules);
4874 propagate_types(cs->elsepart, c, perr, type, rules);
4875 for (cp = cs->casepart; cp ; cp = cp->next)
4876 propagate_types(cp->action, c, perr, type, rules);
4882 ###### interp binode cases
4884 // This just performs one iterration of the loop
4885 rv = interp_exec(c, b->left, &rvtype);
4886 if (rvtype == Tnone ||
4887 (rvtype == Tbool && rv.bool != 0))
4888 // rvtype is Tnone or Tbool, doesn't need to be freed
4889 interp_exec(c, b->right, NULL);
4892 ###### interp exec cases
4893 case Xcond_statement:
4895 struct value v, cnd;
4896 struct type *vtype, *cndtype;
4897 struct casepart *cp;
4898 struct cond_statement *cs = cast(cond_statement, e);
4901 interp_exec(c, cs->forpart, NULL);
4903 while ((cnd = interp_exec(c, cs->looppart, &cndtype)),
4904 cndtype == Tnone || (cndtype == Tbool && cnd.bool != 0))
4905 interp_exec(c, cs->thenpart, NULL);
4907 cnd = interp_exec(c, cs->condpart, &cndtype);
4908 if ((cndtype == Tnone ||
4909 (cndtype == Tbool && cnd.bool != 0))) {
4910 // cnd is Tnone or Tbool, doesn't need to be freed
4911 rv = interp_exec(c, cs->thenpart, &rvtype);
4912 // skip else (and cases)
4916 for (cp = cs->casepart; cp; cp = cp->next) {
4917 v = interp_exec(c, cp->value, &vtype);
4918 if (value_cmp(cndtype, vtype, &v, &cnd) == 0) {
4919 free_value(vtype, &v);
4920 free_value(cndtype, &cnd);
4921 rv = interp_exec(c, cp->action, &rvtype);
4924 free_value(vtype, &v);
4926 free_value(cndtype, &cnd);
4928 rv = interp_exec(c, cs->elsepart, &rvtype);
4935 ### Top level structure
4937 All the language elements so far can be used in various places. Now
4938 it is time to clarify what those places are.
4940 At the top level of a file there will be a number of declarations.
4941 Many of the things that can be declared haven't been described yet,
4942 such as functions, procedures, imports, and probably more.
4943 For now there are two sorts of things that can appear at the top
4944 level. They are predefined constants, `struct` types, and the `main`
4945 function. While the syntax will allow the `main` function to appear
4946 multiple times, that will trigger an error if it is actually attempted.
4948 The various declarations do not return anything. They store the
4949 various declarations in the parse context.
4951 ###### Parser: grammar
4954 Ocean -> OptNL DeclarationList
4956 ## declare terminals
4964 DeclarationList -> Declaration
4965 | DeclarationList Declaration
4967 Declaration -> ERROR Newlines ${
4968 tok_err(c, // UNTESTED
4969 "error: unhandled parse error", &$1);
4975 ## top level grammar
4979 ### The `const` section
4981 As well as being defined in with the code that uses them, constants can
4982 be declared at the top level. These have full-file scope, so they are
4983 always `InScope`, even before(!) they have been declared. The value of
4984 a top level constant can be given as an expression, and this is
4985 evaluated after parsing and before execution.
4987 A function call can be used to evaluate a constant, but it will not have
4988 access to any program state, once such statement becomes meaningful.
4989 e.g. arguments and filesystem will not be visible.
4991 Constants are defined in a section that starts with the reserved word
4992 `const` and then has a block with a list of assignment statements.
4993 For syntactic consistency, these must use the double-colon syntax to
4994 make it clear that they are constants. Type can also be given: if
4995 not, the type will be determined during analysis, as with other
4998 ###### parse context
4999 struct binode *constlist;
5001 ###### top level grammar
5005 DeclareConstant -> const { IN OptNL ConstList OUT OptNL } Newlines
5006 | const { SimpleConstList } Newlines
5007 | const IN OptNL ConstList OUT Newlines
5008 | const SimpleConstList Newlines
5010 ConstList -> ConstList SimpleConstLine
5013 SimpleConstList -> SimpleConstList ; Const
5017 SimpleConstLine -> SimpleConstList Newlines
5018 | ERROR Newlines ${ tok_err(c, "Syntax error in constant", &$1); }$
5021 CType -> Type ${ $0 = $<1; }$
5025 Const -> IDENTIFIER :: CType = Expression ${ {
5027 struct binode *bl, *bv;
5028 struct var *var = new_pos(var, $ID);
5030 v = var_decl(c, $ID.txt);
5032 v->where_decl = var;
5038 v = var_ref(c, $1.txt);
5039 if (v->type == Tnone) {
5040 v->where_decl = var;
5046 tok_err(c, "error: name already declared", &$1);
5047 type_err(c, "info: this is where '%v' was first declared",
5048 v->where_decl, NULL, 0, NULL);
5060 bl->left = c->constlist;
5065 ###### core functions
5066 static void resolve_consts(struct parse_context *c)
5070 enum { none, some, cannot } progress = none;
5072 c->constlist = reorder_bilist(c->constlist);
5075 for (b = cast(binode, c->constlist); b;
5076 b = cast(binode, b->right)) {
5078 struct binode *vb = cast(binode, b->left);
5079 struct var *v = cast(var, vb->left);
5080 if (v->var->frame_pos >= 0)
5084 propagate_types(vb->right, c, &perr,
5086 } while (perr & Eretry);
5088 c->parse_error += 1;
5089 else if (!(perr & Enoconst)) {
5091 struct value res = interp_exec(
5092 c, vb->right, &v->var->type);
5093 global_alloc(c, v->var->type, v->var, &res);
5095 if (progress == cannot)
5096 type_err(c, "error: const %v cannot be resolved.",
5106 progress = cannot; break;
5108 progress = none; break;
5113 ###### print const decls
5118 for (b = cast(binode, context.constlist); b;
5119 b = cast(binode, b->right)) {
5120 struct binode *vb = cast(binode, b->left);
5121 struct var *vr = cast(var, vb->left);
5122 struct variable *v = vr->var;
5128 printf(" %.*s :: ", v->name->name.len, v->name->name.txt);
5129 type_print(v->type, stdout);
5131 print_exec(vb->right, -1, 0);
5136 ###### free const decls
5137 free_binode(context.constlist);
5139 ### Function declarations
5141 The code in an Ocean program is all stored in function declarations.
5142 One of the functions must be named `main` and it must accept an array of
5143 strings as a parameter - the command line arguments.
5145 As this is the top level, several things are handled a bit differently.
5146 The function is not interpreted by `interp_exec` as that isn't passed
5147 the argument list which the program requires. Similarly type analysis
5148 is a bit more interesting at this level.
5150 ###### ast functions
5152 static struct type *handle_results(struct parse_context *c,
5153 struct binode *results)
5155 /* Create a 'struct' type from the results list, which
5156 * is a list for 'struct var'
5158 struct type *t = add_anon_type(c, &structure_prototype,
5163 for (b = results; b; b = cast(binode, b->right))
5165 t->structure.nfields = cnt;
5166 t->structure.fields = calloc(cnt, sizeof(struct field));
5168 for (b = results; b; b = cast(binode, b->right)) {
5169 struct var *v = cast(var, b->left);
5170 struct field *f = &t->structure.fields[cnt++];
5171 int a = v->var->type->align;
5172 f->name = v->var->name->name;
5173 f->type = v->var->type;
5175 f->offset = t->size;
5176 v->var->frame_pos = f->offset;
5177 t->size += ((f->type->size - 1) | (a-1)) + 1;
5180 variable_unlink_exec(v->var);
5182 free_binode(results);
5186 static struct variable *declare_function(struct parse_context *c,
5187 struct variable *name,
5188 struct binode *args,
5190 struct binode *results,
5194 struct value fn = {.function = code};
5196 var_block_close(c, CloseFunction, code);
5197 t = add_anon_type(c, &function_prototype,
5198 "func %.*s", name->name->name.len,
5199 name->name->name.txt);
5201 t->function.params = reorder_bilist(args);
5203 ret = handle_results(c, reorder_bilist(results));
5204 t->function.inline_result = 1;
5205 t->function.local_size = ret->size;
5207 t->function.return_type = ret;
5208 global_alloc(c, t, name, &fn);
5209 name->type->function.scope = c->out_scope;
5214 var_block_close(c, CloseFunction, NULL);
5216 c->out_scope = NULL;
5220 ###### declare terminals
5223 ###### top level grammar
5226 DeclareFunction -> func FuncName ( OpenScope ArgsLine ) Block Newlines ${
5227 $0 = declare_function(c, $<FN, $<Ar, Tnone, NULL, $<Bl);
5229 | func FuncName IN OpenScope Args OUT OptNL do Block Newlines ${
5230 $0 = declare_function(c, $<FN, $<Ar, Tnone, NULL, $<Bl);
5232 | func FuncName NEWLINE OpenScope OptNL do Block Newlines ${
5233 $0 = declare_function(c, $<FN, NULL, Tnone, NULL, $<Bl);
5235 | func FuncName ( OpenScope ArgsLine ) : Type Block Newlines ${
5236 $0 = declare_function(c, $<FN, $<Ar, $<Ty, NULL, $<Bl);
5238 | func FuncName ( OpenScope ArgsLine ) : ( ArgsLine ) Block Newlines ${
5239 $0 = declare_function(c, $<FN, $<AL, NULL, $<AL2, $<Bl);
5241 | func FuncName IN OpenScope Args OUT OptNL return Type Newlines do Block Newlines ${
5242 $0 = declare_function(c, $<FN, $<Ar, $<Ty, NULL, $<Bl);
5244 | func FuncName NEWLINE OpenScope return Type Newlines do Block Newlines ${
5245 $0 = declare_function(c, $<FN, NULL, $<Ty, NULL, $<Bl);
5247 | func FuncName IN OpenScope Args OUT OptNL return IN Args OUT OptNL do Block Newlines ${
5248 $0 = declare_function(c, $<FN, $<Ar, NULL, $<Ar2, $<Bl);
5250 | func FuncName NEWLINE OpenScope return IN Args OUT OptNL do Block Newlines ${
5251 $0 = declare_function(c, $<FN, NULL, NULL, $<Ar, $<Bl);
5254 ###### print func decls
5259 while (target != 0) {
5261 for (v = context.in_scope; v; v=v->in_scope)
5262 if (v->depth == 0 && v->type && v->type->check_args) {
5271 struct value *val = var_value(&context, v);
5272 printf("func %.*s", v->name->name.len, v->name->name.txt);
5273 v->type->print_type_decl(v->type, stdout);
5275 print_exec(val->function, 0, brackets);
5277 print_value(v->type, val, stdout);
5278 printf("/* frame size %d */\n", v->type->function.local_size);
5284 ###### core functions
5286 static int analyse_funcs(struct parse_context *c)
5290 for (v = c->in_scope; v; v = v->in_scope) {
5294 if (v->depth != 0 || !v->type || !v->type->check_args)
5296 ret = v->type->function.inline_result ?
5297 Tnone : v->type->function.return_type;
5298 val = var_value(c, v);
5301 propagate_types(val->function, c, &perr, ret, 0);
5302 } while (!(perr & Efail) && (perr & Eretry));
5303 if (!(perr & Efail))
5304 /* Make sure everything is still consistent */
5305 propagate_types(val->function, c, &perr, ret, 0);
5308 if (!v->type->function.inline_result &&
5309 !v->type->function.return_type->dup) {
5310 type_err(c, "error: function cannot return value of type %1",
5311 v->where_decl, v->type->function.return_type, 0, NULL);
5314 scope_finalize(c, v->type);
5319 static int analyse_main(struct type *type, struct parse_context *c)
5321 struct binode *bp = type->function.params;
5325 struct type *argv_type;
5327 argv_type = add_anon_type(c, &array_prototype, "argv");
5328 argv_type->array.member = Tstr;
5329 argv_type->array.unspec = 1;
5331 for (b = bp; b; b = cast(binode, b->right)) {
5335 propagate_types(b->left, c, &perr, argv_type, 0);
5337 default: /* invalid */ // NOTEST
5338 propagate_types(b->left, c, &perr, Tnone, 0); // NOTEST
5341 c->parse_error += 1;
5344 return !c->parse_error;
5347 static void interp_main(struct parse_context *c, int argc, char **argv)
5349 struct value *progp = NULL;
5350 struct text main_name = { "main", 4 };
5351 struct variable *mainv;
5357 mainv = var_ref(c, main_name);
5359 progp = var_value(c, mainv);
5360 if (!progp || !progp->function) {
5361 fprintf(stderr, "oceani: no main function found.\n");
5362 c->parse_error += 1;
5365 if (!analyse_main(mainv->type, c)) {
5366 fprintf(stderr, "oceani: main has wrong type.\n");
5367 c->parse_error += 1;
5370 al = mainv->type->function.params;
5372 c->local_size = mainv->type->function.local_size;
5373 c->local = calloc(1, c->local_size);
5375 struct var *v = cast(var, al->left);
5376 struct value *vl = var_value(c, v->var);
5386 mpq_set_ui(argcq, argc, 1);
5387 memcpy(var_value(c, t->array.vsize), &argcq, sizeof(argcq));
5388 t->prepare_type(c, t, 0);
5389 array_init(v->var->type, vl);
5390 for (i = 0; i < argc; i++) {
5391 struct value *vl2 = vl->array + i * v->var->type->array.member->size;
5393 arg.str.txt = argv[i];
5394 arg.str.len = strlen(argv[i]);
5395 free_value(Tstr, vl2);
5396 dup_value(Tstr, &arg, vl2);
5400 al = cast(binode, al->right);
5402 v = interp_exec(c, progp->function, &vtype);
5403 free_value(vtype, &v);
5408 ###### ast functions
5409 void free_variable(struct variable *v)
5413 ## And now to test it out.
5415 Having a language requires having a "hello world" program. I'll
5416 provide a little more than that: a program that prints "Hello world"
5417 finds the GCD of two numbers, prints the first few elements of
5418 Fibonacci, performs a binary search for a number, and a few other
5419 things which will likely grow as the languages grows.
5421 ###### File: oceani.mk
5424 @echo "===== DEMO ====="
5425 ./oceani --section "demo: hello" oceani.mdc 55 33
5431 four ::= 2 + 2 ; five ::= 10/2
5432 const pie ::= "I like Pie";
5433 cake ::= "The cake is"
5441 func main(argv:[argc::]string)
5442 print "Hello World, what lovely oceans you have!"
5443 print "Are there", five, "?"
5444 print pi, pie, "but", cake
5446 A := $argv[1]; B := $argv[2]
5448 /* When a variable is defined in both branches of an 'if',
5449 * and used afterwards, the variables are merged.
5455 print "Is", A, "bigger than", B,"? ", bigger
5456 /* If a variable is not used after the 'if', no
5457 * merge happens, so types can be different
5460 double:string = "yes"
5461 print A, "is more than twice", B, "?", double
5464 print "double", B, "is", double
5469 if a > 0 and then b > 0:
5475 print "GCD of", A, "and", B,"is", a
5477 print a, "is not positive, cannot calculate GCD"
5479 print b, "is not positive, cannot calculate GCD"
5484 print "Fibonacci:", f1,f2,
5485 then togo = togo - 1
5493 /* Binary search... */
5498 mid := (lo + hi) / 2
5511 print "Yay, I found", target
5513 print "Closest I found was", lo
5518 // "middle square" PRNG. Not particularly good, but one my
5519 // Dad taught me - the first one I ever heard of.
5520 for i:=1; then i = i + 1; while i < size:
5521 n := list[i-1] * list[i-1]
5522 list[i] = (n / 100) % 10 000
5524 print "Before sort:",
5525 for i:=0; then i = i + 1; while i < size:
5529 for i := 1; then i=i+1; while i < size:
5530 for j:=i-1; then j=j-1; while j >= 0:
5531 if list[j] > list[j+1]:
5535 print " After sort:",
5536 for i:=0; then i = i + 1; while i < size:
5540 if 1 == 2 then print "yes"; else print "no"
5544 bob.alive = (bob.name == "Hello")
5545 print "bob", "is" if bob.alive else "isn't", "alive"