1 # Ocean Interpreter - Jamison Creek version
3 Ocean is intended to be a compiled language, so this interpreter is
4 not targeted at being the final product. It is, rather, an intermediate
5 stage and fills that role in two distinct ways.
7 Firstly, it exists as a platform to experiment with the early language
8 design. An interpreter is easy to write and easy to get working, so
9 the barrier for entry is lower if I aim to start with an interpreter.
11 Secondly, the plan for the Ocean compiler is to write it in the
12 [Ocean language](http://ocean-lang.org). To achieve this we naturally
13 need some sort of boot-strap process and this interpreter - written in
14 portable C - will fill that role. It will be used to bootstrap the
17 Two features that are not needed to fill either of these roles are
18 performance and completeness. The interpreter only needs to be fast
19 enough to run small test programs and occasionally to run the compiler
20 on itself. It only needs to be complete enough to test aspects of the
21 design which are developed before the compiler is working, and to run
22 the compiler on itself. Any features not used by the compiler when
23 compiling itself are superfluous. They may be included anyway, but
26 Nonetheless, the interpreter should end up being reasonably complete,
27 and any performance bottlenecks which appear and are easily fixed, will
32 This third version of the interpreter exists to test out some initial
33 ideas relating to types. Particularly it adds arrays (indexed from
34 zero) and simple structures. Basic control flow and variable scoping
35 are already fairly well established, as are basic numerical and
38 Some operators that have only recently been added, and so have not
39 generated all that much experience yet are "and then" and "or else" as
40 short-circuit Boolean operators, and the "if ... else" trinary
41 operator which can select between two expressions based on a third
42 (which appears syntactically in the middle).
44 The "func" clause currently only allows a "main" function to be
45 declared. That will be extended when proper function support is added.
47 An element that is present purely to make a usable language, and
48 without any expectation that they will remain, is the "print" statement
49 which performs simple output.
51 The current scalar types are "number", "Boolean", and "string".
52 Boolean will likely stay in its current form, the other two might, but
53 could just as easily be changed.
57 Versions of the interpreter which obviously do not support a complete
58 language will be named after creeks and streams. This one is Jamison
61 Once we have something reasonably resembling a complete language, the
62 names of rivers will be used.
63 Early versions of the compiler will be named after seas. Major
64 releases of the compiler will be named after oceans. Hopefully I will
65 be finished once I get to the Pacific Ocean release.
69 As well as parsing and executing a program, the interpreter can print
70 out the program from the parsed internal structure. This is useful
71 for validating the parsing.
72 So the main requirements of the interpreter are:
74 - Parse the program, possibly with tracing,
75 - Analyse the parsed program to ensure consistency,
77 - Execute the "main" function in the program, if no parsing or
78 consistency errors were found.
80 This is all performed by a single C program extracted with
83 There will be two formats for printing the program: a default and one
84 that uses bracketing. So a `--bracket` command line option is needed
85 for that. Normally the first code section found is used, however an
86 alternate section can be requested so that a file (such as this one)
87 can contain multiple programs. This is effected with the `--section`
90 This code must be compiled with `-fplan9-extensions` so that anonymous
91 structures can be used.
93 ###### File: oceani.mk
95 myCFLAGS := -Wall -g -fplan9-extensions
96 CFLAGS := $(filter-out $(myCFLAGS),$(CFLAGS)) $(myCFLAGS)
97 myLDLIBS:= libparser.o libscanner.o libmdcode.o -licuuc
98 LDLIBS := $(filter-out $(myLDLIBS),$(LDLIBS)) $(myLDLIBS)
100 all :: $(LDLIBS) oceani
101 oceani.c oceani.h : oceani.mdc parsergen
102 ./parsergen -o oceani --LALR --tag Parser oceani.mdc
103 oceani.mk: oceani.mdc md2c
106 oceani: oceani.o $(LDLIBS)
107 $(CC) $(CFLAGS) -o oceani oceani.o $(LDLIBS)
109 ###### Parser: header
111 struct parse_context;
113 struct parse_context {
114 struct token_config config;
122 #define container_of(ptr, type, member) ({ \
123 const typeof( ((type *)0)->member ) *__mptr = (ptr); \
124 (type *)( (char *)__mptr - offsetof(type,member) );})
126 #define config2context(_conf) container_of(_conf, struct parse_context, \
129 ###### Parser: reduce
130 struct parse_context *c = config2context(config);
138 #include <sys/mman.h>
157 static char Usage[] =
158 "Usage: oceani --trace --print --noexec --brackets --section=SectionName prog.ocn\n";
159 static const struct option long_options[] = {
160 {"trace", 0, NULL, 't'},
161 {"print", 0, NULL, 'p'},
162 {"noexec", 0, NULL, 'n'},
163 {"brackets", 0, NULL, 'b'},
164 {"section", 1, NULL, 's'},
167 const char *options = "tpnbs";
169 static void pr_err(char *msg) // NOTEST
171 fprintf(stderr, "%s\n", msg); // NOTEST
174 int main(int argc, char *argv[])
179 struct section *s = NULL, *ss;
180 char *section = NULL;
181 struct parse_context context = {
183 .ignored = (1 << TK_mark),
184 .number_chars = ".,_+- ",
189 int doprint=0, dotrace=0, doexec=1, brackets=0;
191 while ((opt = getopt_long(argc, argv, options, long_options, NULL))
194 case 't': dotrace=1; break;
195 case 'p': doprint=1; break;
196 case 'n': doexec=0; break;
197 case 'b': brackets=1; break;
198 case 's': section = optarg; break;
199 default: fprintf(stderr, Usage);
203 if (optind >= argc) {
204 fprintf(stderr, "oceani: no input file given\n");
207 fd = open(argv[optind], O_RDONLY);
209 fprintf(stderr, "oceani: cannot open %s\n", argv[optind]);
212 context.file_name = argv[optind];
213 len = lseek(fd, 0, 2);
214 file = mmap(NULL, len, PROT_READ, MAP_SHARED, fd, 0);
215 s = code_extract(file, file+len, pr_err);
217 fprintf(stderr, "oceani: could not find any code in %s\n",
222 ## context initialization
225 for (ss = s; ss; ss = ss->next) {
226 struct text sec = ss->section;
227 if (sec.len == strlen(section) &&
228 strncmp(sec.txt, section, sec.len) == 0)
232 fprintf(stderr, "oceani: cannot find section %s\n",
239 fprintf(stderr, "oceani: no code found in requested section\n"); // NOTEST
240 goto cleanup; // NOTEST
243 parse_oceani(ss->code, &context.config, dotrace ? stderr : NULL);
245 if (!context.parse_error && !analyse_funcs(&context)) {
246 fprintf(stderr, "oceani: type error in program - not running.\n");
247 context.parse_error = 1;
255 if (doexec && !context.parse_error)
256 interp_main(&context, argc - optind, argv + optind);
259 struct section *t = s->next;
264 // FIXME parser should pop scope even on error
265 while (context.scope_depth > 0)
268 ## free context types
269 ## free context storage
270 exit(context.parse_error ? 1 : 0);
275 The four requirements of parse, analyse, print, interpret apply to
276 each language element individually so that is how most of the code
279 Three of the four are fairly self explanatory. The one that requires
280 a little explanation is the analysis step.
282 The current language design does not require the types of variables to
283 be declared, but they must still have a single type. Different
284 operations impose different requirements on the variables, for example
285 addition requires both arguments to be numeric, and assignment
286 requires the variable on the left to have the same type as the
287 expression on the right.
289 Analysis involves propagating these type requirements around and
290 consequently setting the type of each variable. If any requirements
291 are violated (e.g. a string is compared with a number) or if a
292 variable needs to have two different types, then an error is raised
293 and the program will not run.
295 If the same variable is declared in both branchs of an 'if/else', or
296 in all cases of a 'switch' then the multiple instances may be merged
297 into just one variable if the variable is referenced after the
298 conditional statement. When this happens, the types must naturally be
299 consistent across all the branches. When the variable is not used
300 outside the if, the variables in the different branches are distinct
301 and can be of different types.
303 Undeclared names may only appear in "use" statements and "case" expressions.
304 These names are given a type of "label" and a unique value.
305 This allows them to fill the role of a name in an enumerated type, which
306 is useful for testing the `switch` statement.
308 As we will see, the condition part of a `while` statement can return
309 either a Boolean or some other type. This requires that the expected
310 type that gets passed around comprises a type and a flag to indicate
311 that `Tbool` is also permitted.
313 As there are, as yet, no distinct types that are compatible, there
314 isn't much subtlety in the analysis. When we have distinct number
315 types, this will become more interesting.
319 When analysis discovers an inconsistency it needs to report an error;
320 just refusing to run the code ensures that the error doesn't cascade,
321 but by itself it isn't very useful. A clear understanding of the sort
322 of error message that are useful will help guide the process of
325 At a simplistic level, the only sort of error that type analysis can
326 report is that the type of some construct doesn't match a contextual
327 requirement. For example, in `4 + "hello"` the addition provides a
328 contextual requirement for numbers, but `"hello"` is not a number. In
329 this particular example no further information is needed as the types
330 are obvious from local information. When a variable is involved that
331 isn't the case. It may be helpful to explain why the variable has a
332 particular type, by indicating the location where the type was set,
333 whether by declaration or usage.
335 Using a recursive-descent analysis we can easily detect a problem at
336 multiple locations. In "`hello:= "there"; 4 + hello`" the addition
337 will detect that one argument is not a number and the usage of `hello`
338 will detect that a number was wanted, but not provided. In this
339 (early) version of the language, we will generate error reports at
340 multiple locations, so the use of `hello` will report an error and
341 explain were the value was set, and the addition will report an error
342 and say why numbers are needed. To be able to report locations for
343 errors, each language element will need to record a file location
344 (line and column) and each variable will need to record the language
345 element where its type was set. For now we will assume that each line
346 of an error message indicates one location in the file, and up to 2
347 types. So we provide a `printf`-like function which takes a format, a
348 location (a `struct exec` which has not yet been introduced), and 2
349 types. "`%1`" reports the first type, "`%2`" reports the second. We
350 will need a function to print the location, once we know how that is
351 stored. e As will be explained later, there are sometimes extra rules for
352 type matching and they might affect error messages, we need to pass those
355 As well as type errors, we sometimes need to report problems with
356 tokens, which might be unexpected or might name a type that has not
357 been defined. For these we have `tok_err()` which reports an error
358 with a given token. Each of the error functions sets the flag in the
359 context so indicate that parsing failed.
363 static void fput_loc(struct exec *loc, FILE *f);
364 static void type_err(struct parse_context *c,
365 char *fmt, struct exec *loc,
366 struct type *t1, int rules, struct type *t2);
368 ###### core functions
370 static void type_err(struct parse_context *c,
371 char *fmt, struct exec *loc,
372 struct type *t1, int rules, struct type *t2)
374 fprintf(stderr, "%s:", c->file_name);
375 fput_loc(loc, stderr);
376 for (; *fmt ; fmt++) {
383 case '%': fputc(*fmt, stderr); break; // NOTEST
384 default: fputc('?', stderr); break; // NOTEST
386 type_print(t1, stderr);
389 type_print(t2, stderr);
398 static void tok_err(struct parse_context *c, char *fmt, struct token *t)
400 fprintf(stderr, "%s:%d:%d: %s: %.*s\n", c->file_name, t->line, t->col, fmt,
401 t->txt.len, t->txt.txt);
405 ## Entities: declared and predeclared.
407 There are various "things" that the language and/or the interpreter
408 needs to know about to parse and execute a program. These include
409 types, variables, values, and executable code. These are all lumped
410 together under the term "entities" (calling them "objects" would be
411 confusing) and introduced here. The following section will present the
412 different specific code elements which comprise or manipulate these
417 Values come in a wide range of types, with more likely to be added.
418 Each type needs to be able to print its own values (for convenience at
419 least) as well as to compare two values, at least for equality and
420 possibly for order. For now, values might need to be duplicated and
421 freed, though eventually such manipulations will be better integrated
424 Rather than requiring every numeric type to support all numeric
425 operations (add, multiple, etc), we allow types to be able to present
426 as one of a few standard types: integer, float, and fraction. The
427 existence of these conversion functions eventually enable types to
428 determine if they are compatible with other types, though such types
429 have not yet been implemented.
431 Named type are stored in a simple linked list. Objects of each type are
432 "values" which are often passed around by value.
439 ## value union fields
447 void (*init)(struct type *type, struct value *val);
448 void (*prepare_type)(struct parse_context *c, struct type *type, int parse_time);
449 void (*print)(struct type *type, struct value *val, FILE *f);
450 void (*print_type)(struct type *type, FILE *f);
451 int (*cmp_order)(struct type *t1, struct type *t2,
452 struct value *v1, struct value *v2);
453 int (*cmp_eq)(struct type *t1, struct type *t2,
454 struct value *v1, struct value *v2);
455 void (*dup)(struct type *type, struct value *vold, struct value *vnew);
456 void (*free)(struct type *type, struct value *val);
457 void (*free_type)(struct type *t);
458 long long (*to_int)(struct value *v);
459 double (*to_float)(struct value *v);
460 int (*to_mpq)(mpq_t *q, struct value *v);
469 struct type *typelist;
473 static struct type *find_type(struct parse_context *c, struct text s)
475 struct type *l = c->typelist;
478 text_cmp(l->name, s) != 0)
483 static struct type *add_type(struct parse_context *c, struct text s,
488 n = calloc(1, sizeof(*n));
491 n->next = c->typelist;
496 static void free_type(struct type *t)
498 /* The type is always a reference to something in the
499 * context, so we don't need to free anything.
503 static void free_value(struct type *type, struct value *v)
507 memset(v, 0x5a, type->size);
511 static void type_print(struct type *type, FILE *f)
514 fputs("*unknown*type*", f); // NOTEST
515 else if (type->name.len)
516 fprintf(f, "%.*s", type->name.len, type->name.txt);
517 else if (type->print_type)
518 type->print_type(type, f);
520 fputs("*invalid*type*", f); // NOTEST
523 static void val_init(struct type *type, struct value *val)
525 if (type && type->init)
526 type->init(type, val);
529 static void dup_value(struct type *type,
530 struct value *vold, struct value *vnew)
532 if (type && type->dup)
533 type->dup(type, vold, vnew);
536 static int value_cmp(struct type *tl, struct type *tr,
537 struct value *left, struct value *right)
539 if (tl && tl->cmp_order)
540 return tl->cmp_order(tl, tr, left, right);
541 if (tl && tl->cmp_eq) // NOTEST
542 return tl->cmp_eq(tl, tr, left, right); // NOTEST
546 static void print_value(struct type *type, struct value *v, FILE *f)
548 if (type && type->print)
549 type->print(type, v, f);
551 fprintf(f, "*Unknown*"); // NOTEST
556 static void free_value(struct type *type, struct value *v);
557 static int type_compat(struct type *require, struct type *have, int rules);
558 static void type_print(struct type *type, FILE *f);
559 static void val_init(struct type *type, struct value *v);
560 static void dup_value(struct type *type,
561 struct value *vold, struct value *vnew);
562 static int value_cmp(struct type *tl, struct type *tr,
563 struct value *left, struct value *right);
564 static void print_value(struct type *type, struct value *v, FILE *f);
566 ###### free context types
568 while (context.typelist) {
569 struct type *t = context.typelist;
571 context.typelist = t->next;
577 Type can be specified for local variables, for fields in a structure,
578 for formal parameters to functions, and possibly elsewhere. Different
579 rules may apply in different contexts. As a minimum, a named type may
580 always be used. Currently the type of a formal parameter can be
581 different from types in other contexts, so we have a separate grammar
587 Type -> IDENTIFIER ${
588 $0 = find_type(c, $1.txt);
591 "error: undefined type", &$1);
598 FormalType -> Type ${ $0 = $<1; }$
599 ## formal type grammar
603 Values of the base types can be numbers, which we represent as
604 multi-precision fractions, strings, Booleans and labels. When
605 analysing the program we also need to allow for places where no value
606 is meaningful (type `Tnone`) and where we don't know what type to
607 expect yet (type is `NULL`).
609 Values are never shared, they are always copied when used, and freed
610 when no longer needed.
612 When propagating type information around the program, we need to
613 determine if two types are compatible, where type `NULL` is compatible
614 with anything. There are two special cases with type compatibility,
615 both related to the Conditional Statement which will be described
616 later. In some cases a Boolean can be accepted as well as some other
617 primary type, and in others any type is acceptable except a label (`Vlabel`).
618 A separate function encoding these cases will simplify some code later.
620 ###### type functions
622 int (*compat)(struct type *this, struct type *other);
626 static int type_compat(struct type *require, struct type *have, int rules)
628 if ((rules & Rboolok) && have == Tbool)
630 if ((rules & Rnolabel) && have == Tlabel)
632 if (!require || !have)
636 return require->compat(require, have);
638 return require == have;
643 #include "parse_string.h"
644 #include "parse_number.h"
647 myLDLIBS := libnumber.o libstring.o -lgmp
648 LDLIBS := $(filter-out $(myLDLIBS),$(LDLIBS)) $(myLDLIBS)
650 ###### type union fields
651 enum vtype {Vnone, Vstr, Vnum, Vbool, Vlabel} vtype;
653 ###### value union fields
660 static void _free_value(struct type *type, struct value *v)
664 switch (type->vtype) {
666 case Vstr: free(v->str.txt); break;
667 case Vnum: mpq_clear(v->num); break;
673 ###### value functions
675 static void _val_init(struct type *type, struct value *val)
677 switch(type->vtype) {
678 case Vnone: // NOTEST
681 mpq_init(val->num); break;
683 val->str.txt = malloc(1);
695 static void _dup_value(struct type *type,
696 struct value *vold, struct value *vnew)
698 switch (type->vtype) {
699 case Vnone: // NOTEST
702 vnew->label = vold->label;
705 vnew->bool = vold->bool;
709 mpq_set(vnew->num, vold->num);
712 vnew->str.len = vold->str.len;
713 vnew->str.txt = malloc(vnew->str.len);
714 memcpy(vnew->str.txt, vold->str.txt, vnew->str.len);
719 static int _value_cmp(struct type *tl, struct type *tr,
720 struct value *left, struct value *right)
724 return tl - tr; // NOTEST
726 case Vlabel: cmp = left->label == right->label ? 0 : 1; break;
727 case Vnum: cmp = mpq_cmp(left->num, right->num); break;
728 case Vstr: cmp = text_cmp(left->str, right->str); break;
729 case Vbool: cmp = left->bool - right->bool; break;
730 case Vnone: cmp = 0; // NOTEST
735 static void _print_value(struct type *type, struct value *v, FILE *f)
737 switch (type->vtype) {
738 case Vnone: // NOTEST
739 fprintf(f, "*no-value*"); break; // NOTEST
740 case Vlabel: // NOTEST
741 fprintf(f, "*label-%p*", v->label); break; // NOTEST
743 fprintf(f, "%.*s", v->str.len, v->str.txt); break;
745 fprintf(f, "%s", v->bool ? "True":"False"); break;
750 mpf_set_q(fl, v->num);
751 gmp_fprintf(f, "%Fg", fl);
758 static void _free_value(struct type *type, struct value *v);
760 static struct type base_prototype = {
762 .print = _print_value,
763 .cmp_order = _value_cmp,
764 .cmp_eq = _value_cmp,
769 static struct type *Tbool, *Tstr, *Tnum, *Tnone, *Tlabel;
772 static struct type *add_base_type(struct parse_context *c, char *n,
773 enum vtype vt, int size)
775 struct text txt = { n, strlen(n) };
778 t = add_type(c, txt, &base_prototype);
781 t->align = size > sizeof(void*) ? sizeof(void*) : size;
782 if (t->size & (t->align - 1))
783 t->size = (t->size | (t->align - 1)) + 1; // NOTEST
787 ###### context initialization
789 Tbool = add_base_type(&context, "Boolean", Vbool, sizeof(char));
790 Tstr = add_base_type(&context, "string", Vstr, sizeof(struct text));
791 Tnum = add_base_type(&context, "number", Vnum, sizeof(mpq_t));
792 Tnone = add_base_type(&context, "none", Vnone, 0);
793 Tlabel = add_base_type(&context, "label", Vlabel, sizeof(void*));
797 Variables are scoped named values. We store the names in a linked list
798 of "bindings" sorted in lexical order, and use sequential search and
805 struct binding *next; // in lexical order
809 This linked list is stored in the parse context so that "reduce"
810 functions can find or add variables, and so the analysis phase can
811 ensure that every variable gets a type.
815 struct binding *varlist; // In lexical order
819 static struct binding *find_binding(struct parse_context *c, struct text s)
821 struct binding **l = &c->varlist;
826 (cmp = text_cmp((*l)->name, s)) < 0)
830 n = calloc(1, sizeof(*n));
837 Each name can be linked to multiple variables defined in different
838 scopes. Each scope starts where the name is declared and continues
839 until the end of the containing code block. Scopes of a given name
840 cannot nest, so a declaration while a name is in-scope is an error.
842 ###### binding fields
843 struct variable *var;
847 struct variable *previous;
849 struct binding *name;
850 struct exec *where_decl;// where name was declared
851 struct exec *where_set; // where type was set
855 When a scope closes, the values of the variables might need to be freed.
856 This happens in the context of some `struct exec` and each `exec` will
857 need to know which variables need to be freed when it completes.
860 struct variable *to_free;
862 ####### variable fields
863 struct exec *cleanup_exec;
864 struct variable *next_free;
866 ####### interp exec cleanup
869 for (v = e->to_free; v; v = v->next_free) {
870 struct value *val = var_value(c, v);
871 free_value(v->type, val);
876 static void variable_unlink_exec(struct variable *v)
878 struct variable **vp;
879 if (!v->cleanup_exec)
881 for (vp = &v->cleanup_exec->to_free;
882 *vp; vp = &(*vp)->next_free) {
886 v->cleanup_exec = NULL;
891 While the naming seems strange, we include local constants in the
892 definition of variables. A name declared `var := value` can
893 subsequently be changed, but a name declared `var ::= value` cannot -
896 ###### variable fields
899 Scopes in parallel branches can be partially merged. More
900 specifically, if a given name is declared in both branches of an
901 if/else then its scope is a candidate for merging. Similarly if
902 every branch of an exhaustive switch (e.g. has an "else" clause)
903 declares a given name, then the scopes from the branches are
904 candidates for merging.
906 Note that names declared inside a loop (which is only parallel to
907 itself) are never visible after the loop. Similarly names defined in
908 scopes which are not parallel, such as those started by `for` and
909 `switch`, are never visible after the scope. Only variables defined in
910 both `then` and `else` (including the implicit then after an `if`, and
911 excluding `then` used with `for`) and in all `case`s and `else` of a
912 `switch` or `while` can be visible beyond the `if`/`switch`/`while`.
914 Labels, which are a bit like variables, follow different rules.
915 Labels are not explicitly declared, but if an undeclared name appears
916 in a context where a label is legal, that effectively declares the
917 name as a label. The declaration remains in force (or in scope) at
918 least to the end of the immediately containing block and conditionally
919 in any larger containing block which does not declare the name in some
920 other way. Importantly, the conditional scope extension happens even
921 if the label is only used in one parallel branch of a conditional --
922 when used in one branch it is treated as having been declared in all
925 Merge candidates are tentatively visible beyond the end of the
926 branching statement which creates them. If the name is used, the
927 merge is affirmed and they become a single variable visible at the
928 outer layer. If not - if it is redeclared first - the merge lapses.
930 To track scopes we have an extra stack, implemented as a linked list,
931 which roughly parallels the parse stack and which is used exclusively
932 for scoping. When a new scope is opened, a new frame is pushed and
933 the child-count of the parent frame is incremented. This child-count
934 is used to distinguish between the first of a set of parallel scopes,
935 in which declared variables must not be in scope, and subsequent
936 branches, whether they may already be conditionally scoped.
938 We need a total ordering of scopes so we can easily compare to variables
939 to see if they are concurrently in scope. To achieve this we record a
940 `scope_count` which is actually a count of both beginnings and endings
941 of scopes. Then each variable has a record of the scope count where it
942 enters scope, and where it leaves.
944 To push a new frame *before* any code in the frame is parsed, we need a
945 grammar reduction. This is most easily achieved with a grammar
946 element which derives the empty string, and creates the new scope when
947 it is recognised. This can be placed, for example, between a keyword
948 like "if" and the code following it.
952 struct scope *parent;
959 struct scope *scope_stack;
961 ###### variable fields
962 int scope_start, scope_end;
965 static void scope_pop(struct parse_context *c)
967 struct scope *s = c->scope_stack;
969 c->scope_stack = s->parent;
975 static void scope_push(struct parse_context *c)
977 struct scope *s = calloc(1, sizeof(*s));
979 c->scope_stack->child_count += 1;
980 s->parent = c->scope_stack;
989 OpenScope -> ${ scope_push(c); }$
991 Each variable records a scope depth and is in one of four states:
993 - "in scope". This is the case between the declaration of the
994 variable and the end of the containing block, and also between
995 the usage with affirms a merge and the end of that block.
997 The scope depth is not greater than the current parse context scope
998 nest depth. When the block of that depth closes, the state will
999 change. To achieve this, all "in scope" variables are linked
1000 together as a stack in nesting order.
1002 - "pending". The "in scope" block has closed, but other parallel
1003 scopes are still being processed. So far, every parallel block at
1004 the same level that has closed has declared the name.
1006 The scope depth is the depth of the last parallel block that
1007 enclosed the declaration, and that has closed.
1009 - "conditionally in scope". The "in scope" block and all parallel
1010 scopes have closed, and no further mention of the name has been seen.
1011 This state includes a secondary nest depth (`min_depth`) which records
1012 the outermost scope seen since the variable became conditionally in
1013 scope. If a use of the name is found, the variable becomes "in scope"
1014 and that secondary depth becomes the recorded scope depth. If the
1015 name is declared as a new variable, the old variable becomes "out of
1016 scope" and the recorded scope depth stays unchanged.
1018 - "out of scope". The variable is neither in scope nor conditionally
1019 in scope. It is permanently out of scope now and can be removed from
1020 the "in scope" stack. When a variable becomes out-of-scope it is
1021 moved to a separate list (`out_scope`) of variables which have fully
1022 known scope. This will be used at the end of each function to assign
1023 each variable a place in the stack frame.
1025 ###### variable fields
1026 int depth, min_depth;
1027 enum { OutScope, PendingScope, CondScope, InScope } scope;
1028 struct variable *in_scope;
1030 ###### parse context
1032 struct variable *in_scope;
1033 struct variable *out_scope;
1035 All variables with the same name are linked together using the
1036 'previous' link. Those variable that have been affirmatively merged all
1037 have a 'merged' pointer that points to one primary variable - the most
1038 recently declared instance. When merging variables, we need to also
1039 adjust the 'merged' pointer on any other variables that had previously
1040 been merged with the one that will no longer be primary.
1042 A variable that is no longer the most recent instance of a name may
1043 still have "pending" scope, if it might still be merged with most
1044 recent instance. These variables don't really belong in the
1045 "in_scope" list, but are not immediately removed when a new instance
1046 is found. Instead, they are detected and ignored when considering the
1047 list of in_scope names.
1049 The storage of the value of a variable will be described later. For now
1050 we just need to know that when a variable goes out of scope, it might
1051 need to be freed. For this we need to be able to find it, so assume that
1052 `var_value()` will provide that.
1054 ###### variable fields
1055 struct variable *merged;
1057 ###### ast functions
1059 static void variable_merge(struct variable *primary, struct variable *secondary)
1063 primary = primary->merged;
1065 for (v = primary->previous; v; v=v->previous)
1066 if (v == secondary || v == secondary->merged ||
1067 v->merged == secondary ||
1068 v->merged == secondary->merged) {
1069 v->scope = OutScope;
1070 v->merged = primary;
1071 if (v->scope_start < primary->scope_start)
1072 primary->scope_start = v->scope_start;
1073 if (v->scope_end > primary->scope_end)
1074 primary->scope_end = v->scope_end; // NOTEST
1075 variable_unlink_exec(v);
1079 ###### forward decls
1080 static struct value *var_value(struct parse_context *c, struct variable *v);
1082 ###### free global vars
1084 while (context.varlist) {
1085 struct binding *b = context.varlist;
1086 struct variable *v = b->var;
1087 context.varlist = b->next;
1090 struct variable *next = v->previous;
1093 free_value(v->type, var_value(&context, v));
1095 // This is a global constant
1096 free_exec(v->where_decl);
1103 #### Manipulating Bindings
1105 When a name is conditionally visible, a new declaration discards the old
1106 binding - the condition lapses. Similarly when we reach the end of a
1107 function (outermost non-global scope) any conditional scope must lapse.
1108 Conversely a usage of the name affirms the visibility and extends it to
1109 the end of the containing block - i.e. the block that contains both the
1110 original declaration and the latest usage. This is determined from
1111 `min_depth`. When a conditionally visible variable gets affirmed like
1112 this, it is also merged with other conditionally visible variables with
1115 When we parse a variable declaration we either report an error if the
1116 name is currently bound, or create a new variable at the current nest
1117 depth if the name is unbound or bound to a conditionally scoped or
1118 pending-scope variable. If the previous variable was conditionally
1119 scoped, it and its homonyms becomes out-of-scope.
1121 When we parse a variable reference (including non-declarative assignment
1122 "foo = bar") we report an error if the name is not bound or is bound to
1123 a pending-scope variable; update the scope if the name is bound to a
1124 conditionally scoped variable; or just proceed normally if the named
1125 variable is in scope.
1127 When we exit a scope, any variables bound at this level are either
1128 marked out of scope or pending-scoped, depending on whether the scope
1129 was sequential or parallel. Here a "parallel" scope means the "then"
1130 or "else" part of a conditional, or any "case" or "else" branch of a
1131 switch. Other scopes are "sequential".
1133 When exiting a parallel scope we check if there are any variables that
1134 were previously pending and are still visible. If there are, then
1135 they weren't redeclared in the most recent scope, so they cannot be
1136 merged and must become out-of-scope. If it is not the first of
1137 parallel scopes (based on `child_count`), we check that there was a
1138 previous binding that is still pending-scope. If there isn't, the new
1139 variable must now be out-of-scope.
1141 When exiting a sequential scope that immediately enclosed parallel
1142 scopes, we need to resolve any pending-scope variables. If there was
1143 no `else` clause, and we cannot determine that the `switch` was exhaustive,
1144 we need to mark all pending-scope variable as out-of-scope. Otherwise
1145 all pending-scope variables become conditionally scoped.
1148 enum closetype { CloseSequential, CloseFunction, CloseParallel, CloseElse };
1150 ###### ast functions
1152 static struct variable *var_decl(struct parse_context *c, struct text s)
1154 struct binding *b = find_binding(c, s);
1155 struct variable *v = b->var;
1157 switch (v ? v->scope : OutScope) {
1159 /* Caller will report the error */
1163 v && v->scope == CondScope;
1165 v->scope = OutScope;
1169 v = calloc(1, sizeof(*v));
1170 v->previous = b->var;
1174 v->min_depth = v->depth = c->scope_depth;
1176 v->in_scope = c->in_scope;
1177 v->scope_start = c->scope_count;
1183 static struct variable *var_ref(struct parse_context *c, struct text s)
1185 struct binding *b = find_binding(c, s);
1186 struct variable *v = b->var;
1187 struct variable *v2;
1189 switch (v ? v->scope : OutScope) {
1192 /* Caller will report the error */
1195 /* All CondScope variables of this name need to be merged
1196 * and become InScope
1198 v->depth = v->min_depth;
1200 for (v2 = v->previous;
1201 v2 && v2->scope == CondScope;
1203 variable_merge(v, v2);
1211 static int var_refile(struct parse_context *c, struct variable *v)
1213 /* Variable just went out of scope. Add it to the out_scope
1214 * list, sorted by ->scope_start
1216 struct variable **vp = &c->out_scope;
1217 while ((*vp) && (*vp)->scope_start < v->scope_start)
1218 vp = &(*vp)->in_scope;
1224 static void var_block_close(struct parse_context *c, enum closetype ct,
1227 /* Close off all variables that are in_scope.
1228 * Some variables in c->scope may already be not-in-scope,
1229 * such as when a PendingScope variable is hidden by a new
1230 * variable with the same name.
1231 * So we check for v->name->var != v and drop them.
1232 * If we choose to make a variable OutScope, we drop it
1235 struct variable *v, **vp, *v2;
1238 for (vp = &c->in_scope;
1239 (v = *vp) && v->min_depth > c->scope_depth;
1240 (v->scope == OutScope || v->name->var != v)
1241 ? (*vp = v->in_scope, var_refile(c, v))
1242 : ( vp = &v->in_scope, 0)) {
1243 v->min_depth = c->scope_depth;
1244 if (v->name->var != v)
1245 /* This is still in scope, but we haven't just
1249 v->min_depth = c->scope_depth;
1250 if (v->scope == InScope)
1251 v->scope_end = c->scope_count;
1252 if (v->scope == InScope && e && !v->global) {
1253 /* This variable gets cleaned up when 'e' finishes */
1254 variable_unlink_exec(v);
1255 v->cleanup_exec = e;
1256 v->next_free = e->to_free;
1261 case CloseParallel: /* handle PendingScope */
1265 if (c->scope_stack->child_count == 1)
1266 /* first among parallel branches */
1267 v->scope = PendingScope;
1268 else if (v->previous &&
1269 v->previous->scope == PendingScope)
1270 /* all previous branches used name */
1271 v->scope = PendingScope;
1272 else if (v->type == Tlabel)
1273 /* Labels remain pending even when not used */
1274 v->scope = PendingScope; // UNTESTED
1276 v->scope = OutScope;
1277 if (ct == CloseElse) {
1278 /* All Pending variables with this name
1279 * are now Conditional */
1281 v2 && v2->scope == PendingScope;
1283 v2->scope = CondScope;
1287 /* Not possible as it would require
1288 * parallel scope to be nested immediately
1289 * in a parallel scope, and that never
1293 /* Not possible as we already tested for
1300 if (v->scope == CondScope)
1301 /* Condition cannot continue past end of function */
1304 case CloseSequential:
1305 if (v->type == Tlabel)
1306 v->scope = PendingScope;
1309 v->scope = OutScope;
1312 /* There was no 'else', so we can only become
1313 * conditional if we know the cases were exhaustive,
1314 * and that doesn't mean anything yet.
1315 * So only labels become conditional..
1318 v2 && v2->scope == PendingScope;
1320 if (v2->type == Tlabel)
1321 v2->scope = CondScope;
1323 v2->scope = OutScope;
1326 case OutScope: break;
1335 The value of a variable is store separately from the variable, on an
1336 analogue of a stack frame. There are (currently) two frames that can be
1337 active. A global frame which currently only stores constants, and a
1338 stacked frame which stores local variables. Each variable knows if it
1339 is global or not, and what its index into the frame is.
1341 Values in the global frame are known immediately they are relevant, so
1342 the frame needs to be reallocated as it grows so it can store those
1343 values. The local frame doesn't get values until the interpreted phase
1344 is started, so there is no need to allocate until the size is known.
1346 We initialize the `frame_pos` to an impossible value, so that we can
1347 tell if it was set or not later.
1349 ###### variable fields
1353 ###### variable init
1356 ###### parse context
1358 short global_size, global_alloc;
1360 void *global, *local;
1362 ###### ast functions
1364 static struct value *var_value(struct parse_context *c, struct variable *v)
1367 if (!c->local || !v->type)
1368 return NULL; // NOTEST
1369 if (v->frame_pos + v->type->size > c->local_size) {
1370 printf("INVALID frame_pos\n"); // NOTEST
1373 return c->local + v->frame_pos;
1375 if (c->global_size > c->global_alloc) {
1376 int old = c->global_alloc;
1377 c->global_alloc = (c->global_size | 1023) + 1024;
1378 c->global = realloc(c->global, c->global_alloc);
1379 memset(c->global + old, 0, c->global_alloc - old);
1381 return c->global + v->frame_pos;
1384 static struct value *global_alloc(struct parse_context *c, struct type *t,
1385 struct variable *v, struct value *init)
1388 struct variable scratch;
1390 if (t->prepare_type)
1391 t->prepare_type(c, t, 1); // NOTEST
1393 if (c->global_size & (t->align - 1))
1394 c->global_size = (c->global_size + t->align) & ~(t->align-1);
1399 v->frame_pos = c->global_size;
1401 c->global_size += v->type->size;
1402 ret = var_value(c, v);
1404 memcpy(ret, init, t->size);
1410 As global values are found -- struct field initializers, labels etc --
1411 `global_alloc()` is called to record the value in the global frame.
1413 When the program is fully parsed, each function is analysed, we need to
1414 walk the list of variables local to that function and assign them an
1415 offset in the stack frame. For this we have `scope_finalize()`.
1417 We keep the stack from dense by re-using space for between variables
1418 that are not in scope at the same time. The `out_scope` list is sorted
1419 by `scope_start` and as we process a varible, we move it to an FIFO
1420 stack. For each variable we consider, we first discard any from the
1421 stack anything that went out of scope before the new variable came in.
1422 Then we place the new variable just after the one at the top of the
1425 ###### ast functions
1427 static void scope_finalize(struct parse_context *c, struct type *ft)
1429 int size = ft->function.local_size;
1430 struct variable *next = ft->function.scope;
1431 struct variable *done = NULL;
1434 struct variable *v = next;
1435 struct type *t = v->type;
1442 if (v->frame_pos >= 0)
1444 while (done && done->scope_end < v->scope_start)
1445 done = done->in_scope;
1447 pos = done->frame_pos + done->type->size;
1449 pos = ft->function.local_size;
1450 if (pos & (t->align - 1))
1451 pos = (pos + t->align) & ~(t->align-1);
1453 if (size < pos + v->type->size)
1454 size = pos + v->type->size;
1458 c->out_scope = NULL;
1459 ft->function.local_size = size;
1462 ###### free context storage
1463 free(context.global);
1467 Executables can be lots of different things. In many cases an
1468 executable is just an operation combined with one or two other
1469 executables. This allows for expressions and lists etc. Other times an
1470 executable is something quite specific like a constant or variable name.
1471 So we define a `struct exec` to be a general executable with a type, and
1472 a `struct binode` which is a subclass of `exec`, forms a node in a
1473 binary tree, and holds an operation. There will be other subclasses,
1474 and to access these we need to be able to `cast` the `exec` into the
1475 various other types. The first field in any `struct exec` is the type
1476 from the `exec_types` enum.
1479 #define cast(structname, pointer) ({ \
1480 const typeof( ((struct structname *)0)->type) *__mptr = &(pointer)->type; \
1481 if (__mptr && *__mptr != X##structname) abort(); \
1482 (struct structname *)( (char *)__mptr);})
1484 #define new(structname) ({ \
1485 struct structname *__ptr = ((struct structname *)calloc(1,sizeof(struct structname))); \
1486 __ptr->type = X##structname; \
1487 __ptr->line = -1; __ptr->column = -1; \
1490 #define new_pos(structname, token) ({ \
1491 struct structname *__ptr = ((struct structname *)calloc(1,sizeof(struct structname))); \
1492 __ptr->type = X##structname; \
1493 __ptr->line = token.line; __ptr->column = token.col; \
1502 enum exec_types type;
1511 struct exec *left, *right;
1514 ###### ast functions
1516 static int __fput_loc(struct exec *loc, FILE *f)
1520 if (loc->line >= 0) {
1521 fprintf(f, "%d:%d: ", loc->line, loc->column);
1524 if (loc->type == Xbinode)
1525 return __fput_loc(cast(binode,loc)->left, f) ||
1526 __fput_loc(cast(binode,loc)->right, f); // NOTEST
1529 static void fput_loc(struct exec *loc, FILE *f)
1531 if (!__fput_loc(loc, f))
1532 fprintf(f, "??:??: ");
1535 Each different type of `exec` node needs a number of functions defined,
1536 a bit like methods. We must be able to free it, print it, analyse it
1537 and execute it. Once we have specific `exec` types we will need to
1538 parse them too. Let's take this a bit more slowly.
1542 The parser generator requires a `free_foo` function for each struct
1543 that stores attributes and they will often be `exec`s and subtypes
1544 there-of. So we need `free_exec` which can handle all the subtypes,
1545 and we need `free_binode`.
1547 ###### ast functions
1549 static void free_binode(struct binode *b)
1554 free_exec(b->right);
1558 ###### core functions
1559 static void free_exec(struct exec *e)
1568 ###### forward decls
1570 static void free_exec(struct exec *e);
1572 ###### free exec cases
1573 case Xbinode: free_binode(cast(binode, e)); break;
1577 Printing an `exec` requires that we know the current indent level for
1578 printing line-oriented components. As will become clear later, we
1579 also want to know what sort of bracketing to use.
1581 ###### ast functions
1583 static void do_indent(int i, char *str)
1590 ###### core functions
1591 static void print_binode(struct binode *b, int indent, int bracket)
1595 ## print binode cases
1599 static void print_exec(struct exec *e, int indent, int bracket)
1605 print_binode(cast(binode, e), indent, bracket); break;
1610 do_indent(indent, "/* FREE");
1611 for (v = e->to_free; v; v = v->next_free) {
1612 printf(" %.*s", v->name->name.len, v->name->name.txt);
1613 printf("[%d,%d]", v->scope_start, v->scope_end);
1614 if (v->frame_pos >= 0)
1615 printf("(%d+%d)", v->frame_pos,
1616 v->type ? v->type->size:0);
1622 ###### forward decls
1624 static void print_exec(struct exec *e, int indent, int bracket);
1628 As discussed, analysis involves propagating type requirements around the
1629 program and looking for errors.
1631 So `propagate_types` is passed an expected type (being a `struct type`
1632 pointer together with some `val_rules` flags) that the `exec` is
1633 expected to return, and returns the type that it does return, either
1634 of which can be `NULL` signifying "unknown". An `ok` flag is passed
1635 by reference. It is set to `0` when an error is found, and `2` when
1636 any change is made. If it remains unchanged at `1`, then no more
1637 propagation is needed.
1641 enum val_rules {Rnolabel = 1<<0, Rboolok = 1<<1, Rnoconstant = 1<<2};
1645 if (rules & Rnolabel)
1646 fputs(" (labels not permitted)", stderr);
1649 ###### forward decls
1650 static struct type *propagate_types(struct exec *prog, struct parse_context *c, int *ok,
1651 struct type *type, int rules);
1652 ###### core functions
1654 static struct type *__propagate_types(struct exec *prog, struct parse_context *c, int *ok,
1655 struct type *type, int rules)
1662 switch (prog->type) {
1665 struct binode *b = cast(binode, prog);
1667 ## propagate binode cases
1671 ## propagate exec cases
1676 static struct type *propagate_types(struct exec *prog, struct parse_context *c, int *ok,
1677 struct type *type, int rules)
1679 struct type *ret = __propagate_types(prog, c, ok, type, rules);
1688 Interpreting an `exec` doesn't require anything but the `exec`. State
1689 is stored in variables and each variable will be directly linked from
1690 within the `exec` tree. The exception to this is the `main` function
1691 which needs to look at command line arguments. This function will be
1692 interpreted separately.
1694 Each `exec` can return a value combined with a type in `struct lrval`.
1695 The type may be `Tnone` but must be non-NULL. Some `exec`s will return
1696 the location of a value, which can be updated, in `lval`. Others will
1697 set `lval` to NULL indicating that there is a value of appropriate type
1700 ###### core functions
1704 struct value rval, *lval;
1707 /* If dest is passed, dtype must give the expected type, and
1708 * result can go there, in which case type is returned as NULL.
1710 static struct lrval _interp_exec(struct parse_context *c, struct exec *e,
1711 struct value *dest, struct type *dtype);
1713 static struct value interp_exec(struct parse_context *c, struct exec *e,
1714 struct type **typeret)
1716 struct lrval ret = _interp_exec(c, e, NULL, NULL);
1718 if (!ret.type) abort();
1720 *typeret = ret.type;
1722 dup_value(ret.type, ret.lval, &ret.rval);
1726 static struct value *linterp_exec(struct parse_context *c, struct exec *e,
1727 struct type **typeret)
1729 struct lrval ret = _interp_exec(c, e, NULL, NULL);
1731 if (!ret.type) abort();
1733 *typeret = ret.type;
1735 free_value(ret.type, &ret.rval);
1739 /* dinterp_exec is used when the destination type is certain and
1740 * the value has a place to go.
1742 static void dinterp_exec(struct parse_context *c, struct exec *e,
1743 struct value *dest, struct type *dtype,
1746 struct lrval ret = _interp_exec(c, e, dest, dtype);
1750 free_value(dtype, dest);
1752 dup_value(dtype, ret.lval, dest);
1754 memcpy(dest, &ret.rval, dtype->size);
1757 static struct lrval _interp_exec(struct parse_context *c, struct exec *e,
1758 struct value *dest, struct type *dtype)
1760 /* If the result is copied to dest, ret.type is set to NULL */
1762 struct value rv = {}, *lrv = NULL;
1763 struct type *rvtype;
1765 rvtype = ret.type = Tnone;
1775 struct binode *b = cast(binode, e);
1776 struct value left, right, *lleft;
1777 struct type *ltype, *rtype;
1778 ltype = rtype = Tnone;
1780 ## interp binode cases
1782 free_value(ltype, &left);
1783 free_value(rtype, &right);
1786 ## interp exec cases
1793 ## interp exec cleanup
1799 Now that we have the shape of the interpreter in place we can add some
1800 complex types and connected them in to the data structures and the
1801 different phases of parse, analyse, print, interpret.
1803 Thus far we have arrays and structs.
1807 Arrays can be declared by giving a size and a type, as `[size]type' so
1808 `freq:[26]number` declares `freq` to be an array of 26 numbers. The
1809 size can be either a literal number, or a named constant. Some day an
1810 arbitrary expression will be supported.
1812 As a formal parameter to a function, the array can be declared with a
1813 new variable as the size: `name:[size::number]string`. The `size`
1814 variable is set to the size of the array and must be a constant. As
1815 `number` is the only supported type, it can be left out:
1816 `name:[size::]string`.
1818 Arrays cannot be assigned. When pointers are introduced we will also
1819 introduce array slices which can refer to part or all of an array -
1820 the assignment syntax will create a slice. For now, an array can only
1821 ever be referenced by the name it is declared with. It is likely that
1822 a "`copy`" primitive will eventually be define which can be used to
1823 make a copy of an array with controllable recursive depth.
1825 For now we have two sorts of array, those with fixed size either because
1826 it is given as a literal number or because it is a struct member (which
1827 cannot have a runtime-changing size), and those with a size that is
1828 determined at runtime - local variables with a const size. The former
1829 have their size calculated at parse time, the latter at run time.
1831 For the latter type, the `size` field of the type is the size of a
1832 pointer, and the array is reallocated every time it comes into scope.
1834 We differentiate struct fields with a const size from local variables
1835 with a const size by whether they are prepared at parse time or not.
1837 ###### type union fields
1840 int unspec; // size is unspecified - vsize must be set.
1843 struct variable *vsize;
1844 struct type *member;
1847 ###### value union fields
1848 void *array; // used if not static_size
1850 ###### value functions
1852 static void array_prepare_type(struct parse_context *c, struct type *type,
1855 struct value *vsize;
1857 if (!type->array.vsize || type->array.static_size)
1860 vsize = var_value(c, type->array.vsize);
1862 mpz_tdiv_q(q, mpq_numref(vsize->num), mpq_denref(vsize->num));
1863 type->array.size = mpz_get_si(q);
1867 type->array.static_size = 1;
1868 type->size = type->array.size * type->array.member->size;
1869 type->align = type->array.member->align;
1873 static void array_init(struct type *type, struct value *val)
1876 void *ptr = val->ptr;
1880 if (!type->array.static_size) {
1881 val->array = calloc(type->array.size,
1882 type->array.member->size);
1885 for (i = 0; i < type->array.size; i++) {
1887 v = (void*)ptr + i * type->array.member->size;
1888 val_init(type->array.member, v);
1892 static void array_free(struct type *type, struct value *val)
1895 void *ptr = val->ptr;
1897 if (!type->array.static_size)
1899 for (i = 0; i < type->array.size; i++) {
1901 v = (void*)ptr + i * type->array.member->size;
1902 free_value(type->array.member, v);
1904 if (!type->array.static_size)
1908 static int array_compat(struct type *require, struct type *have)
1910 if (have->compat != require->compat)
1912 /* Both are arrays, so we can look at details */
1913 if (!type_compat(require->array.member, have->array.member, 0))
1915 if (have->array.unspec && require->array.unspec) {
1916 if (have->array.vsize && require->array.vsize &&
1917 have->array.vsize != require->array.vsize) // UNTESTED
1918 /* sizes might not be the same */
1919 return 0; // UNTESTED
1922 if (have->array.unspec || require->array.unspec)
1923 return 1; // UNTESTED
1924 if (require->array.vsize == NULL && have->array.vsize == NULL)
1925 return require->array.size == have->array.size;
1927 return require->array.vsize == have->array.vsize; // UNTESTED
1930 static void array_print_type(struct type *type, FILE *f)
1933 if (type->array.vsize) {
1934 struct binding *b = type->array.vsize->name;
1935 fprintf(f, "%.*s%s]", b->name.len, b->name.txt,
1936 type->array.unspec ? "::" : "");
1938 fprintf(f, "%d]", type->array.size);
1939 type_print(type->array.member, f);
1942 static struct type array_prototype = {
1944 .prepare_type = array_prepare_type,
1945 .print_type = array_print_type,
1946 .compat = array_compat,
1948 .size = sizeof(void*),
1949 .align = sizeof(void*),
1952 ###### declare terminals
1957 | [ NUMBER ] Type ${ {
1960 struct text noname = { "", 0 };
1963 $0 = t = add_type(c, noname, &array_prototype);
1964 t->array.member = $<4;
1965 t->array.vsize = NULL;
1966 if (number_parse(num, tail, $2.txt) == 0)
1967 tok_err(c, "error: unrecognised number", &$2);
1969 tok_err(c, "error: unsupported number suffix", &$2);
1972 t->array.size = mpz_get_ui(mpq_numref(num));
1973 if (mpz_cmp_ui(mpq_denref(num), 1) != 0) {
1974 tok_err(c, "error: array size must be an integer",
1976 } else if (mpz_cmp_ui(mpq_numref(num), 1UL << 30) >= 0)
1977 tok_err(c, "error: array size is too large",
1981 t->array.static_size = 1;
1982 t->size = t->array.size * t->array.member->size;
1983 t->align = t->array.member->align;
1986 | [ IDENTIFIER ] Type ${ {
1987 struct variable *v = var_ref(c, $2.txt);
1988 struct text noname = { "", 0 };
1991 tok_err(c, "error: name undeclared", &$2);
1992 else if (!v->constant)
1993 tok_err(c, "error: array size must be a constant", &$2);
1995 $0 = add_type(c, noname, &array_prototype);
1996 $0->array.member = $<4;
1998 $0->array.vsize = v;
2003 OptType -> Type ${ $0 = $<1; }$
2006 ###### formal type grammar
2008 | [ IDENTIFIER :: OptType ] Type ${ {
2009 struct variable *v = var_decl(c, $ID.txt);
2010 struct text noname = { "", 0 };
2016 $0 = add_type(c, noname, &array_prototype);
2017 $0->array.member = $<6;
2019 $0->array.unspec = 1;
2020 $0->array.vsize = v;
2026 ###### variable grammar
2028 | Variable [ Expression ] ${ {
2029 struct binode *b = new(binode);
2036 ###### print binode cases
2038 print_exec(b->left, -1, bracket);
2040 print_exec(b->right, -1, bracket);
2044 ###### propagate binode cases
2046 /* left must be an array, right must be a number,
2047 * result is the member type of the array
2049 propagate_types(b->right, c, ok, Tnum, 0);
2050 t = propagate_types(b->left, c, ok, NULL, rules & Rnoconstant);
2051 if (!t || t->compat != array_compat) {
2052 type_err(c, "error: %1 cannot be indexed", prog, t, 0, NULL);
2055 if (!type_compat(type, t->array.member, rules)) {
2056 type_err(c, "error: have %1 but need %2", prog,
2057 t->array.member, rules, type);
2059 return t->array.member;
2063 ###### interp binode cases
2069 lleft = linterp_exec(c, b->left, <ype);
2070 right = interp_exec(c, b->right, &rtype);
2072 mpz_tdiv_q(q, mpq_numref(right.num), mpq_denref(right.num));
2076 if (ltype->array.static_size)
2079 ptr = *(void**)lleft;
2080 rvtype = ltype->array.member;
2081 if (i >= 0 && i < ltype->array.size)
2082 lrv = ptr + i * rvtype->size;
2084 val_init(ltype->array.member, &rv); // UNSAFE
2091 A `struct` is a data-type that contains one or more other data-types.
2092 It differs from an array in that each member can be of a different
2093 type, and they are accessed by name rather than by number. Thus you
2094 cannot choose an element by calculation, you need to know what you
2097 The language makes no promises about how a given structure will be
2098 stored in memory - it is free to rearrange fields to suit whatever
2099 criteria seems important.
2101 Structs are declared separately from program code - they cannot be
2102 declared in-line in a variable declaration like arrays can. A struct
2103 is given a name and this name is used to identify the type - the name
2104 is not prefixed by the word `struct` as it would be in C.
2106 Structs are only treated as the same if they have the same name.
2107 Simply having the same fields in the same order is not enough. This
2108 might change once we can create structure initializers from a list of
2111 Each component datum is identified much like a variable is declared,
2112 with a name, one or two colons, and a type. The type cannot be omitted
2113 as there is no opportunity to deduce the type from usage. An initial
2114 value can be given following an equals sign, so
2116 ##### Example: a struct type
2122 would declare a type called "complex" which has two number fields,
2123 each initialised to zero.
2125 Struct will need to be declared separately from the code that uses
2126 them, so we will need to be able to print out the declaration of a
2127 struct when reprinting the whole program. So a `print_type_decl` type
2128 function will be needed.
2130 ###### type union fields
2142 ###### type functions
2143 void (*print_type_decl)(struct type *type, FILE *f);
2145 ###### value functions
2147 static void structure_init(struct type *type, struct value *val)
2151 for (i = 0; i < type->structure.nfields; i++) {
2153 v = (void*) val->ptr + type->structure.fields[i].offset;
2154 if (type->structure.fields[i].init)
2155 dup_value(type->structure.fields[i].type,
2156 type->structure.fields[i].init,
2159 val_init(type->structure.fields[i].type, v);
2163 static void structure_free(struct type *type, struct value *val)
2167 for (i = 0; i < type->structure.nfields; i++) {
2169 v = (void*)val->ptr + type->structure.fields[i].offset;
2170 free_value(type->structure.fields[i].type, v);
2174 static void structure_free_type(struct type *t)
2177 for (i = 0; i < t->structure.nfields; i++)
2178 if (t->structure.fields[i].init) {
2179 free_value(t->structure.fields[i].type,
2180 t->structure.fields[i].init);
2182 free(t->structure.fields);
2185 static struct type structure_prototype = {
2186 .init = structure_init,
2187 .free = structure_free,
2188 .free_type = structure_free_type,
2189 .print_type_decl = structure_print_type,
2203 ###### free exec cases
2205 free_exec(cast(fieldref, e)->left);
2209 ###### declare terminals
2212 ###### variable grammar
2214 | Variable . IDENTIFIER ${ {
2215 struct fieldref *fr = new_pos(fieldref, $2);
2222 ###### print exec cases
2226 struct fieldref *f = cast(fieldref, e);
2227 print_exec(f->left, -1, bracket);
2228 printf(".%.*s", f->name.len, f->name.txt);
2232 ###### ast functions
2233 static int find_struct_index(struct type *type, struct text field)
2236 for (i = 0; i < type->structure.nfields; i++)
2237 if (text_cmp(type->structure.fields[i].name, field) == 0)
2242 ###### propagate exec cases
2246 struct fieldref *f = cast(fieldref, prog);
2247 struct type *st = propagate_types(f->left, c, ok, NULL, 0);
2250 type_err(c, "error: unknown type for field access", f->left, // UNTESTED
2252 else if (st->init != structure_init)
2253 type_err(c, "error: field reference attempted on %1, not a struct",
2254 f->left, st, 0, NULL);
2255 else if (f->index == -2) {
2256 f->index = find_struct_index(st, f->name);
2258 type_err(c, "error: cannot find requested field in %1",
2259 f->left, st, 0, NULL);
2261 if (f->index >= 0) {
2262 struct type *ft = st->structure.fields[f->index].type;
2263 if (!type_compat(type, ft, rules))
2264 type_err(c, "error: have %1 but need %2", prog,
2271 ###### interp exec cases
2274 struct fieldref *f = cast(fieldref, e);
2276 struct value *lleft = linterp_exec(c, f->left, <ype);
2277 lrv = (void*)lleft->ptr + ltype->structure.fields[f->index].offset;
2278 rvtype = ltype->structure.fields[f->index].type;
2284 struct fieldlist *prev;
2288 ###### ast functions
2289 static void free_fieldlist(struct fieldlist *f)
2293 free_fieldlist(f->prev);
2295 free_value(f->f.type, f->f.init); // UNTESTED
2296 free(f->f.init); // UNTESTED
2301 ###### top level grammar
2302 DeclareStruct -> struct IDENTIFIER FieldBlock Newlines ${ {
2304 add_type(c, $2.txt, &structure_prototype);
2306 struct fieldlist *f;
2308 for (f = $3; f; f=f->prev)
2311 t->structure.nfields = cnt;
2312 t->structure.fields = calloc(cnt, sizeof(struct field));
2315 int a = f->f.type->align;
2317 t->structure.fields[cnt] = f->f;
2318 if (t->size & (a-1))
2319 t->size = (t->size | (a-1)) + 1;
2320 t->structure.fields[cnt].offset = t->size;
2321 t->size += ((f->f.type->size - 1) | (a-1)) + 1;
2330 FieldBlock -> { IN OptNL FieldLines OUT OptNL } ${ $0 = $<FL; }$
2331 | { SimpleFieldList } ${ $0 = $<SFL; }$
2332 | IN OptNL FieldLines OUT ${ $0 = $<FL; }$
2333 | SimpleFieldList EOL ${ $0 = $<SFL; }$
2335 FieldLines -> SimpleFieldList Newlines ${ $0 = $<SFL; }$
2336 | FieldLines SimpleFieldList Newlines ${
2341 SimpleFieldList -> Field ${ $0 = $<F; }$
2342 | SimpleFieldList ; Field ${
2346 | SimpleFieldList ; ${
2349 | ERROR ${ tok_err(c, "Syntax error in struct field", &$1); }$
2351 Field -> IDENTIFIER : Type = Expression ${ {
2354 $0 = calloc(1, sizeof(struct fieldlist));
2355 $0->f.name = $1.txt;
2360 propagate_types($<5, c, &ok, $3, 0);
2363 c->parse_error = 1; // UNTESTED
2365 struct value vl = interp_exec(c, $5, NULL);
2366 $0->f.init = global_alloc(c, $0->f.type, NULL, &vl);
2369 | IDENTIFIER : Type ${
2370 $0 = calloc(1, sizeof(struct fieldlist));
2371 $0->f.name = $1.txt;
2373 if ($0->f.type->prepare_type)
2374 $0->f.type->prepare_type(c, $0->f.type, 1);
2377 ###### forward decls
2378 static void structure_print_type(struct type *t, FILE *f);
2380 ###### value functions
2381 static void structure_print_type(struct type *t, FILE *f)
2385 fprintf(f, "struct %.*s\n", t->name.len, t->name.txt);
2387 for (i = 0; i < t->structure.nfields; i++) {
2388 struct field *fl = t->structure.fields + i;
2389 fprintf(f, " %.*s : ", fl->name.len, fl->name.txt);
2390 type_print(fl->type, f);
2391 if (fl->type->print && fl->init) {
2393 if (fl->type == Tstr)
2394 fprintf(f, "\""); // UNTESTED
2395 print_value(fl->type, fl->init, f);
2396 if (fl->type == Tstr)
2397 fprintf(f, "\""); // UNTESTED
2403 ###### print type decls
2408 while (target != 0) {
2410 for (t = context.typelist; t ; t=t->next)
2411 if (t->print_type_decl && !t->check_args && t->name.txt[0] != ' ') {
2420 t->print_type_decl(t, stdout);
2428 A function is a chunk of code which can be passed parameters and can
2429 return results. Each function has a type which includes the set of
2430 parameters and the return value. As yet these types cannot be declared
2431 separately from the function itself.
2433 The parameters can be specified either in parentheses as a ';' separated
2436 ##### Example: function 1
2438 func main(av:[ac::number]string; env:[envc::number]string)
2441 or as an indented list of one parameter per line (though each line can
2442 be a ';' separated list)
2444 ##### Example: function 2
2447 argv:[argc::number]string
2448 env:[envc::number]string
2452 In the first case a return type can follow the parentheses after a colon,
2453 in the second it is given on a line starting with the word `return`.
2455 ##### Example: functions that return
2457 func add(a:number; b:number): number
2467 Rather than returning a type, the function can specify a set of local
2468 variables to return as a struct. The values of these variables when the
2469 function exits will be provided to the caller. For this the return type
2470 is replaced with a block of result declarations, either in parentheses
2471 or bracketed by `return` and `do`.
2473 ##### Example: functions returning multiple variables
2475 func to_cartesian(rho:number; theta:number):(x:number; y:number)
2488 For constructing the lists we use a `List` binode, which will be
2489 further detailed when Expression Lists are introduced.
2491 ###### type union fields
2494 struct binode *params;
2495 struct type *return_type;
2496 struct variable *scope;
2497 int inline_result; // return value is at start of 'local'
2501 ###### value union fields
2502 struct exec *function;
2504 ###### type functions
2505 void (*check_args)(struct parse_context *c, int *ok,
2506 struct type *require, struct exec *args);
2508 ###### value functions
2510 static void function_free(struct type *type, struct value *val)
2512 free_exec(val->function);
2513 val->function = NULL;
2516 static int function_compat(struct type *require, struct type *have)
2518 // FIXME can I do anything here yet?
2522 static void function_check_args(struct parse_context *c, int *ok,
2523 struct type *require, struct exec *args)
2525 /* This should be 'compat', but we don't have a 'tuple' type to
2526 * hold the type of 'args'
2528 struct binode *arg = cast(binode, args);
2529 struct binode *param = require->function.params;
2532 struct var *pv = cast(var, param->left);
2534 type_err(c, "error: insufficient arguments to function.",
2535 args, NULL, 0, NULL);
2539 propagate_types(arg->left, c, ok, pv->var->type, 0);
2540 param = cast(binode, param->right);
2541 arg = cast(binode, arg->right);
2544 type_err(c, "error: too many arguments to function.",
2545 args, NULL, 0, NULL);
2548 static void function_print(struct type *type, struct value *val, FILE *f)
2550 print_exec(val->function, 1, 0);
2553 static void function_print_type_decl(struct type *type, FILE *f)
2557 for (b = type->function.params; b; b = cast(binode, b->right)) {
2558 struct variable *v = cast(var, b->left)->var;
2559 fprintf(f, "%.*s%s", v->name->name.len, v->name->name.txt,
2560 v->constant ? "::" : ":");
2561 type_print(v->type, f);
2566 if (type->function.return_type != Tnone) {
2568 if (type->function.inline_result) {
2570 struct type *t = type->function.return_type;
2572 for (i = 0; i < t->structure.nfields; i++) {
2573 struct field *fl = t->structure.fields + i;
2576 fprintf(f, "%.*s:", fl->name.len, fl->name.txt);
2577 type_print(fl->type, f);
2581 type_print(type->function.return_type, f);
2586 static void function_free_type(struct type *t)
2588 free_exec(t->function.params);
2591 static struct type function_prototype = {
2592 .size = sizeof(void*),
2593 .align = sizeof(void*),
2594 .free = function_free,
2595 .compat = function_compat,
2596 .check_args = function_check_args,
2597 .print = function_print,
2598 .print_type_decl = function_print_type_decl,
2599 .free_type = function_free_type,
2602 ###### declare terminals
2612 FuncName -> IDENTIFIER ${ {
2613 struct variable *v = var_decl(c, $1.txt);
2614 struct var *e = new_pos(var, $1);
2620 v = var_ref(c, $1.txt);
2622 type_err(c, "error: function '%v' redeclared",
2624 type_err(c, "info: this is where '%v' was first declared",
2625 v->where_decl, NULL, 0, NULL);
2631 Args -> ArgsLine NEWLINE ${ $0 = $<AL; }$
2632 | Args ArgsLine NEWLINE ${ {
2633 struct binode *b = $<AL;
2634 struct binode **bp = &b;
2636 bp = (struct binode **)&(*bp)->left;
2641 ArgsLine -> ${ $0 = NULL; }$
2642 | Varlist ${ $0 = $<1; }$
2643 | Varlist ; ${ $0 = $<1; }$
2645 Varlist -> Varlist ; ArgDecl ${
2659 ArgDecl -> IDENTIFIER : FormalType ${ {
2660 struct variable *v = var_decl(c, $1.txt);
2666 ## Executables: the elements of code
2668 Each code element needs to be parsed, printed, analysed,
2669 interpreted, and freed. There are several, so let's just start with
2670 the easy ones and work our way up.
2674 We have already met values as separate objects. When manifest
2675 constants appear in the program text, that must result in an executable
2676 which has a constant value. So the `val` structure embeds a value in
2689 ###### ast functions
2690 struct val *new_val(struct type *T, struct token tk)
2692 struct val *v = new_pos(val, tk);
2703 $0 = new_val(Tbool, $1);
2707 $0 = new_val(Tbool, $1);
2711 $0 = new_val(Tnum, $1);
2714 if (number_parse($0->val.num, tail, $1.txt) == 0)
2715 mpq_init($0->val.num); // UNTESTED
2717 tok_err(c, "error: unsupported number suffix",
2722 $0 = new_val(Tstr, $1);
2725 string_parse(&$1, '\\', &$0->val.str, tail);
2727 tok_err(c, "error: unsupported string suffix",
2732 $0 = new_val(Tstr, $1);
2735 string_parse(&$1, '\\', &$0->val.str, tail);
2737 tok_err(c, "error: unsupported string suffix",
2742 ###### print exec cases
2745 struct val *v = cast(val, e);
2746 if (v->vtype == Tstr)
2748 print_value(v->vtype, &v->val, stdout);
2749 if (v->vtype == Tstr)
2754 ###### propagate exec cases
2757 struct val *val = cast(val, prog);
2758 if (!type_compat(type, val->vtype, rules))
2759 type_err(c, "error: expected %1%r found %2",
2760 prog, type, rules, val->vtype);
2764 ###### interp exec cases
2766 rvtype = cast(val, e)->vtype;
2767 dup_value(rvtype, &cast(val, e)->val, &rv);
2770 ###### ast functions
2771 static void free_val(struct val *v)
2774 free_value(v->vtype, &v->val);
2778 ###### free exec cases
2779 case Xval: free_val(cast(val, e)); break;
2781 ###### ast functions
2782 // Move all nodes from 'b' to 'rv', reversing their order.
2783 // In 'b' 'left' is a list, and 'right' is the last node.
2784 // In 'rv', left' is the first node and 'right' is a list.
2785 static struct binode *reorder_bilist(struct binode *b)
2787 struct binode *rv = NULL;
2790 struct exec *t = b->right;
2794 b = cast(binode, b->left);
2804 Just as we used a `val` to wrap a value into an `exec`, we similarly
2805 need a `var` to wrap a `variable` into an exec. While each `val`
2806 contained a copy of the value, each `var` holds a link to the variable
2807 because it really is the same variable no matter where it appears.
2808 When a variable is used, we need to remember to follow the `->merged`
2809 link to find the primary instance.
2811 When a variable is declared, it may or may not be given an explicit
2812 type. We need to record which so that we can report the parsed code
2821 struct variable *var;
2824 ###### variable fields
2832 VariableDecl -> IDENTIFIER : ${ {
2833 struct variable *v = var_decl(c, $1.txt);
2834 $0 = new_pos(var, $1);
2839 v = var_ref(c, $1.txt);
2841 type_err(c, "error: variable '%v' redeclared",
2843 type_err(c, "info: this is where '%v' was first declared",
2844 v->where_decl, NULL, 0, NULL);
2847 | IDENTIFIER :: ${ {
2848 struct variable *v = var_decl(c, $1.txt);
2849 $0 = new_pos(var, $1);
2855 v = var_ref(c, $1.txt);
2857 type_err(c, "error: variable '%v' redeclared",
2859 type_err(c, "info: this is where '%v' was first declared",
2860 v->where_decl, NULL, 0, NULL);
2863 | IDENTIFIER : Type ${ {
2864 struct variable *v = var_decl(c, $1.txt);
2865 $0 = new_pos(var, $1);
2871 v->explicit_type = 1;
2873 v = var_ref(c, $1.txt);
2875 type_err(c, "error: variable '%v' redeclared",
2877 type_err(c, "info: this is where '%v' was first declared",
2878 v->where_decl, NULL, 0, NULL);
2881 | IDENTIFIER :: Type ${ {
2882 struct variable *v = var_decl(c, $1.txt);
2883 $0 = new_pos(var, $1);
2890 v->explicit_type = 1;
2892 v = var_ref(c, $1.txt);
2894 type_err(c, "error: variable '%v' redeclared",
2896 type_err(c, "info: this is where '%v' was first declared",
2897 v->where_decl, NULL, 0, NULL);
2902 Variable -> IDENTIFIER ${ {
2903 struct variable *v = var_ref(c, $1.txt);
2904 $0 = new_pos(var, $1);
2906 /* This might be a label - allocate a var just in case */
2907 v = var_decl(c, $1.txt);
2914 cast(var, $0)->var = v;
2918 ###### print exec cases
2921 struct var *v = cast(var, e);
2923 struct binding *b = v->var->name;
2924 printf("%.*s", b->name.len, b->name.txt);
2931 if (loc && loc->type == Xvar) {
2932 struct var *v = cast(var, loc);
2934 struct binding *b = v->var->name;
2935 fprintf(stderr, "%.*s", b->name.len, b->name.txt);
2937 fputs("???", stderr); // NOTEST
2939 fputs("NOTVAR", stderr);
2942 ###### propagate exec cases
2946 struct var *var = cast(var, prog);
2947 struct variable *v = var->var;
2949 type_err(c, "%d:BUG: no variable!!", prog, NULL, 0, NULL); // NOTEST
2950 return Tnone; // NOTEST
2953 if (v->constant && (rules & Rnoconstant)) {
2954 type_err(c, "error: Cannot assign to a constant: %v",
2955 prog, NULL, 0, NULL);
2956 type_err(c, "info: name was defined as a constant here",
2957 v->where_decl, NULL, 0, NULL);
2960 if (v->type == Tnone && v->where_decl == prog)
2961 type_err(c, "error: variable used but not declared: %v",
2962 prog, NULL, 0, NULL);
2963 if (v->type == NULL) {
2964 if (type && *ok != 0) {
2966 v->where_set = prog;
2971 if (!type_compat(type, v->type, rules)) {
2972 type_err(c, "error: expected %1%r but variable '%v' is %2", prog,
2973 type, rules, v->type);
2974 type_err(c, "info: this is where '%v' was set to %1", v->where_set,
2975 v->type, rules, NULL);
2982 ###### interp exec cases
2985 struct var *var = cast(var, e);
2986 struct variable *v = var->var;
2989 lrv = var_value(c, v);
2994 ###### ast functions
2996 static void free_var(struct var *v)
3001 ###### free exec cases
3002 case Xvar: free_var(cast(var, e)); break;
3004 ### Expressions: Conditional
3006 Our first user of the `binode` will be conditional expressions, which
3007 is a bit odd as they actually have three components. That will be
3008 handled by having 2 binodes for each expression. The conditional
3009 expression is the lowest precedence operator which is why we define it
3010 first - to start the precedence list.
3012 Conditional expressions are of the form "value `if` condition `else`
3013 other_value". They associate to the right, so everything to the right
3014 of `else` is part of an else value, while only a higher-precedence to
3015 the left of `if` is the if values. Between `if` and `else` there is no
3016 room for ambiguity, so a full conditional expression is allowed in
3028 Expression -> Expression if Expression else Expression $$ifelse ${ {
3029 struct binode *b1 = new(binode);
3030 struct binode *b2 = new(binode);
3039 ## expression grammar
3041 ###### print binode cases
3044 b2 = cast(binode, b->right);
3045 if (bracket) printf("(");
3046 print_exec(b2->left, -1, bracket);
3048 print_exec(b->left, -1, bracket);
3050 print_exec(b2->right, -1, bracket);
3051 if (bracket) printf(")");
3054 ###### propagate binode cases
3057 /* cond must be Tbool, others must match */
3058 struct binode *b2 = cast(binode, b->right);
3061 propagate_types(b->left, c, ok, Tbool, 0);
3062 t = propagate_types(b2->left, c, ok, type, Rnolabel);
3063 t2 = propagate_types(b2->right, c, ok, type ?: t, Rnolabel);
3067 ###### interp binode cases
3070 struct binode *b2 = cast(binode, b->right);
3071 left = interp_exec(c, b->left, <ype);
3073 rv = interp_exec(c, b2->left, &rvtype); // UNTESTED
3075 rv = interp_exec(c, b2->right, &rvtype);
3081 We take a brief detour, now that we have expressions, to describe lists
3082 of expressions. These will be needed for function parameters and
3083 possibly other situations. They seem generic enough to introduce here
3084 to be used elsewhere.
3086 And ExpressionList will use the `List` type of `binode`, building up at
3087 the end. And place where they are used will probably call
3088 `reorder_bilist()` to get a more normal first/next arrangement.
3090 ###### declare terminals
3093 `List` execs have no implicit semantics, so they are never propagated or
3094 interpreted. The can be printed as a comma separate list, which is how
3095 they are parsed. Note they are also used for function formal parameter
3096 lists. In that case a separate function is used to print them.
3098 ###### print binode cases
3102 print_exec(b->left, -1, bracket);
3105 b = cast(binode, b->right);
3109 ###### propagate binode cases
3110 case List: abort(); // NOTEST
3111 ###### interp binode cases
3112 case List: abort(); // NOTEST
3117 ExpressionList -> ExpressionList , Expression ${
3130 ### Expressions: Boolean
3132 The next class of expressions to use the `binode` will be Boolean
3133 expressions. "`and then`" and "`or else`" are similar to `and` and `or`
3134 have same corresponding precendence. The difference is that they don't
3135 evaluate the second expression if not necessary.
3144 ###### expr precedence
3149 ###### expression grammar
3150 | Expression or Expression ${ {
3151 struct binode *b = new(binode);
3157 | Expression or else Expression ${ {
3158 struct binode *b = new(binode);
3165 | Expression and Expression ${ {
3166 struct binode *b = new(binode);
3172 | Expression and then Expression ${ {
3173 struct binode *b = new(binode);
3180 | not Expression ${ {
3181 struct binode *b = new(binode);
3187 ###### print binode cases
3189 if (bracket) printf("(");
3190 print_exec(b->left, -1, bracket);
3192 print_exec(b->right, -1, bracket);
3193 if (bracket) printf(")");
3196 if (bracket) printf("(");
3197 print_exec(b->left, -1, bracket);
3198 printf(" and then ");
3199 print_exec(b->right, -1, bracket);
3200 if (bracket) printf(")");
3203 if (bracket) printf("(");
3204 print_exec(b->left, -1, bracket);
3206 print_exec(b->right, -1, bracket);
3207 if (bracket) printf(")");
3210 if (bracket) printf("(");
3211 print_exec(b->left, -1, bracket);
3212 printf(" or else ");
3213 print_exec(b->right, -1, bracket);
3214 if (bracket) printf(")");
3217 if (bracket) printf("(");
3219 print_exec(b->right, -1, bracket);
3220 if (bracket) printf(")");
3223 ###### propagate binode cases
3229 /* both must be Tbool, result is Tbool */
3230 propagate_types(b->left, c, ok, Tbool, 0);
3231 propagate_types(b->right, c, ok, Tbool, 0);
3232 if (type && type != Tbool)
3233 type_err(c, "error: %1 operation found where %2 expected", prog,
3237 ###### interp binode cases
3239 rv = interp_exec(c, b->left, &rvtype);
3240 right = interp_exec(c, b->right, &rtype);
3241 rv.bool = rv.bool && right.bool;
3244 rv = interp_exec(c, b->left, &rvtype);
3246 rv = interp_exec(c, b->right, NULL);
3249 rv = interp_exec(c, b->left, &rvtype);
3250 right = interp_exec(c, b->right, &rtype);
3251 rv.bool = rv.bool || right.bool;
3254 rv = interp_exec(c, b->left, &rvtype);
3256 rv = interp_exec(c, b->right, NULL);
3259 rv = interp_exec(c, b->right, &rvtype);
3263 ### Expressions: Comparison
3265 Of slightly higher precedence that Boolean expressions are Comparisons.
3266 A comparison takes arguments of any comparable type, but the two types
3269 To simplify the parsing we introduce an `eop` which can record an
3270 expression operator, and the `CMPop` non-terminal will match one of them.
3277 ###### ast functions
3278 static void free_eop(struct eop *e)
3292 ###### expr precedence
3293 $LEFT < > <= >= == != CMPop
3295 ###### expression grammar
3296 | Expression CMPop Expression ${ {
3297 struct binode *b = new(binode);
3307 CMPop -> < ${ $0.op = Less; }$
3308 | > ${ $0.op = Gtr; }$
3309 | <= ${ $0.op = LessEq; }$
3310 | >= ${ $0.op = GtrEq; }$
3311 | == ${ $0.op = Eql; }$
3312 | != ${ $0.op = NEql; }$
3314 ###### print binode cases
3322 if (bracket) printf("(");
3323 print_exec(b->left, -1, bracket);
3325 case Less: printf(" < "); break;
3326 case LessEq: printf(" <= "); break;
3327 case Gtr: printf(" > "); break;
3328 case GtrEq: printf(" >= "); break;
3329 case Eql: printf(" == "); break;
3330 case NEql: printf(" != "); break;
3331 default: abort(); // NOTEST
3333 print_exec(b->right, -1, bracket);
3334 if (bracket) printf(")");
3337 ###### propagate binode cases
3344 /* Both must match but not be labels, result is Tbool */
3345 t = propagate_types(b->left, c, ok, NULL, Rnolabel);
3347 propagate_types(b->right, c, ok, t, 0);
3349 t = propagate_types(b->right, c, ok, NULL, Rnolabel); // UNTESTED
3351 t = propagate_types(b->left, c, ok, t, 0); // UNTESTED
3353 if (!type_compat(type, Tbool, 0))
3354 type_err(c, "error: Comparison returns %1 but %2 expected", prog,
3355 Tbool, rules, type);
3358 ###### interp binode cases
3367 left = interp_exec(c, b->left, <ype);
3368 right = interp_exec(c, b->right, &rtype);
3369 cmp = value_cmp(ltype, rtype, &left, &right);
3372 case Less: rv.bool = cmp < 0; break;
3373 case LessEq: rv.bool = cmp <= 0; break;
3374 case Gtr: rv.bool = cmp > 0; break;
3375 case GtrEq: rv.bool = cmp >= 0; break;
3376 case Eql: rv.bool = cmp == 0; break;
3377 case NEql: rv.bool = cmp != 0; break;
3378 default: rv.bool = 0; break; // NOTEST
3383 ### Expressions: Arithmetic etc.
3385 The remaining expressions with the highest precedence are arithmetic,
3386 string concatenation, and string conversion. String concatenation
3387 (`++`) has the same precedence as multiplication and division, but lower
3390 String conversion is a temporary feature until I get a better type
3391 system. `$` is a prefix operator which expects a string and returns
3394 `+` and `-` are both infix and prefix operations (where they are
3395 absolute value and negation). These have different operator names.
3397 We also have a 'Bracket' operator which records where parentheses were
3398 found. This makes it easy to reproduce these when printing. Possibly I
3399 should only insert brackets were needed for precedence.
3409 ###### expr precedence
3415 ###### expression grammar
3416 | Expression Eop Expression ${ {
3417 struct binode *b = new(binode);
3424 | Expression Top Expression ${ {
3425 struct binode *b = new(binode);
3432 | ( Expression ) ${ {
3433 struct binode *b = new_pos(binode, $1);
3438 | Uop Expression ${ {
3439 struct binode *b = new(binode);
3444 | Value ${ $0 = $<1; }$
3445 | Variable ${ $0 = $<1; }$
3450 Eop -> + ${ $0.op = Plus; }$
3451 | - ${ $0.op = Minus; }$
3453 Uop -> + ${ $0.op = Absolute; }$
3454 | - ${ $0.op = Negate; }$
3455 | $ ${ $0.op = StringConv; }$
3457 Top -> * ${ $0.op = Times; }$
3458 | / ${ $0.op = Divide; }$
3459 | % ${ $0.op = Rem; }$
3460 | ++ ${ $0.op = Concat; }$
3462 ###### print binode cases
3469 if (bracket) printf("(");
3470 print_exec(b->left, indent, bracket);
3472 case Plus: fputs(" + ", stdout); break;
3473 case Minus: fputs(" - ", stdout); break;
3474 case Times: fputs(" * ", stdout); break;
3475 case Divide: fputs(" / ", stdout); break;
3476 case Rem: fputs(" % ", stdout); break;
3477 case Concat: fputs(" ++ ", stdout); break;
3478 default: abort(); // NOTEST
3480 print_exec(b->right, indent, bracket);
3481 if (bracket) printf(")");
3486 if (bracket) printf("(");
3488 case Absolute: fputs("+", stdout); break;
3489 case Negate: fputs("-", stdout); break;
3490 case StringConv: fputs("$", stdout); break;
3491 default: abort(); // NOTEST
3493 print_exec(b->right, indent, bracket);
3494 if (bracket) printf(")");
3498 print_exec(b->right, indent, bracket);
3502 ###### propagate binode cases
3508 /* both must be numbers, result is Tnum */
3511 /* as propagate_types ignores a NULL,
3512 * unary ops fit here too */
3513 propagate_types(b->left, c, ok, Tnum, 0);
3514 propagate_types(b->right, c, ok, Tnum, 0);
3515 if (!type_compat(type, Tnum, 0))
3516 type_err(c, "error: Arithmetic returns %1 but %2 expected", prog,
3521 /* both must be Tstr, result is Tstr */
3522 propagate_types(b->left, c, ok, Tstr, 0);
3523 propagate_types(b->right, c, ok, Tstr, 0);
3524 if (!type_compat(type, Tstr, 0))
3525 type_err(c, "error: Concat returns %1 but %2 expected", prog,
3530 /* op must be string, result is number */
3531 propagate_types(b->left, c, ok, Tstr, 0);
3532 if (!type_compat(type, Tnum, 0))
3533 type_err(c, // UNTESTED
3534 "error: Can only convert string to number, not %1",
3535 prog, type, 0, NULL);
3539 return propagate_types(b->right, c, ok, type, 0);
3541 ###### interp binode cases
3544 rv = interp_exec(c, b->left, &rvtype);
3545 right = interp_exec(c, b->right, &rtype);
3546 mpq_add(rv.num, rv.num, right.num);
3549 rv = interp_exec(c, b->left, &rvtype);
3550 right = interp_exec(c, b->right, &rtype);
3551 mpq_sub(rv.num, rv.num, right.num);
3554 rv = interp_exec(c, b->left, &rvtype);
3555 right = interp_exec(c, b->right, &rtype);
3556 mpq_mul(rv.num, rv.num, right.num);
3559 rv = interp_exec(c, b->left, &rvtype);
3560 right = interp_exec(c, b->right, &rtype);
3561 mpq_div(rv.num, rv.num, right.num);
3566 left = interp_exec(c, b->left, <ype);
3567 right = interp_exec(c, b->right, &rtype);
3568 mpz_init(l); mpz_init(r); mpz_init(rem);
3569 mpz_tdiv_q(l, mpq_numref(left.num), mpq_denref(left.num));
3570 mpz_tdiv_q(r, mpq_numref(right.num), mpq_denref(right.num));
3571 mpz_tdiv_r(rem, l, r);
3572 val_init(Tnum, &rv);
3573 mpq_set_z(rv.num, rem);
3574 mpz_clear(r); mpz_clear(l); mpz_clear(rem);
3579 rv = interp_exec(c, b->right, &rvtype);
3580 mpq_neg(rv.num, rv.num);
3583 rv = interp_exec(c, b->right, &rvtype);
3584 mpq_abs(rv.num, rv.num);
3587 rv = interp_exec(c, b->right, &rvtype);
3590 left = interp_exec(c, b->left, <ype);
3591 right = interp_exec(c, b->right, &rtype);
3593 rv.str = text_join(left.str, right.str);
3596 right = interp_exec(c, b->right, &rvtype);
3600 struct text tx = right.str;
3603 if (tx.txt[0] == '-') {
3604 neg = 1; // UNTESTED
3605 tx.txt++; // UNTESTED
3606 tx.len--; // UNTESTED
3608 if (number_parse(rv.num, tail, tx) == 0)
3609 mpq_init(rv.num); // UNTESTED
3611 mpq_neg(rv.num, rv.num); // UNTESTED
3613 printf("Unsupported suffix: %.*s\n", tx.len, tx.txt); // UNTESTED
3617 ###### value functions
3619 static struct text text_join(struct text a, struct text b)
3622 rv.len = a.len + b.len;
3623 rv.txt = malloc(rv.len);
3624 memcpy(rv.txt, a.txt, a.len);
3625 memcpy(rv.txt+a.len, b.txt, b.len);
3631 A function call can appear either as an expression or as a statement.
3632 As functions cannot yet return values, only the statement version will work.
3633 We use a new 'Funcall' binode type to link the function with a list of
3634 arguments, form with the 'List' nodes.
3639 ###### expression grammar
3640 | Variable ( ExpressionList ) ${ {
3641 struct binode *b = new(binode);
3644 b->right = reorder_bilist($<EL);
3648 struct binode *b = new(binode);
3655 ###### SimpleStatement Grammar
3657 | Variable ( ExpressionList ) ${ {
3658 struct binode *b = new(binode);
3661 b->right = reorder_bilist($<EL);
3665 ###### print binode cases
3668 do_indent(indent, "");
3669 print_exec(b->left, -1, bracket);
3671 for (b = cast(binode, b->right); b; b = cast(binode, b->right)) {
3674 print_exec(b->left, -1, bracket);
3684 ###### propagate binode cases
3687 /* Every arg must match formal parameter, and result
3688 * is return type of function
3690 struct binode *args = cast(binode, b->right);
3691 struct var *v = cast(var, b->left);
3693 if (!v->var->type || v->var->type->check_args == NULL) {
3694 type_err(c, "error: attempt to call a non-function.",
3695 prog, NULL, 0, NULL);
3698 v->var->type->check_args(c, ok, v->var->type, args);
3699 return v->var->type->function.return_type;
3702 ###### interp binode cases
3705 struct var *v = cast(var, b->left);
3706 struct type *t = v->var->type;
3707 void *oldlocal = c->local;
3708 int old_size = c->local_size;
3709 void *local = calloc(1, t->function.local_size);
3710 struct value *fbody = var_value(c, v->var);
3711 struct binode *arg = cast(binode, b->right);
3712 struct binode *param = t->function.params;
3715 struct var *pv = cast(var, param->left);
3716 struct type *vtype = NULL;
3717 struct value val = interp_exec(c, arg->left, &vtype);
3719 c->local = local; c->local_size = t->function.local_size;
3720 lval = var_value(c, pv->var);
3721 c->local = oldlocal; c->local_size = old_size;
3722 memcpy(lval, &val, vtype->size);
3723 param = cast(binode, param->right);
3724 arg = cast(binode, arg->right);
3726 c->local = local; c->local_size = t->function.local_size;
3727 if (t->function.inline_result && dtype) {
3728 _interp_exec(c, fbody->function, NULL, NULL);
3729 memcpy(dest, local, dtype->size);
3730 rvtype = ret.type = NULL;
3732 rv = interp_exec(c, fbody->function, &rvtype);
3733 c->local = oldlocal; c->local_size = old_size;
3738 ### Blocks, Statements, and Statement lists.
3740 Now that we have expressions out of the way we need to turn to
3741 statements. There are simple statements and more complex statements.
3742 Simple statements do not contain (syntactic) newlines, complex statements do.
3744 Statements often come in sequences and we have corresponding simple
3745 statement lists and complex statement lists.
3746 The former comprise only simple statements separated by semicolons.
3747 The later comprise complex statements and simple statement lists. They are
3748 separated by newlines. Thus the semicolon is only used to separate
3749 simple statements on the one line. This may be overly restrictive,
3750 but I'm not sure I ever want a complex statement to share a line with
3753 Note that a simple statement list can still use multiple lines if
3754 subsequent lines are indented, so
3756 ###### Example: wrapped simple statement list
3761 is a single simple statement list. This might allow room for
3762 confusion, so I'm not set on it yet.
3764 A simple statement list needs no extra syntax. A complex statement
3765 list has two syntactic forms. It can be enclosed in braces (much like
3766 C blocks), or it can be introduced by an indent and continue until an
3767 unindented newline (much like Python blocks). With this extra syntax
3768 it is referred to as a block.
3770 Note that a block does not have to include any newlines if it only
3771 contains simple statements. So both of:
3773 if condition: a=b; d=f
3775 if condition { a=b; print f }
3779 In either case the list is constructed from a `binode` list with
3780 `Block` as the operator. When parsing the list it is most convenient
3781 to append to the end, so a list is a list and a statement. When using
3782 the list it is more convenient to consider a list to be a statement
3783 and a list. So we need a function to re-order a list.
3784 `reorder_bilist` serves this purpose.
3786 The only stand-alone statement we introduce at this stage is `pass`
3787 which does nothing and is represented as a `NULL` pointer in a `Block`
3788 list. Other stand-alone statements will follow once the infrastructure
3799 Block -> { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3800 | { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3801 | SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
3802 | SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
3803 | IN OptNL Statementlist OUT ${ $0 = $<Sl; }$
3805 OpenBlock -> OpenScope { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3806 | OpenScope { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3807 | OpenScope SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
3808 | OpenScope SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
3809 | IN OpenScope OptNL Statementlist OUT ${ $0 = $<Sl; }$
3811 UseBlock -> { OpenScope IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3812 | { OpenScope SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3813 | IN OpenScope OptNL Statementlist OUT ${ $0 = $<Sl; }$
3815 ColonBlock -> { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3816 | { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3817 | : SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
3818 | : SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
3819 | : IN OptNL Statementlist OUT ${ $0 = $<Sl; }$
3821 Statementlist -> ComplexStatements ${ $0 = reorder_bilist($<CS); }$
3823 ComplexStatements -> ComplexStatements ComplexStatement ${
3833 | ComplexStatement ${
3845 ComplexStatement -> SimpleStatements Newlines ${
3846 $0 = reorder_bilist($<SS);
3848 | SimpleStatements ; Newlines ${
3849 $0 = reorder_bilist($<SS);
3851 ## ComplexStatement Grammar
3854 SimpleStatements -> SimpleStatements ; SimpleStatement ${
3860 | SimpleStatement ${
3868 SimpleStatement -> pass ${ $0 = NULL; }$
3869 | ERROR ${ tok_err(c, "Syntax error in statement", &$1); }$
3870 ## SimpleStatement Grammar
3872 ###### print binode cases
3876 if (b->left == NULL) // UNTESTED
3877 printf("pass"); // UNTESTED
3879 print_exec(b->left, indent, bracket); // UNTESTED
3880 if (b->right) { // UNTESTED
3881 printf("; "); // UNTESTED
3882 print_exec(b->right, indent, bracket); // UNTESTED
3885 // block, one per line
3886 if (b->left == NULL)
3887 do_indent(indent, "pass\n");
3889 print_exec(b->left, indent, bracket);
3891 print_exec(b->right, indent, bracket);
3895 ###### propagate binode cases
3898 /* If any statement returns something other than Tnone
3899 * or Tbool then all such must return same type.
3900 * As each statement may be Tnone or something else,
3901 * we must always pass NULL (unknown) down, otherwise an incorrect
3902 * error might occur. We never return Tnone unless it is
3907 for (e = b; e; e = cast(binode, e->right)) {
3908 t = propagate_types(e->left, c, ok, NULL, rules);
3909 if ((rules & Rboolok) && (t == Tbool || t == Tnone))
3911 if (t == Tnone && e->right)
3912 /* Only the final statement *must* return a value
3920 type_err(c, "error: expected %1%r, found %2",
3921 e->left, type, rules, t);
3927 ###### interp binode cases
3929 while (rvtype == Tnone &&
3932 rv = interp_exec(c, b->left, &rvtype);
3933 b = cast(binode, b->right);
3937 ### The Print statement
3939 `print` is a simple statement that takes a comma-separated list of
3940 expressions and prints the values separated by spaces and terminated
3941 by a newline. No control of formatting is possible.
3943 `print` uses `ExpressionList` to collect the expressions and stores them
3944 on the left side of a `Print` binode unlessthere is a trailing comma
3945 when the list is stored on the `right` side and no trailing newline is
3951 ##### expr precedence
3954 ###### SimpleStatement Grammar
3956 | print ExpressionList ${
3960 $0->left = reorder_bilist($<EL);
3962 | print ExpressionList , ${ {
3965 $0->right = reorder_bilist($<EL);
3975 ###### print binode cases
3978 do_indent(indent, "print");
3980 print_exec(b->right, -1, bracket);
3983 print_exec(b->left, -1, bracket);
3988 ###### propagate binode cases
3991 /* don't care but all must be consistent */
3993 b = cast(binode, b->left);
3995 b = cast(binode, b->right);
3997 propagate_types(b->left, c, ok, NULL, Rnolabel);
3998 b = cast(binode, b->right);
4002 ###### interp binode cases
4006 struct binode *b2 = cast(binode, b->left);
4008 b2 = cast(binode, b->right);
4009 for (; b2; b2 = cast(binode, b2->right)) {
4010 left = interp_exec(c, b2->left, <ype);
4011 print_value(ltype, &left, stdout);
4012 free_value(ltype, &left);
4016 if (b->right == NULL)
4022 ###### Assignment statement
4024 An assignment will assign a value to a variable, providing it hasn't
4025 been declared as a constant. The analysis phase ensures that the type
4026 will be correct so the interpreter just needs to perform the
4027 calculation. There is a form of assignment which declares a new
4028 variable as well as assigning a value. If a name is assigned before
4029 it is declared, and error will be raised as the name is created as
4030 `Tlabel` and it is illegal to assign to such names.
4036 ###### declare terminals
4039 ###### SimpleStatement Grammar
4040 | Variable = Expression ${
4046 | VariableDecl = Expression ${
4054 if ($1->var->where_set == NULL) {
4056 "Variable declared with no type or value: %v",
4067 ###### print binode cases
4070 do_indent(indent, "");
4071 print_exec(b->left, indent, bracket);
4073 print_exec(b->right, indent, bracket);
4080 struct variable *v = cast(var, b->left)->var;
4081 do_indent(indent, "");
4082 print_exec(b->left, indent, bracket);
4083 if (cast(var, b->left)->var->constant) {
4085 if (v->explicit_type) {
4086 type_print(v->type, stdout);
4091 if (v->explicit_type) {
4092 type_print(v->type, stdout);
4098 print_exec(b->right, indent, bracket);
4105 ###### propagate binode cases
4109 /* Both must match and not be labels,
4110 * Type must support 'dup',
4111 * For Assign, left must not be constant.
4114 t = propagate_types(b->left, c, ok, NULL,
4115 Rnolabel | (b->op == Assign ? Rnoconstant : 0));
4120 if (propagate_types(b->right, c, ok, t, 0) != t)
4121 if (b->left->type == Xvar)
4122 type_err(c, "info: variable '%v' was set as %1 here.",
4123 cast(var, b->left)->var->where_set, t, rules, NULL);
4125 t = propagate_types(b->right, c, ok, NULL, Rnolabel);
4127 propagate_types(b->left, c, ok, t,
4128 (b->op == Assign ? Rnoconstant : 0));
4130 if (t && t->dup == NULL && t->name.txt[0] != ' ') // HACK
4131 type_err(c, "error: cannot assign value of type %1", b, t, 0, NULL);
4136 ###### interp binode cases
4139 lleft = linterp_exec(c, b->left, <ype);
4141 dinterp_exec(c, b->right, lleft, ltype, 1);
4147 struct variable *v = cast(var, b->left)->var;
4150 val = var_value(c, v);
4151 if (v->type->prepare_type)
4152 v->type->prepare_type(c, v->type, 0);
4154 dinterp_exec(c, b->right, val, v->type, 0);
4156 val_init(v->type, val);
4160 ### The `use` statement
4162 The `use` statement is the last "simple" statement. It is needed when a
4163 statement block can return a value. This includes the body of a
4164 function which has a return type, and the "condition" code blocks in
4165 `if`, `while`, and `switch` statements.
4170 ###### expr precedence
4173 ###### SimpleStatement Grammar
4175 $0 = new_pos(binode, $1);
4178 if ($0->right->type == Xvar) {
4179 struct var *v = cast(var, $0->right);
4180 if (v->var->type == Tnone) {
4181 /* Convert this to a label */
4184 v->var->type = Tlabel;
4185 val = global_alloc(c, Tlabel, v->var, NULL);
4191 ###### print binode cases
4194 do_indent(indent, "use ");
4195 print_exec(b->right, -1, bracket);
4200 ###### propagate binode cases
4203 /* result matches value */
4204 return propagate_types(b->right, c, ok, type, 0);
4206 ###### interp binode cases
4209 rv = interp_exec(c, b->right, &rvtype);
4212 ### The Conditional Statement
4214 This is the biggy and currently the only complex statement. This
4215 subsumes `if`, `while`, `do/while`, `switch`, and some parts of `for`.
4216 It is comprised of a number of parts, all of which are optional though
4217 set combinations apply. Each part is (usually) a key word (`then` is
4218 sometimes optional) followed by either an expression or a code block,
4219 except the `casepart` which is a "key word and an expression" followed
4220 by a code block. The code-block option is valid for all parts and,
4221 where an expression is also allowed, the code block can use the `use`
4222 statement to report a value. If the code block does not report a value
4223 the effect is similar to reporting `True`.
4225 The `else` and `case` parts, as well as `then` when combined with
4226 `if`, can contain a `use` statement which will apply to some
4227 containing conditional statement. `for` parts, `do` parts and `then`
4228 parts used with `for` can never contain a `use`, except in some
4229 subordinate conditional statement.
4231 If there is a `forpart`, it is executed first, only once.
4232 If there is a `dopart`, then it is executed repeatedly providing
4233 always that the `condpart` or `cond`, if present, does not return a non-True
4234 value. `condpart` can fail to return any value if it simply executes
4235 to completion. This is treated the same as returning `True`.
4237 If there is a `thenpart` it will be executed whenever the `condpart`
4238 or `cond` returns True (or does not return any value), but this will happen
4239 *after* `dopart` (when present).
4241 If `elsepart` is present it will be executed at most once when the
4242 condition returns `False` or some value that isn't `True` and isn't
4243 matched by any `casepart`. If there are any `casepart`s, they will be
4244 executed when the condition returns a matching value.
4246 The particular sorts of values allowed in case parts has not yet been
4247 determined in the language design, so nothing is prohibited.
4249 The various blocks in this complex statement potentially provide scope
4250 for variables as described earlier. Each such block must include the
4251 "OpenScope" nonterminal before parsing the block, and must call
4252 `var_block_close()` when closing the block.
4254 The code following "`if`", "`switch`" and "`for`" does not get its own
4255 scope, but is in a scope covering the whole statement, so names
4256 declared there cannot be redeclared elsewhere. Similarly the
4257 condition following "`while`" is in a scope the covers the body
4258 ("`do`" part) of the loop, and which does not allow conditional scope
4259 extension. Code following "`then`" (both looping and non-looping),
4260 "`else`" and "`case`" each get their own local scope.
4262 The type requirements on the code block in a `whilepart` are quite
4263 unusal. It is allowed to return a value of some identifiable type, in
4264 which case the loop aborts and an appropriate `casepart` is run, or it
4265 can return a Boolean, in which case the loop either continues to the
4266 `dopart` (on `True`) or aborts and runs the `elsepart` (on `False`).
4267 This is different both from the `ifpart` code block which is expected to
4268 return a Boolean, or the `switchpart` code block which is expected to
4269 return the same type as the casepart values. The correct analysis of
4270 the type of the `whilepart` code block is the reason for the
4271 `Rboolok` flag which is passed to `propagate_types()`.
4273 The `cond_statement` cannot fit into a `binode` so a new `exec` is
4274 defined. As there are two scopes which cover multiple parts - one for
4275 the whole statement and one for "while" and "do" - and as we will use
4276 the 'struct exec' to track scopes, we actually need two new types of
4277 exec. One is a `binode` for the looping part, the rest is the
4278 `cond_statement`. The `cond_statement` will use an auxilliary `struct
4279 casepart` to track a list of case parts.
4290 struct exec *action;
4291 struct casepart *next;
4293 struct cond_statement {
4295 struct exec *forpart, *condpart, *thenpart, *elsepart;
4296 struct binode *looppart;
4297 struct casepart *casepart;
4300 ###### ast functions
4302 static void free_casepart(struct casepart *cp)
4306 free_exec(cp->value);
4307 free_exec(cp->action);
4314 static void free_cond_statement(struct cond_statement *s)
4318 free_exec(s->forpart);
4319 free_exec(s->condpart);
4320 free_exec(s->looppart);
4321 free_exec(s->thenpart);
4322 free_exec(s->elsepart);
4323 free_casepart(s->casepart);
4327 ###### free exec cases
4328 case Xcond_statement: free_cond_statement(cast(cond_statement, e)); break;
4330 ###### ComplexStatement Grammar
4331 | CondStatement ${ $0 = $<1; }$
4333 ###### expr precedence
4334 $TERM for then while do
4341 // A CondStatement must end with EOL, as does CondSuffix and
4343 // ForPart, ThenPart, SwitchPart, CasePart are non-empty and
4344 // may or may not end with EOL
4345 // WhilePart and IfPart include an appropriate Suffix
4347 // ForPart, SwitchPart, and IfPart open scopes, o we have to close
4348 // them. WhilePart opens and closes its own scope.
4349 CondStatement -> ForPart OptNL ThenPart OptNL WhilePart CondSuffix ${
4352 $0->thenpart = $<TP;
4353 $0->looppart = $<WP;
4354 var_block_close(c, CloseSequential, $0);
4356 | ForPart OptNL WhilePart CondSuffix ${
4359 $0->looppart = $<WP;
4360 var_block_close(c, CloseSequential, $0);
4362 | WhilePart CondSuffix ${
4364 $0->looppart = $<WP;
4366 | SwitchPart OptNL CasePart CondSuffix ${
4368 $0->condpart = $<SP;
4369 $CP->next = $0->casepart;
4370 $0->casepart = $<CP;
4371 var_block_close(c, CloseSequential, $0);
4373 | SwitchPart : IN OptNL CasePart CondSuffix OUT Newlines ${
4375 $0->condpart = $<SP;
4376 $CP->next = $0->casepart;
4377 $0->casepart = $<CP;
4378 var_block_close(c, CloseSequential, $0);
4380 | IfPart IfSuffix ${
4382 $0->condpart = $IP.condpart; $IP.condpart = NULL;
4383 $0->thenpart = $IP.thenpart; $IP.thenpart = NULL;
4384 // This is where we close an "if" statement
4385 var_block_close(c, CloseSequential, $0);
4388 CondSuffix -> IfSuffix ${
4391 | Newlines CasePart CondSuffix ${
4393 $CP->next = $0->casepart;
4394 $0->casepart = $<CP;
4396 | CasePart CondSuffix ${
4398 $CP->next = $0->casepart;
4399 $0->casepart = $<CP;
4402 IfSuffix -> Newlines ${ $0 = new(cond_statement); }$
4403 | Newlines ElsePart ${ $0 = $<EP; }$
4404 | ElsePart ${$0 = $<EP; }$
4406 ElsePart -> else OpenBlock Newlines ${
4407 $0 = new(cond_statement);
4408 $0->elsepart = $<OB;
4409 var_block_close(c, CloseElse, $0->elsepart);
4411 | else OpenScope CondStatement ${
4412 $0 = new(cond_statement);
4413 $0->elsepart = $<CS;
4414 var_block_close(c, CloseElse, $0->elsepart);
4418 CasePart -> case Expression OpenScope ColonBlock ${
4419 $0 = calloc(1,sizeof(struct casepart));
4422 var_block_close(c, CloseParallel, $0->action);
4426 // These scopes are closed in CondStatement
4427 ForPart -> for OpenBlock ${
4431 ThenPart -> then OpenBlock ${
4433 var_block_close(c, CloseSequential, $0);
4437 // This scope is closed in CondStatement
4438 WhilePart -> while UseBlock OptNL do OpenBlock ${
4443 var_block_close(c, CloseSequential, $0->right);
4444 var_block_close(c, CloseSequential, $0);
4446 | while OpenScope Expression OpenScope ColonBlock ${
4451 var_block_close(c, CloseSequential, $0->right);
4452 var_block_close(c, CloseSequential, $0);
4456 IfPart -> if UseBlock OptNL then OpenBlock ${
4459 var_block_close(c, CloseParallel, $0.thenpart);
4461 | if OpenScope Expression OpenScope ColonBlock ${
4464 var_block_close(c, CloseParallel, $0.thenpart);
4466 | if OpenScope Expression OpenScope OptNL then Block ${
4469 var_block_close(c, CloseParallel, $0.thenpart);
4473 // This scope is closed in CondStatement
4474 SwitchPart -> switch OpenScope Expression ${
4477 | switch UseBlock ${
4481 ###### print binode cases
4483 if (b->left && b->left->type == Xbinode &&
4484 cast(binode, b->left)->op == Block) {
4486 do_indent(indent, "while {\n");
4488 do_indent(indent, "while\n");
4489 print_exec(b->left, indent+1, bracket);
4491 do_indent(indent, "} do {\n");
4493 do_indent(indent, "do\n");
4494 print_exec(b->right, indent+1, bracket);
4496 do_indent(indent, "}\n");
4498 do_indent(indent, "while ");
4499 print_exec(b->left, 0, bracket);
4504 print_exec(b->right, indent+1, bracket);
4506 do_indent(indent, "}\n");
4510 ###### print exec cases
4512 case Xcond_statement:
4514 struct cond_statement *cs = cast(cond_statement, e);
4515 struct casepart *cp;
4517 do_indent(indent, "for");
4518 if (bracket) printf(" {\n"); else printf("\n");
4519 print_exec(cs->forpart, indent+1, bracket);
4522 do_indent(indent, "} then {\n");
4524 do_indent(indent, "then\n");
4525 print_exec(cs->thenpart, indent+1, bracket);
4527 if (bracket) do_indent(indent, "}\n");
4530 print_exec(cs->looppart, indent, bracket);
4534 do_indent(indent, "switch");
4536 do_indent(indent, "if");
4537 if (cs->condpart && cs->condpart->type == Xbinode &&
4538 cast(binode, cs->condpart)->op == Block) {
4543 print_exec(cs->condpart, indent+1, bracket);
4545 do_indent(indent, "}\n");
4547 do_indent(indent, "then\n");
4548 print_exec(cs->thenpart, indent+1, bracket);
4552 print_exec(cs->condpart, 0, bracket);
4558 print_exec(cs->thenpart, indent+1, bracket);
4560 do_indent(indent, "}\n");
4565 for (cp = cs->casepart; cp; cp = cp->next) {
4566 do_indent(indent, "case ");
4567 print_exec(cp->value, -1, 0);
4572 print_exec(cp->action, indent+1, bracket);
4574 do_indent(indent, "}\n");
4577 do_indent(indent, "else");
4582 print_exec(cs->elsepart, indent+1, bracket);
4584 do_indent(indent, "}\n");
4589 ###### propagate binode cases
4591 t = propagate_types(b->right, c, ok, Tnone, 0);
4592 if (!type_compat(Tnone, t, 0))
4593 *ok = 0; // UNTESTED
4594 return propagate_types(b->left, c, ok, type, rules);
4596 ###### propagate exec cases
4597 case Xcond_statement:
4599 // forpart and looppart->right must return Tnone
4600 // thenpart must return Tnone if there is a loopart,
4601 // otherwise it is like elsepart.
4603 // be bool if there is no casepart
4604 // match casepart->values if there is a switchpart
4605 // either be bool or match casepart->value if there
4607 // elsepart and casepart->action must match the return type
4608 // expected of this statement.
4609 struct cond_statement *cs = cast(cond_statement, prog);
4610 struct casepart *cp;
4612 t = propagate_types(cs->forpart, c, ok, Tnone, 0);
4613 if (!type_compat(Tnone, t, 0))
4614 *ok = 0; // UNTESTED
4617 t = propagate_types(cs->thenpart, c, ok, Tnone, 0);
4618 if (!type_compat(Tnone, t, 0))
4619 *ok = 0; // UNTESTED
4621 if (cs->casepart == NULL) {
4622 propagate_types(cs->condpart, c, ok, Tbool, 0);
4623 propagate_types(cs->looppart, c, ok, Tbool, 0);
4625 /* Condpart must match case values, with bool permitted */
4627 for (cp = cs->casepart;
4628 cp && !t; cp = cp->next)
4629 t = propagate_types(cp->value, c, ok, NULL, 0);
4630 if (!t && cs->condpart)
4631 t = propagate_types(cs->condpart, c, ok, NULL, Rboolok); // UNTESTED
4632 if (!t && cs->looppart)
4633 t = propagate_types(cs->looppart, c, ok, NULL, Rboolok); // UNTESTED
4634 // Now we have a type (I hope) push it down
4636 for (cp = cs->casepart; cp; cp = cp->next)
4637 propagate_types(cp->value, c, ok, t, 0);
4638 propagate_types(cs->condpart, c, ok, t, Rboolok);
4639 propagate_types(cs->looppart, c, ok, t, Rboolok);
4642 // (if)then, else, and case parts must return expected type.
4643 if (!cs->looppart && !type)
4644 type = propagate_types(cs->thenpart, c, ok, NULL, rules);
4646 type = propagate_types(cs->elsepart, c, ok, NULL, rules);
4647 for (cp = cs->casepart;
4649 cp = cp->next) // UNTESTED
4650 type = propagate_types(cp->action, c, ok, NULL, rules); // UNTESTED
4653 propagate_types(cs->thenpart, c, ok, type, rules);
4654 propagate_types(cs->elsepart, c, ok, type, rules);
4655 for (cp = cs->casepart; cp ; cp = cp->next)
4656 propagate_types(cp->action, c, ok, type, rules);
4662 ###### interp binode cases
4664 // This just performs one iterration of the loop
4665 rv = interp_exec(c, b->left, &rvtype);
4666 if (rvtype == Tnone ||
4667 (rvtype == Tbool && rv.bool != 0))
4668 // rvtype is Tnone or Tbool, doesn't need to be freed
4669 interp_exec(c, b->right, NULL);
4672 ###### interp exec cases
4673 case Xcond_statement:
4675 struct value v, cnd;
4676 struct type *vtype, *cndtype;
4677 struct casepart *cp;
4678 struct cond_statement *cs = cast(cond_statement, e);
4681 interp_exec(c, cs->forpart, NULL);
4683 while ((cnd = interp_exec(c, cs->looppart, &cndtype)),
4684 cndtype == Tnone || (cndtype == Tbool && cnd.bool != 0))
4685 interp_exec(c, cs->thenpart, NULL);
4687 cnd = interp_exec(c, cs->condpart, &cndtype);
4688 if ((cndtype == Tnone ||
4689 (cndtype == Tbool && cnd.bool != 0))) {
4690 // cnd is Tnone or Tbool, doesn't need to be freed
4691 rv = interp_exec(c, cs->thenpart, &rvtype);
4692 // skip else (and cases)
4696 for (cp = cs->casepart; cp; cp = cp->next) {
4697 v = interp_exec(c, cp->value, &vtype);
4698 if (value_cmp(cndtype, vtype, &v, &cnd) == 0) {
4699 free_value(vtype, &v);
4700 free_value(cndtype, &cnd);
4701 rv = interp_exec(c, cp->action, &rvtype);
4704 free_value(vtype, &v);
4706 free_value(cndtype, &cnd);
4708 rv = interp_exec(c, cs->elsepart, &rvtype);
4715 ### Top level structure
4717 All the language elements so far can be used in various places. Now
4718 it is time to clarify what those places are.
4720 At the top level of a file there will be a number of declarations.
4721 Many of the things that can be declared haven't been described yet,
4722 such as functions, procedures, imports, and probably more.
4723 For now there are two sorts of things that can appear at the top
4724 level. They are predefined constants, `struct` types, and the `main`
4725 function. While the syntax will allow the `main` function to appear
4726 multiple times, that will trigger an error if it is actually attempted.
4728 The various declarations do not return anything. They store the
4729 various declarations in the parse context.
4731 ###### Parser: grammar
4734 Ocean -> OptNL DeclarationList
4736 ## declare terminals
4743 DeclarationList -> Declaration
4744 | DeclarationList Declaration
4746 Declaration -> ERROR Newlines ${
4747 tok_err(c, // UNTESTED
4748 "error: unhandled parse error", &$1);
4754 ## top level grammar
4758 ### The `const` section
4760 As well as being defined in with the code that uses them, constants
4761 can be declared at the top level. These have full-file scope, so they
4762 are always `InScope`. The value of a top level constant can be given
4763 as an expression, and this is evaluated immediately rather than in the
4764 later interpretation stage. Once we add functions to the language, we
4765 will need rules concern which, if any, can be used to define a top
4768 Constants are defined in a section that starts with the reserved word
4769 `const` and then has a block with a list of assignment statements.
4770 For syntactic consistency, these must use the double-colon syntax to
4771 make it clear that they are constants. Type can also be given: if
4772 not, the type will be determined during analysis, as with other
4775 As the types constants are inserted at the head of a list, printing
4776 them in the same order that they were read is not straight forward.
4777 We take a quadratic approach here and count the number of constants
4778 (variables of depth 0), then count down from there, each time
4779 searching through for the Nth constant for decreasing N.
4781 ###### top level grammar
4785 DeclareConstant -> const { IN OptNL ConstList OUT OptNL } Newlines
4786 | const { SimpleConstList } Newlines
4787 | const IN OptNL ConstList OUT Newlines
4788 | const SimpleConstList Newlines
4790 ConstList -> ConstList SimpleConstLine
4792 SimpleConstList -> SimpleConstList ; Const
4795 SimpleConstLine -> SimpleConstList Newlines
4796 | ERROR Newlines ${ tok_err(c, "Syntax error in constant", &$1); }$
4799 CType -> Type ${ $0 = $<1; }$
4802 Const -> IDENTIFIER :: CType = Expression ${ {
4806 v = var_decl(c, $1.txt);
4808 struct var *var = new_pos(var, $1);
4809 v->where_decl = var;
4815 struct variable *vorig = var_ref(c, $1.txt);
4816 tok_err(c, "error: name already declared", &$1);
4817 type_err(c, "info: this is where '%v' was first declared",
4818 vorig->where_decl, NULL, 0, NULL);
4822 propagate_types($5, c, &ok, $3, 0);
4827 struct value res = interp_exec(c, $5, &v->type);
4828 global_alloc(c, v->type, v, &res);
4832 ###### print const decls
4837 while (target != 0) {
4839 for (v = context.in_scope; v; v=v->in_scope)
4840 if (v->depth == 0 && v->constant) {
4851 struct value *val = var_value(&context, v);
4852 printf(" %.*s :: ", v->name->name.len, v->name->name.txt);
4853 type_print(v->type, stdout);
4855 if (v->type == Tstr)
4857 print_value(v->type, val, stdout);
4858 if (v->type == Tstr)
4866 ### Function declarations
4868 The code in an Ocean program is all stored in function declarations.
4869 One of the functions must be named `main` and it must accept an array of
4870 strings as a parameter - the command line arguments.
4872 As this is the top level, several things are handled a bit differently.
4873 The function is not interpreted by `interp_exec` as that isn't passed
4874 the argument list which the program requires. Similarly type analysis
4875 is a bit more interesting at this level.
4877 ###### ast functions
4879 static struct type *handle_results(struct parse_context *c,
4880 struct binode *results)
4882 /* Create a 'struct' type from the results list, which
4883 * is a list for 'struct var'
4885 struct text result_type_name = { " function_result", 5 };
4886 struct type *t = add_type(c, result_type_name, &structure_prototype);
4890 for (b = results; b; b = cast(binode, b->right))
4892 t->structure.nfields = cnt;
4893 t->structure.fields = calloc(cnt, sizeof(struct field));
4895 for (b = results; b; b = cast(binode, b->right)) {
4896 struct var *v = cast(var, b->left);
4897 struct field *f = &t->structure.fields[cnt++];
4898 int a = v->var->type->align;
4899 f->name = v->var->name->name;
4900 f->type = v->var->type;
4902 f->offset = t->size;
4903 v->var->frame_pos = f->offset;
4904 t->size += ((f->type->size - 1) | (a-1)) + 1;
4907 variable_unlink_exec(v->var);
4909 free_binode(results);
4913 static struct variable *declare_function(struct parse_context *c,
4914 struct variable *name,
4915 struct binode *args,
4917 struct binode *results,
4920 struct text funcname = {" func", 5};
4922 struct value fn = {.function = code};
4924 var_block_close(c, CloseFunction, code);
4925 t = add_type(c, funcname, &function_prototype);
4927 t->function.params = reorder_bilist(args);
4929 ret = handle_results(c, reorder_bilist(results));
4930 t->function.inline_result = 1;
4931 t->function.local_size = ret->size;
4933 t->function.return_type = ret;
4934 global_alloc(c, t, name, &fn);
4935 name->type->function.scope = c->out_scope;
4940 var_block_close(c, CloseFunction, NULL);
4942 c->out_scope = NULL;
4946 ###### declare terminals
4949 ###### top level grammar
4952 DeclareFunction -> func FuncName ( OpenScope ArgsLine ) Block Newlines ${
4953 $0 = declare_function(c, $<FN, $<Ar, Tnone, NULL, $<Bl);
4955 | func FuncName IN OpenScope Args OUT OptNL do Block Newlines ${
4956 $0 = declare_function(c, $<FN, $<Ar, Tnone, NULL, $<Bl);
4958 | func FuncName NEWLINE OpenScope OptNL do Block Newlines ${
4959 $0 = declare_function(c, $<FN, NULL, Tnone, NULL, $<Bl);
4961 | func FuncName ( OpenScope ArgsLine ) : Type Block Newlines ${
4962 $0 = declare_function(c, $<FN, $<Ar, $<Ty, NULL, $<Bl);
4964 | func FuncName ( OpenScope ArgsLine ) : ( ArgsLine ) Block Newlines ${
4965 $0 = declare_function(c, $<FN, $<AL, NULL, $<AL2, $<Bl);
4967 | func FuncName IN OpenScope Args OUT OptNL return Type Newlines do Block Newlines ${
4968 $0 = declare_function(c, $<FN, $<Ar, $<Ty, NULL, $<Bl);
4970 | func FuncName NEWLINE OpenScope return Type Newlines do Block Newlines ${
4971 $0 = declare_function(c, $<FN, NULL, $<Ty, NULL, $<Bl);
4973 | func FuncName IN OpenScope Args OUT OptNL return IN Args OUT OptNL do Block Newlines ${
4974 $0 = declare_function(c, $<FN, $<Ar, NULL, $<Ar2, $<Bl);
4976 | func FuncName NEWLINE OpenScope return IN Args OUT OptNL do Block Newlines ${
4977 $0 = declare_function(c, $<FN, NULL, NULL, $<Ar, $<Bl);
4980 ###### print func decls
4985 while (target != 0) {
4987 for (v = context.in_scope; v; v=v->in_scope)
4988 if (v->depth == 0 && v->type && v->type->check_args) {
4997 struct value *val = var_value(&context, v);
4998 printf("func %.*s", v->name->name.len, v->name->name.txt);
4999 v->type->print_type_decl(v->type, stdout);
5001 print_exec(val->function, 0, brackets);
5003 print_value(v->type, val, stdout);
5004 printf("/* frame size %d */\n", v->type->function.local_size);
5010 ###### core functions
5012 static int analyse_funcs(struct parse_context *c)
5016 for (v = c->in_scope; v; v = v->in_scope) {
5020 if (v->depth != 0 || !v->type || !v->type->check_args)
5022 ret = v->type->function.inline_result ?
5023 Tnone : v->type->function.return_type;
5024 val = var_value(c, v);
5027 propagate_types(val->function, c, &ok, ret, 0);
5030 /* Make sure everything is still consistent */
5031 propagate_types(val->function, c, &ok, ret, 0);
5034 if (!v->type->function.inline_result &&
5035 !v->type->function.return_type->dup) {
5036 type_err(c, "error: function cannot return value of type %1",
5037 v->where_decl, v->type->function.return_type, 0, NULL);
5040 scope_finalize(c, v->type);
5045 static int analyse_main(struct type *type, struct parse_context *c)
5047 struct binode *bp = type->function.params;
5051 struct type *argv_type;
5052 struct text argv_type_name = { " argv", 5 };
5054 argv_type = add_type(c, argv_type_name, &array_prototype);
5055 argv_type->array.member = Tstr;
5056 argv_type->array.unspec = 1;
5058 for (b = bp; b; b = cast(binode, b->right)) {
5062 propagate_types(b->left, c, &ok, argv_type, 0);
5064 default: /* invalid */ // NOTEST
5065 propagate_types(b->left, c, &ok, Tnone, 0); // NOTEST
5071 return !c->parse_error;
5074 static void interp_main(struct parse_context *c, int argc, char **argv)
5076 struct value *progp = NULL;
5077 struct text main_name = { "main", 4 };
5078 struct variable *mainv;
5084 mainv = var_ref(c, main_name);
5086 progp = var_value(c, mainv);
5087 if (!progp || !progp->function) {
5088 fprintf(stderr, "oceani: no main function found.\n");
5092 if (!analyse_main(mainv->type, c)) {
5093 fprintf(stderr, "oceani: main has wrong type.\n");
5097 al = mainv->type->function.params;
5099 c->local_size = mainv->type->function.local_size;
5100 c->local = calloc(1, c->local_size);
5102 struct var *v = cast(var, al->left);
5103 struct value *vl = var_value(c, v->var);
5113 mpq_set_ui(argcq, argc, 1);
5114 memcpy(var_value(c, t->array.vsize), &argcq, sizeof(argcq));
5115 t->prepare_type(c, t, 0);
5116 array_init(v->var->type, vl);
5117 for (i = 0; i < argc; i++) {
5118 struct value *vl2 = vl->array + i * v->var->type->array.member->size;
5120 arg.str.txt = argv[i];
5121 arg.str.len = strlen(argv[i]);
5122 free_value(Tstr, vl2);
5123 dup_value(Tstr, &arg, vl2);
5127 al = cast(binode, al->right);
5129 v = interp_exec(c, progp->function, &vtype);
5130 free_value(vtype, &v);
5135 ###### ast functions
5136 void free_variable(struct variable *v)
5140 ## And now to test it out.
5142 Having a language requires having a "hello world" program. I'll
5143 provide a little more than that: a program that prints "Hello world"
5144 finds the GCD of two numbers, prints the first few elements of
5145 Fibonacci, performs a binary search for a number, and a few other
5146 things which will likely grow as the languages grows.
5148 ###### File: oceani.mk
5151 @echo "===== DEMO ====="
5152 ./oceani --section "demo: hello" oceani.mdc 55 33
5158 four ::= 2 + 2 ; five ::= 10/2
5159 const pie ::= "I like Pie";
5160 cake ::= "The cake is"
5168 func main(argv:[argc::]string)
5169 print "Hello World, what lovely oceans you have!"
5170 print "Are there", five, "?"
5171 print pi, pie, "but", cake
5173 A := $argv[1]; B := $argv[2]
5175 /* When a variable is defined in both branches of an 'if',
5176 * and used afterwards, the variables are merged.
5182 print "Is", A, "bigger than", B,"? ", bigger
5183 /* If a variable is not used after the 'if', no
5184 * merge happens, so types can be different
5187 double:string = "yes"
5188 print A, "is more than twice", B, "?", double
5191 print "double", B, "is", double
5196 if a > 0 and then b > 0:
5202 print "GCD of", A, "and", B,"is", a
5204 print a, "is not positive, cannot calculate GCD"
5206 print b, "is not positive, cannot calculate GCD"
5211 print "Fibonacci:", f1,f2,
5212 then togo = togo - 1
5220 /* Binary search... */
5225 mid := (lo + hi) / 2
5238 print "Yay, I found", target
5240 print "Closest I found was", lo
5245 // "middle square" PRNG. Not particularly good, but one my
5246 // Dad taught me - the first one I ever heard of.
5247 for i:=1; then i = i + 1; while i < size:
5248 n := list[i-1] * list[i-1]
5249 list[i] = (n / 100) % 10 000
5251 print "Before sort:",
5252 for i:=0; then i = i + 1; while i < size:
5256 for i := 1; then i=i+1; while i < size:
5257 for j:=i-1; then j=j-1; while j >= 0:
5258 if list[j] > list[j+1]:
5262 print " After sort:",
5263 for i:=0; then i = i + 1; while i < size:
5267 if 1 == 2 then print "yes"; else print "no"
5271 bob.alive = (bob.name == "Hello")
5272 print "bob", "is" if bob.alive else "isn't", "alive"